/[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 226 by ph10, Tue Aug 21 11:46:08 2007 UTC revision 1055 by chpe, Tue Oct 16 15:53:30 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 37  POSSIBILITY OF SUCH DAMAGE. Line 37  POSSIBILITY OF SUCH DAMAGE.
37  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
38  */  */
39    
   
40  /* This module contains pcre_exec(), the externally visible function that does  /* This module contains pcre_exec(), the externally visible function that does
41  pattern matching using an NFA algorithm, trying to mimic Perl as closely as  pattern matching using an NFA algorithm, trying to mimic Perl as closely as
42  possible. There are also some static supporting functions. */  possible. There are also some static supporting functions. */
43    
44  #ifdef HAVE_CONFIG_H  #ifdef HAVE_CONFIG_H
45  #include <config.h>  #include "config.h"
46  #endif  #endif
47    
48  #define NLBLOCK md             /* Block containing newline information */  #define NLBLOCK md             /* Block containing newline information */
# Line 57  possible. There are also some static sup Line 56  possible. There are also some static sup
56  #undef min  #undef min
57  #undef max  #undef max
58    
59  /* Flag bits for the match() function */  /* Values for setting in md->match_function_type to indicate two special types
60    of call to match(). We do it this way to save on using another stack variable,
61    as stack usage is to be discouraged. */
62    
63  #define match_condassert     0x01  /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
64  #define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
65    
66  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
67  defined PCRE_ERROR_xxx codes, which are all negative. */  defined PCRE_ERROR_xxx codes, which are all negative. */
# Line 71  defined PCRE_ERROR_xxx codes, which are Line 72  defined PCRE_ERROR_xxx codes, which are
72  /* Special internal returns from the match() function. Make them sufficiently  /* Special internal returns from the match() function. Make them sufficiently
73  negative to avoid the external error codes. */  negative to avoid the external error codes. */
74    
75  #define MATCH_COMMIT       (-999)  #define MATCH_ACCEPT       (-999)
76  #define MATCH_PRUNE        (-998)  #define MATCH_COMMIT       (-998)
77  #define MATCH_SKIP         (-997)  #define MATCH_KETRPOS      (-997)
78  #define MATCH_THEN         (-996)  #define MATCH_ONCE         (-996)
79    #define MATCH_PRUNE        (-995)
80    #define MATCH_SKIP         (-994)
81    #define MATCH_SKIP_ARG     (-993)
82    #define MATCH_THEN         (-992)
83    
84  /* Maximum number of ints of offset to save on the stack for recursive calls.  /* Maximum number of ints of offset to save on the stack for recursive calls.
85  If the offset vector is bigger, malloc is used. This should be a multiple of 3,  If the offset vector is bigger, malloc is used. This should be a multiple of 3,
# Line 89  static const char rep_max[] = { 0, 0, 0, Line 94  static const char rep_max[] = { 0, 0, 0,
94    
95    
96    
97  #ifdef DEBUG  #ifdef PCRE_DEBUG
98  /*************************************************  /*************************************************
99  *        Debugging function to print chars       *  *        Debugging function to print chars       *
100  *************************************************/  *************************************************/
# Line 107  Returns:     nothing Line 112  Returns:     nothing
112  */  */
113    
114  static void  static void
115  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)  pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md)
116  {  {
117  unsigned int c;  unsigned int c;
118  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
# Line 122  while (length-- > 0) Line 127  while (length-- > 0)
127  *          Match a back-reference                *  *          Match a back-reference                *
128  *************************************************/  *************************************************/
129    
130  /* If a back reference hasn't been set, the length that is passed is greater  /* Normally, if a back reference hasn't been set, the length that is passed is
131  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
132    the length passed is zero. Note that in caseless UTF-8 mode, the number of
133    subject bytes matched may be different to the number of reference bytes.
134    
135  Arguments:  Arguments:
136    offset      index into the offset vector    offset      index into the offset vector
137    eptr        points into the subject    eptr        pointer into the subject
138    length      length to be matched    length      length of reference to be matched (number of bytes)
139    md          points to match data block    md          points to match data block
140    ims         the ims flags    caseless    TRUE if caseless
141    
142  Returns:      TRUE if matched  Returns:      >= 0 the number of subject bytes matched
143                  -1 no match
144                  -2 partial match; always given if at end subject
145  */  */
146    
147  static BOOL  static int
148  match_ref(int offset, register USPTR eptr, int length, match_data *md,  match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
149    unsigned long int ims)    BOOL caseless)
150  {  {
151  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
152    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
153    
154  #ifdef DEBUG  #ifdef PCRE_DEBUG
155  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
156    printf("matching subject <null>");    printf("matching subject <null>");
157  else  else
# Line 154  pchars(p, length, FALSE, md); Line 164  pchars(p, length, FALSE, md);
164  printf("\n");  printf("\n");
165  #endif  #endif
166    
167  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible - in that
168    case the length is passed as zero). */
169    
170  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
171    
172  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
173    properly if Unicode properties are supported. Otherwise, we can check only
174    ASCII characters. */
175    
176  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
177    {    {
178    while (length-- > 0)  #ifdef SUPPORT_UTF
179      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
180      if (md->utf)
181        {
182        /* Match characters up to the end of the reference. NOTE: the number of
183        data units matched may differ, because in UTF-8 there are some characters
184        whose upper and lower case versions code have different numbers of bytes.
185        For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65
186        (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a
187        sequence of two of the latter. It is important, therefore, to check the
188        length along the reference, not along the subject (earlier code did this
189        wrong). */
190    
191        PCRE_PUCHAR endptr = p + length;
192        while (p < endptr)
193          {
194          unsigned int c, d;
195          const ucd_record *ur;
196          if (eptr >= md->end_subject) return -2;   /* Partial match */
197          GETCHARINC(c, eptr);
198          GETCHARINC(d, p);
199          ur = GET_UCD(d);
200          if (c != d && c != d + ur->other_case)
201            {
202            const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
203            for (;;)
204              {
205              if (c < *pp) return -1;
206              if (c == *pp++) break;
207              }
208            }
209          }
210        }
211      else
212    #endif
213    #endif
214    
215      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
216      is no UCP support. */
217        {
218        while (length-- > 0)
219          {
220          if (eptr >= md->end_subject) return -2;   /* Partial match */
221          if (TABLE_GET(*p, md->lcc, *p) != TABLE_GET(*eptr, md->lcc, *eptr)) return -1;
222          p++;
223          eptr++;
224          }
225        }
226    }    }
227    
228    /* In the caseful case, we can just compare the bytes, whether or not we
229    are in UTF-8 mode. */
230    
231  else  else
232    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
233      while (length-- > 0)
234        {
235        if (eptr >= md->end_subject) return -2;   /* Partial match */
236        if (*p++ != *eptr++) return -1;
237        }
238      }
239    
240  return TRUE;  return (int)(eptr - eptr_start);
241  }  }
242    
243    
# Line 219  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM Line 288  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM
288         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
289         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
290         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
291         RM51,  RM52, RM53, RM54 };         RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
292           RM61,  RM62, RM63, RM64, RM65, RM66 };
293    
294  /* These versions of the macros use the stack, as normal. There are debugging  /* These versions of the macros use the stack, as normal. There are debugging
295  versions and production versions. Note that the "rw" argument of RMATCH isn't  versions and production versions. Note that the "rw" argument of RMATCH isn't
296  actuall used in this definition. */  actually used in this definition. */
297    
298  #ifndef NO_RECURSE  #ifndef NO_RECURSE
299  #define REGISTER register  #define REGISTER register
300    
301  #ifdef DEBUG  #ifdef PCRE_DEBUG
302  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
303    { \    { \
304    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
305    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
306    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
307    }    }
308  #define RRETURN(ra) \  #define RRETURN(ra) \
# Line 241  actuall used in this definition. */ Line 311  actuall used in this definition. */
311    return ra; \    return ra; \
312    }    }
313  #else  #else
314  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
315    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
316  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
317  #endif  #endif
318    
# Line 255  argument of match(), which never changes Line 325  argument of match(), which never changes
325    
326  #define REGISTER  #define REGISTER
327    
328  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
329    {\    {\
330    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = frame->Xnextframe;\
331    frame->Xwhere = rw; \    if (newframe == NULL)\
332        {\
333        newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
334        if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
335        newframe->Xnextframe = NULL;\
336        frame->Xnextframe = newframe;\
337        }\
338      frame->Xwhere = rw;\
339    newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
340    newframe->Xecode = rb;\    newframe->Xecode = rb;\
341    newframe->Xmstart = mstart;\    newframe->Xmstart = mstart;\
342    newframe->Xoffset_top = rc;\    newframe->Xoffset_top = rc;\
343    newframe->Xims = re;\    newframe->Xeptrb = re;\
   newframe->Xeptrb = rf;\  
   newframe->Xflags = rg;\  
344    newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xrdepth = frame->Xrdepth + 1;\
345    newframe->Xprevframe = frame;\    newframe->Xprevframe = frame;\
346    frame = newframe;\    frame = newframe;\
# Line 277  argument of match(), which never changes Line 352  argument of match(), which never changes
352    
353  #define RRETURN(ra)\  #define RRETURN(ra)\
354    {\    {\
355    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
356    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
   (pcre_stack_free)(newframe);\  
357    if (frame != NULL)\    if (frame != NULL)\
358      {\      {\
359      rrc = ra;\      rrc = ra;\
# Line 293  argument of match(), which never changes Line 367  argument of match(), which never changes
367    
368  typedef struct heapframe {  typedef struct heapframe {
369    struct heapframe *Xprevframe;    struct heapframe *Xprevframe;
370      struct heapframe *Xnextframe;
371    
372    /* Function arguments that may change */    /* Function arguments that may change */
373    
374    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
375    const uschar *Xecode;    const pcre_uchar *Xecode;
376    const uschar *Xmstart;    PCRE_PUCHAR Xmstart;
377    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
378    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
379    unsigned int Xrdepth;    unsigned int Xrdepth;
380    
381    /* Function local variables */    /* Function local variables */
382    
383    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
384    const uschar *Xcharptr;  #ifdef SUPPORT_UTF
385    const uschar *Xdata;    PCRE_PUCHAR Xcharptr;
386    const uschar *Xnext;  #endif
387    const uschar *Xpp;    PCRE_PUCHAR Xdata;
388    const uschar *Xprev;    PCRE_PUCHAR Xnext;
389    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xpp;
390      PCRE_PUCHAR Xprev;
391      PCRE_PUCHAR Xsaved_eptr;
392    
393    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
394    
# Line 321  typedef struct heapframe { Line 396  typedef struct heapframe {
396    BOOL Xcondition;    BOOL Xcondition;
397    BOOL Xprev_is_word;    BOOL Xprev_is_word;
398    
   unsigned long int Xoriginal_ims;  
   
399  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
400    int Xprop_type;    int Xprop_type;
401    int Xprop_value;    int Xprop_value;
402    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
403    int Xoclength;    int Xoclength;
404    uschar Xocchars[8];    pcre_uchar Xocchars[6];
405  #endif  #endif
406    
407      int Xcodelink;
408    int Xctype;    int Xctype;
409    unsigned int Xfc;    unsigned int Xfc;
410    int Xfi;    int Xfi;
# Line 369  typedef struct heapframe { Line 440  typedef struct heapframe {
440    
441  /* This function is called recursively in many circumstances. Whenever it  /* This function is called recursively in many circumstances. Whenever it
442  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
443  same response.  same response. */
444    
445    /* These macros pack up tests that are used for partial matching, and which
446    appear several times in the code. We set the "hit end" flag if the pointer is
447    at the end of the subject and also past the start of the subject (i.e.
448    something has been matched). For hard partial matching, we then return
449    immediately. The second one is used when we already know we are past the end of
450    the subject. */
451    
452    #define CHECK_PARTIAL()\
453      if (md->partial != 0 && eptr >= md->end_subject && \
454          eptr > md->start_used_ptr) \
455        { \
456        md->hitend = TRUE; \
457        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
458        }
459    
460  Performance note: It might be tempting to extract commonly used fields from the  #define SCHECK_PARTIAL()\
461  md structure (e.g. utf8, end_subject) into individual variables to improve    if (md->partial != 0 && eptr > md->start_used_ptr) \
462        { \
463        md->hitend = TRUE; \
464        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
465        }
466    
467    
468    /* Performance note: It might be tempting to extract commonly used fields from
469    the md structure (e.g. utf, end_subject) into individual variables to improve
470  performance. Tests using gcc on a SPARC disproved this; in the first case, it  performance. Tests using gcc on a SPARC disproved this; in the first case, it
471  made performance worse.  made performance worse.
472    
# Line 383  Arguments: Line 477  Arguments:
477                   by encountering \K)                   by encountering \K)
478     offset_top  current top pointer     offset_top  current top pointer
479     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
480     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
481                   brackets - for testing for empty matches                   brackets - for testing for empty matches
    flags       can contain  
                  match_condassert - this is an assertion condition  
                  match_cbegroup - this is the start of an unlimited repeat  
                    group that can match an empty string  
482     rdepth      the recursion depth     rdepth      the recursion depth
483    
484  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
485                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
486                   a negative MATCH_xxx value for PRUNE, SKIP, etc
487                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
488                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
489  */  */
490    
491  static int  static int
492  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, const uschar *mstart,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
493    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
494    int flags, unsigned int rdepth)    unsigned int rdepth)
495  {  {
496  /* These variables do not need to be preserved over recursion in this function,  /* These variables do not need to be preserved over recursion in this function,
497  so they can be ordinary variables in all cases. Mark some of them with  so they can be ordinary variables in all cases. Mark some of them with
# Line 410  so they can be ordinary variables in all Line 500  so they can be ordinary variables in all
500  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
501  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
502  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register unsigned int c;   /* Character values not kept over RMATCH() calls */
503  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
504    
505  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
506    BOOL caseless;
507    int condcode;
508    
509  /* When recursion is not being used, all "local" variables that have to be  /* When recursion is not being used, all "local" variables that have to be
510  preserved over calls to RMATCH() are part of a "frame" which is obtained from  preserved over calls to RMATCH() are part of a "frame". We set up the top-level
511  heap storage. Set up the top-level frame here; others are obtained from the  frame on the stack here; subsequent instantiations are obtained from the heap
512  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
513    the top-level on the stack rather than malloc-ing them all gives a performance
514    boost in many cases where there is not much "recursion". */
515    
516  #ifdef NO_RECURSE  #ifdef NO_RECURSE
517  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)md->match_frames_base;
 frame->Xprevframe = NULL;            /* Marks the top level */  
518    
519  /* Copy in the original argument variables */  /* Copy in the original argument variables */
520    
# Line 429  frame->Xeptr = eptr; Line 522  frame->Xeptr = eptr;
522  frame->Xecode = ecode;  frame->Xecode = ecode;
523  frame->Xmstart = mstart;  frame->Xmstart = mstart;
524  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
525  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
526  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
527    
528  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 444  HEAP_RECURSE: Line 535  HEAP_RECURSE:
535  #define ecode              frame->Xecode  #define ecode              frame->Xecode
536  #define mstart             frame->Xmstart  #define mstart             frame->Xmstart
537  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
538  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
539  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
540    
541  /* Ditto for the local variables */  /* Ditto for the local variables */
542    
543  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
544  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
545  #endif  #endif
546  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
547    #define codelink           frame->Xcodelink
548  #define data               frame->Xdata  #define data               frame->Xdata
549  #define next               frame->Xnext  #define next               frame->Xnext
550  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 467  HEAP_RECURSE: Line 557  HEAP_RECURSE:
557  #define condition          frame->Xcondition  #define condition          frame->Xcondition
558  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
559    
 #define original_ims       frame->Xoriginal_ims  
   
