/[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 87 by nigel, Sat Feb 24 21:41:21 2007 UTC revision 1376 by ph10, Sat Oct 12 18:02:11 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) 1997-2006 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
45    #include "config.h"
46    #endif
47    
48    #define NLBLOCK md             /* Block containing newline information */
49    #define PSSTART start_subject  /* Field containing processed string start */
50    #define PSEND   end_subject    /* Field containing processed string end */
51    
52  #include "pcre_internal.h"  #include "pcre_internal.h"
53    
54    /* Undefine some potentially clashing cpp symbols */
55    
56  /* Structure for building a chain of data that actually lives on the  #undef min
57  stack, for holding the values of the subject pointer at the start of each  #undef max
 subpattern, so as to detect when an empty string has been matched by a  
 subpattern - to break infinite loops. When NO_RECURSE is set, these blocks  
 are on the heap, not on the stack. */  
   
 typedef struct eptrblock {  
   struct eptrblock *epb_prev;  
   USPTR epb_saved_eptr;  
 } eptrblock;  
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_isgroup      0x02    /* Set if start of bracketed 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  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 186  calls by keeping local variables that ne Line 286  calls by keeping local variables that ne
286  obtained from malloc() instead instead of on the stack. Macros are used to  obtained from malloc() instead instead of on the stack. Macros are used to
287  achieve this so that the actual code doesn't look very different to what it  achieve this so that the actual code doesn't look very different to what it
288  always used to.  always used to.
289    
290    The original heap-recursive code used longjmp(). However, it seems that this
291    can be very slow on some operating systems. Following a suggestion from Stan
292    Switzer, the use of longjmp() has been abolished, at the cost of having to
293    provide a unique number for each call to RMATCH. There is no way of generating
294    a sequence of numbers at compile time in C. I have given them names, to make
295    them stand out more clearly.
296    
297    Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
298    FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
299    tests. Furthermore, not using longjmp() means that local dynamic variables
300    don't have indeterminate values; this has meant that the frame size can be
301    reduced because the result can be "passed back" by straight setting of the
302    variable instead of being passed in the frame.
303  ****************************************************************************  ****************************************************************************
304  ***************************************************************************/  ***************************************************************************/
305    
306    /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
307    below must be updated in sync.  */
308    
309    enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
310           RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
311           RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
312           RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
313           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. */  versions and production versions. Note that the "rw" argument of RMATCH isn't
319    actually used in this definition. */
320    
321  #ifndef NO_RECURSE  #ifndef NO_RECURSE
322  #define REGISTER register  #define REGISTER register
323  #ifdef DEBUG  
324  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #ifdef PCRE_DEBUG
325    #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    rx = match(ra,rb,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(rx,ra,rb,rc,rd,re,rf,rg) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
338    rx = match(ra,rb,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    
342  #else  #else
343    
344    
345  /* These versions of the macros manage a private stack on the heap. Note  /* These versions of the macros manage a private stack on the heap. Note that
346  that the rd argument of RMATCH isn't actually used. It's the md argument of  the "rd" argument of RMATCH isn't actually used in this definition. It's the md
347  match(), which never changes. */  argument of match(), which never changes. */
348    
349  #define REGISTER  #define REGISTER
350    
351  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
352    {\    {\
353    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = frame->Xnextframe;\
