/[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 199 by ph10, Tue Jul 31 14:39:09 2007 UTC revision 1361 by ph10, Fri Sep 6 17:47:32 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 37  POSSIBILITY OF SUCH DAMAGE. Line 37  POSSIBILITY OF SUCH DAMAGE.
37  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
38  */  */
39    
   
40  /* This module contains pcre_exec(), the externally visible function that does  /* This module contains pcre_exec(), the externally visible function that does
41  pattern matching using an NFA algorithm, trying to mimic Perl as closely as  pattern matching using an NFA algorithm, trying to mimic Perl as closely as
42  possible. There are also some static supporting functions. */  possible. There are also some static supporting functions. */
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 */
49  #define PSSTART start_subject  /* Field containing processed string start */  #define PSSTART start_subject  /* Field containing processed string start */
# Line 57  possible. There are also some static sup Line 56  possible. There are also some static sup
56  #undef min  #undef min
57  #undef max  #undef max
58    
59  /* Flag bits for the match() function */  /* The md->capture_last field uses the lower 16 bits for the last captured
60    substring (which can never be greater than 65535) and a bit in the top half
61    to mean "capture vector overflowed". This odd way of doing things was
62    implemented when it was realized that preserving and restoring the overflow bit
63    whenever the last capture number was saved/restored made for a neater
64    interface, and doing it this way saved on (a) another variable, which would
65    have increased the stack frame size (a big NO-NO in PCRE) and (b) another
66    separate set of save/restore instructions. The following defines are used in
67    implementing this. */
68    
69    #define CAPLMASK    0x0000ffff    /* The bits used for last_capture */
70    #define OVFLMASK    0xffff0000    /* The bits used for the overflow flag */
71    #define OVFLBIT     0x00010000    /* The bit that is set for overflow */
72    
73    /* Values for setting in md->match_function_type to indicate two special types
74    of call to match(). We do it this way to save on using another stack variable,
75    as stack usage is to be discouraged. */
76    
77  #define match_condassert     0x01  /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
78  #define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
79    
80  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
81  defined PCRE_ERROR_xxx codes, which are all negative. */  defined PCRE_ERROR_xxx codes, which are all negative. */
# Line 68  defined PCRE_ERROR_xxx codes, which are Line 83  defined PCRE_ERROR_xxx codes, which are
83  #define MATCH_MATCH        1  #define MATCH_MATCH        1
84  #define MATCH_NOMATCH      0  #define MATCH_NOMATCH      0
85    
86    /* Special internal returns from the match() function. Make them sufficiently
87    negative to avoid the external error codes. */
88    
89    #define MATCH_ACCEPT       (-999)
90    #define MATCH_KETRPOS      (-998)
91    #define MATCH_ONCE         (-997)
92    /* The next 5 must be kept together and in sequence so that a test that checks
93    for any one of them can use a range. */
94    #define MATCH_COMMIT       (-996)
95    #define MATCH_PRUNE        (-995)
96    #define MATCH_SKIP         (-994)
97    #define MATCH_SKIP_ARG     (-993)
98    #define MATCH_THEN         (-992)
99    #define MATCH_BACKTRACK_MAX MATCH_THEN
100    #define MATCH_BACKTRACK_MIN MATCH_COMMIT
101    
102  /* Maximum number of ints of offset to save on the stack for recursive calls.  /* Maximum number of ints of offset to save on the stack for recursive calls.
103  If the offset vector is bigger, malloc is used. This should be a multiple of 3,  If the offset vector is bigger, malloc is used. This should be a multiple of 3,
104  because the offset vector is always a multiple of 3 long. */  because the offset vector is always a multiple of 3 long. */
# Line 79  because the offset vector is always a mu Line 110  because the offset vector is always a mu
110  static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };  static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };
111  static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };  static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };
112    
113    #ifdef PCRE_DEBUG
   
 #ifdef DEBUG  
114  /*************************************************  /*************************************************
115  *        Debugging function to print chars       *  *        Debugging function to print chars       *
116  *************************************************/  *************************************************/
# Line 99  Returns:     nothing Line 128  Returns:     nothing
128  */  */
129    
130  static void  static void
131  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)  pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md)
132  {  {
133  unsigned int c;  pcre_uint32 c;
134    BOOL utf = md->utf;
135  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
136  while (length-- > 0)  while (length-- > 0)
137    if (isprint(c = *(p++))) printf("%c", c); else printf("\\x%02x", c);    if (isprint(c = RAWUCHARINCTEST(p))) printf("%c", (char)c); else printf("\\x{%02x}", c);
138  }  }
139  #endif  #endif
140    
# Line 114  while (length-- > 0) Line 144  while (length-- > 0)
144  *          Match a back-reference                *  *          Match a back-reference                *
145  *************************************************/  *************************************************/
146    
147  /* If a back reference hasn't been set, the length that is passed is greater  /* Normally, if a back reference hasn't been set, the length that is passed is
148  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
149    the length passed is zero. Note that in caseless UTF-8 mode, the number of
150    subject bytes matched may be different to the number of reference bytes.
151    
152  Arguments:  Arguments:
153    offset      index into the offset vector    offset      index into the offset vector
154    eptr        points into the subject    eptr        pointer into the subject
155    length      length to be matched    length      length of reference to be matched (number of bytes)
156    md          points to match data block    md          points to match data block
157    ims         the ims flags    caseless    TRUE if caseless
158    
159  Returns:      TRUE if matched  Returns:      >= 0 the number of subject bytes matched
160                  -1 no match
161                  -2 partial match; always given if at end subject
162  */  */
163    
164  static BOOL  static int
165  match_ref(int offset, register USPTR eptr, int length, match_data *md,  match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
166    unsigned long int ims)    BOOL caseless)
167  {  {
168  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
169    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
170    #ifdef SUPPORT_UTF
171    BOOL utf = md->utf;
172    #endif
173    
174  #ifdef DEBUG  #ifdef PCRE_DEBUG
175  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
176    printf("matching subject <null>");    printf("matching subject <null>");
177  else  else
# Line 146  pchars(p, length, FALSE, md); Line 184  pchars(p, length, FALSE, md);
184  printf("\n");  printf("\n");
185  #endif  #endif
186    
187  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible - in that
188    case the length is passed as zero). */
189    
190  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
191    
192  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
193    properly if Unicode properties are supported. Otherwise, we can check only
194    ASCII characters. */
195    
196  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
197    {    {
198    while (length-- > 0)  #ifdef SUPPORT_UTF
199      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
200      if (utf)
201        {
202        /* Match characters up to the end of the reference. NOTE: the number of
203        data units matched may differ, because in UTF-8 there are some characters
204        whose upper and lower case versions code have different numbers of bytes.
205        For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65
206        (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a
207        sequence of two of the latter. It is important, therefore, to check the
208        length along the reference, not along the subject (earlier code did this
209        wrong). */
210    
211        PCRE_PUCHAR endptr = p + length;
212        while (p < endptr)
213          {
214          pcre_uint32 c, d;
215          const ucd_record *ur;
216          if (eptr >= md->end_subject) return -2;   /* Partial match */
217          GETCHARINC(c, eptr);
218          GETCHARINC(d, p);
219          ur = GET_UCD(d);
220          if (c != d && c != d + ur->other_case)
221            {
222            const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
223            for (;;)
224              {
225              if (c < *pp) return -1;
226              if (c == *pp++) break;
227              }
228            }
229          }
230        }
231      else
232    #endif
233    #endif
234    
235      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
236      is no UCP support. */
237        {
238        while (length-- > 0)
239          {
240          pcre_uint32 cc, cp;
241          if (eptr >= md->end_subject) return -2;   /* Partial match */
242          cc = RAWUCHARTEST(eptr);
243          cp = RAWUCHARTEST(p);
244          if (TABLE_GET(cp, md->lcc, cp) != TABLE_GET(cc, md->lcc, cc)) return -1;
245          p++;
246          eptr++;
247          }
248        }
249    }    }
250    
251    /* In the caseful case, we can just compare the bytes, whether or not we
252    are in UTF-8 mode. */
253    
254  else  else
255    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
256      while (length-- > 0)
257        {
258        if (eptr >= md->end_subject) return -2;   /* Partial match */
259        if (RAWUCHARINCTEST(p) != RAWUCHARINCTEST(eptr)) return -1;
260        }
261      }
262    
263  return TRUE;  return (int)(eptr - eptr_start);
264  }  }
265    
266    
# Line 203  variable instead of being passed in the Line 303  variable instead of being passed in the
303  ****************************************************************************  ****************************************************************************
304  ***************************************************************************/  ***************************************************************************/
305    
306    /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
307  /* Numbers for RMATCH calls */  below must be updated in sync.  */
308    
309  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
310         RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,         RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
311         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
312         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
313         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50 };         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
314           RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
315           RM61,  RM62, RM63, RM64, RM65, RM66, RM67, RM68 };
316    
317  /* These versions of the macros use the stack, as normal. There are debugging  /* These versions of the macros use the stack, as normal. There are debugging
318  versions and production versions. Note that the "rw" argument of RMATCH isn't  versions and production versions. Note that the "rw" argument of RMATCH isn't
319  actuall used in this definition. */  actually used in this definition. */
320    
321  #ifndef NO_RECURSE  #ifndef NO_RECURSE
322  #define REGISTER register  #define REGISTER register
323    
324  #ifdef DEBUG  #ifdef PCRE_DEBUG
325  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
326    { \    { \
327    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
328    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
329    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
330    }    }
331  #define RRETURN(ra) \  #define RRETURN(ra) \
332    { \    { \
333    printf("match() returned %d from line %d ", ra, __LINE__); \    printf("match() returned %d from line %d\n", ra, __LINE__); \
334    return ra; \    return ra; \
335    }    }
336  #else  #else
337  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
338    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
339  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
340  #endif  #endif
341    
# Line 247  argument of match(), which never changes Line 348  argument of match(), which never changes
348    
349  #define REGISTER  #define REGISTER
350    
351  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
352    {\    {\
353    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = frame->Xnextframe;\
354    frame->Xwhere = rw; \    if (newframe == NULL)\
355        {\
356        newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
357        if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
358        newframe->Xnextframe = NULL;\
359        frame->Xnextframe = newframe;\
360        }\
361      frame->Xwhere = rw;\
362    newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
363    newframe->Xecode = rb;\    newframe->Xecode = rb;\
364    newframe->Xmstart = mstart;\    newframe->Xmstart = mstart;\
365    newframe->Xoffset_top = rc;\    newframe->Xoffset_top = rc;\
366    newframe->Xims = re;\    newframe->Xeptrb = re;\
   newframe->Xeptrb = rf;\  
   newframe->Xflags = rg;\  
367    newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xrdepth = frame->Xrdepth + 1;\
368    newframe->Xprevframe = frame;\    newframe->Xprevframe = frame;\
369    frame = newframe;\    frame = newframe;\
# Line 269  argument of match(), which never changes Line 375  argument of match(), which never changes
375    
376  #define RRETURN(ra)\  #define RRETURN(ra)\
377    {\    {\
378    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
379    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
   (pcre_stack_free)(newframe);\  
380    if (frame != NULL)\    if (frame != NULL)\
381      {\      {\
382      rrc = ra;\      rrc = ra;\
# Line 285  argument of match(), which never changes Line 390  argument of match(), which never changes
390    
391  typedef struct heapframe {  typedef struct heapframe {
392    struct heapframe *Xprevframe;    struct heapframe *Xprevframe;
393      struct heapframe *Xnextframe;
394    
395    /* Function arguments that may change */    /* Function arguments that may change */
396    
397    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
398    const uschar *Xecode;    const pcre_uchar *Xecode;
399    const uschar *Xmstart;    PCRE_PUCHAR Xmstart;
400    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
401    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
402    unsigned int Xrdepth;    unsigned int Xrdepth;
403    
404    /* Function local variables */    /* Function local variables */
405    
406    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
407    const uschar *Xcharptr;  #ifdef SUPPORT_UTF
408    const uschar *Xdata;    PCRE_PUCHAR Xcharptr;
409    const uschar *Xnext;  #endif
410    const uschar *Xpp;    PCRE_PUCHAR Xdata;
411    const uschar *Xprev;    PCRE_PUCHAR Xnext;
412    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xpp;
413      PCRE_PUCHAR Xprev;
414      PCRE_PUCHAR Xsaved_eptr;
415    
416    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
417    
# Line 313  typedef struct heapframe { Line 419  typedef struct heapframe {
419    BOOL Xcondition;    BOOL Xcondition;
420    BOOL Xprev_is_word;    BOOL Xprev_is_word;
421    
   unsigned long int Xoriginal_ims;  
   
422  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
423    int Xprop_type;    int Xprop_type;
424    int Xprop_value;    unsigned int Xprop_value;
425    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
426    int Xoclength;    int Xoclength;
427    uschar Xocchars[8];    pcre_uchar Xocchars[6];
428  #endif  #endif
429    
430      int Xcodelink;
431    int Xctype;    int Xctype;
432    unsigned int Xfc;    unsigned int Xfc;
433    int Xfi;    int Xfi;
434    int Xlength;    int Xlength;
435    int Xmax;    int Xmax;
436    int Xmin;    int Xmin;
437    int Xnumber;    unsigned int Xnumber;
438    int Xoffset;    int Xoffset;
439    int Xop;    unsigned int Xop;
440    int Xsave_capture_last;    pcre_int32 Xsave_capture_last;
441    int Xsave_offset1, Xsave_offset2, Xsave_offset3;    int Xsave_offset1, Xsave_offset2, Xsave_offset3;
442    int Xstacksave[REC_STACK_SAVE_MAX];    int Xstacksave[REC_STACK_SAVE_MAX];
443    
# Line 361  typedef struct heapframe { Line 463  typedef struct heapframe {
463    
464  /* This function is called recursively in many circumstances. Whenever it  /* This function is called recursively in many circumstances. Whenever it
465  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
466  same response.  same response. */
467    
468    /* These macros pack up tests that are used for partial matching, and which
469    appear several times in the code. We set the "hit end" flag if the pointer is
470    at the end of the subject and also past the start of the subject (i.e.