560  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
561  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
562  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
563  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
 #define prop_category      frame->Xprop_category  
 #define prop_chartype      frame->Xprop_chartype  
 #define prop_script        frame->Xprop_script  
564  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
565  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
566  #endif  #endif
# Line 505  i, and fc and c, can be the same variabl Line 590  i, and fc and c, can be the same variabl
590  #define fi i  #define fi i
591  #define fc c  #define fc c
592    
593    /* Many of the following variables are used only in small blocks of the code.
594    My normal style of coding would have declared them within each of those blocks.
595    However, in order to accommodate the version of this code that uses an external
596    "stack" implemented on the heap, it is easier to declare them all here, so the
597    declarations can be cut out in a block. The only declarations within blocks
598    below are for variables that do not have to be preserved over a recursive call
599    to RMATCH(). */
600    
601  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  #ifdef SUPPORT_UTF
602  const uschar *charptr;             /* in small blocks of the code. My normal */  const pcre_uchar *charptr;
603  #endif                             /* style of coding would have declared    */  #endif
604  const uschar *callpat;             /* them within each of those blocks.      */  const pcre_uchar *callpat;
605  const uschar *data;                /* However, in order to accommodate the   */  const pcre_uchar *data;
606  const uschar *next;                /* version of this code that uses an      */  const pcre_uchar *next;
607  USPTR         pp;                  /* external "stack" implemented on the    */  PCRE_PUCHAR       pp;
608  const uschar *prev;                /* heap, it is easier to declare them all */  const pcre_uchar *prev;
609  USPTR         saved_eptr;          /* here, so the declarations can be cut   */  PCRE_PUCHAR       saved_eptr;
610                                     /* out in a block. The only declarations  */  
611  recursion_info new_recursive;      /* within blocks below are for variables  */  recursion_info new_recursive;
612                                     /* that do not have to be preserved over  */  
613  BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  BOOL cur_is_word;
614  BOOL condition;  BOOL condition;
615  BOOL prev_is_word;  BOOL prev_is_word;
616    
 unsigned long int original_ims;  
   
617  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
618  int prop_type;  int prop_type;
619  int prop_value;  int prop_value;
620  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
621  int oclength;  int oclength;
622  uschar occhars[8];  pcre_uchar occhars[6];
623  #endif  #endif
624    
625    int codelink;
626  int ctype;  int ctype;
627  int length;  int length;
628  int max;  int max;
# Line 547  int save_offset1, save_offset2, save_off Line 635  int save_offset1, save_offset2, save_off
635  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
636    
637  eptrblock newptrb;  eptrblock newptrb;
638    
639    /* There is a special fudge for calling match() in a way that causes it to
640    measure the size of its basic stack frame when the stack is being used for
641    recursion. The second argument (ecode) being NULL triggers this behaviour. It
642    cannot normally ever be NULL. The return is the negated value of the frame
643    size. */
644    
645    if (ecode == NULL)
646      {
647      if (rdepth == 0)
648        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
649      else
650        {
651        int len = (char *)&rdepth - (char *)eptr;
652        return (len > 0)? -len : len;
653        }
654      }
655  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
656    
657    /* To save space on the stack and in the heap frame, I have doubled up on some
658    of the local variables that are used only in localised parts of the code, but
659    still need to be preserved over recursive calls of match(). These macros define
660    the alternative names that are used. */
661    
662    #define allow_zero    cur_is_word
663    #define cbegroup      condition
664    #define code_offset   codelink
665    #define condassert    condition
666    #define matched_once  prev_is_word
667    #define foc           number
668    #define save_mark     data
669    
670  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
671  variables. */  variables. */
672    
# Line 568  TAIL_RECURSE: Line 686  TAIL_RECURSE:
686  /* 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
687  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
688  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()
689  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
690  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
691  complicated macro. It has to be used in one particular way. This shouldn't,  complicated macro. It has to be used in one particular way. This shouldn't,
692  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
693    
694  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
695  utf8 = md->utf8;       /* Local copy of the flag */  utf = md->utf;       /* Local copy of the flag */
696  #else  #else
697  utf8 = FALSE;  utf = FALSE;
698  #endif  #endif
699    
700  /* First check that we haven't called match() too many times, or that we  /* First check that we haven't called match() too many times, or that we
# Line 585  haven't exceeded the recursive call limi Line 703  haven't exceeded the recursive call limi
703  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
704  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
705    
 original_ims = ims;    /* Save for resetting on ')' */  
   
706  /* At the start of a group with an unlimited repeat that may match an empty  /* At the start of a group with an unlimited repeat that may match an empty
707  string, the match_cbegroup flag is set. When this is the case, add the current  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
708  subject pointer to the chain of such remembered pointers, to be checked when we  done this way to save having to use another function argument, which would take
709  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
710  When match() is called in other circumstances, don't add to the chain. The  
711  match_cbegroup flag must NOT be used with tail recursion, because the memory  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
712  block that is used is on the stack, so a new one may be required for each  such remembered pointers, to be checked when we hit the closing ket, in order
713  match(). */  to break infinite loops that match no characters. When match() is called in
714    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
715    NOT be used with tail recursion, because the memory block that is used is on
716    the stack, so a new one may be required for each match(). */
717    
718  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
719    {    {
720    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
721    newptrb.epb_prev = eptrb;    newptrb.epb_prev = eptrb;
722    eptrb = &newptrb;    eptrb = &newptrb;
723      md->match_function_type = 0;
724    }    }
725    
726  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 610  for (;;) Line 730  for (;;)
730    minimize = possessive = FALSE;    minimize = possessive = FALSE;
731    op = *ecode;    op = *ecode;
732    
   /* 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;  
   