354    if (setjmp(frame->Xwhere) == 0)\    if (newframe == NULL)\
     {\  
     newframe->Xeptr = ra;\  
     newframe->Xecode = rb;\  
     newframe->Xoffset_top = rc;\  
     newframe->Xims = re;\  
     newframe->Xeptrb = rf;\  
     newframe->Xflags = rg;\  
     newframe->Xrdepth = frame->Xrdepth + 1;\  
     newframe->Xprevframe = frame;\  
     frame = newframe;\  
     DPRINTF(("restarting from line %d\n", __LINE__));\  
     goto HEAP_RECURSE;\  
     }\  
   else\  
355      {\      {\
356      DPRINTF(("longjumped back to line %d\n", __LINE__));\      newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
357      frame = md->thisframe;\      if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
358      rx = frame->Xresult;\      newframe->Xnextframe = NULL;\
359        frame->Xnextframe = newframe;\
360      }\      }\
361      frame->Xwhere = rw;\
362      newframe->Xeptr = ra;\
363      newframe->Xecode = rb;\
364      newframe->Xmstart = mstart;\
365      newframe->Xoffset_top = rc;\
366      newframe->Xeptrb = re;\
367      newframe->Xrdepth = frame->Xrdepth + 1;\
368      newframe->Xprevframe = frame;\
369      frame = newframe;\
370      DPRINTF(("restarting from line %d\n", __LINE__));\
371      goto HEAP_RECURSE;\
372      L_##rw:\
373      DPRINTF(("jumped back to line %d\n", __LINE__));\
374    }    }
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      frame->Xresult = ra;\      rrc = ra;\
383      md->thisframe = frame;\      goto HEAP_RETURN;\
     longjmp(frame->Xwhere, 1);\  
384      }\      }\
385    return ra;\    return ra;\
386    }    }
# Line 266  match(), which never changes. */ Line 390  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      PCRE_PUCHAR Xmstart;
400    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
401    eptrblock *Xeptrb;    eptrblock *Xeptrb;
402    int Xflags;    unsigned int Xrdepth;
   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    
418    BOOL Xcur_is_word;    BOOL Xcur_is_word;
419    BOOL Xcondition;    BOOL Xcondition;
   BOOL Xminimize;  
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;
426    int Xprop_category;    int Xoclength;
427    int Xprop_chartype;    pcre_uchar Xocchars[6];
   int Xprop_script;  
   int *Xprop_test_variable;  
428  #endif  #endif
429    
430      int Xcodelink;
431    int Xctype;    int Xctype;
432    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    
444    eptrblock Xnewptrb;    eptrblock Xnewptrb;
445    
446    /* Place to pass back result, and where to jump back to */    /* Where to jump back to */
447    
448    int  Xresult;    int Xwhere;
   jmp_buf Xwhere;  
449    
450  } heapframe;  } heapframe;
451    
# Line 340  typedef struct heapframe { Line 461  typedef struct heapframe {
461  *         Match from current position            *  *         Match from current position            *
462  *************************************************/  *************************************************/
463    
464  /* On entry ecode points to the first opcode, and eptr to the first character  /* This function is called recursively in many circumstances. Whenever it
 in the subject string, while eptrb holds the value of eptr at the start of the  
 last bracketed group - used for breaking infinite loops matching zero-length  
 strings. 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    #define SCHECK_PARTIAL()\
484      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 the  /* Performance note: It might be tempting to extract commonly used fields from
492  md structure (e.g. utf8, end_subject) into individual variables to improve  the md structure (e.g. utf, end_subject) into individual variables to improve
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    
496  Arguments:  Arguments:
497     eptr        pointer in subject     eptr        pointer to current character in subject
498     ecode       position in code     ecode       pointer to current position in compiled code
499       mstart      pointer to the current match start position (can be modified
500                     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_isgroup - this is the start of a bracketed group  
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,  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, 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 them with "register"  so they can be ordinary variables in all cases. Mark some of them with
521  because they are used a lot in loops. */  "register" because they are used a lot in loops. */
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 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 */
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    
544  frame->Xeptr = eptr;  frame->Xeptr = eptr;