471    something has been matched). For hard partial matching, we then return
472    immediately. The second one is used when we already know we are past the end of
473    the subject. */
474    
475    #define CHECK_PARTIAL()\
476      if (md->partial != 0 && eptr >= md->end_subject && \
477          eptr > md->start_used_ptr) \
478        { \
479        md->hitend = TRUE; \
480        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
481        }
482    
483  Performance note: It might be tempting to extract commonly used fields from the  #define SCHECK_PARTIAL()\
484  md structure (e.g. utf8, end_subject) into individual variables to improve    if (md->partial != 0 && eptr > md->start_used_ptr) \
485        { \
486        md->hitend = TRUE; \
487        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
488        }
489    
490    
491    /* Performance note: It might be tempting to extract commonly used fields from
492    the md structure (e.g. utf, end_subject) into individual variables to improve
493  performance. Tests using gcc on a SPARC disproved this; in the first case, it  performance. Tests using gcc on a SPARC disproved this; in the first case, it
494  made performance worse.  made performance worse.
495    
# Line 375  Arguments: Line 500  Arguments:
500                   by encountering \K)                   by encountering \K)
501     offset_top  current top pointer     offset_top  current top pointer
502     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
503     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
504                   brackets - for testing for empty matches                   brackets - for testing for empty matches
    flags       can contain  
                  match_condassert - this is an assertion condition  
                  match_cbegroup - this is the start of an unlimited repeat  
                    group that can match an empty string  
505     rdepth      the recursion depth     rdepth      the recursion depth
506    
507  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
508                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
509                   a negative MATCH_xxx value for PRUNE, SKIP, etc
510                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
511                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
512  */  */
513    
514  static int  static int
515  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, const uschar *mstart,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
516    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
517    int flags, unsigned int rdepth)    unsigned int rdepth)
518  {  {
519  /* These variables do not need to be preserved over recursion in this function,  /* These variables do not need to be preserved over recursion in this function,
520  so they can be ordinary variables in all cases. Mark some of them with  so they can be ordinary variables in all cases. Mark some of them with
# Line 401  so they can be ordinary variables in all Line 522  so they can be ordinary variables in all
522    
523  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
524  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
525  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register pcre_uint32 c;    /* Character values not kept over RMATCH() calls */
526  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
527    
528  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
529    BOOL caseless;
530    int condcode;
531    
532  /* When recursion is not being used, all "local" variables that have to be  /* When recursion is not being used, all "local" variables that have to be
533  preserved over calls to RMATCH() are part of a "frame" which is obtained from  preserved over calls to RMATCH() are part of a "frame". We set up the top-level
534  heap storage. Set up the top-level frame here; others are obtained from the  frame on the stack here; subsequent instantiations are obtained from the heap
535  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
536    the top-level on the stack rather than malloc-ing them all gives a performance
537    boost in many cases where there is not much "recursion". */
538    
539  #ifdef NO_RECURSE  #ifdef NO_RECURSE
540  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)md->match_frames_base;
 frame->Xprevframe = NULL;            /* Marks the top level */  
541    
542  /* Copy in the original argument variables */  /* Copy in the original argument variables */
543    
# Line 421  frame->Xeptr = eptr; Line 545  frame->Xeptr = eptr;
545  frame->Xecode = ecode;  frame->Xecode = ecode;
546  frame->Xmstart = mstart;  frame->Xmstart = mstart;
547  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
548  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
549  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
550    
551  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 436  HEAP_RECURSE: Line 558  HEAP_RECURSE:
558  #define ecode              frame->Xecode  #define ecode              frame->Xecode
559  #define mstart             frame->Xmstart  #define mstart             frame->Xmstart
560  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
561  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
562  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
563    
564  /* Ditto for the local variables */  /* Ditto for the local variables */
565    
566  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
567  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
568  #endif  #endif
569  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
570    #define codelink           frame->Xcodelink
571  #define data               frame->Xdata  #define data               frame->Xdata
572  #define next               frame->Xnext  #define next               frame->Xnext
573  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 459  HEAP_RECURSE: Line 580  HEAP_RECURSE:
580  #define condition          frame->Xcondition  #define condition          frame->Xcondition
581  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
582    
 #define original_ims       frame->Xoriginal_ims  
   
583  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
584  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
585  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
586  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
 #define prop_category      frame->Xprop_category  
 #define prop_chartype      frame->Xprop_chartype  
 #define prop_script        frame->Xprop_script  
587  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
588  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
589  #endif  #endif
# Line 497  i, and fc and c, can be the same variabl Line 613  i, and fc and c, can be the same variabl
613  #define fi i  #define fi i
614  #define fc c  #define fc c
615    
616    /* Many of the following variables are used only in small blocks of the code.
617    My normal style of coding would have declared them within each of those blocks.
618    However, in order to accommodate the version of this code that uses an external
619    "stack" implemented on the heap, it is easier to declare them all here, so the
620    declarations can be cut out in a block. The only declarations within blocks
621    below are for variables that do not have to be preserved over a recursive call
622    to RMATCH(). */
623    
624  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  #ifdef SUPPORT_UTF
625  const uschar *charptr;             /* in small blocks of the code. My normal */  const pcre_uchar *charptr;
626  #endif                             /* style of coding would have declared    */  #endif
627  const uschar *callpat;             /* them within each of those blocks.      */  const pcre_uchar *callpat;
628  const uschar *data;                /* However, in order to accommodate the   */  const pcre_uchar *data;
629  const uschar *next;                /* version of this code that uses an      */  const pcre_uchar *next;
630  USPTR         pp;                  /* external "stack" implemented on the    */  PCRE_PUCHAR       pp;
631  const uschar *prev;                /* heap, it is easier to declare them all */  const pcre_uchar *prev;
632  USPTR         saved_eptr;          /* here, so the declarations can be cut   */  PCRE_PUCHAR       saved_eptr;
633                                     /* out in a block. The only declarations  */  
634  recursion_info new_recursive;      /* within blocks below are for variables  */  recursion_info new_recursive;
635                                     /* that do not have to be preserved over  */  
636  BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  BOOL cur_is_word;
637  BOOL condition;  BOOL condition;
638  BOOL prev_is_word;  BOOL prev_is_word;
639    
 unsigned long int original_ims;  
   
640  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
641  int prop_type;  int prop_type;
642  int prop_value;  unsigned int prop_value;
643  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
644  int oclength;  int oclength;
645  uschar occhars[8];  pcre_uchar occhars[6];
646  #endif  #endif
647    
648    int codelink;
649  int ctype;  int ctype;
650  int length;  int length;
651  int max;  int max;
652  int min;  int min;
653  int number;  unsigned int number;
654  int offset;  int offset;
655  int op;  unsigned int op;
656  int save_capture_last;  pcre_int32 save_capture_last;
657  int save_offset1, save_offset2, save_offset3;  int save_offset1, save_offset2, save_offset3;
658  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
659    
660  eptrblock newptrb;  eptrblock newptrb;
661    
662    /* There is a special fudge for calling match() in a way that causes it to
663    measure the size of its basic stack frame when the stack is being used for
664    recursion. The second argument (ecode) being NULL triggers this behaviour. It
665    cannot normally ever be NULL. The return is the negated value of the frame
666    size. */
667    
668    if (ecode == NULL)
669      {
670      if (rdepth == 0)
671        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
672      else
673        {
674        int len = (char *)&rdepth - (char *)eptr;
675        return (len > 0)? -len : len;
676        }
677      }
678  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
679    
680    /* To save space on the stack and in the heap frame, I have doubled up on some
681    of the local variables that are used only in localised parts of the code, but
682    still need to be preserved over recursive calls of match(). These macros define
683    the alternative names that are used. */
684    
685    #define allow_zero    cur_is_word
686    #define cbegroup      condition
687    #define code_offset   codelink
688    #define condassert    condition
689    #define matched_once  prev_is_word
690    #define foc           number
691    #define save_mark     data
692    
693  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
694  variables. */  variables. */
695    
# Line 560  TAIL_RECURSE: Line 709  TAIL_RECURSE:
709  /* OK, now we can get on with the real code of the function. Recursive calls  /* OK, now we can get on with the real code of the function. Recursive calls
710  are specified by the macro RMATCH and RRETURN is used to return. When  are specified by the macro RMATCH and RRETURN is used to return. When
711  NO_RECURSE is *not* defined, these just turn into a recursive call to match()  NO_RECURSE is *not* defined, these just turn into a recursive call to match()
712  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
713  defined). However, RMATCH isn't like a function call because it's quite a  defined). However, RMATCH isn't like a function call because it's quite a
714  complicated macro. It has to be used in one particular way. This shouldn't,  complicated macro. It has to be used in one particular way. This shouldn't,
715  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
716    
717  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
718  utf8 = md->utf8;       /* Local copy of the flag */  utf = md->utf;       /* Local copy of the flag */
719  #else  #else
720  utf8 = FALSE;  utf = FALSE;
721  #endif  #endif
722    
723  /* First check that we haven't called match() too many times, or that we  /* First check that we haven't called match() too many times, or that we
# Line 577  haven't exceeded the recursive call limi Line 726  haven't exceeded the recursive call limi
726  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
727  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
728    
 original_ims = ims;    /* Save for resetting on ')' */  
   
729  /* At the start of a group with an unlimited repeat that may match an empty  /* At the start of a group with an unlimited repeat that may match an empty
730  string, the match_cbegroup flag is set. When this is the case, add the current  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
731  subject pointer to the chain of such remembered pointers, to be checked when we  done this way to save having to use another function argument, which would take
732  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
733  When match() is called in other circumstances, don't add to the chain. The  
734  match_cbegroup flag must NOT be used with tail recursion, because the memory  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
735  block that is used is on the stack, so a new one may be required for each  such remembered pointers, to be checked when we hit the closing ket, in order
736  match(). */  to break infinite loops that match no characters. When match() is called in
737    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
738    NOT be used with tail recursion, because the memory block that is used is on
739    the stack, so a new one may be required for each match(). */
740    
741  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
742    {    {
743    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
744    newptrb.epb_prev = eptrb;    newptrb.epb_prev = eptrb;
745    eptrb = &newptrb;    eptrb = &newptrb;
746      md->match_function_type = 0;
747    }    }
748    
749  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 602  for (;;) Line 753  for (;;)
753    minimize = possessive = FALSE;    minimize = possessive = FALSE;
754    op = *ecode;    op = *ecode;
755    
   /* 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;  
   
756    switch(op)    switch(op)
757      {      {
758      /* Handle a capturing bracket. If there is space in the offset vector, save      case OP_MARK:
759      the current subject position in the working slot at the top of the vector.      md->nomatch_mark = ecode + 2;
760      We mustn't change the current values of the data slot, because they may be      md->mark = NULL;    /* In case previously set by assertion */
761      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, md,
762      reference inside the group.        eptrb, RM55);
763        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
764      If the bracket fails to match, we need to restore this value and also the           md->mark == NULL) md->mark = ecode + 2;
765      values of the final offsets, in case they were set by a previous iteration  
766      of the same bracket.      /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
767        argument, and we must check whether that argument matches this MARK's
768        argument. It is passed back in md->start_match_ptr (an overloading of that
769        variable). If it does match, we reset that variable to the current subject
770        position and return MATCH_SKIP. Otherwise, pass back the return code
771        unaltered. */
772    
773        else if (rrc == MATCH_SKIP_ARG &&
774            STRCMP_UC_UC_TEST(ecode + 2, md->start_match_ptr) == 0)
775          {
776          md->start_match_ptr = eptr;
777          RRETURN(MATCH_SKIP);
778          }
779        RRETURN(rrc);
780    
781        case OP_FAIL:
782        RRETURN(MATCH_NOMATCH);
783    
784        case OP_COMMIT:
785        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
786          eptrb, RM52);
787        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
788        RRETURN(MATCH_COMMIT);
789    
790        case OP_PRUNE:
791        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
792          eptrb, RM51);
793        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
794        RRETURN(MATCH_PRUNE);
795    
796        case OP_PRUNE_ARG:
797        md->nomatch_mark = ecode + 2;
798        md->mark = NULL;    /* In case previously set by assertion */
799        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
800          eptrb, RM56);
801        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
802             md->mark == NULL) md->mark = ecode + 2;
803        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
804        RRETURN(MATCH_PRUNE);
805    
806        case OP_SKIP:
807        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
808          eptrb, RM53);
809        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
810        md->start_match_ptr = eptr;   /* Pass back current position */
811        RRETURN(MATCH_SKIP);
812    
813        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
814        nomatch_mark. When a pattern match ends with a SKIP_ARG for which there was
815        not a matching mark, we have to re-run the match, ignoring the SKIP_ARG
816        that failed and any that precede it (either they also failed, or were not
817        triggered). To do this, we maintain a count of executed SKIP_ARGs. If a
818        SKIP_ARG gets to top level, the match is re-run with md->ignore_skip_arg
819        set to the count of the one that failed. */
820    
821        case OP_SKIP_ARG:
822        md->skip_arg_count++;
823        if (md->skip_arg_count <= md->ignore_skip_arg)
824          {
825          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
826          break;
827          }
828        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
829          eptrb, RM57);
830        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
831    
832        /* Pass back the current skip name by overloading md->start_match_ptr and
833        returning the special MATCH_SKIP_ARG return code. This will either be
834        caught by a matching MARK, or get to the top, where it causes a rematch
835        with md->ignore_skip_arg set to the value of md->skip_arg_count. */
836    
837        md->start_match_ptr = ecode + 2;
838        RRETURN(MATCH_SKIP_ARG);
839    
840        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
841        the branch in which it occurs can be determined. Overload the start of
842        match pointer to do this. */
843    
844        case OP_THEN:
845        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
846          eptrb, RM54);
847        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
848        md->start_match_ptr = ecode;
849        RRETURN(MATCH_THEN);
850    
851        case OP_THEN_ARG:
852        md->nomatch_mark = ecode + 2;
853        md->mark = NULL;    /* In case previously set by assertion */
854        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
855          md, eptrb, RM58);
856        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
857             md->mark == NULL) md->mark = ecode + 2;
858        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
859        md->start_match_ptr = ecode;
860        RRETURN(MATCH_THEN);
861    
862        /* Handle an atomic group that does not contain any capturing parentheses.