733    switch(op)    switch(op)
734      {      {
735        case OP_MARK:
736        md->nomatch_mark = ecode + 2;
737        md->mark = NULL;    /* In case previously set by assertion */
738        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
739          eptrb, RM55);
740        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
741             md->mark == NULL) md->mark = ecode + 2;
742    
743        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
744        argument, and we must check whether that argument matches this MARK's
745        argument. It is passed back in md->start_match_ptr (an overloading of that
746        variable). If it does match, we reset that variable to the current subject
747        position and return MATCH_SKIP. Otherwise, pass back the return code
748        unaltered. */
749    
750        else if (rrc == MATCH_SKIP_ARG &&
751            STRCMP_UC_UC(ecode + 2, md->start_match_ptr) == 0)
752          {
753          md->start_match_ptr = eptr;
754          RRETURN(MATCH_SKIP);
755          }
756        RRETURN(rrc);
757    
758      case OP_FAIL:      case OP_FAIL:
759      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
760    
761      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);  
762    
763      case OP_COMMIT:      case OP_COMMIT:
764      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
765        ims, eptrb, flags, RM52);        eptrb, RM52);
766      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
767            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
768            rrc != MATCH_THEN)
769          RRETURN(rrc);
770      RRETURN(MATCH_COMMIT);      RRETURN(MATCH_COMMIT);
771    
772        /* PRUNE overrides THEN */
773    
774        case OP_PRUNE:
775        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
776          eptrb, RM51);
777        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
778        RRETURN(MATCH_PRUNE);
779    
780        case OP_PRUNE_ARG:
781        md->nomatch_mark = ecode + 2;
782        md->mark = NULL;    /* In case previously set by assertion */
783        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
784          eptrb, RM56);
785        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
786             md->mark == NULL) md->mark = ecode + 2;
787        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
788        RRETURN(MATCH_PRUNE);
789    
790        /* SKIP overrides PRUNE and THEN */
791    
792      case OP_SKIP:      case OP_SKIP:
793      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
794        ims, eptrb, flags, RM53);        eptrb, RM53);
795      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
796          RRETURN(rrc);
797      md->start_match_ptr = eptr;   /* Pass back current position */      md->start_match_ptr = eptr;   /* Pass back current position */
798      RRETURN(MATCH_SKIP);      RRETURN(MATCH_SKIP);
799    
800        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
801        nomatch_mark. There is a flag that disables this opcode when re-matching a
802        pattern that ended with a SKIP for which there was not a matching MARK. */
803    
804        case OP_SKIP_ARG:
805        if (md->ignore_skip_arg)
806          {
807          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
808          break;
809          }
810        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
811          eptrb, RM57);
812        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
813          RRETURN(rrc);
814    
815        /* Pass back the current skip name by overloading md->start_match_ptr and
816        returning the special MATCH_SKIP_ARG return code. This will either be
817        caught by a matching MARK, or get to the top, where it causes a rematch
818        with the md->ignore_skip_arg flag set. */
819    
820        md->start_match_ptr = ecode + 2;
821        RRETURN(MATCH_SKIP_ARG);
822    
823        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
824        the branch in which it occurs can be determined. Overload the start of
825        match pointer to do this. */
826    
827      case OP_THEN:      case OP_THEN:
828      RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
829        ims, eptrb, flags, RM54);        eptrb, RM54);
830      if (rrc != MATCH_NOMATCH) RRETURN(rrc);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
831        md->start_match_ptr = ecode;
832      RRETURN(MATCH_THEN);      RRETURN(MATCH_THEN);
833    
834      /* Handle a capturing bracket. If there is space in the offset vector, save      case OP_THEN_ARG:
835      the current subject position in the working slot at the top of the vector.      md->nomatch_mark = ecode + 2;
836      We mustn't change the current values of the data slot, because they may be      md->mark = NULL;    /* In case previously set by assertion */
837      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,
838      reference inside the group.        md, eptrb, RM58);
839        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
840      If the bracket fails to match, we need to restore this value and also the           md->mark == NULL) md->mark = ecode + 2;
841      values of the final offsets, in case they were set by a previous iteration      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
842      of the same bracket.      md->start_match_ptr = ecode;
843        RRETURN(MATCH_THEN);
844    
845        /* Handle an atomic group that does not contain any capturing parentheses.
846        This can be handled like an assertion. Prior to 8.13, all atomic groups
847        were handled this way. In 8.13, the code was changed as below for ONCE, so
848        that backups pass through the group and thereby reset captured values.
849        However, this uses a lot more stack, so in 8.20, atomic groups that do not
850        contain any captures generate OP_ONCE_NC, which can be handled in the old,
851        less stack intensive way.
852    
853        Check the alternative branches in turn - the matching won't pass the KET
854        for this kind of subpattern. If any one branch matches, we carry on as at
855        the end of a normal bracket, leaving the subject pointer, but resetting
856        the start-of-match value in case it was changed by \K. */
857    
858        case OP_ONCE_NC:
859        prev = ecode;
860        saved_eptr = eptr;
861        save_mark = md->mark;
862        do
863          {
864          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
865          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
866            {
867            mstart = md->start_match_ptr;
868            break;
869            }
870          if (rrc == MATCH_THEN)
871            {
872            next = ecode + GET(ecode,1);
873            if (md->start_match_ptr < next &&
874                (*ecode == OP_ALT || *next == OP_ALT))
875              rrc = MATCH_NOMATCH;
876            }
877    
878          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
879          ecode += GET(ecode,1);
880          md->mark = save_mark;
881          }
882        while (*ecode == OP_ALT);
883    
884        /* If hit the end of the group (which could be repeated), fail */
885    
886        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
887    
888        /* Continue as from after the group, updating the offsets high water
889        mark, since extracts may have been taken. */
890    
891        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
892    
893        offset_top = md->end_offset_top;
894        eptr = md->end_match_ptr;
895    
896        /* For a non-repeating ket, just continue at this level. This also
897        happens for a repeating ket if no characters were matched in the group.
898        This is the forcible breaking of infinite loops as implemented in Perl
899        5.005. */
900    
901        if (*ecode == OP_KET || eptr == saved_eptr)
902          {
903          ecode += 1+LINK_SIZE;
904          break;
905          }
906    
907        /* The repeating kets try the rest of the pattern or restart from the
908        preceding bracket, in the appropriate order. The second "call" of match()
909        uses tail recursion, to avoid using another stack frame. */
910    
911        if (*ecode == OP_KETRMIN)
912          {
913          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
914          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
915          ecode = prev;
916          goto TAIL_RECURSE;
917          }
918        else  /* OP_KETRMAX */
919          {
920          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
921          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
922          ecode += 1 + LINK_SIZE;
923          goto TAIL_RECURSE;
924          }
925        /* Control never gets here */
926    
927        /* Handle a capturing bracket, other than those that are possessive with an
928        unlimited repeat. If there is space in the offset vector, save the current
929        subject position in the working slot at the top of the vector. We mustn't
930        change the current values of the data slot, because they may be set from a
931        previous iteration of this group, and be referred to by a reference inside
932        the group. A failure to match might occur after the group has succeeded,
933        if something later on doesn't match. For this reason, we need to restore
934        the working value and also the values of the final offsets, in case they
935        were set by a previous iteration of the same bracket.
936    
937      If there isn't enough space in the offset vector, treat this as if it were      If there isn't enough space in the offset vector, treat this as if it were
938      a non-capturing bracket. Don't worry about setting the flag for the error      a non-capturing bracket. Don't worry about setting the flag for the error
# Line 667  for (;;) Line 943  for (;;)
943      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
944      offset = number << 1;      offset = number << 1;
945    
946  #ifdef DEBUG  #ifdef PCRE_DEBUG
947      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
948      printf("subject=");      printf("subject=");
949      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 680  for (;;) Line 956  for (;;)
956        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
957        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
958        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
959          save_mark = md->mark;
960    
961        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
962        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
963            (int)(eptr - md->start_subject);
964    
965        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
966          {          {
967          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
968            ims, eptrb, flags, RM1);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
969          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);            eptrb, RM1);
970            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
971    
972            /* If we backed up to a THEN, check whether it is within the current
973            branch by comparing the address of the THEN that is passed back with
974            the end of the branch. If it is within the current branch, and the
975            branch is one of two or more alternatives (it either starts or ends
976            with OP_ALT), we have reached the limit of THEN's action, so convert
977            the return code to NOMATCH, which will cause normal backtracking to
978            happen from now on. Otherwise, THEN is passed back to an outer
979            alternative. This implements Perl's treatment of parenthesized groups,
980            where a group not containing | does not affect the current alternative,
981            that is, (X) is NOT the same as (X|(*F)). */
982    
983            if (rrc == MATCH_THEN)
984              {
985              next = ecode + GET(ecode,1);
986              if (md->start_match_ptr < next &&
987                  (*ecode == OP_ALT || *next == OP_ALT))
988                rrc = MATCH_NOMATCH;
989              }
990    
991            /* Anything other than NOMATCH is passed back. */
992    
993            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
994          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
995          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
996            md->mark = save_mark;
997            if (*ecode != OP_ALT) break;
998          }          }
       while (*ecode == OP_ALT);  
999    
1000        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
1001        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
1002        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
1003        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
1004    
1005        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1006    
1007          RRETURN(rrc);
1008        }        }
1009    
1010      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
# Line 715  for (;;) Line 1018  for (;;)
1018      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1019      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1020    
1021      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* Non-capturing or atomic group, except for possessive with unlimited
1022      final alternative within the brackets, we would return the result of a      repeat and ONCE group with no captures. Loop for all the alternatives.
     recursive call to match() whatever happened. We can reduce stack usage by  
     turning this into a tail recursion, except in the case when match_cbegroup  
     is set.*/  
1023    
1024        When we get to the final alternative within the brackets, we used to return
1025        the result of a recursive call to match() whatever happened so it was
1026        possible to reduce stack usage by turning this into a tail recursion,
1027        except in the case of a possibly empty group. However, now that there is
1028        the possiblity of (*THEN) occurring in the final alternative, this
1029        optimization is no longer always possible.
1030    
1031        We can optimize if we know there are no (*THEN)s in the pattern; at present
1032        this is the best that can be done.
1033    
1034        MATCH_ONCE is returned when the end of an atomic group is successfully
1035        reached, but subsequent matching fails. It passes back up the tree (causing
1036        captured values to be reset) until the original atomic group level is
1037        reached. This is tested by comparing md->once_target with the start of the
1038        group. At this point, the return is converted into MATCH_NOMATCH so that
1039        previous backup points can be taken. */
1040    
1041        case OP_ONCE:
1042      case OP_BRA:      case OP_BRA:
1043      case OP_SBRA:      case OP_SBRA:
1044      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1045      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
1046      for (;;)      for (;;)
1047        {        {
1048        if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */        if (op >= OP_SBRA || op == OP_ONCE)
1049          {          md->match_function_type = MATCH_CBEGROUP;
         if (flags == 0)    /* Not a possibly empty group */  
           {  
           ecode += _pcre_OP_lengths[*ecode];  
           DPRINTF(("bracket 0 tail recursion\n"));  
           goto TAIL_RECURSE;  
           }  
1050    
1051          /* Possibly empty group; can't use tail recursion. */        /* If this is not a possibly empty group, and there are no (*THEN)s in
1052          the pattern, and this is the final alternative, optimize as described
1053          above. */
1054    
1055          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1056            eptrb, flags, RM48);          {
1057          RRETURN(rrc);          ecode += PRIV(OP_lengths)[*ecode];
1058            goto TAIL_RECURSE;
1059          }          }
1060    
1061        /* For non-final alternatives, continue the loop for a NOMATCH result;        /* In all other cases, we have to make another call to match(). */
       otherwise return. */  
1062    
1063        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        save_mark = md->mark;
1064          eptrb, flags, RM2);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1065        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);          RM2);
1066    
1067          /* See comment in the code for capturing groups above about handling
1068          THEN. */
1069    
1070          if (rrc == MATCH_THEN)
1071            {
1072            next = ecode + GET(ecode,1);
1073            if (md->start_match_ptr < next &&
1074                (*ecode == OP_ALT || *next == OP_ALT))
1075              rrc = MATCH_NOMATCH;
1076            }
1077    
1078          if (rrc != MATCH_NOMATCH)
1079            {
1080            if (rrc == MATCH_ONCE)
1081              {
1082              const pcre_uchar *scode = ecode;
1083              if (*scode != OP_ONCE)           /* If not at start, find it */
1084                {
1085                while (*scode == OP_ALT) scode += GET(scode, 1);
1086                scode -= GET(scode, 1);
1087                }
1088              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1089              }
1090            RRETURN(rrc);
1091            }
1092        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1093          md->mark = save_mark;
1094          if (*ecode != OP_ALT) break;
1095        }        }
     /* Control never reaches here. */  
1096    
1097      /* Conditional group: compilation checked that there are no more than      RRETURN(MATCH_NOMATCH);
     two branches. If the condition is false, skipping the first branch takes us  
     past the end if there is only one branch, but that's OK because that is  
     exactly what going to the ket would do. As there is only one branch to be  
     obeyed, we can use tail recursion to avoid using another stack frame. */  
1098    
1099      case OP_COND:      /* Handle possessive capturing brackets with an unlimited repeat. We come
1100      case OP_SCOND:      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1101      if (ecode[LINK_SIZE+1] == OP_RREF)         /* Recursion test */      handled similarly to the normal case above. However, the matching is
1102        {      different. The end of these brackets will always be OP_KETRPOS, which
1103        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/      returns MATCH_KETRPOS without going further in the pattern. By this means
1104        condition = md->recursive != NULL &&      we can handle the group by iteration rather than recursion, thereby
1105          (offset == RREF_ANY || offset == md->recursive->group_num);      reducing the amount of stack needed. */
1106        ecode += condition? 3 : GET(ecode, 1);  
1107        }      case OP_CBRAPOS:
1108        case OP_SCBRAPOS:
1109        allow_zero = FALSE;
1110    
1111      else if (ecode[LINK_SIZE+1] == OP_CREF)    /* Group used test */      POSSESSIVE_CAPTURE:
1112        {      number = GET2(ecode, 1+LINK_SIZE);
1113        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */      offset = number << 1;
1114        condition = offset < offset_top && md->offset_vector[offset] >= 0;  
1115        ecode += condition? 3 : GET(ecode, 1);  #ifdef PCRE_DEBUG
1116        }      printf("start possessive bracket %d\n", number);
1117        printf("subject=");
1118        pchars(eptr, 16, TRUE, md);
1119        printf("\n");
1120    #endif
1121    
1122      else if (ecode[LINK_SIZE+1] == OP_DEF)     /* DEFINE - always false */      if (offset < md->offset_max)
1123        {        {
1124        condition = FALSE;        matched_once = FALSE;
1125        ecode += GET(ecode, 1);        code_offset = (int)(ecode - md->start_code);
       }  
1126    
1127      /* The condition is an assertion. Call match() to evaluate it - setting        save_offset1 = md->offset_vector[offset];
1128      the final argument match_condassert causes it to stop at the end of an        save_offset2 = md->offset_vector[offset+1];
1129      assertion. */        save_offset3 = md->offset_vector[md->offset_end - number];
1130          save_capture_last = md->capture_last;
1131    
1132      else        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1133        {  
1134        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        /* Each time round the loop, save the current subject position for use
1135            match_condassert, RM3);        when the group matches. For MATCH_MATCH, the group has matched, so we
1136        if (rrc == MATCH_MATCH)        restart it with a new subject starting position, remembering that we had
1137          {        at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1138          condition = TRUE;        usual. If we haven't matched any alternatives in any iteration, check to
1139          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);        see if a previous iteration matched. If so, the group has matched;
1140          while (*ecode == OP_ALT) ecode += GET(ecode, 1);        continue from afterwards. Otherwise it has failed; restore the previous
1141          capture values before returning NOMATCH. */
1142    
1143          for (;;)
1144            {
1145            md->offset_vector[md->offset_end - number] =
1146              (int)(eptr - md->start_subject);
1147            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1148            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1149              eptrb, RM63);
1150            if (rrc == MATCH_KETRPOS)
1151              {
1152              offset_top = md->end_offset_top;
1153              eptr = md->end_match_ptr;
1154              ecode = md->start_code + code_offset;
1155              save_capture_last = md->capture_last;
1156              matched_once = TRUE;
1157              continue;
1158              }
1159    
1160            /* See comment in the code for capturing groups above about handling
1161            THEN. */
1162    
1163            if (rrc == MATCH_THEN)
1164              {
1165              next = ecode + GET(ecode,1);
1166              if (md->start_match_ptr < next &&
1167                  (*ecode == OP_ALT || *next == OP_ALT))
1168                rrc = MATCH_NOMATCH;
1169              }
1170    
1171            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1172            md->capture_last = save_capture_last;
1173            ecode += GET(ecode, 1);
1174            if (*ecode != OP_ALT) break;
1175          }          }
1176        else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)  
1177          if (!matched_once)
1178          {          {
1179          RRETURN(rrc);         /* Need braces because of following else */          md->offset_vector[offset] = save_offset1;
1180            md->offset_vector[offset+1] = save_offset2;
1181            md->offset_vector[md->offset_end - number] = save_offset3;
1182          }          }
1183        else  
1184          if (allow_zero || matched_once)
1185          {          {
1186          condition = FALSE;          ecode += 1 + LINK_SIZE;
1187          ecode += GET(ecode, 1);          break;
1188          }          }
1189    
1190          RRETURN(MATCH_NOMATCH);
1191        }        }
1192    
1193      /* We are now at the branch that is to be obeyed. As there is only one,      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1194      we can use tail recursion to avoid using another stack frame, except when      as a non-capturing bracket. */
     match_cbegroup is required for an unlimited repeat of a possibly empty  
     group. If the second alternative doesn't exist, we can just plough on. */  
1195    
1196      if (condition || *ecode == OP_ALT)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1197        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1198    
1199        DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1200    
1201        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1202        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1203    
1204        /* Non-capturing possessive bracket with unlimited repeat. We come here
1205        from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1206        without the capturing complication. It is written out separately for speed
1207        and cleanliness. */
1208    
1209        case OP_BRAPOS:
1210        case OP_SBRAPOS:
1211        allow_zero = FALSE;
1212    
1213        POSSESSIVE_NON_CAPTURE:
1214        matched_once = FALSE;
1215        code_offset = (int)(ecode - md->start_code);
1216    
1217        for (;;)
1218        {        {
1219        ecode += 1 + LINK_SIZE;        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1220        if (op == OP_SCOND)        /* Possibly empty group */        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1221          {          eptrb, RM48);
1222          RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);        if (rrc == MATCH_KETRPOS)
1223          RRETURN(rrc);          {
1224          }          offset_top = md->end_offset_top;
1225        else                       /* Group must match something */          eptr = md->end_match_ptr;
1226          {          ecode = md->start_code + code_offset;
1227          flags = 0;          matched_once = TRUE;
1228          goto TAIL_RECURSE;          continue;
1229            }
1230    
1231          /* See comment in the code for capturing groups above about handling
1232          THEN. */
1233    
1234          if (rrc == MATCH_THEN)
1235            {
1236            next = ecode + GET(ecode,1);
1237            if (md->start_match_ptr < next &&
1238                (*ecode == OP_ALT || *next == OP_ALT))
1239              rrc = MATCH_NOMATCH;
1240          }          }
1241    
1242          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1243          ecode += GET(ecode, 1);
1244          if (*ecode != OP_ALT) break;
1245        }        }
1246      else                         /* Condition false & no 2nd alternative */  
1247        if (matched_once || allow_zero)
1248        {        {
1249        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
1250          break;
1251        }        }
1252      break;      RRETURN(MATCH_NOMATCH);
1253    
1254        /* Control never reaches here. */
1255    
1256        /* Conditional group: compilation checked that there are no more than
1257        two branches. If the condition is false, skipping the first branch takes us
1258        past the end if there is only one branch, but that's OK because that is
1259        exactly what going to the ket would do. */
1260    
1261        case OP_COND:
1262        case OP_SCOND:
1263        codelink = GET(ecode, 1);
1264    
1265      /* End of the pattern, either real or forced. If we are in a top-level      /* Because of the way auto-callout works during compile, a callout item is
1266      recursion, we should restore the offsets appropriately and continue from      inserted between OP_COND and an assertion condition. */
     after the call. */  
1267    
1268      case OP_ACCEPT:      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
     case OP_END:  
     if (md->recursive != NULL && md->recursive->group_num == 0)  
1269        {        {
1270        recursion_info *rec = md->recursive;        if (PUBL(callout) != NULL)
1271        DPRINTF(("End of pattern in a (?0) recursion\n"));          {
1272        md->recursive = rec->prevrec;          PUBL(callout_block) cb;
1273        memmove(md->offset_vector, rec->offset_save,          cb.version          = 2;   /* Version 1 of the callout block */
1274          rec->saved_max * sizeof(int));          cb.callout_number   = ecode[LINK_SIZE+2];
1275        mstart = rec->save_start;          cb.offset_vector    = md->offset_vector;
1276        ims = original_ims;  #if defined COMPILE_PCRE8
1277        ecode = rec->after_call;          cb.subject          = (PCRE_SPTR)md->start_subject;
1278        break;  #elif defined COMPILE_PCRE16
1279            cb.subject          = (PCRE_SPTR16)md->start_subject;
1280    #elif defined COMPILE_PCRE32
1281            cb.subject          = (PCRE_SPTR32)md->start_subject;
1282    #endif
1283            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1284            cb.start_match      = (int)(mstart - md->start_subject);
1285            cb.current_position = (int)(eptr - md->start_subject);
1286            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1287            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1288            cb.capture_top      = offset_top/2;
1289            cb.capture_last     = md->capture_last;
1290            cb.callout_data     = md->callout_data;
1291            cb.mark             = md->nomatch_mark;
1292            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1293            if (rrc < 0) RRETURN(rrc);
1294            }
1295          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1296        }        }
1297    
1298      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      condcode = ecode[LINK_SIZE+1];
     string - backtracking will then try other alternatives, if any. */  
   