545  frame->Xecode = ecode;  frame->Xecode = ecode;
546    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 412  HEAP_RECURSE: Line 556  HEAP_RECURSE:
556    
557  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
558  #define ecode              frame->Xecode  #define ecode              frame->Xecode
559    #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 434  HEAP_RECURSE: Line 578  HEAP_RECURSE:
578    
579  #define cur_is_word        frame->Xcur_is_word  #define cur_is_word        frame->Xcur_is_word
580  #define condition          frame->Xcondition  #define condition          frame->Xcondition
 #define minimize           frame->Xminimize  
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
587  #define prop_category      frame->Xprop_category  #define oclength           frame->Xoclength
588  #define prop_chartype      frame->Xprop_chartype  #define occhars            frame->Xocchars
 #define prop_script        frame->Xprop_script  
 #define prop_test_variable frame->Xprop_test_variable  
589  #endif  #endif
590    
591  #define ctype              frame->Xctype  #define ctype              frame->Xctype
# Line 470  HEAP_RECURSE: Line 609  HEAP_RECURSE:
609  get preserved during recursion in the normal way. In this environment, fi and  get preserved during recursion in the normal way. In this environment, fi and
610  i, and fc and c, can be the same variables. */  i, and fc and c, can be the same variables. */
611    
612  #else  #else         /* NO_RECURSE not defined */
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_UTF
625    const pcre_uchar *charptr;
626    #endif
627    const pcre_uchar *callpat;
628    const pcre_uchar *data;
629    const pcre_uchar *next;
630    PCRE_PUCHAR       pp;
631    const pcre_uchar *prev;
632    PCRE_PUCHAR       saved_eptr;
633    
634  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  recursion_info new_recursive;
635  const uschar *charptr;             /* in small blocks of the code. My normal */  
636  #endif                             /* style of coding would have declared    */  BOOL cur_is_word;
 const uschar *callpat;             /* them within each of those blocks.      */  
 const uschar *data;                /* However, in order to accommodate the   */  
 const uschar *next;                /* version of this code that uses an      */  
 USPTR         pp;                  /* external "stack" implemented on the    */  
 const uschar *prev;                /* heap, it is easier to declare them all */  
 USPTR         saved_eptr;          /* here, so the declarations can be cut   */  
                                    /* out in a block. The only declarations  */  
 recursion_info new_recursive;      /* within blocks below are for variables  */  
                                    /* that do not have to be preserved over  */  
 BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  
637  BOOL condition;  BOOL condition;
 BOOL minimize;  
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;
644  int prop_category;  int oclength;
645  int prop_chartype;  pcre_uchar occhars[6];
 int prop_script;  
 int *prop_test_variable;  
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  #endif  
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 */
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. */
# Line 524  variables. */ Line 696  variables. */
696  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
697  prop_value = 0;  prop_value = 0;
698  prop_fail_result = 0;  prop_fail_result = 0;
 prop_test_variable = NULL;  
699  #endif  #endif
700    
701    
702    /* This label is used for tail recursion, which is used in a few cases even
703    when NO_RECURSE is not defined, in order to reduce the amount of stack that is
704    used. Thanks to Ian Taylor for noticing this possibility and sending the
705    original patch. */
706    
707    TAIL_RECURSE:
708    
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_UTF
718    utf = md->utf;       /* Local copy of the flag */
719    #else
720    utf = FALSE;
721    #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
724  haven't exceeded the recursive call limit. */  haven't exceeded the recursive call limit. */
725    
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    
729  original_ims = ims;    /* Save for resetting on ')' */  /* At the start of a group with an unlimited repeat that may match an empty
730  utf8 = md->utf8;       /* Local copy of the flag */  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
731    done this way to save having to use another function argument, which would take
732  /* At the start of a bracketed group, add the current subject pointer to the  up space on the stack. See also MATCH_CONDASSERT below.