863        This can be handled like an assertion. Prior to 8.13, all atomic groups
864        were handled this way. In 8.13, the code was changed as below for ONCE, so
865        that backups pass through the group and thereby reset captured values.
866        However, this uses a lot more stack, so in 8.20, atomic groups that do not
867        contain any captures generate OP_ONCE_NC, which can be handled in the old,
868        less stack intensive way.
869    
870        Check the alternative branches in turn - the matching won't pass the KET
871        for this kind of subpattern. If any one branch matches, we carry on as at
872        the end of a normal bracket, leaving the subject pointer, but resetting
873        the start-of-match value in case it was changed by \K. */
874    
875        case OP_ONCE_NC:
876        prev = ecode;
877        saved_eptr = eptr;
878        save_mark = md->mark;
879        do
880          {
881          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
882          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
883            {
884            mstart = md->start_match_ptr;
885            break;
886            }
887          if (rrc == MATCH_THEN)
888            {
889            next = ecode + GET(ecode,1);
890            if (md->start_match_ptr < next &&
891                (*ecode == OP_ALT || *next == OP_ALT))
892              rrc = MATCH_NOMATCH;
893            }
894    
895          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
896          ecode += GET(ecode,1);
897          md->mark = save_mark;
898          }
899        while (*ecode == OP_ALT);
900    
901        /* If hit the end of the group (which could be repeated), fail */
902    
903        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
904    
905        /* Continue as from after the group, updating the offsets high water
906        mark, since extracts may have been taken. */
907    
908        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
909    
910        offset_top = md->end_offset_top;
911        eptr = md->end_match_ptr;
912    
913        /* For a non-repeating ket, just continue at this level. This also
914        happens for a repeating ket if no characters were matched in the group.
915        This is the forcible breaking of infinite loops as implemented in Perl
916        5.005. */
917    
918        if (*ecode == OP_KET || eptr == saved_eptr)
919          {
920          ecode += 1+LINK_SIZE;
921          break;
922          }
923    
924        /* The repeating kets try the rest of the pattern or restart from the
925        preceding bracket, in the appropriate order. The second "call" of match()
926        uses tail recursion, to avoid using another stack frame. */
927    
928        if (*ecode == OP_KETRMIN)
929          {
930          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
931          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
932          ecode = prev;
933          goto TAIL_RECURSE;
934          }
935        else  /* OP_KETRMAX */
936          {
937          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
938          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
939          ecode += 1 + LINK_SIZE;
940          goto TAIL_RECURSE;
941          }
942        /* Control never gets here */
943    
944        /* Handle a capturing bracket, other than those that are possessive with an
945        unlimited repeat. If there is space in the offset vector, save the current
946        subject position in the working slot at the top of the vector. We mustn't
947        change the current values of the data slot, because they may be set from a
948        previous iteration of this group, and be referred to by a reference inside
949        the group. A failure to match might occur after the group has succeeded,
950        if something later on doesn't match. For this reason, we need to restore
951        the working value and also the values of the final offsets, in case they
952        were set by a previous iteration of the same bracket.
953    
954      If there isn't enough space in the offset vector, treat this as if it were      If there isn't enough space in the offset vector, treat this as if it were
955      a non-capturing bracket. Don't worry about setting the flag for the error      a non-capturing bracket. Don't worry about setting the flag for the error
# Line 631  for (;;) Line 960  for (;;)
960      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
961      offset = number << 1;      offset = number << 1;
962    
963  #ifdef DEBUG  #ifdef PCRE_DEBUG
964      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
965      printf("subject=");      printf("subject=");
966      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 644  for (;;) Line 973  for (;;)
973        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
974        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
975        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
976          save_mark = md->mark;
977    
978        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
979        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
980            (int)(eptr - md->start_subject);
981    
982        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
983          {          {
984          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
985            ims, eptrb, flags, RM1);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
986              eptrb, RM1);
987            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
988    
989            /* If we backed up to a THEN, check whether it is within the current
990            branch by comparing the address of the THEN that is passed back with
991            the end of the branch. If it is within the current branch, and the
992            branch is one of two or more alternatives (it either starts or ends
993            with OP_ALT), we have reached the limit of THEN's action, so convert
994            the return code to NOMATCH, which will cause normal backtracking to
995            happen from now on. Otherwise, THEN is passed back to an outer
996            alternative. This implements Perl's treatment of parenthesized groups,
997            where a group not containing | does not affect the current alternative,
998            that is, (X) is NOT the same as (X|(*F)). */
999    
1000            if (rrc == MATCH_THEN)
1001              {
1002              next = ecode + GET(ecode,1);
1003              if (md->start_match_ptr < next &&
1004                  (*ecode == OP_ALT || *next == OP_ALT))
1005                rrc = MATCH_NOMATCH;
1006              }
1007    
1008            /* Anything other than NOMATCH is passed back. */
1009    
1010          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1011          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
1012          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
1013            md->mark = save_mark;
1014            if (*ecode != OP_ALT) break;
1015          }          }
       while (*ecode == OP_ALT);  
1016    
1017        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
1018        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
1019        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
1020        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
1021    
1022        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1023    
1024          RRETURN(rrc);
1025        }        }
1026    
1027      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
# Line 679  for (;;) Line 1035  for (;;)
1035      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1036      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1037    
1038      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* Non-capturing or atomic group, except for possessive with unlimited
1039      final alternative within the brackets, we would return the result of a      repeat and ONCE group with no captures. Loop for all the alternatives.
1040      recursive call to match() whatever happened. We can reduce stack usage by  
1041      turning this into a tail recursion, except in the case when match_cbegroup      When we get to the final alternative within the brackets, we used to return
1042      is set.*/      the result of a recursive call to match() whatever happened so it was
1043        possible to reduce stack usage by turning this into a tail recursion,
1044        except in the case of a possibly empty group. However, now that there is
1045        the possiblity of (*THEN) occurring in the final alternative, this
1046        optimization is no longer always possible.
1047    
1048        We can optimize if we know there are no (*THEN)s in the pattern; at present
1049        this is the best that can be done.
1050    
1051        MATCH_ONCE is returned when the end of an atomic group is successfully
1052        reached, but subsequent matching fails. It passes back up the tree (causing
1053        captured values to be reset) until the original atomic group level is
1054        reached. This is tested by comparing md->once_target with the start of the
1055        group. At this point, the return is converted into MATCH_NOMATCH so that
1056        previous backup points can be taken. */
1057    
1058        case OP_ONCE:
1059      case OP_BRA:      case OP_BRA:
1060      case OP_SBRA:      case OP_SBRA:
1061      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1062      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
1063      for (;;)      for (;;)
1064        {        {
1065        if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */        if (op >= OP_SBRA || op == OP_ONCE)
1066          {          md->match_function_type = MATCH_CBEGROUP;
         if (flags == 0)    /* Not a possibly empty group */  
           {  
           ecode += _pcre_OP_lengths[*ecode];  
           DPRINTF(("bracket 0 tail recursion\n"));  
           goto TAIL_RECURSE;  
           }  
1067    
1068          /* Possibly empty group; can't use tail recursion. */        /* If this is not a possibly empty group, and there are no (*THEN)s in
1069          the pattern, and this is the final alternative, optimize as described
1070          above. */
1071    
1072          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1073            eptrb, flags, RM48);          {
1074          RRETURN(rrc);          ecode += PRIV(OP_lengths)[*ecode];
1075            goto TAIL_RECURSE;
1076          }          }
1077    
1078        /* 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. */  
1079    
1080        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        save_mark = md->mark;
1081          eptrb, flags, RM2);        save_capture_last = md->capture_last;
1082        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1083            RM2);
1084    
1085          /* See comment in the code for capturing groups above about handling
1086          THEN. */
1087    
1088          if (rrc == MATCH_THEN)
1089            {
1090            next = ecode + GET(ecode,1);
1091            if (md->start_match_ptr < next &&
1092                (*ecode == OP_ALT || *next == OP_ALT))
1093              rrc = MATCH_NOMATCH;
1094            }
1095    
1096          if (rrc != MATCH_NOMATCH)
1097            {
1098            if (rrc == MATCH_ONCE)
1099              {
1100              const pcre_uchar *scode = ecode;
1101              if (*scode != OP_ONCE)           /* If not at start, find it */
1102                {
1103                while (*scode == OP_ALT) scode += GET(scode, 1);
1104                scode -= GET(scode, 1);
1105                }
1106              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1107              }
1108            RRETURN(rrc);
1109            }
1110        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1111        }        md->mark = save_mark;
1112      /* Control never reaches here. */        if (*ecode != OP_ALT) break;
1113          md->capture_last = save_capture_last;
1114          }
1115    
1116        RRETURN(MATCH_NOMATCH);
1117    
1118        /* Handle possessive capturing brackets with an unlimited repeat. We come
1119        here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1120        handled similarly to the normal case above. However, the matching is
1121        different. The end of these brackets will always be OP_KETRPOS, which
1122        returns MATCH_KETRPOS without going further in the pattern. By this means
1123        we can handle the group by iteration rather than recursion, thereby
1124        reducing the amount of stack needed. */
1125    
1126        case OP_CBRAPOS:
1127        case OP_SCBRAPOS:
1128        allow_zero = FALSE;
1129    
1130      /* Conditional group: compilation checked that there are no more than      POSSESSIVE_CAPTURE:
1131      two branches. If the condition is false, skipping the first branch takes us      number = GET2(ecode, 1+LINK_SIZE);
1132      past the end if there is only one branch, but that's OK because that is      offset = number << 1;
     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. */  
1133    
1134      case OP_COND:  #ifdef PCRE_DEBUG
1135      case OP_SCOND:      printf("start possessive bracket %d\n", number);
1136      if (ecode[LINK_SIZE+1] == OP_RREF)         /* Recursion test */      printf("subject=");
1137        {      pchars(eptr, 16, TRUE, md);
1138        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/      printf("\n");
1139        condition = md->recursive != NULL &&  #endif
         (offset == RREF_ANY || offset == md->recursive->group_num);  
       ecode += condition? 3 : GET(ecode, 1);  
       }  
1140    
1141      else if (ecode[LINK_SIZE+1] == OP_CREF)    /* Group used test */      if (offset < md->offset_max)
1142        {        {
1143        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */        matched_once = FALSE;
1144        condition = offset < offset_top && md->offset_vector[offset] >= 0;        code_offset = (int)(ecode - md->start_code);
       ecode += condition? 3 : GET(ecode, 1);  
       }  
1145    
1146      else if (ecode[LINK_SIZE+1] == OP_DEF)     /* DEFINE - always false */        save_offset1 = md->offset_vector[offset];
1147        {        save_offset2 = md->offset_vector[offset+1];
1148        condition = FALSE;        save_offset3 = md->offset_vector[md->offset_end - number];
1149        ecode += GET(ecode, 1);        save_capture_last = md->capture_last;
       }  
1150    
1151      /* The condition is an assertion. Call match() to evaluate it - setting        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
     the final argument match_condassert causes it to stop at the end of an  
     assertion. */  
1152    
1153      else        /* Each time round the loop, save the current subject position for use
1154        {        when the group matches. For MATCH_MATCH, the group has matched, so we
1155        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        restart it with a new subject starting position, remembering that we had
1156            match_condassert, RM3);        at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1157        if (rrc == MATCH_MATCH)        usual. If we haven't matched any alternatives in any iteration, check to
1158          {        see if a previous iteration matched. If so, the group has matched;
1159          condition = TRUE;        continue from afterwards. Otherwise it has failed; restore the previous
1160          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);        capture values before returning NOMATCH. */
1161          while (*ecode == OP_ALT) ecode += GET(ecode, 1);  
1162          }        for (;;)
1163        else if (rrc != MATCH_NOMATCH)          {
1164          {          md->offset_vector[md->offset_end - number] =
1165          RRETURN(rrc);         /* Need braces because of following else */            (int)(eptr - md->start_subject);
1166          }          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1167        else          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1168          {            eptrb, RM63);
1169          condition = FALSE;          if (rrc == MATCH_KETRPOS)
1170              {
1171              offset_top = md->end_offset_top;
1172              eptr = md->end_match_ptr;
1173              ecode = md->start_code + code_offset;
1174              save_capture_last = md->capture_last;
1175              matched_once = TRUE;
1176              continue;
1177              }
1178    
1179            /* See comment in the code for capturing groups above about handling
1180            THEN. */
1181    
1182            if (rrc == MATCH_THEN)
1183              {
1184              next = ecode + GET(ecode,1);
1185              if (md->start_match_ptr < next &&
1186                  (*ecode == OP_ALT || *next == OP_ALT))
1187                rrc = MATCH_NOMATCH;
1188              }
1189    
1190            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1191            md->capture_last = save_capture_last;
1192          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
1193            if (*ecode != OP_ALT) break;
1194          }          }
       }  
1195    
1196      /* We are now at the branch that is to be obeyed. As there is only one,        if (!matched_once)
     we can use tail recursion to avoid using another stack frame, except when  
     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. */  
   
     if (condition || *ecode == OP_ALT)  
       {  
       ecode += 1 + LINK_SIZE;  
       if (op == OP_SCOND)        /* Possibly empty group */  
1197          {          {
1198          RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);          md->offset_vector[offset] = save_offset1;
1199          RRETURN(rrc);          md->offset_vector[offset+1] = save_offset2;
1200            md->offset_vector[md->offset_end - number] = save_offset3;
1201          }          }
1202        else                       /* Group must match something */  
1203          if (allow_zero || matched_once)
1204          {          {
1205          flags = 0;          ecode += 1 + LINK_SIZE;
1206          goto TAIL_RECURSE;          break;
1207          }          }
       }  
     else                         /* Condition false & no 2nd alternative */  
       {  
       ecode += 1 + LINK_SIZE;  
       }  
     break;  
   
1208    
1209      /* End of the pattern. If we are in a top-level recursion, we should        RRETURN(MATCH_NOMATCH);
     restore the offsets appropriately and continue from after the call. */  
   
     case OP_END:  
     if (md->recursive != NULL && md->recursive->group_num == 0)  
       {  
       recursion_info *rec = md->recursive;  
       DPRINTF(("End of pattern in a (?0) recursion\n"));  
       md->recursive = rec->prevrec;  
       memmove(md->offset_vector, rec->offset_save,  
         rec->saved_max * sizeof(int));  
       mstart = rec->save_start;  
       ims = original_ims;  
       ecode = rec->after_call;  
       break;  
1210        }        }
1211    
1212      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1213      string - backtracking will then try other alternatives, if any. */      as a non-capturing bracket. */
1214    
1215      if (md->notempty && eptr == mstart) RRETURN(MATCH_NOMATCH);      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1216      md->end_match_ptr = eptr;           /* Record where we ended */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     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);  
1217    
1218      /* Change option settings */      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1219    
1220      case OP_OPT:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1221      ims = ecode[1];      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1222    
1223      /* Assertion brackets. Check the alternative branches in turn - the      /* Non-capturing possessive bracket with unlimited repeat. We come here
1224      matching won't pass the KET for an assertion. If any one branch matches,      from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1225      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      without the capturing complication. It is written out separately for speed
1226      start of each branch to move the current point backwards, so the code at      and cleanliness. */