     if (md->notempty && eptr == mstart) RRETURN(MATCH_NOMATCH);  
     md->end_match_ptr = eptr;           /* Record where we ended */  
     md->end_offset_top = offset_top;    /* and how many extracts were taken */  
     md->start_match_ptr = mstart;       /* and the start (\K can modify) */  
     RRETURN(MATCH_MATCH);  
1299    
1300      /* Change option settings */      /* Now see what the actual condition is */
1301    
1302      case OP_OPT:      if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1303      ims = ecode[1];        {
1304      ecode += 2;        if (md->recursive == NULL)                /* Not recursing => FALSE */
1305      DPRINTF(("ims set to %02lx\n", ims));          {
1306            condition = FALSE;
1307            ecode += GET(ecode, 1);
1308            }
1309          else
1310            {
1311            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1312            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1313    
1314            /* If the test is for recursion into a specific subpattern, and it is
1315            false, but the test was set up by name, scan the table to see if the
1316            name refers to any other numbers, and test them. The condition is true
1317            if any one is set. */
1318    
1319            if (!condition && condcode == OP_NRREF)
1320              {
1321              pcre_uchar *slotA = md->name_table;
1322              for (i = 0; i < md->name_count; i++)
1323                {
1324                if (GET2(slotA, 0) == recno) break;
1325                slotA += md->name_entry_size;
1326                }
1327    
1328              /* Found a name for the number - there can be only one; duplicate
1329              names for different numbers are allowed, but not vice versa. First
1330              scan down for duplicates. */
1331    
1332              if (i < md->name_count)
1333                {
1334                pcre_uchar *slotB = slotA;
1335                while (slotB > md->name_table)
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                /* Scan up for duplicates */
1347    
1348                if (!condition)
1349                  {
1350                  slotB = slotA;
1351                  for (i++; i < md->name_count; i++)
1352                    {
1353                    slotB += md->name_entry_size;
1354                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1355                      {
1356                      condition = GET2(slotB, 0) == md->recursive->group_num;
1357                      if (condition) break;
1358                      }
1359                    else break;
1360                    }
1361                  }
1362                }
1363              }
1364    
1365            /* Chose branch according to the condition */
1366    
1367            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1368            }
1369          }
1370    
1371        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1372          {
1373          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1374          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1375    
1376          /* If the numbered capture is unset, but the reference was by name,
1377          scan the table to see if the name refers to any other numbers, and test
1378          them. The condition is true if any one is set. This is tediously similar
1379          to the code above, but not close enough to try to amalgamate. */
1380    
1381          if (!condition && condcode == OP_NCREF)
1382            {
1383            int refno = offset >> 1;
1384            pcre_uchar *slotA = md->name_table;
1385    
1386            for (i = 0; i < md->name_count; i++)
1387              {
1388              if (GET2(slotA, 0) == refno) break;
1389              slotA += md->name_entry_size;
1390              }
1391    
1392            /* Found a name for the number - there can be only one; duplicate names
1393            for different numbers are allowed, but not vice versa. First scan down
1394            for duplicates. */
1395    
1396            if (i < md->name_count)
1397              {
1398              pcre_uchar *slotB = slotA;
1399              while (slotB > md->name_table)
1400                {
1401                slotB -= md->name_entry_size;
1402                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1403                  {
1404                  offset = GET2(slotB, 0) << 1;
1405                  condition = offset < offset_top &&
1406                    md->offset_vector[offset] >= 0;
1407                  if (condition) break;
1408                  }
1409                else break;
1410                }
1411    
1412              /* Scan up for duplicates */
1413    
1414              if (!condition)
1415                {
1416                slotB = slotA;
1417                for (i++; i < md->name_count; i++)
1418                  {
1419                  slotB += md->name_entry_size;
1420                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1421                    {
1422                    offset = GET2(slotB, 0) << 1;
1423                    condition = offset < offset_top &&
1424                      md->offset_vector[offset] >= 0;
1425                    if (condition) break;
1426                    }
1427                  else break;
1428                  }
1429                }
1430              }
1431            }
1432    
1433          /* Chose branch according to the condition */
1434    
1435          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1436          }
1437    
1438        else if (condcode == OP_DEF)     /* DEFINE - always false */
1439          {
1440          condition = FALSE;
1441          ecode += GET(ecode, 1);
1442          }
1443    
1444        /* The condition is an assertion. Call match() to evaluate it - setting
1445        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1446        an assertion. */
1447    
1448        else
1449          {
1450          md->match_function_type = MATCH_CONDASSERT;
1451          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1452          if (rrc == MATCH_MATCH)
1453            {
1454            if (md->end_offset_top > offset_top)
1455              offset_top = md->end_offset_top;  /* Captures may have happened */
1456            condition = TRUE;
1457            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1458            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1459            }
1460    
1461          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1462          assertion; it is therefore treated as NOMATCH. */
1463    
1464          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1465            {
1466            RRETURN(rrc);         /* Need braces because of following else */
1467            }
1468          else
1469            {
1470            condition = FALSE;
1471            ecode += codelink;
1472            }
1473          }
1474    
1475        /* We are now at the branch that is to be obeyed. As there is only one, can
1476        use tail recursion to avoid using another stack frame, except when there is
1477        unlimited repeat of a possibly empty group. In the latter case, a recursive
1478        call to match() is always required, unless the second alternative doesn't
1479        exist, in which case we can just plough on. Note that, for compatibility
1480        with Perl, the | in a conditional group is NOT treated as creating two
1481        alternatives. If a THEN is encountered in the branch, it propagates out to
1482        the enclosing alternative (unless nested in a deeper set of alternatives,
1483        of course). */
1484    
1485        if (condition || *ecode == OP_ALT)
1486          {
1487          if (op != OP_SCOND)
1488            {
1489            ecode += 1 + LINK_SIZE;
1490            goto TAIL_RECURSE;
1491            }
1492    
1493          md->match_function_type = MATCH_CBEGROUP;
1494          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1495          RRETURN(rrc);
1496          }
1497    
1498         /* Condition false & no alternative; continue after the group. */
1499    
1500        else
1501          {
1502          ecode += 1 + LINK_SIZE;
1503          }
1504        break;
1505    
1506    
1507        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1508        to close any currently open capturing brackets. */
1509    
1510        case OP_CLOSE:
1511        number = GET2(ecode, 1);
1512        offset = number << 1;
1513    
1514    #ifdef PCRE_DEBUG
1515          printf("end bracket %d at *ACCEPT", number);
1516          printf("\n");
1517    #endif
1518    
1519        md->capture_last = number;
1520        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1521          {
1522          md->offset_vector[offset] =
1523            md->offset_vector[md->offset_end - number];
1524          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1525          if (offset_top <= offset) offset_top = offset + 2;
1526          }
1527        ecode += 1 + IMM2_SIZE;
1528      break;      break;
1529    
1530    
1531        /* End of the pattern, either real or forced. */
1532    
1533        case OP_END:
1534        case OP_ACCEPT:
1535        case OP_ASSERT_ACCEPT:
1536    
1537        /* If we have matched an empty string, fail if not in an assertion and not
1538        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1539        is set and we have matched at the start of the subject. In both cases,
1540        backtracking will then try other alternatives, if any. */
1541    
1542        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1543             md->recursive == NULL &&
1544             (md->notempty ||
1545               (md->notempty_atstart &&
1546                 mstart == md->start_subject + md->start_offset)))
1547          RRETURN(MATCH_NOMATCH);
1548    
1549        /* Otherwise, we have a match. */
1550    
1551        md->end_match_ptr = eptr;           /* Record where we ended */
1552        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1553        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1554    
1555        /* For some reason, the macros don't work properly if an expression is
1556        given as the argument to RRETURN when the heap is in use. */
1557    
1558        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1559        RRETURN(rrc);
1560    
1561      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
1562      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,
1563      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
1564      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
1565      this level is identical to the lookahead case. */      this level is identical to the lookahead case. When the assertion is part
1566        of a condition, we want to return immediately afterwards. The caller of
1567        this incarnation of the match() function will have set MATCH_CONDASSERT in
1568        md->match_function type, and one of these opcodes will be the first opcode
1569        that is processed. We use a local variable that is preserved over calls to
1570        match() to remember this case. */
1571    
1572      case OP_ASSERT:      case OP_ASSERT:
1573      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1574        save_mark = md->mark;
1575        if (md->match_function_type == MATCH_CONDASSERT)
1576          {
1577          condassert = TRUE;
1578          md->match_function_type = 0;
1579          }
1580        else condassert = FALSE;
1581    
1582      do      do
1583        {        {
1584        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1585          RM4);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1586        if (rrc == MATCH_MATCH) break;          {
1587            mstart = md->start_match_ptr;   /* In case \K reset it */
1588            break;
1589            }
1590          md->mark = save_mark;
1591    
1592          /* A COMMIT failure must fail the entire assertion, without trying any
1593          subsequent branches. */
1594    
1595          if (rrc == MATCH_COMMIT) RRETURN(MATCH_NOMATCH);
1596    
1597          /* PCRE does not allow THEN to escape beyond an assertion; it
1598          is treated as NOMATCH. */
1599    
1600        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1601        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1602        }        }
1603      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1604    
1605      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);      if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1606    
1607      /* If checking an assertion for a condition, return MATCH_MATCH. */      /* If checking an assertion for a condition, return MATCH_MATCH. */
1608    
1609      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);
1610    
1611      /* Continue from after the assertion, updating the offsets high water      /* Continue from after the assertion, updating the offsets high water
1612      mark, since extracts may have been taken during the assertion. */      mark, since extracts may have been taken during the assertion. */
# Line 900  for (;;) Line 1616  for (;;)
1616      offset_top = md->end_offset_top;      offset_top = md->end_offset_top;
1617      continue;      continue;
1618    
1619      /* Negative assertion: all branches must fail to match */      /* Negative assertion: all branches must fail to match. Encountering SKIP,
1620        PRUNE, or COMMIT means we must assume failure without checking subsequent
1621        branches. */
1622    
1623      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1624      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1625        save_mark = md->mark;
1626        if (md->match_function_type == MATCH_CONDASSERT)
1627          {
1628          condassert = TRUE;
1629          md->match_function_type = 0;
1630          }
1631        else condassert = FALSE;
1632    
1633      do      do
1634        {        {
1635        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1636          RM5);        md->mark = save_mark;
1637        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);        if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1638          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1639            {
1640            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1641            break;
1642            }
1643    
1644          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1645          as NOMATCH. */
1646    
1647        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1648        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1649        }        }
1650      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1651    
1652      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1653    
1654      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1655      continue;      continue;
# Line 925  for (;;) Line 1660  for (;;)
1660      back a number of characters, not bytes. */      back a number of characters, not bytes. */
1661    
1662      case OP_REVERSE:      case OP_REVERSE:
1663  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1664      if (utf8)      if (utf)
1665        {        {
1666        i = GET(ecode, 1);        i = GET(ecode, 1);
1667        while (i-- > 0)        while (i-- > 0)
# Line 946  for (;;) Line 1681  for (;;)
1681        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1682        }        }
1683    
1684      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1685    
1686        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1687      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1688      break;      break;
1689    
# Line 956  for (;;) Line 1692  for (;;)
1692      function is able to force a failure. */      function is able to force a failure. */
1693    
1694      case OP_CALLOUT:      case OP_CALLOUT:
1695      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1696        {        {
1697        pcre_callout_block cb;        PUBL(callout_block) cb;
1698        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1699        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1700        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1701    #if defined COMPILE_PCRE8
1702        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1703        cb.subject_length   = md->end_subject - md->start_subject;  #elif defined COMPILE_PCRE16
1704        cb.start_match      = mstart - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1705        cb.current_position = eptr - md->start_subject;  #elif defined COMPILE_PCRE32
1706          cb.subject          = (PCRE_SPTR32)md->start_subject;
1707    #endif
1708          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1709          cb.start_match      = (int)(mstart - md->start_subject);
1710          cb.current_position = (int)(eptr - md->start_subject);
1711        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1712        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1713        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1714        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1715        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1716        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1717          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1718        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1719        }        }
1720      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 981  for (;;) Line 1724  for (;;)
1724      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
1725      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1726    
1727      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1728      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
1729      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
1730      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
1731      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
1732      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
1733      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.  
1734    
1735      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
1736      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
1737      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1738        a lot, so he is not to blame for the current way it works. */
1739    
1740      case OP_RECURSE:      case OP_RECURSE:
1741        {        {
1742          recursion_info *ri;
1743          int recno;
1744    
1745        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1746        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1747          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1748    
1749          /* Check for repeating a recursion without advancing the subject pointer.
1750          This should catch convoluted mutual recursions. (Some simple cases are
1751          caught at compile time.) */
1752    
1753          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1754            if (recno == ri->group_num && eptr == ri->subject_position)
1755              RRETURN(PCRE_ERROR_RECURSELOOP);
1756    
1757        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1758    
1759          new_recursive.group_num = recno;
1760          new_recursive.subject_position = eptr;
1761        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1762        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1763    
1764        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1765    
1766        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1767    
1768        /* Now save the offset data. */        /* Now save the offset data */
1769    
1770        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1771        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 1020  for (;;) Line 1773  for (;;)
1773        else        else
1774          {          {
1775          new_recursive.offset_save =          new_recursive.offset_save =
1776            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1777          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1778          }          }
   