733  stack of such pointers, to be re-instated at the end of the group when we hit  
734  the closing ket. When match() is called in other circumstances, we don't add to  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
735  this stack. */  such remembered pointers, to be checked when we hit the closing ket, in order
736    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_isgroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
742    {    {
   newptrb.epb_prev = eptrb;  
743    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
744      newptrb.epb_prev = eptrb;
745    eptrb = &newptrb;    eptrb = &newptrb;
746      md->match_function_type = 0;
747    }    }
748    
749  /* Now start processing the operations. */  /* Now start processing the opcodes. */
750    
751  for (;;)  for (;;)
752    {    {
753      minimize = possessive = FALSE;
754    op = *ecode;    op = *ecode;
   minimize = FALSE;  
755    
756    /* For partial matching, remember if we ever hit the end of the subject after    switch(op)
757    matching at least one subject character. */      {
758        case OP_MARK:
759        md->nomatch_mark = ecode + 2;
760        md->mark = NULL;    /* In case previously set by assertion */
761        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
762          eptrb, RM55);
763        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
764             md->mark == NULL) md->mark = ecode + 2;
765    
766        /* 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    if (md->partial &&      case OP_COMMIT:
785        eptr >= md->end_subject &&      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
786        eptr > md->start_match)        eptrb, RM52);
787      md->hitend = TRUE;      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
788        RRETURN(MATCH_COMMIT);
789    /* Opening capturing bracket. If there is space in the offset vector, save  
790    the current subject position in the working slot at the top of the vector. We      case OP_PRUNE:
791    mustn't change the current values of the data slot, because they may be set      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
792    from a previous iteration of this group, and be referred to by a reference        eptrb, RM51);
793    inside the group.      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
794        RRETURN(MATCH_PRUNE);
795    If the bracket fails to match, we need to restore this value and also the  
796    values of the final offsets, in case they were set by a previous iteration of      case OP_PRUNE_ARG:
797    the same bracket.      md->nomatch_mark = ecode + 2;
798        md->mark = NULL;    /* In case previously set by assertion */
799    If there isn't enough space in the offset vector, treat this as if it were a      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
800    non-capturing bracket. Don't worry about setting the flag for the error case        eptrb, RM56);
801    here; that is handled in the code for KET. */      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    if (op > OP_BRA)      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      number = op - OP_BRA;      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      /* For extended extraction brackets (large number), we have to fish out the      do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
     number from a dummy opcode at the start. */  
909    
910      if (number > EXTRACT_BASIC_MAX)      offset_top = md->end_offset_top;
911        number = GET2(ecode, 2+LINK_SIZE);      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
955        a non-capturing bracket. Don't worry about setting the flag for the error
956        case here; that is handled in the code for KET. */
957    
958        case OP_CBRA:
959        case OP_SCBRA:
960        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 subject=", number);      printf("start bracket %d\n", number);
965        printf("subject=");
966      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
967      printf("\n");      printf("\n");
968  #endif  #endif
# Line 608  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        do        for (;;)
983          {          {
984          RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
985            match_isgroup);          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      /* Insufficient room for saving captured contents */      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1028        as a non-capturing bracket. */
1029    
1030      else op = OP_BRA;      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1031      }      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1032    
1033    /* Other types of node can be handled by a switch */      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1034    
1035    switch(op)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1036      {      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1037      case OP_BRA:     /* Non-capturing bracket: optimized */  
1038      DPRINTF(("start bracket 0\n"));      /* Non-capturing or atomic group, except for possessive with unlimited
1039      do      repeat and ONCE group with no captures. Loop for all the alternatives.