1227      this level is identical to the lookahead case. */  
1228        case OP_BRAPOS:
1229        case OP_SBRAPOS:
1230        allow_zero = FALSE;
1231    
1232        POSSESSIVE_NON_CAPTURE:
1233        matched_once = FALSE;
1234        code_offset = (int)(ecode - md->start_code);
1235        save_capture_last = md->capture_last;
1236    
1237      case OP_ASSERT:      for (;;)
     case OP_ASSERTBACK:  
     do  
1238        {        {
1239        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1240          RM4);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1241        if (rrc == MATCH_MATCH) break;          eptrb, RM48);
1242          if (rrc == MATCH_KETRPOS)
1243            {
1244            offset_top = md->end_offset_top;
1245            eptr = md->end_match_ptr;
1246            ecode = md->start_code + code_offset;
1247            matched_once = TRUE;
1248            continue;
1249            }
1250    
1251          /* See comment in the code for capturing groups above about handling
1252          THEN. */
1253    
1254          if (rrc == MATCH_THEN)
1255            {
1256            next = ecode + GET(ecode,1);
1257            if (md->start_match_ptr < next &&
1258                (*ecode == OP_ALT || *next == OP_ALT))
1259              rrc = MATCH_NOMATCH;
1260            }
1261    
1262        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1263        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1264          if (*ecode != OP_ALT) break;
1265          md->capture_last = save_capture_last;
1266        }        }
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1267    
1268      /* If checking an assertion for a condition, return MATCH_MATCH. */      if (matched_once || allow_zero)
1269          {
1270          ecode += 1 + LINK_SIZE;
1271          break;
1272          }
1273        RRETURN(MATCH_NOMATCH);
1274    
1275      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* Control never reaches here. */
1276    
1277      /* Continue from after the assertion, updating the offsets high water      /* Conditional group: compilation checked that there are no more than
1278      mark, since extracts may have been taken during the assertion. */      two branches. If the condition is false, skipping the first branch takes us
1279        past the end if there is only one branch, but that's OK because that is
1280        exactly what going to the ket would do. */
1281    
1282      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_COND:
1283      ecode += 1 + LINK_SIZE;      case OP_SCOND:
1284      offset_top = md->end_offset_top;      codelink = GET(ecode, 1);
     continue;  
1285    
1286      /* Negative assertion: all branches must fail to match */      /* Because of the way auto-callout works during compile, a callout item is
1287        inserted between OP_COND and an assertion condition. */
1288    
1289      case OP_ASSERT_NOT:      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
     case OP_ASSERTBACK_NOT:  
     do  
1290        {        {
1291        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        if (PUBL(callout) != NULL)
1292          RM5);          {
1293        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);          PUBL(callout_block) cb;
1294        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          cb.version          = 2;   /* Version 1 of the callout block */
1295            cb.callout_number   = ecode[LINK_SIZE+2];
1296            cb.offset_vector    = md->offset_vector;
1297    #if defined COMPILE_PCRE8
1298            cb.subject          = (PCRE_SPTR)md->start_subject;
1299    #elif defined COMPILE_PCRE16
1300            cb.subject          = (PCRE_SPTR16)md->start_subject;
1301    #elif defined COMPILE_PCRE32
1302            cb.subject          = (PCRE_SPTR32)md->start_subject;
1303    #endif
1304            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1305            cb.start_match      = (int)(mstart - md->start_subject);
1306            cb.current_position = (int)(eptr - md->start_subject);
1307            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1308            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1309            cb.capture_top      = offset_top/2;
1310            cb.capture_last     = md->capture_last & CAPLMASK;
1311            /* Internal change requires this for API compatibility. */
1312            if (cb.capture_last == 0) cb.capture_last = -1;
1313            cb.callout_data     = md->callout_data;
1314            cb.mark             = md->nomatch_mark;
1315            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1316            if (rrc < 0) RRETURN(rrc);
1317            }
1318          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1319          codelink -= PRIV(OP_lengths)[OP_CALLOUT];
1320          }
1321    
1322        condcode = ecode[LINK_SIZE+1];
1323    
1324        /* Now see what the actual condition is */
1325    
1326        if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1327          {
1328          if (md->recursive == NULL)                /* Not recursing => FALSE */
1329            {
1330            condition = FALSE;
1331            ecode += GET(ecode, 1);
1332            }
1333          else
1334            {
1335            unsigned int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1336            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1337    
1338            /* If the test is for recursion into a specific subpattern, and it is
1339            false, but the test was set up by name, scan the table to see if the
1340            name refers to any other numbers, and test them. The condition is true
1341            if any one is set. */
1342    
1343            if (!condition && condcode == OP_NRREF)
1344              {
1345              pcre_uchar *slotA = md->name_table;
1346              for (i = 0; i < md->name_count; i++)
1347                {
1348                if (GET2(slotA, 0) == recno) break;
1349                slotA += md->name_entry_size;
1350                }
1351    
1352              /* Found a name for the number - there can be only one; duplicate
1353              names for different numbers are allowed, but not vice versa. First
1354              scan down for duplicates. */
1355    
1356              if (i < md->name_count)
1357                {
1358                pcre_uchar *slotB = slotA;
1359                while (slotB > md->name_table)
1360                  {
1361                  slotB -= md->name_entry_size;
1362                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1363                    {
1364                    condition = GET2(slotB, 0) == md->recursive->group_num;
1365                    if (condition) break;
1366                    }
1367                  else break;
1368                  }
1369    
1370                /* Scan up for duplicates */
1371    
1372                if (!condition)
1373                  {
1374                  slotB = slotA;
1375                  for (i++; i < md->name_count; i++)
1376                    {
1377                    slotB += md->name_entry_size;
1378                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1379                      {
1380                      condition = GET2(slotB, 0) == md->recursive->group_num;
1381                      if (condition) break;
1382                      }
1383                    else break;
1384                    }
1385                  }
1386                }
1387              }
1388    
1389            /* Chose branch according to the condition */
1390    
1391            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1392            }
1393          }
1394    
1395        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1396          {
1397          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1398          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1399    
1400          /* If the numbered capture is unset, but the reference was by name,
1401          scan the table to see if the name refers to any other numbers, and test
1402          them. The condition is true if any one is set. This is tediously similar
1403          to the code above, but not close enough to try to amalgamate. */
1404    
1405          if (!condition && condcode == OP_NCREF)
1406            {
1407            unsigned int refno = offset >> 1;
1408            pcre_uchar *slotA = md->name_table;
1409    
1410            for (i = 0; i < md->name_count; i++)
1411              {
1412              if (GET2(slotA, 0) == refno) break;
1413              slotA += md->name_entry_size;
1414              }
1415    
1416            /* Found a name for the number - there can be only one; duplicate names
1417            for different numbers are allowed, but not vice versa. First scan down
1418            for duplicates. */
1419    
1420            if (i < md->name_count)
1421              {
1422              pcre_uchar *slotB = slotA;
1423              while (slotB > md->name_table)
1424                {
1425                slotB -= md->name_entry_size;
1426                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1427                  {
1428                  offset = GET2(slotB, 0) << 1;
1429                  condition = offset < offset_top &&
1430                    md->offset_vector[offset] >= 0;
1431                  if (condition) break;
1432                  }
1433                else break;
1434                }
1435    
1436              /* Scan up for duplicates */
1437    
1438              if (!condition)
1439                {
1440                slotB = slotA;
1441                for (i++; i < md->name_count; i++)
1442                  {
1443                  slotB += md->name_entry_size;
1444                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1445                    {
1446                    offset = GET2(slotB, 0) << 1;
1447                    condition = offset < offset_top &&
1448                      md->offset_vector[offset] >= 0;
1449                    if (condition) break;
1450                    }
1451                  else break;
1452                  }
1453                }
1454              }
1455            }
1456    
1457          /* Chose branch according to the condition */
1458    
1459          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1460          }
1461    
1462        else if (condcode == OP_DEF)     /* DEFINE - always false */
1463          {
1464          condition = FALSE;
1465          ecode += GET(ecode, 1);
1466          }
1467    
1468        /* The condition is an assertion. Call match() to evaluate it - setting
1469        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1470        an assertion. */
1471    
1472        else
1473          {
1474          md->match_function_type = MATCH_CONDASSERT;
1475          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1476          if (rrc == MATCH_MATCH)
1477            {
1478            if (md->end_offset_top > offset_top)
1479              offset_top = md->end_offset_top;  /* Captures may have happened */
1480            condition = TRUE;
1481            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1482            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1483            }
1484    
1485          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1486          assertion; it is therefore treated as NOMATCH. */
1487    
1488          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1489            {
1490            RRETURN(rrc);         /* Need braces because of following else */
1491            }
1492          else
1493            {
1494            condition = FALSE;
1495            ecode += codelink;
1496            }
1497          }
1498    
1499        /* We are now at the branch that is to be obeyed. As there is only one, can
1500        use tail recursion to avoid using another stack frame, except when there is
1501        unlimited repeat of a possibly empty group. In the latter case, a recursive
1502        call to match() is always required, unless the second alternative doesn't
1503        exist, in which case we can just plough on. Note that, for compatibility
1504        with Perl, the | in a conditional group is NOT treated as creating two
1505        alternatives. If a THEN is encountered in the branch, it propagates out to
1506        the enclosing alternative (unless nested in a deeper set of alternatives,
1507        of course). */
1508    
1509        if (condition || *ecode == OP_ALT)
1510          {
1511          if (op != OP_SCOND)
1512            {
1513            ecode += 1 + LINK_SIZE;
1514            goto TAIL_RECURSE;
1515            }
1516    
1517          md->match_function_type = MATCH_CBEGROUP;
1518          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1519          RRETURN(rrc);
1520          }
1521    
1522         /* Condition false & no alternative; continue after the group. */
1523    
1524        else
1525          {
1526          ecode += 1 + LINK_SIZE;
1527          }
1528        break;
1529    
1530    
1531        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1532        to close any currently open capturing brackets. */
1533    
1534        case OP_CLOSE:
1535        number = GET2(ecode, 1);   /* Must be less than 65536 */
1536        offset = number << 1;
1537    
1538    #ifdef PCRE_DEBUG
1539          printf("end bracket %d at *ACCEPT", number);
1540          printf("\n");
1541    #endif
1542    
1543        md->capture_last = (md->capture_last & OVFLMASK) | number;
1544        if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1545          {
1546          md->offset_vector[offset] =
1547            md->offset_vector[md->offset_end - number];
1548          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1549          if (offset_top <= offset) offset_top = offset + 2;
1550          }
1551        ecode += 1 + IMM2_SIZE;
1552        break;
1553    
1554    
1555        /* End of the pattern, either real or forced. */
1556    
1557        case OP_END:
1558        case OP_ACCEPT:
1559        case OP_ASSERT_ACCEPT:
1560    
1561        /* If we have matched an empty string, fail if not in an assertion and not
1562        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1563        is set and we have matched at the start of the subject. In both cases,
1564        backtracking will then try other alternatives, if any. */
1565    
1566        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1567             md->recursive == NULL &&
1568             (md->notempty ||
1569               (md->notempty_atstart &&
1570                 mstart == md->start_subject + md->start_offset)))
1571          RRETURN(MATCH_NOMATCH);
1572    
1573        /* Otherwise, we have a match. */
1574    
1575        md->end_match_ptr = eptr;           /* Record where we ended */
1576        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1577        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1578    
1579        /* For some reason, the macros don't work properly if an expression is
1580        given as the argument to RRETURN when the heap is in use. */
1581    
1582        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1583        RRETURN(rrc);
1584    
1585        /* Assertion brackets. Check the alternative branches in turn - the
1586        matching won't pass the KET for an assertion. If any one branch matches,
1587        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1588        start of each branch to move the current point backwards, so the code at
1589        this level is identical to the lookahead case. When the assertion is part
1590        of a condition, we want to return immediately afterwards. The caller of
1591        this incarnation of the match() function will have set MATCH_CONDASSERT in
1592        md->match_function type, and one of these opcodes will be the first opcode
1593        that is processed. We use a local variable that is preserved over calls to
1594        match() to remember this case. */
1595    
1596        case OP_ASSERT:
1597        case OP_ASSERTBACK:
1598        save_mark = md->mark;
1599        if (md->match_function_type == MATCH_CONDASSERT)
1600          {
1601          condassert = TRUE;
1602          md->match_function_type = 0;
1603          }
1604        else condassert = FALSE;
1605    
1606        /* Loop for each branch */
1607    
1608        do
1609          {
1610          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1611    
1612          /* A match means that the assertion is true; break out of the loop
1613          that matches its alternatives. */
1614    
1615          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1616            {
1617            mstart = md->start_match_ptr;   /* In case \K reset it */
1618            break;
1619            }
1620    
1621          /* If not matched, restore the previous mark setting. */
1622    
1623          md->mark = save_mark;
1624    
1625          /* See comment in the code for capturing groups above about handling
1626          THEN. */
1627    
1628          if (rrc == MATCH_THEN)
1629            {
1630            next = ecode + GET(ecode,1);
1631            if (md->start_match_ptr < next &&
1632                (*ecode == OP_ALT || *next == OP_ALT))
1633              rrc = MATCH_NOMATCH;
1634            }
1635    
1636          /* Anything other than NOMATCH causes the entire assertion to fail,
1637          passing back the return code. This includes COMMIT, SKIP, PRUNE and an
1638          uncaptured THEN, which means they take their normal effect. This
1639          consistent approach does not always have exactly the same effect as in
1640          Perl. */
1641    
1642          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1643          ecode += GET(ecode, 1);
1644          }
1645        while (*ecode == OP_ALT);   /* Continue for next alternative */
1646    
1647        /* If we have tried all the alternative branches, the assertion has
1648        failed. If not, we broke out after a match. */
1649    
1650        if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1651    
1652        /* If checking an assertion for a condition, return MATCH_MATCH. */
1653    
1654        if (condassert) RRETURN(MATCH_MATCH);
1655    
1656        /* Continue from after a successful assertion, updating the offsets high
1657        water mark, since extracts may have been taken during the assertion. */
1658    
1659        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1660        ecode += 1 + LINK_SIZE;
1661        offset_top = md->end_offset_top;
1662        continue;
1663    
1664        /* Negative assertion: all branches must fail to match for the assertion to
1665        succeed. */
1666    
1667        case OP_ASSERT_NOT:
1668        case OP_ASSERTBACK_NOT:
1669        save_mark = md->mark;
1670        if (md->match_function_type == MATCH_CONDASSERT)
1671          {
1672          condassert = TRUE;
1673          md->match_function_type = 0;
1674          }
1675        else condassert = FALSE;
1676    
1677        /* Loop for each alternative branch. */
1678    
1679        do
1680          {
1681          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1682          md->mark = save_mark;   /* Always restore the mark setting */
1683    
1684          switch(rrc)
1685            {
1686            case MATCH_MATCH:            /* A successful match means */
1687            case MATCH_ACCEPT:           /* the assertion has failed. */
1688            RRETURN(MATCH_NOMATCH);
1689    
1690            case MATCH_NOMATCH:          /* Carry on with next branch */
1691            break;
1692    
1693            /* See comment in the code for capturing groups above about handling
1694            THEN. */
1695    
1696            case MATCH_THEN:
1697            next = ecode + GET(ecode,1);
1698            if (md->start_match_ptr < next &&
1699                (*ecode == OP_ALT || *next == OP_ALT))
1700              {
1701              rrc = MATCH_NOMATCH;
1702              break;
1703              }
1704            /* Otherwise fall through. */
1705    
1706            /* COMMIT, SKIP, PRUNE, and an uncaptured THEN cause the whole
1707            assertion to fail to match, without considering any more alternatives.