1779        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1780              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = mstart;  
       mstart = eptr;  
1781    
1782        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1783        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1784          might be changed, so reset it before looping. */
1785    
1786        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1787        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1788        do        do
1789          {          {
1790          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1791            md, ims, eptrb, flags, RM6);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1792          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1793            memcpy(md->offset_vector, new_recursive.offset_save,
1794                new_recursive.saved_max * sizeof(int));
1795            md->recursive = new_recursive.prevrec;
1796            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1797            {            {
1798            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1799            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1800              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1801            RRETURN(MATCH_MATCH);  
1802              /* Set where we got to in the subject, and reset the start in case
1803              it was changed by \K. This *is* propagated back out of a recursion,
1804              for Perl compatibility. */
1805    
1806              eptr = md->end_match_ptr;
1807              mstart = md->start_match_ptr;
1808              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1809            }            }
1810          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)  
1811            /* PCRE does not allow THEN or COMMIT to escape beyond a recursion; it
1812            is treated as NOMATCH. */
1813    
1814            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN &&
1815                     rrc != MATCH_COMMIT)
1816            {            {
1817            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1818              if (new_recursive.offset_save != stacksave)
1819                (PUBL(free))(new_recursive.offset_save);
1820            RRETURN(rrc);            RRETURN(rrc);
1821            }            }
1822    
1823          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1824          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1825          }          }
1826        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1062  for (;;) Line 1828  for (;;)
1828        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1829        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1830        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1831          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1832        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1833        }        }
     /* Control never reaches here */  
1834    
1835      /* "Once" brackets are like assertion brackets except that after a match,      RECURSION_MATCHED:
1836      the point in the subject string is not moved back. Thus there can never be      break;
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM7);  
       if (rrc == MATCH_MATCH) break;  
       if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
   
     /* If hit the end of the group (which could be repeated), fail */  
   
     if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
   
     /* Continue as from after the assertion, updating the offsets high water  
     mark, since extracts may have been taken. */  
   
     do ecode += GET(ecode, 1); while (*ecode == OP_ALT);  
   
     offset_top = md->end_offset_top;  
     eptr = md->end_match_ptr;  
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
       {  
       ecode += 1+LINK_SIZE;  
       break;  
       }  
   
     /* The repeating kets try the rest of the pattern or restart from the  
     preceding bracket, in the appropriate order. The second "call" of match()  
     uses tail recursion, to avoid using another stack frame. We need to reset  
     any options that changed within the bracket before re-running it, so  
     check the next opcode. */  
   
     if (ecode[1+LINK_SIZE] == OP_OPT)  
       {  
       ims = (ims & ~PCRE_IMS) | ecode[4];  
       DPRINTF(("ims set to %02lx at group repeat\n", ims));  
       }  
   