1040    
1041        When we get to the final alternative within the brackets, we used to return
1042        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:
1060        case OP_SBRA:
1061        DPRINTF(("start non-capturing bracket\n"));
1062    
1063        for (;;)
1064        {        {
1065        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,        if (op >= OP_SBRA || op == OP_ONCE)
1066          match_isgroup);          md->match_function_type = MATCH_CBEGROUP;
1067        if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
1068          /* 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          else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1073            {
1074            ecode += PRIV(OP_lengths)[*ecode];
1075            goto TAIL_RECURSE;
1076            }
1077    
1078          /* In all other cases, we have to make another call to match(). */
1079    
1080          save_mark = md->mark;
1081          save_capture_last = md->capture_last;
1082          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          if (*ecode != OP_ALT) break;
1113          md->capture_last = save_capture_last;
1114        }        }
1115      while (*ecode == OP_ALT);  
     DPRINTF(("bracket 0 failed\n"));  
1116      RRETURN(MATCH_NOMATCH);      RRETURN(MATCH_NOMATCH);
1117    
1118      /* Conditional group: compilation checked that there are no more than      /* Handle possessive capturing brackets with an unlimited repeat. We come
1119      two branches. If the condition is false, skipping the first branch takes us      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1120      past the end if there is only one branch, but that's OK because that is      handled similarly to the normal case above. However, the matching is
1121      exactly what going to the ket would do. */      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      case OP_COND:      POSSESSIVE_CAPTURE:
1131      if (ecode[LINK_SIZE+1] == OP_CREF) /* Condition extract or recurse test */      number = GET2(ecode, 1+LINK_SIZE);
1132        {      offset = number << 1;
       offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */  
       condition = (offset == CREF_RECURSE * 2)?  
         (md->recursive != NULL) :  
         (offset < offset_top && md->offset_vector[offset] >= 0);  
       RMATCH(rrc, eptr, ecode + (condition?  
         (LINK_SIZE + 4) : (LINK_SIZE + 1 + GET(ecode, 1))),  
         offset_top, md, ims, eptrb, match_isgroup);  
       RRETURN(rrc);  
       }  
1133    
1134      /* The condition is an assertion. Call match() to evaluate it - setting  #ifdef PCRE_DEBUG
1135      the final argument TRUE causes it to stop at the end of an assertion. */      printf("start possessive bracket %d\n", number);
1136        printf("subject=");
1137        pchars(eptr, 16, TRUE, md);
1138        printf("\n");
1139    #endif
1140    
1141      else      if (offset < md->offset_max)
1142        {        {
1143        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        matched_once = FALSE;
1144            match_condassert | match_isgroup);        code_offset = (int)(ecode - md->start_code);
1145        if (rrc == MATCH_MATCH)  
1146          save_offset1 = md->offset_vector[offset];
1147          save_offset2 = md->offset_vector[offset+1];
1148          save_offset3 = md->offset_vector[md->offset_end - number];
1149          save_capture_last = md->capture_last;
1150    
1151          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1152    
1153          /* 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          restart it with a new subject starting position, remembering that we had
1156          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1157          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          continue from afterwards. Otherwise it has failed; restore the previous
1160          capture values before returning NOMATCH. */
1161    
1162          for (;;)
1163            {
1164            md->offset_vector[md->offset_end - number] =
1165              (int)(eptr - md->start_subject);
1166            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1167            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1168              eptrb, RM63);
1169            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);
1193            if (*ecode != OP_ALT) break;
1194            }
1195    
1196          if (!matched_once)
1197          {          {
1198          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE+2);          md->offset_vector[offset] = save_offset1;
1199          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          md->offset_vector[offset+1] = save_offset2;
1200            md->offset_vector[md->offset_end - number] = save_offset3;
1201          }          }
1202        else if (rrc != MATCH_NOMATCH)  
1203          if (allow_zero || matched_once)
1204          {          {
1205          RRETURN(rrc);         /* Need braces because of following else */          ecode += 1 + LINK_SIZE;