1708            Failing to match means the assertion is true. This is a consistent
1709            approach, but does not always have the same effect as in Perl. */
1710    
1711            case MATCH_COMMIT:
1712            case MATCH_SKIP:
1713            case MATCH_SKIP_ARG:
1714            case MATCH_PRUNE:
1715            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1716            goto NEG_ASSERT_TRUE;   /* Break out of alternation loop */
1717    
1718            /* Anything else is an error */
1719    
1720            default:
1721            RRETURN(rrc);
1722            }
1723    
1724          /* Continue with next branch */
1725    
1726        ecode += GET(ecode,1);        ecode += GET(ecode,1);
1727        }        }
1728      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1729    
1730      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* All branches in the assertion failed to match. */
1731    
1732      ecode += 1 + LINK_SIZE;      NEG_ASSERT_TRUE:
1733        if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1734        ecode += 1 + LINK_SIZE;                /* Continue with current branch */
1735      continue;      continue;
1736    
1737      /* Move the subject pointer back. This occurs only at the start of      /* Move the subject pointer back. This occurs only at the start of
# Line 887  for (;;) Line 1740  for (;;)
1740      back a number of characters, not bytes. */      back a number of characters, not bytes. */
1741    
1742      case OP_REVERSE:      case OP_REVERSE:
1743  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1744      if (utf8)      if (utf)
1745        {        {
1746        i = GET(ecode, 1);        i = GET(ecode, 1);
1747        while (i-- > 0)        while (i-- > 0)
1748          {          {
1749          eptr--;          eptr--;
1750          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1751          BACKCHAR(eptr)          BACKCHAR(eptr);
1752          }          }
1753        }        }
1754      else      else
# Line 908  for (;;) Line 1761  for (;;)
1761        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1762        }        }
1763    
1764      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1765    
1766        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1767      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1768      break;      break;
1769    
# Line 918  for (;;) Line 1772  for (;;)
1772      function is able to force a failure. */      function is able to force a failure. */
1773    
1774      case OP_CALLOUT:      case OP_CALLOUT:
1775      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1776        {        {
1777        pcre_callout_block cb;        PUBL(callout_block) cb;
1778        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1779        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1780        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1781    #if defined COMPILE_PCRE8
1782        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1783        cb.subject_length   = md->end_subject - md->start_subject;  #elif defined COMPILE_PCRE16
1784        cb.start_match      = mstart - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1785        cb.current_position = eptr - md->start_subject;  #elif defined COMPILE_PCRE32
1786          cb.subject          = (PCRE_SPTR32)md->start_subject;
1787    #endif
1788          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1789          cb.start_match      = (int)(mstart - md->start_subject);
1790          cb.current_position = (int)(eptr - md->start_subject);
1791        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1792        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1793        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1794        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last & CAPLMASK;
1795          /* Internal change requires this for API compatibility. */
1796          if (cb.capture_last == 0) cb.capture_last = -1;
1797        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1798        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1799          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1800        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1801        }        }
1802      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 943  for (;;) Line 1806  for (;;)
1806      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
1807      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1808    
1809      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1810      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
1811      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
1812      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
1813      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
1814      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
1815      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.  
1816    
1817      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
1818      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
1819      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1820        a lot, so he is not to blame for the current way it works. */
1821    
1822      case OP_RECURSE:      case OP_RECURSE:
1823        {        {
1824          recursion_info *ri;
1825          unsigned int recno;
1826    
1827        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1828        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1829          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1830    
1831          /* Check for repeating a recursion without advancing the subject pointer.
1832          This should catch convoluted mutual recursions. (Some simple cases are
1833          caught at compile time.) */
1834    
1835          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1836            if (recno == ri->group_num && eptr == ri->subject_position)
1837              RRETURN(PCRE_ERROR_RECURSELOOP);
1838    
1839        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1840    
1841          new_recursive.group_num = recno;
1842          new_recursive.saved_capture_last = md->capture_last;
1843          new_recursive.subject_position = eptr;
1844        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1845        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1846    
1847        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1848    
1849        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1850    
1851        /* Now save the offset data. */        /* Now save the offset data */
1852    
1853        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1854        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 982  for (;;) Line 1856  for (;;)
1856        else        else
1857          {          {
1858          new_recursive.offset_save =          new_recursive.offset_save =
1859            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1860          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1861          }          }
   
1862        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1863              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = mstart;  
       mstart = eptr;  
1864    
1865        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1866        restore the offset and recursion data. */        restore the offset data and the last captured value. If there were nested
1867          recursions, md->recursive might be changed, so reset it before looping.
1868          */
1869    
1870        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1871        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1872        do        do
1873          {          {
1874          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1875            md, ims, eptrb, flags, RM6);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1876          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1877            memcpy(md->offset_vector, new_recursive.offset_save,
1878                new_recursive.saved_max * sizeof(int));
1879            md->capture_last = new_recursive.saved_capture_last;
1880            md->recursive = new_recursive.prevrec;
1881            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1882            {            {
1883            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1884            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1885              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1886            RRETURN(MATCH_MATCH);  
1887              /* Set where we got to in the subject, and reset the start in case
1888              it was changed by \K. This *is* propagated back out of a recursion,
1889              for Perl compatibility. */
1890    
1891              eptr = md->end_match_ptr;
1892              mstart = md->start_match_ptr;
1893              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1894            }            }
1895          else if (rrc != MATCH_NOMATCH)  
1896            /* PCRE does not allow THEN, SKIP, PRUNE or COMMIT to escape beyond a
1897            recursion; they cause a NOMATCH for the entire recursion. These codes
1898            are defined in a range that can be tested for. */
1899    
1900            if (rrc >= MATCH_BACKTRACK_MIN && rrc <= MATCH_BACKTRACK_MAX)
1901              RRETURN(MATCH_NOMATCH);
1902    
1903            /* Any return code other than NOMATCH is an error. */
1904    
1905            if (rrc != MATCH_NOMATCH)
1906            {            {
1907            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1908              if (new_recursive.offset_save != stacksave)
1909                (PUBL(free))(new_recursive.offset_save);
1910            RRETURN(rrc);            RRETURN(rrc);
1911            }            }
1912    
1913          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1914          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1915          }          }
1916        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1024  for (;;) Line 1918  for (;;)
1918        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1919        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1920        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1921          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1922        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1923        }        }
     /* Control never reaches here */  
   
     /* "Once" brackets are like assertion brackets except that after a match,  
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM7);  
       if (rrc == MATCH_MATCH) break;  
       if (rrc != MATCH_NOMATCH) 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));  
       }  
1924    
1925      if (*ecode == OP_KETRMIN)      RECURSION_MATCHED:
1926        {      break;
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM8);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       RMATCH(eptr, prev, offset_top, md, ims, eptrb, match_cbegroup, RM9);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1927    
1928      /* 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
1929      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1110  for (;;) Line 1932  for (;;)
1932      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1933      break;      break;
1934    
1935      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1936      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
1937      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
1938      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
1939      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1940    
1941      case OP_BRAZERO:      case OP_BRAZERO:
1942        {      next = ecode + 1;
1943        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1944        RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1945        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1946        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1947      break;      break;
1948    
1949      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1950        {      next = ecode + 1;
1951        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1952        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1953        RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1954        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
       ecode++;  
       }  
1955      break;      break;
1956    
1957        case OP_SKIPZERO:
1958        next = ecode+1;
1959        do next += GET(next,1); while (*next == OP_ALT);
1960        ecode = next + 1 + LINK_SIZE;
1961        break;
1962    
1963        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1964        here; just jump to the group, with allow_zero set TRUE. */
1965    
1966        case OP_BRAPOSZERO:
1967        op = *(++ecode);
1968        allow_zero = TRUE;
1969        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1970          goto POSSESSIVE_NON_CAPTURE;
1971    
1972      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1973    
1974      case OP_KET:      case OP_KET:
1975      case OP_KETRMIN:      case OP_KETRMIN:
1976      case OP_KETRMAX:      case OP_KETRMAX:
1977        case OP_KETRPOS:
1978      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1979    
1980      /* 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
1981      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1982      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1983    
1984      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1985        {        {
1986        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1987        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1988        }        }
1989      else saved_eptr = NULL;      else saved_eptr = NULL;
1990    
1991      /* 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
1992      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
1993      assertions. Do this also for the "once" (atomic) groups. */      water mark for use by positive assertions. We also need to record the match
1994        start in case it was changed by \K. */
1995      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
1996          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1997          *prev == OP_ONCE)           *prev == OP_ONCE_NC)
1998        {        {
1999        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
2000        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
2001        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
2002          RRETURN(MATCH_MATCH);         /* Sets md->mark */
2003        }        }
2004    
2005      /* 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
2006      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
2007      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
2008      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
2009      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
2010        the current subject position and start match pointer and give a MATCH
2011        return. */
2012    
2013      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
2014            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
2015        {        {
2016        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
2017        offset = number << 1;        offset = number << 1;
2018    
2019  #ifdef DEBUG  #ifdef PCRE_DEBUG
2020        printf("end bracket %d", number);        printf("end bracket %d", number);
2021        printf("\n");        printf("\n");
2022  #endif  #endif
2023    
2024        md->capture_last = number;        /* Handle a recursively called group. */
2025        if (offset >= md->offset_max) md->offset_overflow = TRUE; else  
2026          if (md->recursive != NULL && md->recursive->group_num == number)
2027          {          {
2028          md->offset_vector[offset] =          md->end_match_ptr = eptr;
2029            md->offset_vector[md->offset_end - number];          md->start_match_ptr = mstart;
2030          md->offset_vector[offset+1] = eptr - md->start_subject;          RRETURN(MATCH_MATCH);
         if (offset_top <= offset) offset_top = offset + 2;  
2031          }          }
2032    
2033        /* Handle a recursively called group. Restore the offsets        /* Deal with capturing */
       appropriately and continue from after the call. */  
2034    
2035        if (md->recursive != NULL && md->recursive->group_num == number)        md->capture_last = (md->capture_last & OVFLMASK) | number;
2036          if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
2037          {          {
2038          recursion_info *rec = md->recursive;          /* If offset is greater than offset_top, it means that we are
2039          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          "skipping" a capturing group, and that group's offsets must be marked
2040          md->recursive = rec->prevrec;          unset. In earlier versions of PCRE, all the offsets were unset at the
2041          mstart = rec->save_start;          start of matching, but this doesn't work because atomic groups and
2042          memcpy(md->offset_vector, rec->offset_save,          assertions can cause a value to be set that should later be unset.