     if (*ecode == OP_KETRMIN)  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM8);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       RMATCH(eptr, prev, offset_top, md, ims, eptrb, match_cbegroup, RM9);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1837    
1838      /* 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
1839      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1148  for (;;) Line 1842  for (;;)
1842      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1843      break;      break;
1844    
1845      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1846      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
1847      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
1848      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
1849      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1850    
1851      case OP_BRAZERO:      case OP_BRAZERO:
1852        {      next = ecode + 1;
1853        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1854        RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1855        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1856        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1857      break;      break;
1858    
1859      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1860        {      next = ecode + 1;
1861        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1862        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1863        RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1864        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
       ecode++;  
       }  
1865      break;      break;
1866    
1867        case OP_SKIPZERO:
1868        next = ecode+1;
1869        do next += GET(next,1); while (*next == OP_ALT);
1870        ecode = next + 1 + LINK_SIZE;
1871        break;
1872    
1873        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1874        here; just jump to the group, with allow_zero set TRUE. */
1875    
1876        case OP_BRAPOSZERO:
1877        op = *(++ecode);
1878        allow_zero = TRUE;
1879        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1880          goto POSSESSIVE_NON_CAPTURE;
1881    
1882      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1883    
1884      case OP_KET:      case OP_KET:
1885      case OP_KETRMIN:      case OP_KETRMIN:
1886      case OP_KETRMAX:      case OP_KETRMAX:
1887        case OP_KETRPOS:
1888      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1889    
1890      /* 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
1891      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1892      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1893    
1894      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1895        {        {
1896        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1897        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1898        }        }
1899      else saved_eptr = NULL;      else saved_eptr = NULL;
1900    
1901      /* 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
1902      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
1903      assertions. Do this also for the "once" (atomic) groups. */      water mark for use by positive assertions. We also need to record the match
1904        start in case it was changed by \K. */
1905      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
1906          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1907          *prev == OP_ONCE)           *prev == OP_ONCE_NC)
1908        {        {
1909        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1910        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1911        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1912          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1913        }        }
1914    
1915      /* 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
1916      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1917      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
1918      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
1919      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
1920        the current subject position and start match pointer and give a MATCH
1921        return. */
1922    
1923      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1924            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1925        {        {
1926        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1927        offset = number << 1;        offset = number << 1;
1928    
1929  #ifdef DEBUG  #ifdef PCRE_DEBUG
1930        printf("end bracket %d", number);        printf("end bracket %d", number);
1931        printf("\n");        printf("\n");
1932  #endif  #endif
1933    
1934          /* Handle a recursively called group. */
1935    
1936          if (md->recursive != NULL && md->recursive->group_num == number)
1937            {
1938            md->end_match_ptr = eptr;
1939            md->start_match_ptr = mstart;
1940            RRETURN(MATCH_MATCH);
1941            }
1942    
1943          /* Deal with capturing */
1944    
1945        md->capture_last = number;        md->capture_last = number;
1946        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1947          {          {
1948            /* If offset is greater than offset_top, it means that we are
1949            "skipping" a capturing group, and that group's offsets must be marked
1950            unset. In earlier versions of PCRE, all the offsets were unset at the
1951            start of matching, but this doesn't work because atomic groups and
1952            assertions can cause a value to be set that should later be unset.
1953            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1954            part of the atomic group, but this is not on the final matching path,
1955            so must be unset when 2 is set. (If there is no group 2, there is no
1956            problem, because offset_top will then be 2, indicating no capture.) */
1957    
1958            if (offset > offset_top)
1959              {
1960              register int *iptr = md->offset_vector + offset_top;
1961              register int *iend = md->offset_vector + offset;
1962              while (iptr < iend) *iptr++ = -1;
1963              }
1964    
1965            /* Now make the extraction */
1966    
1967          md->offset_vector[offset] =          md->offset_vector[offset] =
1968            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
1969          md->offset_vector[offset+1] = eptr - md->start_subject;          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1970          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
1971          }          }
1972          }
1973    
1974        /* Handle a recursively called group. Restore the offsets      /* For an ordinary non-repeating ket, just continue at this level. This
1975        appropriately and continue from after the call. */      also happens for a repeating ket if no characters were matched in the
1976        group. This is the forcible breaking of infinite loops as implemented in
1977        Perl 5.005. For a non-repeating atomic group that includes captures,
1978        establish a backup point by processing the rest of the pattern at a lower
1979        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1980        original OP_ONCE level, thereby bypassing intermediate backup points, but
1981        resetting any captures that happened along the way. */
1982    
1983        if (md->recursive != NULL && md->recursive->group_num == number)      if (*ecode == OP_KET || eptr == saved_eptr)
1984          {
1985          if (*prev == OP_ONCE)
1986          {          {
1987          recursion_info *rec = md->recursive;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1988          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1989          md->recursive = rec->prevrec;          md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1990          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;  
1991          }          }
1992          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1993          break;
1994        }        }
1995    
1996      /* For both capturing and non-capturing groups, reset the value of the ims      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
1997      flags, in case they got changed during the group. */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
1998        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. */  
1999    
2000      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KETRPOS)
2001        {        {
2002        ecode += 1 + LINK_SIZE;        md->end_match_ptr = eptr;
2003        break;        md->end_offset_top = offset_top;
2004          RRETURN(MATCH_KETRPOS);
2005        }        }
2006    
2007      /* 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
2008      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
2009      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
2010      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
2011        string. */
     flags = (*prev >= OP_SBRA)? match_cbegroup : 0;  
2012    
2013      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
2014        {        {
2015        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2016        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2017        if (flags != 0)    /* Could match an empty string */        if (*prev == OP_ONCE)
2018            {
2019            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2020            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2021            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2022            RRETURN(MATCH_ONCE);
2023            }
2024          if (*prev >= OP_SBRA)    /* Could match an empty string */
2025          {          {
2026          RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2027          RRETURN(rrc);          RRETURN(rrc);
2028          }          }
2029        ecode = prev;        ecode = prev;
# Line 1286  for (;;) Line 2031  for (;;)
2031        }        }
2032      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
2033        {        {
2034        RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2035          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2036        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2037          if (*prev == OP_ONCE)
2038            {
2039            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2040            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2041            md->once_target = prev;
2042            RRETURN(MATCH_ONCE);
2043            }
2044        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
2045        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2046        }        }
2047      /* Control never gets here */      /* Control never gets here */
2048    
2049      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2050    
2051      case OP_CIRC:      case OP_CIRC:
2052      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 */  
2053    
2054      /* Start of subject assertion */      /* Start of subject assertion */
2055    
# Line 1315  for (;;) Line 2058  for (;;)
2058      ecode++;      ecode++;
2059      break;      break;
2060    
2061        /* Multiline mode: start of subject unless notbol, or after any newline. */
2062    
2063        case OP_CIRCM:
2064        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2065        if (eptr != md->start_subject &&
2066            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2067          RRETURN(MATCH_NOMATCH);
2068        ecode++;
2069        break;
2070    
2071      /* Start of match assertion */      /* Start of match assertion */
2072    
2073      case OP_SOM:      case OP_SOM:
# Line 1329  for (;;) Line 2082  for (;;)
2082      ecode++;      ecode++;
2083      break;      break;
2084    
2085      /* Assert before internal newline if multiline, or before a terminating      /* Multiline mode: assert before any newline, or before end of subject
2086      newline unless endonly is set, else end of subject unless noteol is set. */      unless noteol is set. */
2087    
2088      case OP_DOLL:      case OP_DOLLM:
2089      if ((ims & PCRE_MULTILINE) != 0)      if (eptr < md->end_subject)
2090        {        {
2091        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2092          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }          {
2093        else          if (md->partial != 0 &&
2094          { if (md->noteol) RRETURN(MATCH_NOMATCH); }              eptr + 1 >= md->end_subject &&
2095        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2096        break;              NLBLOCK->nllen == 2 &&
2097                *eptr == NLBLOCK->nl[0])
2098              {
2099              md->hitend = TRUE;
2100              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2101              }
2102            RRETURN(MATCH_NOMATCH);
2103            }
2104        }        }
2105      else      else
2106        {        {
2107        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2108        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2109        }        }
2110        ecode++;
2111        break;
2112    
2113        /* Not multiline mode: assert before a terminating newline or before end of
2114        subject unless noteol is set. */
2115    
2116        case OP_DOLL:
2117        if (md->noteol) RRETURN(MATCH_NOMATCH);
2118        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2119    
2120      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2121    
2122      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2123    
2124      case OP_EOD:      case OP_EOD:
2125      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2126        SCHECK_PARTIAL();
2127      ecode++;      ecode++;
2128      break;      break;
2129    
2130      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2131    
2132      case OP_EODN:      case OP_EODN:
2133      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2134        if (eptr < md->end_subject &&
2135          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2136          {
2137          if (md->partial != 0 &&
2138              eptr + 1 >= md->end_subject &&
2139              NLBLOCK->nltype == NLTYPE_FIXED &&
2140              NLBLOCK->nllen == 2 &&
2141              *eptr == NLBLOCK->nl[0])
2142            {
2143            md->hitend = TRUE;
2144            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2145            }
2146        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2147          }
2148    
2149        /* Either at end of string or \n before end. */
2150    
2151        SCHECK_PARTIAL();
2152      ecode++;      ecode++;
2153      break;      break;
2154    
# Line 1380  for (;;) Line 2160  for (;;)
2160    
2161        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2162        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
2163        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2164          partial matching. */
2165    
2166  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2167        if (utf8)        if (utf)
2168          {          {
2169            /* Get status of previous character */
2170    
2171          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2172            {            {
2173            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2174            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2175              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2176            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2177    #ifdef SUPPORT_UCP
2178              if (md->use_ucp)
2179                {
2180                if (c == '_') prev_is_word = TRUE; else
2181                  {
2182                  int cat = UCD_CATEGORY(c);
2183                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2184                  }
2185                }
2186              else
2187    #endif
2188            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2189            }            }
2190          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2191            /* Get status of next character */
2192    
2193            if (eptr >= md->end_subject)
2194              {
2195              SCHECK_PARTIAL();
2196              cur_is_word = FALSE;
2197              }
2198            else
2199            {            {
2200            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2201    #ifdef SUPPORT_UCP
2202              if (md->use_ucp)
2203                {
2204                if (c == '_') cur_is_word = TRUE; else
2205                  {
2206                  int cat = UCD_CATEGORY(c);
2207                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2208                  }
2209                }
2210              else
2211    #endif
2212            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2213            }            }
2214          }          }
2215        else        else
2216  #endif  #endif
2217    
2218        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2219          consistency with the behaviour of \w we do use it in this case. */
2220    
2221          {          {
2222          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2223            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2224          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2225            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2226              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2227    #ifdef SUPPORT_UCP
2228              if (md->use_ucp)
2229                {
2230                c = eptr[-1];
2231                if (c == '_') prev_is_word = TRUE; else
2232                  {
2233                  int cat = UCD_CATEGORY(c);
2234                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2235                  }
2236                }
2237              else
2238    #endif
2239              prev_is_word = MAX_255(eptr[-1])
2240                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2241              }
2242    
2243            /* Get status of next character */
2244    
2245            if (eptr >= md->end_subject)
2246              {
2247              SCHECK_PARTIAL();
2248              cur_is_word = FALSE;
2249              }
2250            else
2251    #ifdef SUPPORT_UCP
2252            if (md->use_ucp)
2253              {
2254              c = *eptr;
2255              if (c == '_') cur_is_word = TRUE; else
2256                {
2257                int cat = UCD_CATEGORY(c);
2258                cur_is_word = (cat == ucp_L || cat == ucp_N);
2259                }
2260              }
2261            else
2262    #endif
2263            cur_is_word = MAX_255(*eptr)
2264              && ((md->ctypes[*eptr] & ctype_word) != 0);
2265          }          }
2266    
2267        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1418  for (;;) Line 2272  for (;;)
2272        }        }
2273      break;      break;
2274    
2275      /* Match a single character type; inline for speed */      /* Match any single character type except newline; have to take care with
2276        CRLF newlines and partial matching. */
2277    
2278      case OP_ANY:      case OP_ANY:
2279      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2280        if (md->partial != 0 &&
2281            eptr + 1 >= md->end_subject &&
2282            NLBLOCK->nltype == NLTYPE_FIXED &&
2283            NLBLOCK->nllen == 2 &&
2284            *eptr == NLBLOCK->nl[0])
2285        {        {
2286        if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);        md->hitend = TRUE;
2287          if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2288        }        }
2289      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);  
2290      if (utf8)      /* Fall through */
2291        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;  
2292        /* Match any single character whatsoever. */
2293    
2294        case OP_ALLANY:
2295        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2296          {                            /* not be updated before SCHECK_PARTIAL. */
2297          SCHECK_PARTIAL();
2298          RRETURN(MATCH_NOMATCH);
2299          }
2300        eptr++;
2301    #ifdef SUPPORT_UTF
2302        if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2303    #endif
2304      ecode++;      ecode++;
2305      break;      break;
2306    
# Line 1435  for (;;) Line 2308  for (;;)
2308      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2309    
2310      case OP_ANYBYTE:      case OP_ANYBYTE:
2311      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2312          {                            /* not be updated before SCHECK_PARTIAL. */
2313          SCHECK_PARTIAL();
2314          RRETURN(MATCH_NOMATCH);
2315          }
2316        eptr++;
2317      ecode++;      ecode++;
2318      break;      break;
2319    
2320      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2321      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2322          {
2323          SCHECK_PARTIAL();
2324          RRETURN(MATCH_NOMATCH);
2325          }
2326      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2327      if (      if (
2328  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2329         c < 256 &&         c < 256 &&
2330  #endif  #endif
2331         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1453  for (;;) Line 2335  for (;;)
2335      break;      break;
2336    
2337      case OP_DIGIT:      case OP_DIGIT:
2338      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2339          {
2340          SCHECK_PARTIAL();
2341          RRETURN(MATCH_NOMATCH);
2342          }
2343      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2344      if (      if (
2345  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2346         c >= 256 ||         c > 255 ||
2347  #endif  #endif
2348         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2349         )         )
# Line 1466  for (;;) Line 2352  for (;;)
2352      break;      break;
2353    
2354      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2355      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2356          {
2357          SCHECK_PARTIAL();
2358          RRETURN(MATCH_NOMATCH);
2359          }
2360      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2361      if (      if (
2362  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2363         c < 256 &&         c < 256 &&
2364  #endif  #endif
2365         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1479  for (;;) Line 2369  for (;;)
2369      break;      break;
2370    
2371      case OP_WHITESPACE:      case OP_WHITESPACE:
2372      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2373          {
2374          SCHECK_PARTIAL();
2375          RRETURN(MATCH_NOMATCH);
2376          }
2377      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2378      if (      if (
2379  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2380         c >= 256 ||         c > 255 ||
2381  #endif  #endif
2382         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2383         )         )
# Line 1492  for (;;) Line 2386  for (;;)
2386      break;      break;
2387    
2388      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2389      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2390          {
2391          SCHECK_PARTIAL();
2392          RRETURN(MATCH_NOMATCH);
2393          }
2394      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2395      if (      if (
2396  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2397         c < 256 &&         c < 256 &&
2398  #endif  #endif
2399         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1505  for (;;) Line 2403  for (;;)
2403      break;      break;
2404    
2405      case OP_WORDCHAR:      case OP_WORDCHAR:
2406      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2407          {
2408          SCHECK_PARTIAL();
2409          RRETURN(MATCH_NOMATCH);
2410          }
2411      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2412      if (      if (
2413  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2414         c >= 256 ||         c > 255 ||
2415  #endif  #endif
2416         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2417         )         )
# Line 1518  for (;;) Line 2420  for (;;)
2420      break;      break;
2421    
2422      case OP_ANYNL:      case OP_ANYNL:
2423      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2424          {
2425          SCHECK_PARTIAL();
2426          RRETURN(MATCH_NOMATCH);
2427          }
2428      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2429      switch(c)      switch(c)
2430        {        {
2431        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2432        case 0x000d:  
2433        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        case CHAR_CR:
2434          if (eptr >= md->end_subject)
2435            {
2436            SCHECK_PARTIAL();
2437            }
2438          else if (*eptr == CHAR_LF) eptr++;
2439          break;
2440    
2441          case CHAR_LF:
2442        break;        break;
2443        case 0x000a:  
2444        case 0x000b:        case CHAR_VT:
2445        case 0x000c:        case CHAR_FF:
2446        case 0x0085:        case CHAR_NEL:
2447    #ifndef EBCDIC
2448        case 0x2028:        case 0x2028:
2449        case 0x2029:        case 0x2029:
2450    #endif  /* Not EBCDIC */
2451          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2452        break;        break;
2453        }        }
2454      ecode++;      ecode++;
2455      break;      break;
2456    
2457      case OP_NOT_HSPACE:      case OP_NOT_HSPACE:
2458      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2459          {
2460          SCHECK_PARTIAL();
2461          RRETURN(MATCH_NOMATCH);
2462          }
2463      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2464      switch(c)      switch(c)
2465        {        {
2466          HSPACE_CASES: RRETURN(MATCH_NOMATCH);  /* Byte and multibyte cases */
2467        default: break;        default: break;
       case 0x09:      /* HT */  
       case 0x20:      /* SPACE */  
       case 0xa0:      /* NBSP */  
       case 0x1680:    /* OGHAM SPACE MARK */  
       case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */  
       case 0x2000:    /* EN QUAD */  
       case 0x2001:    /* EM QUAD */  
       case 0x2002:    /* EN SPACE */  
       case 0x2003:    /* EM SPACE */  
       case 0x2004:    /* THREE-PER-EM SPACE */  
       case 0x2005:    /* FOUR-PER-EM SPACE */  
       case 0x2006:    /* SIX-PER-EM SPACE */  
       case 0x2007:    /* FIGURE SPACE */  
       case 0x2008:    /* PUNCTUATION SPACE */  
       case 0x2009:    /* THIN SPACE */  
       case 0x200A:    /* HAIR SPACE */  
       case 0x202f:    /* NARROW NO-BREAK SPACE */  
       case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */  
       case 0x3000:    /* IDEOGRAPHIC SPACE */  
       RRETURN(MATCH_NOMATCH);  
2468        }        }
2469      ecode++;      ecode++;
2470      break;      break;
2471    
2472      case OP_HSPACE:      case OP_HSPACE:
2473      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2474          {
2475          SCHECK_PARTIAL();
2476          RRETURN(MATCH_NOMATCH);
2477          }
2478      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2479      switch(c)      switch(c)
2480        {        {
2481          HSPACE_CASES: break;  /* Byte and multibyte cases */
2482        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
       case 0x09:      /* HT */  
       case 0x20:      /* SPACE */  
       case 0xa0:      /* NBSP */  
       case 0x1680:    /* OGHAM SPACE MARK */  
       case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */  
       case 0x2000:    /* EN QUAD */  
       case 0x2001:    /* EM QUAD */  
       case 0x2002:    /* EN SPACE */  
       case 0x2003:    /* EM SPACE */  
       case 0x2004:    /* THREE-PER-EM SPACE */  
       case 0x2005:    /* FOUR-PER-EM SPACE */  
       case 0x2006:    /* SIX-PER-EM SPACE */  
       case 0x2007:    /* FIGURE SPACE */  
       case 0x2008:    /* PUNCTUATION SPACE */  
       case 0x2009:    /* THIN SPACE */  
       case 0x200A:    /* HAIR SPACE */  
       case 0x202f:    /* NARROW NO-BREAK SPACE */  
       case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */  
       case 0x3000:    /* IDEOGRAPHIC SPACE */  
       break;  
2483        }        }
2484      ecode++;      ecode++;
2485      break;      break;
2486    
2487      case OP_NOT_VSPACE:      case OP_NOT_VSPACE:
2488      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2489          {
2490          SCHECK_PARTIAL();
2491          RRETURN(MATCH_NOMATCH);
2492          }
2493      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2494      switch(c)      switch(c)
2495        {        {
2496          VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2497        default: break;        default: break;
       case 0x0a:      /* LF */  
       case 0x0b:      /* VT */  
       case 0x0c:      /* FF */  
       case 0x0d:      /* CR */  
       case 0x85:      /* NEL */  
       case 0x2028:    /* LINE SEPARATOR */  
       case 0x2029:    /* PARAGRAPH SEPARATOR */  
       RRETURN(MATCH_NOMATCH);  
2498        }        }
2499      ecode++;      ecode++;
2500      break;      break;
2501    
2502      case OP_VSPACE:      case OP_VSPACE:
2503      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2504          {
2505          SCHECK_PARTIAL();
2506          RRETURN(MATCH_NOMATCH);
2507          }
2508      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2509      switch(c)      switch(c)
2510        {        {
2511          VSPACE_CASES: break;
2512        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
       case 0x0a:      /* LF */  
       case 0x0b:      /* VT */  
       case 0x0c:      /* FF */  
       case 0x0d:      /* CR */  
       case 0x85:      /* NEL */  
       case 0x2028:    /* LINE SEPARATOR */  
       case 0x2029:    /* PARAGRAPH SEPARATOR */  
       break;  
2513        }        }
2514      ecode++;      ecode++;
2515      break;      break;
# Line 1639  for (;;) Line 2520  for (;;)
2520    
2521      case OP_PROP:      case OP_PROP:
2522      case OP_NOTPROP:      case OP_NOTPROP:
2523      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2524          {
2525          SCHECK_PARTIAL();
2526          RRETURN(MATCH_NOMATCH);
2527          }
2528      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2529        {        {
2530        int chartype, script;        const pcre_uint32 *cp;
2531        int category = _pcre_ucp_findprop(c, &chartype, &script);        const ucd_record *prop = GET_UCD(c);
2532    
2533        switch(ecode[1])        switch(ecode[1])
2534          {          {
# Line 1652  for (;;) Line 2537  for (;;)
2537          break;          break;
2538    
2539          case PT_LAMP:          case PT_LAMP:
2540          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2541               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2542               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2543            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2544           break;          break;
2545    
2546          case PT_GC:          case PT_GC:
2547          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2548            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2549          break;          break;
2550    
2551          case PT_PC:          case PT_PC:
2552          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2553            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2554          break;          break;
2555    
2556          case PT_SC:          case PT_SC:
2557          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2558            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2559          break;          break;
2560    
2561            /* These are specials */
2562    
2563            case PT_ALNUM:
2564            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2565                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2566              RRETURN(MATCH_NOMATCH);
2567            break;
2568    
2569            case PT_SPACE:    /* Perl space */
2570            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2571                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2572                   == (op == OP_NOTPROP))
2573              RRETURN(MATCH_NOMATCH);
2574            break;
2575    
2576            case PT_PXSPACE:  /* POSIX space */
2577            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2578                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2579                 c == CHAR_FF || c == CHAR_CR)
2580                   == (op == OP_NOTPROP))
2581              RRETURN(MATCH_NOMATCH);
2582            break;
2583    
2584            case PT_WORD:
2585            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2586                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2587                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2588              RRETURN(MATCH_NOMATCH);
2589            break;
2590    
2591            case PT_CLIST:
2592            cp = PRIV(ucd_caseless_sets) + prop->caseset;
2593            for (;;)
2594              {
2595              if (c < *cp)
2596                { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; }
2597              if (c == *cp++)
2598                { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } }
2599              }
2600            break;
2601    
2602            /* This should never occur */
2603    
2604          default:          default:
2605          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
2606          }          }
# Line 1685  for (;;) Line 2613  for (;;)
2613      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2614    
2615      case OP_EXTUNI:      case OP_EXTUNI:
2616      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
     GETCHARINCTEST(c, eptr);  
2617        {        {
2618        int chartype, script;        SCHECK_PARTIAL();
2619        int category = _pcre_ucp_findprop(c, &chartype, &script);        RRETURN(MATCH_NOMATCH);
2620        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        }
2621        else
2622          {
2623          int lgb, rgb;
2624          GETCHARINCTEST(c, eptr);
2625          lgb = UCD_GRAPHBREAK(c);
2626        while (eptr < md->end_subject)        while (eptr < md->end_subject)
2627          {          {
2628          int len = 1;          int len = 1;
2629          if (!utf8) c = *eptr; else          if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2630            {          rgb = UCD_GRAPHBREAK(c);
2631            GETCHARLEN(c, eptr, len);          if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2632            }          lgb = rgb;
         category = _pcre_ucp_findprop(c, &chartype, &script);  
         if (category != ucp_M) break;  
2633          eptr += len;          eptr += len;
2634          }          }
2635        }        }
2636        CHECK_PARTIAL();
2637      ecode++;      ecode++;
2638      break;      break;
2639  #endif  #endif  /* SUPPORT_UCP */
2640    
2641    
2642      /* Match a back reference, possibly repeatedly. Look past the end of the      /* Match a back reference, possibly repeatedly. Look past the end of the
# Line 1717  for (;;) Line 2648  for (;;)
2648      loops). */      loops). */
2649    
2650      case OP_REF:      case OP_REF:
2651        {      case OP_REFI:
2652        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2653        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2654        ecode += 1 + IMM2_SIZE;
2655    
2656        /* If the reference is unset, there are two possibilities:
2657    
2658        (a) In the default, Perl-compatible state, set the length negative;
2659        this ensures that every attempt at a match fails. We can't just fail
2660        here, because of the possibility of quantifiers with zero minima.
2661    
2662        (b) If the JavaScript compatibility flag is set, set the length to zero
2663        so that the back reference matches an empty string.
2664    
2665        Otherwise, set the length to the length of what was matched by the
2666        referenced subpattern. */
2667    
2668        /* If the reference is unset, set the length to be longer than the amount      if (offset >= offset_top || md->offset_vector[offset] < 0)
2669        of subject left; this ensures that every attempt at a match fails. We        length = (md->jscript_compat)? 0 : -1;
2670        can't just fail here, because of the possibility of quantifiers with zero      else
2671        minima. */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
   