1206            break;
1207          }          }
1208        else ecode += GET(ecode, 1);  
1209        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,        RRETURN(MATCH_NOMATCH);
         match_isgroup);  
       RRETURN(rrc);  
1210        }        }
     /* Control never reaches here */  
1211    
1212      /* Skip over conditional reference or large extraction number data if      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1213      encountered. */      as a non-capturing bracket. */
1214    
1215      case OP_CREF:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1216      case OP_BRANUMBER:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     ecode += 3;  
     break;  
1217    
1218      /* End of the pattern. If we are in a recursion, we should restore the      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
     offsets appropriately and continue from after the call. */  
1219    
1220      case OP_END:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1221      if (md->recursive != NULL && md->recursive->group_num == 0)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
       {  
       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));  
       md->start_match = rec->save_start;  
       ims = original_ims;  
       ecode = rec->after_call;  
       break;  
       }  
1222    
1223      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      /* Non-capturing possessive bracket with unlimited repeat. We come here
1224      string - backtracking will then try other alternatives, if any. */      from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1225        without the capturing complication. It is written out separately for speed
1226        and cleanliness. */
1227    
1228      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      case OP_BRAPOS:
1229      md->end_match_ptr = eptr;          /* Record where we ended */      case OP_SBRAPOS:
1230      md->end_offset_top = offset_top;   /* and how many extracts were taken */      allow_zero = FALSE;
     RRETURN(MATCH_MATCH);  
   
     /* Change option settings */  
   
     case OP_OPT:  
     ims = ecode[1];  
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1231    
1232      /* Assertion brackets. Check the alternative branches in turn - the      POSSESSIVE_NON_CAPTURE:
1233      matching won't pass the KET for an assertion. If any one branch matches,      matched_once = FALSE;
1234      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      code_offset = (int)(ecode - md->start_code);
1235      start of each branch to move the current point backwards, so the code at      save_capture_last = md->capture_last;
     this level is identical to the lookahead case. */  
1236    
1237      case OP_ASSERT:      for (;;)
     case OP_ASSERTBACK:  
     do  
1238        {        {
1239        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1240          match_isgroup);        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 two
1278      mark, since extracts may have been taken during the assertion. */      branches. If the condition is false, skipping the first branch takes us
1279        past the end of the item if there is only one branch, but that's exactly
1280        what we want. */
1281    
1282      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_COND:
1283      ecode += 1 + LINK_SIZE;      case OP_SCOND:
     offset_top = md->end_offset_top;  
     continue;  
1284    
1285      /* Negative assertion: all branches must fail to match */      /* The variable codelink will be added to ecode when the condition is
1286        false, to get to the second branch. Setting it to the offset to the ALT
1287        or KET, then incrementing ecode achieves this effect. We now have ecode
1288        pointing to the condition or callout. */
1289    
1290        codelink = GET(ecode, 1);   /* Offset to the second branch */
1291        ecode += 1 + LINK_SIZE;     /* From this opcode */
1292    
1293        /* Because of the way auto-callout works during compile, a callout item is
1294        inserted between OP_COND and an assertion condition. */
1295    
1296        if (*ecode == OP_CALLOUT)
1297          {
1298          if (PUBL(callout) != NULL)
1299            {
1300            PUBL(callout_block) cb;
1301            cb.version          = 2;   /* Version 1 of the callout block */
1302            cb.callout_number   = ecode[1];
1303            cb.offset_vector    = md->offset_vector;
1304    #if defined COMPILE_PCRE8
1305            cb.subject          = (PCRE_SPTR)md->start_subject;
1306    #elif defined COMPILE_PCRE16
1307            cb.subject          = (PCRE_SPTR16)md->start_subject;
1308    #elif defined COMPILE_PCRE32
1309            cb.subject          = (PCRE_SPTR32)md->start_subject;
1310    #endif
1311            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1312            cb.start_match      = (int)(mstart - md->start_subject);
1313            cb.current_position = (int)(eptr - md->start_subject);
1314            cb.pattern_position = GET(ecode, 2);