2043            rec->saved_max * sizeof(int));          Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
2044          ecode = rec->after_call;          part of the atomic group, but this is not on the final matching path,
2045          ims = original_ims;          so must be unset when 2 is set. (If there is no group 2, there is no
2046          break;          problem, because offset_top will then be 2, indicating no capture.) */
         }  
       }  
2047    
2048      /* For both capturing and non-capturing groups, reset the value of the ims          if (offset > offset_top)
2049      flags, in case they got changed during the group. */            {
2050              register int *iptr = md->offset_vector + offset_top;
2051              register int *iend = md->offset_vector + offset;
2052              while (iptr < iend) *iptr++ = -1;
2053              }
2054    
2055      ims = original_ims;          /* Now make the extraction */
     DPRINTF(("ims reset to %02lx\n", ims));  
2056    
2057      /* For a non-repeating ket, just continue at this level. This also          md->offset_vector[offset] =
2058      happens for a repeating ket if no characters were matched in the group.            md->offset_vector[md->offset_end - number];
2059      This is the forcible breaking of infinite loops as implemented in Perl          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
2060      5.005. If there is an options reset, it will get obeyed in the normal          if (offset_top <= offset) offset_top = offset + 2;
2061      course of events. */          }
2062          }
2063    
2064        /* For an ordinary non-repeating ket, just continue at this level. This
2065        also happens for a repeating ket if no characters were matched in the
2066        group. This is the forcible breaking of infinite loops as implemented in
2067        Perl 5.005. For a non-repeating atomic group that includes captures,
2068        establish a backup point by processing the rest of the pattern at a lower
2069        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
2070        original OP_ONCE level, thereby bypassing intermediate backup points, but
2071        resetting any captures that happened along the way. */
2072    
2073      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
2074        {        {
2075        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
2076            {
2077            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
2078            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2079            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2080            RRETURN(MATCH_ONCE);
2081            }
2082          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
2083        break;        break;
2084        }        }
2085    
2086      /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
2087      preceding bracket, in the appropriate order. In the second case, we can use      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
2088      tail recursion to avoid using another stack frame, unless we have an      at a time from the outer level, thus saving stack. */
2089      unlimited repeat of a group that can match an empty string. */  
2090        if (*ecode == OP_KETRPOS)
2091          {
2092          md->end_match_ptr = eptr;
2093          md->end_offset_top = offset_top;
2094          RRETURN(MATCH_KETRPOS);
2095          }
2096    
2097      flags = (*prev >= OP_SBRA)? match_cbegroup : 0;      /* The normal repeating kets try the rest of the pattern or restart from
2098        the preceding bracket, in the appropriate order. In the second case, we can
2099        use tail recursion to avoid using another stack frame, unless we have an
2100        an atomic group or an unlimited repeat of a group that can match an empty
2101        string. */
2102    
2103      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
2104        {        {
2105        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2106        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2107        if (flags != 0)    /* Could match an empty string */        if (*prev == OP_ONCE)
2108          {          {
2109          RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2110            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2111            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2112            RRETURN(MATCH_ONCE);
2113            }
2114          if (*prev >= OP_SBRA)    /* Could match an empty string */
2115            {
2116            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2117          RRETURN(rrc);          RRETURN(rrc);
2118          }          }
2119        ecode = prev;        ecode = prev;
# Line 1248  for (;;) Line 2121  for (;;)
2121        }        }
2122      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
2123        {        {
2124        RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2125          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2126        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2127          if (*prev == OP_ONCE)
2128            {
2129            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2130            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2131            md->once_target = prev;
2132            RRETURN(MATCH_ONCE);
2133            }
2134        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
2135        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2136        }        }
2137      /* Control never gets here */      /* Control never gets here */
2138    
2139      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2140    
2141      case OP_CIRC:      case OP_CIRC:
2142      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 */  
2143    
2144      /* Start of subject assertion */      /* Start of subject assertion */
2145    
# Line 1277  for (;;) Line 2148  for (;;)
2148      ecode++;      ecode++;
2149      break;      break;
2150    
2151        /* Multiline mode: start of subject unless notbol, or after any newline. */
2152    
2153        case OP_CIRCM:
2154        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2155        if (eptr != md->start_subject &&
2156            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2157          RRETURN(MATCH_NOMATCH);
2158        ecode++;
2159        break;
2160    
2161      /* Start of match assertion */      /* Start of match assertion */
2162    
2163      case OP_SOM:      case OP_SOM:
# Line 1291  for (;;) Line 2172  for (;;)
2172      ecode++;      ecode++;
2173      break;      break;
2174    
2175      /* Assert before internal newline if multiline, or before a terminating      /* Multiline mode: assert before any newline, or before end of subject
2176      newline unless endonly is set, else end of subject unless noteol is set. */      unless noteol is set. */
2177    
2178      case OP_DOLL:      case OP_DOLLM:
2179      if ((ims & PCRE_MULTILINE) != 0)      if (eptr < md->end_subject)
2180        {        {
2181        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2182          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }          {
2183        else          if (md->partial != 0 &&
2184          { if (md->noteol) RRETURN(MATCH_NOMATCH); }              eptr + 1 >= md->end_subject &&
2185        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2186        break;              NLBLOCK->nllen == 2 &&
2187                RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2188              {
2189              md->hitend = TRUE;
2190              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2191              }
2192            RRETURN(MATCH_NOMATCH);
2193            }
2194        }        }
2195      else      else
2196        {        {
2197        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2198        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2199        }        }
2200        ecode++;
2201        break;
2202    
2203        /* Not multiline mode: assert before a terminating newline or before end of
2204        subject unless noteol is set. */
2205    
2206        case OP_DOLL:
2207        if (md->noteol) RRETURN(MATCH_NOMATCH);
2208        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2209    
2210      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2211    
2212      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2213    
2214      case OP_EOD:      case OP_EOD:
2215      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2216        SCHECK_PARTIAL();
2217      ecode++;      ecode++;
2218      break;      break;
2219    
2220      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2221    
2222      case OP_EODN:      case OP_EODN:
2223      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2224        if (eptr < md->end_subject &&
2225          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2226          {
2227          if (md->partial != 0 &&
2228              eptr + 1 >= md->end_subject &&
2229              NLBLOCK->nltype == NLTYPE_FIXED &&
2230              NLBLOCK->nllen == 2 &&
2231              RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2232            {
2233            md->hitend = TRUE;
2234            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2235            }
2236        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2237          }
2238    
2239        /* Either at end of string or \n before end. */
2240    
2241        SCHECK_PARTIAL();
2242      ecode++;      ecode++;
2243      break;      break;
2244    
# Line 1342  for (;;) Line 2250  for (;;)
2250    
2251        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2252        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
2253        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2254          partial matching. */
2255    
2256  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2257        if (utf8)        if (utf)
2258          {          {
2259            /* Get status of previous character */
2260    
2261          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2262            {            {
2263            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2264            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2265              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2266            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2267    #ifdef SUPPORT_UCP
2268              if (md->use_ucp)
2269                {
2270                if (c == '_') prev_is_word = TRUE; else
2271                  {
2272                  int cat = UCD_CATEGORY(c);
2273                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2274                  }
2275                }
2276              else
2277    #endif
2278            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2279            }            }
2280          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2281            /* Get status of next character */
2282    
2283            if (eptr >= md->end_subject)
2284              {
2285              SCHECK_PARTIAL();
2286              cur_is_word = FALSE;
2287              }
2288            else
2289            {            {
2290            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2291    #ifdef SUPPORT_UCP
2292              if (md->use_ucp)
2293                {
2294                if (c == '_') cur_is_word = TRUE; else
2295                  {
2296                  int cat = UCD_CATEGORY(c);
2297                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2298                  }
2299                }
2300              else
2301    #endif
2302            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2303            }            }
2304          }          }
2305        else        else
2306  #endif  #endif
2307    
2308        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2309          consistency with the behaviour of \w we do use it in this case. */
2310    
2311          {          {
2312          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2313            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2314          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2315            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2316              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2317    #ifdef SUPPORT_UCP
2318              if (md->use_ucp)
2319                {
2320                c = eptr[-1];
2321                if (c == '_') prev_is_word = TRUE; else
2322                  {
2323                  int cat = UCD_CATEGORY(c);
2324                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2325                  }
2326                }
2327              else
2328    #endif
2329              prev_is_word = MAX_255(eptr[-1])
2330                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2331              }
2332    
2333            /* Get status of next character */
2334    
2335            if (eptr >= md->end_subject)
2336              {
2337              SCHECK_PARTIAL();
2338              cur_is_word = FALSE;
2339              }
2340            else
2341    #ifdef SUPPORT_UCP
2342            if (md->use_ucp)
2343              {
2344              c = *eptr;
2345              if (c == '_') cur_is_word = TRUE; else
2346                {
2347                int cat = UCD_CATEGORY(c);
2348                cur_is_word = (cat == ucp_L || cat == ucp_N);
2349                }
2350              }
2351            else
2352    #endif
2353            cur_is_word = MAX_255(*eptr)
2354              && ((md->ctypes[*eptr] & ctype_word) != 0);
2355          }          }
2356    
2357        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1380  for (;;) Line 2362  for (;;)
2362        }        }
2363      break;      break;
2364    
2365      /* Match a single character type; inline for speed */      /* Match any single character type except newline; have to take care with
2366        CRLF newlines and partial matching. */
2367    
2368      case OP_ANY:      case OP_ANY:
2369      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2370        if (md->partial != 0 &&
2371            eptr + 1 >= md->end_subject &&
2372            NLBLOCK->nltype == NLTYPE_FIXED &&
2373            NLBLOCK->nllen == 2 &&
2374            RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2375        {        {
2376        if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);        md->hitend = TRUE;
2377          if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2378          }
2379    
2380        /* Fall through */
2381    
2382        /* Match any single character whatsoever. */
2383    
2384        case OP_ALLANY:
2385        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2386          {                            /* not be updated before SCHECK_PARTIAL. */
2387          SCHECK_PARTIAL();
2388          RRETURN(MATCH_NOMATCH);
2389        }        }
2390      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      eptr++;
2391      if (utf8)  #ifdef SUPPORT_UTF
2392        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2393    #endif
2394      ecode++;      ecode++;
2395      break;      break;
2396    
# Line 1397  for (;;) Line 2398  for (;;)
2398      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2399    
2400      case OP_ANYBYTE:      case OP_ANYBYTE:
2401      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2402          {                            /* not be updated before SCHECK_PARTIAL. */
2403          SCHECK_PARTIAL();
2404          RRETURN(MATCH_NOMATCH);
2405          }
2406        eptr++;
2407      ecode++;      ecode++;
2408      break;      break;
2409    
2410      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2411      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2412          {
2413          SCHECK_PARTIAL();
2414          RRETURN(MATCH_NOMATCH);
2415          }
2416      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2417      if (      if (
2418  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2419         c < 256 &&         c < 256 &&
2420  #endif  #endif
2421         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1415  for (;;) Line 2425  for (;;)
2425      break;      break;
2426    
2427      case OP_DIGIT:      case OP_DIGIT:
2428      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2429          {
2430          SCHECK_PARTIAL();
2431          RRETURN(MATCH_NOMATCH);
2432          }
2433      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2434      if (      if (
2435  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2436         c >= 256 ||         c > 255 ||
2437  #endif  #endif
2438         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2439         )         )
# Line 1428  for (;;) Line 2442  for (;;)
2442      break;      break;
2443    
2444      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2445      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2446          {
2447          SCHECK_PARTIAL();
2448          RRETURN(MATCH_NOMATCH);
2449          }
2450      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2451      if (      if (
2452  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2453         c < 256 &&         c < 256 &&
2454  #endif  #endif
2455         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1441  for (;;) Line 2459  for (;;)
2459      break;      break;
2460    
2461      case OP_WHITESPACE:      case OP_WHITESPACE:
2462      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2463          {
2464          SCHECK_PARTIAL();
2465          RRETURN(MATCH_NOMATCH);
2466          }
2467      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2468      if (      if (
2469  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2470         c >= 256 ||         c > 255 ||
2471  #endif  #endif
2472         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2473         )         )
# Line 1454  for (;;) Line 2476  for (;;)
2476      break;      break;
2477    
2478      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2479      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2480          {
2481          SCHECK_PARTIAL();
2482          RRETURN(MATCH_NOMATCH);
2483          }
2484      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2485      if (      if (
2486  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2487         c < 256 &&         c < 256 &&
2488  #endif  #endif
2489         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1467  for (;;) Line 2493  for (;;)
2493      break;      break;
2494    
2495      case OP_WORDCHAR:      case OP_WORDCHAR:
2496      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2497          {
2498          SCHECK_PARTIAL();
2499          RRETURN(MATCH_NOMATCH);
2500          }
2501      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2502      if (      if (
2503  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2504         c >= 256 ||         c > 255 ||
2505  #endif  #endif
2506         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2507         )         )
# Line 1480  for (;;) Line 2510  for (;;)
2510      break;      break;
2511    
2512      case OP_ANYNL:      case OP_ANYNL:
2513      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2514          {
2515          SCHECK_PARTIAL();
2516          RRETURN(MATCH_NOMATCH);
2517          }
2518      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2519      switch(c)      switch(c)
2520        {        {
2521        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2522        case 0x000d:  
2523        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        case CHAR_CR:
2524          if (eptr >= md->end_subject)
2525            {
2526            SCHECK_PARTIAL();
2527            }
2528          else if (RAWUCHARTEST(eptr) == CHAR_LF) eptr++;
2529          break;
2530    
2531          case CHAR_LF:
2532        break;        break;
2533        case 0x000a:  
2534        case 0x000b:        case CHAR_VT:
2535        case 0x000c:        case CHAR_FF:
2536        case 0x0085:        case CHAR_NEL:
2537    #ifndef EBCDIC
2538        case 0x2028:        case 0x2028:
2539        case 0x2029:        case 0x2029:
2540    #endif  /* Not EBCDIC */
2541          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2542        break;        break;
2543        }        }
2544      ecode++;      ecode++;
2545      break;      break;
2546    
2547      case OP_NOT_HSPACE:      case OP_NOT_HSPACE:
2548      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2549          {
2550          SCHECK_PARTIAL();
2551          RRETURN(MATCH_NOMATCH);
2552          }
2553      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2554      switch(c)      switch(c)
2555        {        {
2556          HSPACE_CASES: RRETURN(MATCH_NOMATCH);  /* Byte and multibyte cases */
2557        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);  
2558        }        }
2559      ecode++;      ecode++;
2560      break;      break;
2561    
2562      case OP_HSPACE:      case OP_HSPACE:
2563      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2564          {
2565          SCHECK_PARTIAL();
2566          RRETURN(MATCH_NOMATCH);
2567          }
2568      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2569      switch(c)      switch(c)