       length = (offset >= offset_top || md->offset_vector[offset] < 0)?  
         md->end_subject - eptr + 1 :  
         md->offset_vector[offset+1] - md->offset_vector[offset];  
2672    
2673        /* Set up for repetition, or handle the non-repeated case */      /* Set up for repetition, or handle the non-repeated case */
2674    
2675        switch (*ecode)      switch (*ecode)
2676          {        {
2677          case OP_CRSTAR:        case OP_CRSTAR:
2678          case OP_CRMINSTAR:        case OP_CRMINSTAR:
2679          case OP_CRPLUS:        case OP_CRPLUS:
2680          case OP_CRMINPLUS:        case OP_CRMINPLUS:
2681          case OP_CRQUERY:        case OP_CRQUERY:
2682          case OP_CRMINQUERY:        case OP_CRMINQUERY:
2683          c = *ecode++ - OP_CRSTAR;        c = *ecode++ - OP_CRSTAR;
2684          minimize = (c & 1) != 0;        minimize = (c & 1) != 0;
2685          min = rep_min[c];                 /* Pick up values from tables; */        min = rep_min[c];                 /* Pick up values from tables; */
2686          max = rep_max[c];                 /* zero for max => infinity */        max = rep_max[c];                 /* zero for max => infinity */
2687          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2688          break;        break;
2689    
2690          case OP_CRRANGE:        case OP_CRRANGE:
2691          case OP_CRMINRANGE:        case OP_CRMINRANGE:
2692          minimize = (*ecode == OP_CRMINRANGE);        minimize = (*ecode == OP_CRMINRANGE);
2693          min = GET2(ecode, 1);        min = GET2(ecode, 1);
2694          max = GET2(ecode, 3);        max = GET2(ecode, 1 + IMM2_SIZE);
2695          if (max == 0) max = INT_MAX;        if (max == 0) max = INT_MAX;
2696          ecode += 5;        ecode += 1 + 2 * IMM2_SIZE;
2697          break;        break;
2698    
2699          default:               /* No repeat follows */        default:               /* No repeat follows */
2700          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2701          eptr += length;          {
2702          continue;              /* With the main loop */          if (length == -2) eptr = md->end_subject;   /* Partial match */
2703            CHECK_PARTIAL();
2704            RRETURN(MATCH_NOMATCH);
2705          }          }
2706          eptr += length;
2707          continue;              /* With the main loop */
2708          }
2709    
2710        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2711        main loop. */      zero, just continue with the main loop. If the length is negative, it
2712        means the reference is unset in non-Java-compatible mode. If the minimum is
2713        zero, we can continue at the same level without recursion. For any other
2714        minimum, carrying on will result in NOMATCH. */
2715    
2716        if (length == 0) continue;      if (length == 0) continue;
2717        if (length < 0 && min == 0) continue;
2718    
2719        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2720        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2721        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2722    
2723        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2724          {
2725          int slength;
2726          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2727          {          {
2728          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          if (slength == -2) eptr = md->end_subject;   /* Partial match */
2729          eptr += length;          CHECK_PARTIAL();
2730            RRETURN(MATCH_NOMATCH);
2731          }          }
2732          eptr += slength;
2733          }
2734    
2735        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2736        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2737    
2738        if (min == max) continue;      if (min == max) continue;
2739    
2740        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2741    
2742        if (minimize)      if (minimize)
2743          {
2744          for (fi = min;; fi++)
2745          {          {
2746          for (fi = min;; fi++)          int slength;
2747            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2748            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2749            if (fi >= max) RRETURN(MATCH_NOMATCH);
2750            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2751            {            {
2752            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2753            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2754            if (fi >= max || !match_ref(offset, eptr, length, md, ims))            RRETURN(MATCH_NOMATCH);
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2755            }            }
2756          /* Control never gets here */          eptr += slength;
2757          }          }
2758          /* Control never gets here */
2759          }
2760    
2761        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2762    
2763        else      else
2764          {
2765          pp = eptr;
2766          for (i = min; i < max; i++)
2767          {          {
2768          pp = eptr;          int slength;
2769          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)  
2770            {            {
2771            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2772            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            the soft partial matching case. */
2773            eptr -= length;  
2774              if (slength == -2 && md->partial != 0 &&
2775                  md->end_subject > md->start_used_ptr)
2776                {
2777                md->hitend = TRUE;
2778                if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2779                }
2780              break;
2781            }            }
2782          RRETURN(MATCH_NOMATCH);          eptr += slength;
2783            }
2784    
2785          while (eptr >= pp)
2786            {
2787            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2788            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2789            eptr -= length;
2790          }          }
2791          RRETURN(MATCH_NOMATCH);
2792        }        }
2793      /* Control never gets here */      /* Control never gets here */
2794    
   