1315            cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1316            cb.capture_top      = offset_top/2;
1317            cb.capture_last     = md->capture_last & CAPLMASK;
1318            /* Internal change requires this for API compatibility. */
1319            if (cb.capture_last == 0) cb.capture_last = -1;
1320            cb.callout_data     = md->callout_data;
1321            cb.mark             = md->nomatch_mark;
1322            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1323            if (rrc < 0) RRETURN(rrc);
1324            }
1325    
1326          /* Advance ecode past the callout, so it now points to the condition. We
1327          must adjust codelink so that the value of ecode+codelink is unchanged. */
1328    
1329          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1330          codelink -= PRIV(OP_lengths)[OP_CALLOUT];
1331          }
1332    
1333        /* Test the various possible conditions */
1334    
1335      case OP_ASSERT_NOT:      condition = FALSE;
1336      case OP_ASSERTBACK_NOT:      switch(condcode = *ecode)
     do  
1337        {        {
1338        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        case OP_RREF:         /* Numbered group recursion test */
1339          match_isgroup);        if (md->recursive != NULL)     /* Not recursing => FALSE */
1340        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);          {
1341        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          unsigned int recno = GET2(ecode, 1);   /* Recursion group number*/
1342        ecode += GET(ecode,1);          condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1343        }          }
1344      while (*ecode == OP_ALT);        break;
   
     if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);  
1345    
1346      ecode += 1 + LINK_SIZE;        case OP_DNRREF:       /* Duplicate named group recursion test */
1347      continue;        if (md->recursive != NULL)
1348            {
1349            int count = GET2(ecode, 1 + IMM2_SIZE);
1350            pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
1351            while (count-- > 0)
1352              {
1353              unsigned int recno = GET2(slot, 0);
1354              condition = recno == md->recursive->group_num;
1355              if (condition) break;
1356              slot += md->name_entry_size;
1357              }
1358            }
1359          break;
1360    
1361      /* Move the subject pointer back. This occurs only at the start of        case OP_CREF:         /* Numbered group used test */
1362      each branch of a lookbehind assertion. If we are too close to the start to        offset = GET2(ecode, 1) << 1;  /* Doubled ref number */
1363      move back, this match function fails. When working with UTF-8 we move        condition = offset < offset_top && md->offset_vector[offset] >= 0;
1364      back a number of characters, not bytes. */        break;
1365    
1366      case OP_REVERSE:        case OP_DNCREF:      /* Duplicate named group used test */
 #ifdef SUPPORT_UTF8  
     if (utf8)  
       {  
       c = GET(ecode,1);  
       for (i = 0; i < c; i++)  
1367          {          {
1368          eptr--;          int count = GET2(ecode, 1 + IMM2_SIZE);
1369          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
1370          BACKCHAR(eptr)          while (count-- > 0)
1371              {
1372              offset = GET2(slot, 0) << 1;
1373              condition = offset < offset_top && md->offset_vector[offset] >= 0;
1374              if (condition) break;
1375              slot += md->name_entry_size;
1376              }
1377          }          }
1378        }        break;
     else  
 #endif  
1379    
1380      /* No UTF-8 support, or not in UTF-8 mode: count is byte count */        case OP_DEF:     /* DEFINE - always false */
1381          break;
1382    
1383        {        /* The condition is an assertion. Call match() to evaluate it - setting
1384        eptr -= GET(ecode,1);        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end
1385        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        of an assertion. */
1386        }  
1387          default:
1388          md->match_function_type = MATCH_CONDASSERT;
1389          RMATCH(eptr, ecode, offset_top, md, NULL, RM3);
1390          if (rrc == MATCH_MATCH)
1391            {
1392            if (md->end_offset_top > offset_top)
1393              offset_top = md->end_offset_top;  /* Captures may have happened */
1394            condition = TRUE;
1395    
1396      /* Skip to next op code */          /* Advance ecode past the assertion to the start of the first branch,
1397            but adjust it so that the general choosing code below works. */
1398    
1399      ecode += 1 + LINK_SIZE;          ecode += GET(ecode, 1);
1400      break;          while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1401            ecode += 1 + LINK_SIZE - PRIV(OP_lengths)[condcode];
1402            }
1403    