2570        {        {
2571          HSPACE_CASES: break;  /* Byte and multibyte cases */
2572        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;  
2573        }        }
2574      ecode++;      ecode++;
2575      break;      break;
2576    
2577      case OP_NOT_VSPACE:      case OP_NOT_VSPACE:
2578      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2579          {
2580          SCHECK_PARTIAL();
2581          RRETURN(MATCH_NOMATCH);
2582          }
2583      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2584      switch(c)      switch(c)
2585        {        {
2586          VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2587        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);  
2588        }        }
2589      ecode++;      ecode++;
2590      break;      break;
2591    
2592      case OP_VSPACE:      case OP_VSPACE:
2593      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2594          {
2595          SCHECK_PARTIAL();
2596          RRETURN(MATCH_NOMATCH);
2597          }
2598      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2599      switch(c)      switch(c)
2600        {        {
2601          VSPACE_CASES: break;
2602        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;  
2603        }        }
2604      ecode++;      ecode++;
2605      break;      break;
# Line 1601  for (;;) Line 2610  for (;;)
2610    
2611      case OP_PROP:      case OP_PROP:
2612      case OP_NOTPROP:      case OP_NOTPROP:
2613      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2614          {
2615          SCHECK_PARTIAL();
2616          RRETURN(MATCH_NOMATCH);
2617          }
2618      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2619        {        {
2620        int chartype, script;        const pcre_uint32 *cp;
2621        int category = _pcre_ucp_findprop(c, &chartype, &script);        const ucd_record *prop = GET_UCD(c);
2622    
2623        switch(ecode[1])        switch(ecode[1])
2624          {          {
# Line 1614  for (;;) Line 2627  for (;;)
2627          break;          break;
2628    
2629          case PT_LAMP:          case PT_LAMP:
2630          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2631               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2632               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2633            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2634           break;          break;
2635    
2636          case PT_GC:          case PT_GC:
2637          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2638            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2639          break;          break;
2640    
2641          case PT_PC:          case PT_PC:
2642          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2643            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2644          break;          break;
2645    
2646          case PT_SC:          case PT_SC:
2647          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2648              RRETURN(MATCH_NOMATCH);
2649            break;
2650    
2651            /* These are specials */
2652    
2653            case PT_ALNUM:
2654            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2655                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2656              RRETURN(MATCH_NOMATCH);
2657            break;
2658    
2659            case PT_SPACE:    /* Perl space */
2660            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2661                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2662                   == (op == OP_NOTPROP))
2663              RRETURN(MATCH_NOMATCH);
2664            break;
2665    
2666            case PT_PXSPACE:  /* POSIX space */
2667            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2668                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2669                 c == CHAR_FF || c == CHAR_CR)
2670                   == (op == OP_NOTPROP))
2671              RRETURN(MATCH_NOMATCH);
2672            break;
2673    
2674            case PT_WORD:
2675            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2676                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2677                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2678              RRETURN(MATCH_NOMATCH);
2679            break;
2680    
2681            case PT_CLIST:
2682            cp = PRIV(ucd_caseless_sets) + ecode[2];
2683            for (;;)
2684              {
2685              if (c < *cp)
2686                { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; }
2687              if (c == *cp++)
2688                { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } }
2689              }
2690            break;
2691    
2692            case PT_UCNC:
2693            if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
2694                 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
2695                 c >= 0xe000) == (op == OP_NOTPROP))
2696            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2697          break;          break;
2698    
2699            /* This should never occur */
2700    
2701          default:          default:
2702          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
2703          }          }
# Line 1647  for (;;) Line 2710  for (;;)
2710      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2711    
2712      case OP_EXTUNI:      case OP_EXTUNI:
2713      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
     GETCHARINCTEST(c, eptr);  
2714        {        {
2715        int chartype, script;        SCHECK_PARTIAL();
2716        int category = _pcre_ucp_findprop(c, &chartype, &script);        RRETURN(MATCH_NOMATCH);
2717        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        }
2718        else
2719          {
2720          int lgb, rgb;
2721          GETCHARINCTEST(c, eptr);
2722          lgb = UCD_GRAPHBREAK(c);
2723        while (eptr < md->end_subject)        while (eptr < md->end_subject)
2724          {          {
2725          int len = 1;          int len = 1;
2726          if (!utf8) c = *eptr; else          if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2727            {          rgb = UCD_GRAPHBREAK(c);
2728            GETCHARLEN(c, eptr, len);          if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2729            }          lgb = rgb;
         category = _pcre_ucp_findprop(c, &chartype, &script);  
         if (category != ucp_M) break;  
2730          eptr += len;          eptr += len;
2731          }          }
2732        }        }
2733        CHECK_PARTIAL();
2734      ecode++;      ecode++;
2735      break;      break;
2736  #endif  #endif  /* SUPPORT_UCP */
2737    
2738    
2739      /* 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 1676  for (;;) Line 2742  for (;;)
2742      similar code to character type repeats - written out again for speed.      similar code to character type repeats - written out again for speed.
2743      However, if the referenced string is the empty string, always treat      However, if the referenced string is the empty string, always treat
2744      it as matched, any number of times (otherwise there could be infinite      it as matched, any number of times (otherwise there could be infinite
2745      loops). */      loops). If the reference is unset, there are two possibilities:
2746    
2747        (a) In the default, Perl-compatible state, set the length negative;
2748        this ensures that every attempt at a match fails. We can't just fail
2749        here, because of the possibility of quantifiers with zero minima.
2750    
2751        (b) If the JavaScript compatibility flag is set, set the length to zero
2752        so that the back reference matches an empty string.
2753    
2754        Otherwise, set the length to the length of what was matched by the
2755        referenced subpattern.
2756    
2757        The OP_REF and OP_REFI opcodes are used for a reference to a numbered group
2758        or to a non-duplicated named group. For a duplicated named group, OP_DNREF
2759        and OP_DNREFI are used. In this case we must scan the list of groups to
2760        which the name refers, and use the first one that is set. */
2761    
2762        case OP_DNREF:
2763        case OP_DNREFI:
2764        caseless = op == OP_DNREFI;
2765          {
2766          int count = GET2(ecode, 1+IMM2_SIZE);
2767          pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
2768          ecode += 1 + 2*IMM2_SIZE;
2769    
2770          while (count-- > 0)
2771            {
2772            offset = GET2(slot, 0) << 1;
2773            if (offset < offset_top && md->offset_vector[offset] >= 0) break;
2774            slot += md->name_entry_size;
2775            }
2776          if (count < 0)
2777            length = (md->jscript_compat)? 0 : -1;
2778          else
2779            length = md->offset_vector[offset+1] - md->offset_vector[offset];
2780          }
2781        goto REF_REPEAT;
2782    
2783      case OP_REF:      case OP_REF:
2784        {      case OP_REFI:
2785        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2786        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2787        ecode += 1 + IMM2_SIZE;
2788    
       /* If the reference is unset, set the length to be longer than the amount  
       of subject left; this ensures that every attempt at a match fails. We  
       can't just fail here, because of the possibility of quantifiers with zero  
       minima. */  
   
       length = (offset >= offset_top || md->offset_vector[offset] < 0)?  
         md->end_subject - eptr + 1 :  
         md->offset_vector[offset+1] - md->offset_vector[offset];  
2789    
2790        /* Set up for repetition, or handle the non-repeated case */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2791          length = (md->jscript_compat)? 0 : -1;
2792        else
2793          length = md->offset_vector[offset+1] - md->offset_vector[offset];
2794    
2795        switch (*ecode)      /* Set up for repetition, or handle the non-repeated case */
         {  
         case OP_CRSTAR:  
         case OP_CRMINSTAR:  
         case OP_CRPLUS:  
         case OP_CRMINPLUS:  
         case OP_CRQUERY:  
         case OP_CRMINQUERY:  
         c = *ecode++ - OP_CRSTAR;  
         minimize = (c & 1) != 0;  
         min = rep_min[c];                 /* Pick up values from tables; */  
         max = rep_max[c];                 /* zero for max => infinity */  
         if (max == 0) max = INT_MAX;  
         break;  
2796    
2797          case OP_CRRANGE:      REF_REPEAT:
2798          case OP_CRMINRANGE:      switch (*ecode)
2799          minimize = (*ecode == OP_CRMINRANGE);        {
2800          min = GET2(ecode, 1);        case OP_CRSTAR:
2801          max = GET2(ecode, 3);        case OP_CRMINSTAR:
2802          if (max == 0) max = INT_MAX;        case OP_CRPLUS:
2803          ecode += 5;        case OP_CRMINPLUS:
2804          break;        case OP_CRQUERY:
2805          case OP_CRMINQUERY:
2806          c = *ecode++ - OP_CRSTAR;
2807          minimize = (c & 1) != 0;
2808          min = rep_min[c];                 /* Pick up values from tables; */
2809          max = rep_max[c];                 /* zero for max => infinity */
2810          if (max == 0) max = INT_MAX;
2811          break;
2812    
2813          default:               /* No repeat follows */        case OP_CRRANGE:
2814          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        case OP_CRMINRANGE:
2815          eptr += length;        minimize = (*ecode == OP_CRMINRANGE);
2816          continue;              /* With the main loop */        min = GET2(ecode, 1);
2817          max = GET2(ecode, 1 + IMM2_SIZE);
2818          if (max == 0) max = INT_MAX;
2819          ecode += 1 + 2 * IMM2_SIZE;
2820          break;
2821    
2822          default:               /* No repeat follows */
2823          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2824            {
2825            if (length == -2) eptr = md->end_subject;   /* Partial match */
2826            CHECK_PARTIAL();
2827            RRETURN(MATCH_NOMATCH);
2828          }          }
2829          eptr += length;
2830          continue;              /* With the main loop */
2831          }
2832    
2833        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2834        main loop. */      zero, just continue with the main loop. If the length is negative, it
2835        means the reference is unset in non-Java-compatible mode. If the minimum is
2836        zero, we can continue at the same level without recursion. For any other
2837        minimum, carrying on will result in NOMATCH. */
2838    
2839        if (length == 0) continue;      if (length == 0) continue;
2840        if (length < 0 && min == 0) continue;
2841    
2842        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2843        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2844        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2845    
2846        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2847          {
2848          int slength;
2849          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2850          {          {
2851          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          if (slength == -2) eptr = md->end_subject;   /* Partial match */
2852          eptr += length;          CHECK_PARTIAL();
2853            RRETURN(MATCH_NOMATCH);
2854          }          }
2855          eptr += slength;
2856          }
2857    
2858        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2859        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2860    
2861        if (min == max) continue;      if (min == max) continue;
2862    
2863        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2864    
2865        if (minimize)      if (minimize)
2866          {
2867          for (fi = min;; fi++)
2868          {          {
2869          for (fi = min;; fi++)          int slength;
2870            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2871            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2872            if (fi >= max) RRETURN(MATCH_NOMATCH);
2873            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2874            {            {
2875            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2876            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2877            if (fi >= max || !match_ref(offset, eptr, length, md, ims))            RRETURN(MATCH_NOMATCH);
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2878            }            }
2879          /* Control never gets here */          eptr += slength;
2880          }          }
2881          /* Control never gets here */
2882          }
2883    
2884        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2885    
2886        else      else
2887          {
2888          pp = eptr;
2889          for (i = min; i < max; i++)
2890          {          {
2891          pp = eptr;          int slength;
2892          for (i = min; i < max; i++)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2893            {            {
2894            if (!match_ref(offset, eptr, length, md, ims)) break;            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2895            eptr += length;            the soft partial matching case. */
2896            }  
2897          while (eptr >= pp)            if (slength == -2 && md->partial != 0 &&
2898            {                md->end_subject > md->start_used_ptr)
2899            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);              {
2900            if (rrc != MATCH_NOMATCH) RRETURN(rrc);              md->hitend = TRUE;
2901            eptr -= length;              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2902                }
2903              break;
2904            }            }
2905          RRETURN(MATCH_NOMATCH);          eptr += slength;
2906          }          }
2907    
2908          while (eptr >= pp)
2909            {
2910            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2911            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2912            eptr -= length;
2913            }
2914          RRETURN(MATCH_NOMATCH);
2915        }        }
2916      /* Control never gets here */      /* Control never gets here */
2917    
   
   
2918      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2919      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,
2920      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 1796  for (;;) Line 2929  for (;;)
2929      case OP_NCLASS:      case OP_NCLASS:
2930      case OP_CLASS:      case OP_CLASS:
2931        {        {
2932          /* The data variable is saved across frames, so the byte map needs to
2933          be stored there. */
2934    #define BYTE_MAP ((pcre_uint8 *)data)
2935        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2936        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2937    
2938        switch (*ecode)        switch (*ecode)
2939          {          {
# Line 1818  for (;;) Line 2954  for (;;)
2954          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2955          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2956          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2957          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2958          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2959          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2960          break;          break;
2961    
2962          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1830  for (;;) Line 2966  for (;;)
2966    
2967        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2968    
2969  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2970        /* UTF-8 mode */        if (utf)
       if (utf8)  
2971          {          {
2972          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2973            {            {
2974            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2975                {
2976                SCHECK_PARTIAL();
2977                RRETURN(MATCH_NOMATCH);
2978                }
2979            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2980            if (c > 255)            if (c > 255)
2981              {              {
2982              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2983              }              }
2984            else            else
2985              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2986            }            }
2987          }          }
2988        else        else
2989  #endif  #endif
2990        /* Not UTF-8 mode */        /* Not UTF mode */
2991          {          {
2992          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2993            {            {
2994            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2995            c = *eptr++;              {
2996            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);              SCHECK_PARTIAL();
2997                RRETURN(MATCH_NOMATCH);
2998                }
2999              c = *eptr++;
3000    #ifndef COMPILE_PCRE8
3001              if (c > 255)
3002                {
3003                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
3004                }
3005              else
3006    #endif
3007                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
3008            }            }
3009          }          }
3010    
# Line 1870  for (;;) Line 3018  for (;;)
3018    
3019        if (minimize)        if (minimize)
3020          {          {
3021  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3022          /* UTF-8 mode */          if (utf)
         if (utf8)  
3023            {            {
3024            for (fi = min;; fi++)            for (fi = min;; fi++)
3025              {              {
3026              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
3027              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3028              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3029                if (eptr >= md->end_subject)
3030                  {
3031                  SCHECK_PARTIAL();
3032                  RRETURN(MATCH_NOMATCH);
3033                  }
3034              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
3035              if (c > 255)              if (c > 255)
3036                {                {
3037                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
3038                }                }
3039              else              else
3040                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
3041              }              }
3042            }            }
3043          else          else
3044  #endif  #endif
3045          /* Not UTF-8 mode */          /* Not UTF mode */
3046            {            {
3047            for (fi = min;; fi++)            for (fi = min;; fi++)
3048              {              {
3049              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
3050              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3051              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3052                if (eptr >= md->end_subject)
3053                  {