   
2795      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2796      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,
2797      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 1834  for (;;) Line 2806  for (;;)
2806      case OP_NCLASS:      case OP_NCLASS:
2807      case OP_CLASS:      case OP_CLASS:
2808        {        {
2809          /* The data variable is saved across frames, so the byte map needs to
2810          be stored there. */
2811    #define BYTE_MAP ((pcre_uint8 *)data)
2812        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2813        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2814    
2815        switch (*ecode)        switch (*ecode)
2816          {          {
# Line 1856  for (;;) Line 2831  for (;;)
2831          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2832          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2833          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2834          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2835          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2836          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2837          break;          break;
2838    
2839          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1868  for (;;) Line 2843  for (;;)
2843    
2844        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2845    
2846  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2847        /* UTF-8 mode */        if (utf)
       if (utf8)  
2848          {          {
2849          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2850            {            {
2851            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2852                {
2853                SCHECK_PARTIAL();
2854                RRETURN(MATCH_NOMATCH);
2855                }
2856            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2857            if (c > 255)            if (c > 255)
2858              {              {
2859              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2860              }              }
2861            else            else
2862              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2863            }            }
2864          }          }
2865        else        else
2866  #endif  #endif
2867        /* Not UTF-8 mode */        /* Not UTF mode */
2868          {          {
2869          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2870            {            {
2871            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2872                {
2873                SCHECK_PARTIAL();
2874                RRETURN(MATCH_NOMATCH);
2875                }
2876            c = *eptr++;            c = *eptr++;
2877            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2878              if (c > 255)
2879                {
2880                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2881                }
2882              else
2883    #endif
2884                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2885            }            }
2886          }          }
2887    
# Line 1908  for (;;) Line 2895  for (;;)
2895    
2896        if (minimize)        if (minimize)
2897          {          {
2898  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2899          /* UTF-8 mode */          if (utf)
         if (utf8)  
2900            {            {
2901            for (fi = min;; fi++)            for (fi = min;; fi++)
2902              {              {
2903              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2904              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2905              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2906                if (eptr >= md->end_subject)
2907                  {
2908                  SCHECK_PARTIAL();
2909                  RRETURN(MATCH_NOMATCH);
2910                  }
2911              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2912              if (c > 255)              if (c > 255)
2913                {                {
2914                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2915                }                }
2916              else              else
2917                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2918              }              }
2919            }            }
2920          else          else
2921  #endif  #endif
2922          /* Not UTF-8 mode */          /* Not UTF mode */
2923            {            {
2924            for (fi = min;; fi++)            for (fi = min;; fi++)
2925              {              {
2926              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2927              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2928              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2929                if (eptr >= md->end_subject)
2930                  {
2931                  SCHECK_PARTIAL();
2932                  RRETURN(MATCH_NOMATCH);
2933                  }
2934              c = *eptr++;              c = *eptr++;
2935              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2936                if (c > 255)
2937                  {
2938                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2939                  }
2940                else
2941    #endif
2942                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2943              }              }
2944            }            }
2945          /* Control never gets here */          /* Control never gets here */
# Line 1950  for (;;) Line 2951  for (;;)
2951          {          {
2952          pp = eptr;          pp = eptr;
2953    
2954  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2955          /* UTF-8 mode */          if (utf)
         if (utf8)  
2956            {            {
2957            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2958              {              {
2959              int len = 1;              int len = 1;
2960              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2961                  {
2962                  SCHECK_PARTIAL();
2963                  break;
2964                  }
2965              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2966              if (c > 255)              if (c > 255)
2967                {                {
2968                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2969                }                }
2970              else              else
2971                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
2972              eptr += len;              eptr += len;
2973              }              }
2974            for (;;)            for (;;)
2975              {              {
2976              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2977              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2978              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2979              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1979  for (;;) Line 2981  for (;;)
2981            }            }
2982          else          else
2983  #endif  #endif
2984            /* Not UTF-8 mode */            /* Not UTF mode */
2985            {            {
2986            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2987              {              {
2988              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2989                  {
2990                  SCHECK_PARTIAL();
2991                  break;
2992                  }
2993              c = *eptr;              c = *eptr;
2994              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
2995                if (c > 255)
2996                  {
2997                  if (op == OP_CLASS) break;
2998                  }
2999                else
3000    #endif
3001                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3002              eptr++;              eptr++;
3003              }              }
3004            while (eptr >= pp)            while (eptr >= pp)
3005              {              {
3006              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3007              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3008              eptr--;              eptr--;
3009              }              }
# Line 1998  for (;;) Line 3011  for (;;)
3011    
3012          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3013          }          }
3014    #undef BYTE_MAP
3015        }        }
3016      /* Control never gets here */      /* Control never gets here */
3017    
3018    
3019      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
3020      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
3021        mode, because Unicode properties are supported in non-UTF-8 mode. */
3022    
3023  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3024      case OP_XCLASS:      case OP_XCLASS:
3025        {        {
3026        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 2030  for (;;) Line 3045  for (;;)
3045          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3046          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3047          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3048          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3049          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3050          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3051          break;          break;
3052    
3053          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 2044  for (;;) Line 3059  for (;;)
3059    
3060        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3061          {          {
3062          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3063          GETCHARINC(c, eptr);            {
3064          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
3065              RRETURN(MATCH_NOMATCH);
3066              }
3067            GETCHARINCTEST(c, eptr);
3068            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3069          }          }
3070    
3071        /* 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 2061  for (;;) Line 3080  for (;;)
3080          {          {
3081          for (fi = min;; fi++)          for (fi = min;; fi++)
3082            {            {
3083            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3084            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3085            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3086            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3087            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3088                SCHECK_PARTIAL();
3089                RRETURN(MATCH_NOMATCH);
3090                }
3091              GETCHARINCTEST(c, eptr);
3092              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3093            }            }
3094          /* Control never gets here */          /* Control never gets here */
3095          }          }
# Line 2078  for (;;) Line 3102  for (;;)
3102          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3103            {            {
3104            int len = 1;            int len = 1;
3105            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3106            GETCHARLEN(c, eptr, len);              {
3107            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3108                break;
3109                }
3110    #ifdef SUPPORT_UTF
3111              GETCHARLENTEST(c, eptr, len);
3112    #else
3113              c = *eptr;
3114    #endif
3115              if (!PRIV(xclass)(c, data, utf)) break;
3116            eptr += len;            eptr += len;
3117            }            }
3118          for(;;)          for(;;)
3119            {            {
3120            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3121            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3122            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3123            if (utf8) BACKCHAR(eptr);  #ifdef SUPPORT_UTF
3124              if (utf) BACKCHAR(eptr);
3125    #endif
3126            }            }
3127          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3128          }          }
# Line 2100  for (;;) Line 3134  for (;;)
3134      /* Match a single character, casefully */      /* Match a single character, casefully */
3135    
3136      case OP_CHAR:      case OP_CHAR:
3137  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3138      if (utf8)      if (utf)
3139        {        {
3140        length = 1;        length = 1;
3141        ecode++;        ecode++;
3142        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3143        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3144            {
3145            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3146            RRETURN(MATCH_NOMATCH);
3147            }
3148        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);
3149        }        }
3150      else      else
3151  #endif  #endif
3152        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3153        {        {
3154        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3155            {
3156            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3157            RRETURN(MATCH_NOMATCH);
3158            }
3159        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3160        ecode += 2;        ecode += 2;
3161        }        }
3162      break;      break;
3163    
3164      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3165        subject, give up immediately. */
3166    
3167        case OP_CHARI:
3168        if (eptr >= md->end_subject)
3169          {
3170          SCHECK_PARTIAL();
3171          RRETURN(MATCH_NOMATCH);
3172          }
3173    
3174      case OP_CHARNC:  #ifdef SUPPORT_UTF
3175  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3176        {        {
3177        length = 1;        length = 1;
3178        ecode++;        ecode++;
3179        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3180    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3181        /* 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
3182        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3183          fast lookup table. We know that there is at least one byte left in the
3184          subject. */
3185    
3186        if (fc < 128)        if (fc < 128)
3187          {          {
3188          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          if (md->lcc[fc]
3189                != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3190            ecode++;
3191            eptr++;
3192          }          }
3193    
3194        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3195          use the value of "length" to check for sufficient bytes left, because the
3196          other case of the character may have more or fewer bytes.  */
3197    
3198        else        else
3199          {          {
# Line 2154  for (;;) Line 3207  for (;;)
3207          if (fc != dc)          if (fc != dc)
3208            {            {
3209  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3210            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3211  #endif  #endif
3212              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3213            }            }
3214          }          }
3215        }        }
3216      else      else
3217  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3218    
3219      /* Non-UTF-8 mode */      /* Not UTF mode */
3220        {        {
3221        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3222        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3223          eptr++;
3224        ecode += 2;        ecode += 2;
3225        }        }
3226      break;      break;
# Line 2174  for (;;) Line 3228  for (;;)
3228      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3229    
3230      case OP_EXACT:      case OP_EXACT:
3231        case OP_EXACTI:
3232      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3233      ecode += 3;      ecode += 1 + IMM2_SIZE;
3234      goto REPEATCHAR;      goto REPEATCHAR;
3235    
3236      case OP_POSUPTO:      case OP_POSUPTO:
3237        case OP_POSUPTOI:
3238      possessive = TRUE;      possessive = TRUE;
3239      /* Fall through */      /* Fall through */
3240    
3241      case OP_UPTO:      case OP_UPTO:
3242        case OP_UPTOI:
3243      case OP_MINUPTO:      case OP_MINUPTO:
3244        case OP_MINUPTOI:
3245      min = 0;      min = 0;
3246      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3247      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3248      ecode += 3;      ecode += 1 + IMM2_SIZE;
3249      goto REPEATCHAR;      goto REPEATCHAR;
3250    
3251      case OP_POSSTAR:      case OP_POSSTAR:
3252        case OP_POSSTARI:
3253      possessive = TRUE;      possessive = TRUE;
3254      min = 0;      min = 0;
3255      max = INT_MAX;      max = INT_MAX;
# Line 2198  for (;;) Line 3257  for (;;)
3257      goto REPEATCHAR;      goto REPEATCHAR;
3258    
3259      case OP_POSPLUS:      case OP_POSPLUS:
3260        case OP_POSPLUSI:
3261      possessive = TRUE;      possessive = TRUE;
3262      min = 1;      min = 1;
3263      max = INT_MAX;      max = INT_MAX;
# Line 2205  for (;;) Line 3265  for (;;)
3265      goto REPEATCHAR;      goto REPEATCHAR;
3266    
3267      case OP_POSQUERY:      case OP_POSQUERY:
3268        case OP_POSQUERYI:
3269      possessive = TRUE;      possessive = TRUE;
3270      min = 0;      min = 0;
3271      max = 1;      max = 1;
# Line 2212  for (;;) Line 3273  for (;;)
3273      goto REPEATCHAR;      goto REPEATCHAR;
3274    
3275      case OP_STAR:      case OP_STAR:
3276        case OP_STARI:
3277      case OP_MINSTAR:      case OP_MINSTAR:
3278        case OP_MINSTARI:
3279      case OP_PLUS:      case OP_PLUS:
3280        case OP_PLUSI:
3281      case OP_MINPLUS:      case OP_MINPLUS:
3282        case OP_MINPLUSI:
3283      case OP_QUERY:      case OP_QUERY:
3284        case OP_QUERYI:
3285      case OP_MINQUERY:      case OP_MINQUERY:
3286      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3287        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3288      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3289      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3290      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3291      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3292    
3293      /* 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. */  
3294    
3295      REPEATCHAR:      REPEATCHAR:
3296  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3297      if (utf8)      if (utf)
3298        {        {
3299        length = 1;        length = 1;
3300        charptr = ecode;        charptr = ecode;
3301        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3302        ecode += length;        ecode += length;
3303    
3304        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 2244  for (;;) Line 3308  for (;;)
3308          {          {
3309  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3310          unsigned int othercase;          unsigned int othercase;
3311          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3312              (othercase = _pcre_ucp_othercase(fc)) != NOTACHAR)              (othercase = UCD_OTHERCASE(fc)) != fc)
3313            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3314          else oclength = 0;          else oclength = 0;
3315  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3316    
3317          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3318            {            {
3319            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3320                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3321  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3322            /* Need braces because of following else */            else if (oclength > 0 &&
3323            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                     eptr <= md->end_subject - oclength &&
3324                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3325    #endif  /* SUPPORT_UCP */
3326            else            else
3327              {              {
3328              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3329              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3330              }              }
 #else   /* without SUPPORT_UCP */  
           else { RRETURN(MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3331            }            }
3332    
3333          if (min == max) continue;          if (min == max) continue;
# Line 2272  for (;;) Line 3336  for (;;)
3336            {            {
3337            for (fi = min;; fi++)            for (fi = min;; fi++)
3338              {              {
3339              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM22);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3340              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3341              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3342              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3343                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3344  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3345              /* Need braces because of following else */              else if (oclength > 0 &&
3346              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                       eptr <= md->end_subject - oclength &&
3347                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3348    #endif  /* SUPPORT_UCP */
3349              else              else
3350                {                {
3351                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3352                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3353                }                }
 #else   /* without SUPPORT_UCP */  
             else { RRETURN (MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3354              }              }
3355            /* Control never gets here */            /* Control never gets here */
3356            }            }
# Line 2296  for (;;) Line 3360  for (;;)
3360            pp = eptr;            pp = eptr;
3361            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3362              {              {
3363              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3364              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3365  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3366              else if (oclength == 0) break;              else if (oclength > 0 &&
3367                         eptr <= md->end_subject - oclength &&
3368                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3369    #endif  /* SUPPORT_UCP */
3370              else              else
3371                {                {
3372                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3373                eptr += oclength;                break;
3374                }                }
 #else   /* without SUPPORT_UCP */  
             else break;  
 #endif  /* SUPPORT_UCP */  
3375              }              }
3376    
3377            if (possessive) continue;            if (possessive) continue;
3378    
3379            for(;;)            for(;;)
3380             {              {
3381             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM23);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3382             if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3383             if (eptr == pp) RRETURN(MATCH_NOMATCH);              if (eptr == pp) { RRETURN(MATCH_NOMATCH); }
3384  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3385             eptr--;              eptr--;
3386             BACKCHAR(eptr);              BACKCHAR(eptr);
3387  #else   /* without SUPPORT_UCP */  #else   /* without SUPPORT_UCP */
3388             eptr -= length;              eptr -= length;
3389  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3390             }              }
3391            }            }
3392          /* Control never gets here */          /* Control never gets here */
3393          }          }
# Line 2332  for (;;) Line 3397  for (;;)
3397        value of fc will always be < 128. */        value of fc will always be < 128. */
3398        }        }
3399      else      else
3400  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3401          /* 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);  
3402        fc = *ecode++;        fc = *ecode++;
       }  
3403    
3404      /* 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
3405      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
3406      caseful cases, for speed, since matching characters is likely to be quite      caseful cases, for speed, since matching characters is likely to be quite
3407      common. First, ensure the minimum number of matches are present. If min =      common. First, ensure the minimum number of matches are present. If min =
3408      max, continue at the same level without recursing. Otherwise, if      max, continue at the same level without recursing. Otherwise, if
# Line 2350  for (;;) Line 3411  for (;;)
3411      maximizing, find the maximum number of characters and work backwards. */      maximizing, find the maximum number of characters and work backwards. */
3412    
3413      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", fc, min, max,
3414        max, eptr));        max, (char *)eptr));
3415    
3416      if ((ims & PCRE_CASELESS) != 0)      if (op >= OP_STARI)  /* Caseless */
3417        {        {
3418        fc = md->lcc[fc];  #ifdef COMPILE_PCRE8
3419          /* fc must be < 128 if UTF is enabled. */
3420          foc = md->fcc[fc];
3421    #else
3422    #ifdef SUPPORT_UTF
3423    #ifdef SUPPORT_UCP
3424          if (utf && fc > 127)
3425            foc = UCD_OTHERCASE(fc);
3426    #else
3427          if (utf && fc > 127)
3428            foc = fc;
3429    #endif /* SUPPORT_UCP */
3430          else
3431    #endif /* SUPPORT_UTF */
3432            foc = TABLE_GET(fc, md->fcc, fc);
3433    #endif /* COMPILE_PCRE8 */
3434    
3435        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3436          if (fc != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          {
3437            if (eptr >= md->end_subject)
3438              {
3439              SCHECK_PARTIAL();
3440              RRETURN(MATCH_NOMATCH);
3441              }
3442            if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3443            eptr++;
3444            }
3445        if (min == max) continue;        if (min == max) continue;
3446        if (minimize)        if (minimize)
3447          {          {
3448          for (fi = min;; fi++)          for (fi = min;; fi++)
3449            {            {
3450            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM24);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM24);
3451            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3452            if (fi >= max || eptr >= md->end_subject ||            if (fi >= max) RRETURN(MATCH_NOMATCH);
3453                fc != md->lcc[*eptr++])            if (eptr >= md->end_subject)
3454                {
3455                SCHECK_PARTIAL();
3456              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3457                }
3458              if (fc != *eptr && foc != *eptr) RRETURN(MATCH_NOMATCH);
3459              eptr++;
3460            }            }
3461          /* Control never gets here */          /* Control never gets here */
3462          }          }
# Line 2375  for (;;) Line 3465  for (;;)
3465          pp = eptr;          pp = eptr;
3466          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3467            {            {
3468            if (eptr >= md->end_subject || fc != md->lcc[*eptr]) break;            if (eptr >= md->end_subject)
3469                {
3470                SCHECK_PARTIAL();
3471                break;
3472                }
3473              if (fc != *eptr && foc != *eptr) break;
3474            eptr++;            eptr++;
3475            }            }
3476    
3477          if (possessive) continue;          if (possessive) continue;
3478    
3479          while (eptr >= pp)          while (eptr >= pp)
3480            {            {
3481            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM25);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM25);
3482            eptr--;            eptr--;
3483            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3484            }            }
# Line 2394  for (;;) Line 3491  for (;;)
3491    
3492      else      else
3493        {        {
3494        for (i = 1; i <= min; i++) if (fc != *eptr++) RRETURN(MATCH_NOMATCH);        for (i = 1; i <= min; i++)
3495            {
3496            if (eptr >= md->end_subject)
3497              {
3498              SCHECK_PARTIAL();
3499              RRETURN(MATCH_NOMATCH);
3500              }
3501            if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
3502            }
3503    
3504        if (min == max) continue;        if (min == max) continue;
3505    
3506        if (minimize)        if (minimize)
3507          {          {
3508          for (fi = min;; fi++)          for (fi = min;; fi++)
3509            {            {
3510            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM26);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM26);
3511            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3512            if (fi >= max || eptr >= md->end_subject || fc != *eptr++)            if (fi >= max) RRETURN(MATCH_NOMATCH);
3513              if (eptr >= md->end_subject)
3514                {
3515                SCHECK_PARTIAL();
3516              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3517                }
3518              if (fc != *eptr++) RRETURN(MATCH_NOMATCH);
3519            }            }
3520          /* Control never gets here */          /* Control never gets here */
3521          }          }
# Line 2412  for (;;) Line 3524  for (;;)
3524          pp = eptr;          pp = eptr;
3525          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3526            {            {
3527            if (eptr >= md->end_subject || fc != *eptr) break;            if (eptr >= md->end_subject)
3528                {
3529                SCHECK_PARTIAL();
3530                break;
3531                }
3532              if (fc != *eptr) break;
3533            eptr++;            eptr++;
3534            }            }
3535          if (possessive) continue;          if (possessive) continue;
3536    
3537          while (eptr >= pp)          while (eptr >= pp)
3538            {            {
3539            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM27);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM27);
3540            eptr--;            eptr--;
3541            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3542            }            }
# Line 2431  for (;;) Line 3549  for (;;)
3549      checking can be multibyte. */      checking can be multibyte. */
3550    
3551      case OP_NOT:      case OP_NOT:
3552      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      case OP_NOTI:
3553      ecode++;      if (eptr >= md->end_subject)
     GETCHARINCTEST(c, eptr);  
     if ((ims & PCRE_CASELESS) != 0)  
3554        {        {
3555  #ifdef SUPPORT_UTF8        SCHECK_PARTIAL();
3556        if (c < 256)        RRETURN(MATCH_NOMATCH);
3557  #endif        }
3558        c = md->lcc[c];  #ifdef SUPPORT_UTF
3559        if (md->lcc[*ecode++] == c) RRETURN(MATCH_NOMATCH);      if (utf)
3560          {
3561          register unsigned int ch, och;
3562    
3563          ecode++;
3564          GETCHARINC(ch, ecode);
3565          GETCHARINC(c, eptr);
3566    
3567          if (op == OP_NOT)
3568            {
3569            if (ch == c) RRETURN(MATCH_NOMATCH);
3570            }
3571          else
3572            {
3573    #ifdef SUPPORT_UCP
3574            if (ch > 127)
3575              och = UCD_OTHERCASE(ch);
3576    #else
3577            if (ch > 127)
3578              och = ch;
3579    #endif /* SUPPORT_UCP */
3580            else
3581              och = TABLE_GET(ch, md->fcc, ch);
3582            if (ch == c || och == c) RRETURN(MATCH_NOMATCH);
3583            }
3584        }        }
3585      else      else
3586    #endif
3587        {        {
3588        if (*ecode++ == c) RRETURN(MATCH_NOMATCH);        register unsigned int ch = ecode[1];
3589          c = *eptr++;
3590          if (ch == c || (op == OP_NOTI && TABLE_GET(ch, md->fcc, ch) == c))
3591            RRETURN(MATCH_NOMATCH);
3592          ecode += 2;
3593        }        }
3594      break;      break;
3595    
# Line 2456  for (;;) Line 3601  for (;;)
3601      about... */      about... */
3602    
3603      case OP_NOTEXACT:      case OP_NOTEXACT:
3604        case OP_NOTEXACTI:
3605      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3606      ecode += 3;      ecode += 1 + IMM2_SIZE;
3607      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3608    
3609      case OP_NOTUPTO:      case OP_NOTUPTO:
3610        case OP_NOTUPTOI:
3611      case OP_NOTMINUPTO:      case OP_NOTMINUPTO:
3612        case OP_NOTMINUPTOI:
3613      min = 0;      min = 0;
3614      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3615      minimize = *ecode == OP_NOTMINUPTO;      minimize = *ecode == OP_NOTMINUPTO || *ecode == OP_NOTMINUPTOI;
3616      ecode += 3;      ecode += 1 + IMM2_SIZE;
3617      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3618    
3619      case OP_NOTPOSSTAR:      case OP_NOTPOSSTAR:
3620        case OP_NOTPOSSTARI:
3621      possessive = TRUE;      possessive = TRUE;
3622      min = 0;      min = 0;
3623      max = INT_MAX;      max = INT_MAX;
# Line 2476  for (;;) Line 3625  for (;;)
3625      goto REPEATNOTCHAR;      goto REPEATNOTCHAR;
3626    
3627      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
3628        case OP_NOTPOSPLUSI:
3629      possessive = TRUE;      possessive = TRUE;
3630 &nb