3054                  SCHECK_PARTIAL();
3055                  RRETURN(MATCH_NOMATCH);
3056                  }
3057              c = *eptr++;              c = *eptr++;
3058              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
3059                if (c > 255)
3060                  {
3061                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
3062                  }
3063                else
3064    #endif
3065                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
3066              }              }
3067            }            }
3068          /* Control never gets here */          /* Control never gets here */
# Line 1912  for (;;) Line 3074  for (;;)
3074          {          {
3075          pp = eptr;          pp = eptr;
3076    
3077  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3078          /* UTF-8 mode */          if (utf)
         if (utf8)  
3079            {            {
3080            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3081              {              {
3082              int len = 1;              int len = 1;
3083              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3084                  {
3085                  SCHECK_PARTIAL();
3086                  break;
3087                  }
3088              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
3089              if (c > 255)              if (c > 255)
3090                {                {
3091                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
3092                }                }
3093              else              else
3094                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
3095              eptr += len;              eptr += len;
3096              }              }
3097            for (;;)            for (;;)
3098              {              {
3099              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
3100              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3101              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3102              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1941  for (;;) Line 3104  for (;;)
3104            }            }
3105          else          else
3106  #endif  #endif
3107            /* Not UTF-8 mode */            /* Not UTF mode */
3108            {            {
3109            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3110              {              {
3111              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3112                  {
3113                  SCHECK_PARTIAL();
3114                  break;
3115                  }
3116              c = *eptr;              c = *eptr;
3117              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
3118                if (c > 255)
3119                  {
3120                  if (op == OP_CLASS) break;
3121                  }
3122                else
3123    #endif
3124                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3125              eptr++;              eptr++;
3126              }              }
3127            while (eptr >= pp)            while (eptr >= pp)
3128              {              {
3129              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3130              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3131              eptr--;              eptr--;
3132              }              }
# Line 1960  for (;;) Line 3134  for (;;)
3134    
3135          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3136          }          }
3137    #undef BYTE_MAP
3138        }        }
3139      /* Control never gets here */      /* Control never gets here */
3140    
3141    
3142      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
3143      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
3144        mode, because Unicode properties are supported in non-UTF-8 mode. */
3145    
3146  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3147      case OP_XCLASS:      case OP_XCLASS:
3148        {        {
3149        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 1992  for (;;) Line 3168  for (;;)
3168          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3169          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3170          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3171          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3172          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3173          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3174          break;          break;
3175    
3176          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 2006  for (;;) Line 3182  for (;;)
3182    
3183        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3184          {          {
3185          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3186          GETCHARINC(c, eptr);            {
3187          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
3188              RRETURN(MATCH_NOMATCH);
3189              }
3190            GETCHARINCTEST(c, eptr);
3191            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3192          }          }
3193    
3194        /* 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 2023  for (;;) Line 3203  for (;;)
3203          {          {
3204          for (fi = min;; fi++)          for (fi = min;; fi++)
3205            {            {
3206            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3207            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3208            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3209            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3210            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3211                SCHECK_PARTIAL();
3212                RRETURN(MATCH_NOMATCH);
3213                }
3214              GETCHARINCTEST(c, eptr);
3215              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3216            }            }
3217          /* Control never gets here */          /* Control never gets here */
3218          }          }
# Line 2040  for (;;) Line 3225  for (;;)
3225          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3226            {            {
3227            int len = 1;            int len = 1;
3228            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3229            GETCHARLEN(c, eptr, len);              {
3230            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3231                break;
3232                }
3233    #ifdef SUPPORT_UTF
3234              GETCHARLENTEST(c, eptr, len);
3235    #else
3236              c = *eptr;
3237    #endif
3238              if (!PRIV(xclass)(c, data, utf)) break;
3239            eptr += len;            eptr += len;
3240            }            }
3241          for(;;)          for(;;)
3242            {            {
3243            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3244            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3245            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3246            BACKCHAR(eptr)  #ifdef SUPPORT_UTF
3247              if (utf) BACKCHAR(eptr);
3248    #endif
3249            }            }
3250          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3251          }          }
# Line 2062  for (;;) Line 3257  for (;;)
3257      /* Match a single character, casefully */      /* Match a single character, casefully */
3258    
3259      case OP_CHAR:      case OP_CHAR:
3260  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3261      if (utf8)      if (utf)
3262        {        {
3263        length = 1;        length = 1;
3264        ecode++;        ecode++;
3265        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3266        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3267        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);          {
3268            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3269            RRETURN(MATCH_NOMATCH);
3270            }
3271          while (length-- > 0) if (*ecode++ != RAWUCHARINC(eptr)) RRETURN(MATCH_NOMATCH);
3272        }        }
3273      else      else
3274  #endif  #endif
3275        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3276        {        {
3277        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3278            {
3279            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3280            RRETURN(MATCH_NOMATCH);
3281            }
3282        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3283        ecode += 2;        ecode += 2;
3284        }        }
3285      break;      break;
3286    
3287      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3288        subject, give up immediately. */
3289    
3290        case OP_CHARI:
3291        if (eptr >= md->end_subject)
3292          {
3293          SCHECK_PARTIAL();
3294          RRETURN(MATCH_NOMATCH);
3295          }
3296    
3297      case OP_CHARNC:  #ifdef SUPPORT_UTF
3298  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3299        {        {
3300        length = 1;        length = 1;
3301        ecode++;        ecode++;
3302        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3303    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3304        /* 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
3305        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3306          fast lookup table. We know that there is at least one byte left in the
3307          subject. */
3308    
3309        if (fc < 128)        if (fc < 128)
3310          {          {
3311          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          pcre_uint32 cc = RAWUCHAR(eptr);
3312            if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH);
3313            ecode++;
3314            eptr++;
3315          }          }
3316    
3317        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3318          use the value of "length" to check for sufficient bytes left, because the
3319          other case of the character may have more or fewer bytes.  */
3320    
3321        else        else
3322          {          {
3323          unsigned int dc;          pcre_uint32 dc;
3324          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
3325          ecode += length;          ecode += length;
3326    
# Line 2116  for (;;) Line 3330  for (;;)
3330          if (fc != dc)          if (fc != dc)
3331            {            {
3332  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3333            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3334  #endif  #endif
3335              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3336            }            }
3337          }          }
3338        }        }
3339      else      else
3340  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3341    
3342      /* Non-UTF-8 mode */      /* Not UTF mode */
3343        {        {
3344        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3345        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3346          eptr++;
3347        ecode += 2;        ecode += 2;
3348        }        }
3349      break;      break;
# Line 2136  for (;;) Line 3351  for (;;)
3351      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3352    
3353      case OP_EXACT:      case OP_EXACT:
3354        case OP_EXACTI:
3355      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3356      ecode += 3;      ecode += 1 + IMM2_SIZE;
3357      goto REPEATCHAR;      goto REPEATCHAR;
3358    
3359      case OP_POSUPTO:      case OP_POSUPTO:
3360        case OP_POSUPTOI:
3361      possessive = TRUE;      possessive = TRUE;
3362      /* Fall through */      /* Fall through */
3363    
3364      case OP_UPTO:      case OP_UPTO:
3365        case OP_UPTOI:
3366      case OP_MINUPTO:      case OP_MINUPTO:
3367        case OP_MINUPTOI:
3368      min = 0;      min = 0;
3369      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3370      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3371      ecode += 3;      ecode += 1 + IMM2_SIZE;
3372      goto REPEATCHAR;      goto REPEATCHAR;
3373    
3374      case OP_POSSTAR:      case OP_POSSTAR:
3375        case OP_POSSTARI:
3376      possessive = TRUE;      possessive = TRUE;
3377      min = 0;      min = 0;
3378      max = INT_MAX;      max = INT_MAX;
# Line 2160  for (;;) Line 3380  for (;;)
3380      goto REPEATCHAR;      goto REPEATCHAR;
3381    
3382      case OP_POSPLUS:      case OP_POSPLUS:
3383        case OP_POSPLUSI:
3384      possessive = TRUE;      possessive = TRUE;
3385      min = 1;      min = 1;
3386      max = INT_MAX;      max = INT_MAX;
# Line 2167  for (;;) Line 3388  for (;;)
3388      goto REPEATCHAR;      goto REPEATCHAR;
3389    
3390      case OP_POSQUERY:      case OP_POSQUERY:
3391        case OP_POSQUERYI:
3392      possessive = TRUE;      possessive = TRUE;
3393      min = 0;      min = 0;
3394      max = 1;      max = 1;
# Line 2174  for (;;) Line 3396  for (;;)
3396      goto REPEATCHAR;      goto REPEATCHAR;
3397    
3398      case OP_STAR:      case OP_STAR:
3399        case OP_STARI:
3400      case OP_MINSTAR:      case OP_MINSTAR:
3401        case OP_MINSTARI:
3402      case OP_PLUS:      case OP_PLUS:
3403        case OP_PLUSI:
3404      case OP_MINPLUS:      case OP_MINPLUS:
3405        case OP_MINPLUSI:
3406      case OP_QUERY:      case OP_QUERY:
3407        case OP_QUERYI:
3408      case OP_MINQUERY:      case OP_MINQUERY:
3409      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3410        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3411      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3412      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3413      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3414      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3415    
3416      /* Common code for all repeated single-character matches. We can give      /* Common code for all repeated single-character matches. We first check
3417      up quickly if there are fewer than the minimum number of characters left in      for the minimum number of characters. If the minimum equals the maximum, we
3418      the subject. */      are done. Otherwise, if minimizing, check the rest of the pattern for a
3419        match; if there isn't one, advance up to the maximum, one character at a
3420        time.
3421    
3422        If maximizing, advance up to the maximum number of matching characters,
3423        until eptr is past the end of the maximum run. If possessive, we are
3424        then done (no backing up). Otherwise, match at this position; anything
3425        other than no match is immediately returned. For nomatch, back up one
3426        character, unless we are matching \R and the last thing matched was
3427        \r\n, in which case, back up two bytes. When we reach the first optional
3428        character position, we can save stack by doing a tail recurse.
3429    
3430        The various UTF/non-UTF and caseful/caseless cases are handled separately,
3431        for speed. */
3432    
3433      REPEATCHAR:      REPEATCHAR:
3434  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3435      if (utf8)      if (utf)
3436        {        {
3437        length = 1;        length = 1;
3438        charptr = ecode;        charptr = ecode;
3439        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3440        ecode += length;        ecode += length;
3441    
3442        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 2205  for (;;) Line 3445  for (;;)
3445        if (length > 1)        if (length > 1)
3446          {          {
3447  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3448          unsigned int othercase;          pcre_uint32 othercase;
3449          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3450              (othercase = _pcre_ucp_othercase(fc)) != NOTACHAR)              (othercase = UCD_OTHERCASE(fc)) != fc)
3451            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3452          else oclength = 0;          else oclength = 0;
3453  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3454    
3455          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3456            {            {
3457            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3458                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3459  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3460            /* Need braces because of following else */            else if (oclength > 0 &&
3461            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                     eptr <= md->end_subject - oclength &&
3462                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3463    #endif  /* SUPPORT_UCP */
3464            else            else
3465              {              {
3466              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3467              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3468              }              }
 #else   /* without SUPPORT_UCP */  
           else { RRETURN(MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3469            }            }
3470    
3471          if (min == max) continue;          if (min == max) continue;
# Line 2234  for (;;) Line 3474  for (;;)
3474            {            {
3475            for (fi = min;; fi++)            for (fi = min;; fi++)
3476              {              {
3477              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM22);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3478              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3479              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3480              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3481                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3482  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3483              /* Need braces because of following else */              else if (oclength > 0 &&
3484              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                       eptr <= md->end_subject - oclength &&
3485                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3486    #endif  /* SUPPORT_UCP */
3487              else              else
3488                {                {
3489                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3490                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3491                }                }
 #else   /* without SUPPORT_UCP */  
             else { RRETURN (MATCH_NOMATCH); }  
 #endif  /* SUPPORT_UCP */  
3492              }              }
3493            /* Control never gets here */            /* Control never gets here */
3494            }            }
# Line 2258  for (;;) Line 3498  for (;;)
3498            pp = eptr;            pp = eptr;
3499            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3500              {              {
3501              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3502              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3503  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3504              else if (oclength == 0) break;              else if (oclength > 0 &&
3505                         eptr <= md->end_subject - oclength &&
3506                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3507    #endif  /* SUPPORT_UCP */
3508              else              else
3509                {                {
3510                if (memcmp(eptr, occhars, oclength) != 0) break;                CHECK_PARTIAL();
3511                eptr += oclength;                break;
3512                }                }
 #else   /* without SUPPORT_UCP */  
             else break;  
 #endif  /* SUPPORT_UCP */  
3513              }              }
3514    
3515            if (possessive) continue;            if (possessive) continue;    /* No backtracking */
3516            for(;;)            for(;;)
3517             {              {
3518             RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM23);              if (eptr == pp) goto TAIL_RECURSE;
3519             if (rrc != MATCH_NOMATCH) RRETURN(rrc);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM23);
3520             if (eptr == pp) RRETURN(MATCH_NOMATCH);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3521  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3522             eptr--;              eptr--;
3523             BACKCHAR(eptr);              BACKCHAR(eptr);
3524  #else   /* without SUPPORT_UCP */  #else   /* without SUPPORT_UCP */
3525             eptr -= length;              eptr -= length;
3526  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3527             }              }
3528            }            }
3529          /* Control never gets here */          /* Control never gets here */
3530          }          }
# Line 2294  for (;;) Line 3534  for (;;)
3534        value of fc will always be < 128. */        value of fc will always be < 128. */
3535        }        }
3536      else      else
3537  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF */
3538          /* 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);  
3539        fc = *ecode++;        fc = *ecode++;
       }  
3540    
3541      /* 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
3542      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
3543