1404          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1405          assertion; it is therefore treated as NOMATCH. Any other return is an
1406          error. */
1407    
1408          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1409            {
1410            RRETURN(rrc);         /* Need braces because of following else */
1411            }
1412          break;
1413          }
1414    
1415        /* Choose branch according to the condition */
1416    
1417        ecode += condition? PRIV(OP_lengths)[condcode] : codelink;
1418    
1419        /* We are now at the branch that is to be obeyed. As there is only one, we
1420        can use tail recursion to avoid using another stack frame, except when
1421        there is unlimited repeat of a possibly empty group. In the latter case, a
1422        recursive call to match() is always required, unless the second alternative
1423        doesn't exist, in which case we can just plough on. Note that, for
1424        compatibility with Perl, the | in a conditional group is NOT treated as
1425        creating two alternatives. If a THEN is encountered in the branch, it
1426        propagates out to the enclosing alternative (unless nested in a deeper set
1427        of alternatives, of course). */
1428    
1429        if (condition || ecode[-(1+LINK_SIZE)] == OP_ALT)
1430          {
1431          if (op != OP_SCOND)
1432            {
1433            goto TAIL_RECURSE;
1434            }
1435    
1436          md->match_function_type = MATCH_CBEGROUP;
1437          RMATCH(eptr, ecode, offset_top, md, eptrb, RM49);
1438          RRETURN(rrc);
1439          }
1440    
1441         /* Condition false & no alternative; continue after the group. */
1442    
1443        else
1444          {
1445          }
1446        break;
1447    
1448    
1449        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1450        to close any currently open capturing brackets. */
1451    
1452        case OP_CLOSE:
1453        number = GET2(ecode, 1);   /* Must be less than 65536 */
1454        offset = number << 1;
1455    
1456    #ifdef PCRE_DEBUG
1457          printf("end bracket %d at *ACCEPT", number);
1458          printf("\n");
1459    #endif
1460    
1461        md->capture_last = (md->capture_last & OVFLMASK) | number;
1462        if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1463          {
1464          md->offset_vector[offset] =
1465            md->offset_vector[md->offset_end - number];
1466          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1467          if (offset_top <= offset) offset_top = offset + 2;
1468          }
1469        ecode += 1 + IMM2_SIZE;
1470        break;
1471    
1472    
1473        /* End of the pattern, either real or forced. */
1474    
1475        case OP_END:
1476        case OP_ACCEPT:
1477        case OP_ASSERT_ACCEPT:
1478    
1479        /* If we have matched an empty string, fail if not in an assertion and not
1480        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1481        is set and we have matched at the start of the subject. In both cases,
1482        backtracking will then try other alternatives, if any. */
1483    
1484        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1485             md->recursive == NULL &&
1486             (md->notempty ||
1487               (md->notempty_atstart &&
1488                 mstart == md->start_subject + md->start_offset)))
1489          RRETURN(MATCH_NOMATCH);
1490    
1491        /* Otherwise, we have a match. */
1492    
1493        md->end_match_ptr = eptr;           /* Record where we ended */
1494        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1495        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1496    
1497        /* For some reason, the macros don't work properly if an expression is
1498        given as the argument to RRETURN when the heap is in use. */
1499    
1500        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1501        RRETURN(rrc);
1502    
1503        /* Assertion brackets. Check the alternative branches in turn - the
1504        matching won't pass the KET for an assertion. If any one branch matches,
1505        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1506        start of each branch to move the current point backwards, so the code at
1507        this level is identical to the lookahead case. When the assertion is part
1508        of a condition, we want to return immediately afterwards. The caller of
1509        this incarnation of the match() function will have set MATCH_CONDASSERT in
1510        md->match_function type, and one of these opcodes will be the first opcode
1511        that is processed. We use a local variable that is preserved over calls to
1512        match() to remember this case. */
1513    
1514        case OP_ASSERT:
1515        case OP_ASSERTBACK:
1516        save_mark = md->mark;
1517        if (md->match_function_type == MATCH_CONDASSERT)
1518          {
1519          condassert = TRUE;
1520          md->match_function_type = 0;
1521          }
1522        else condassert = FALSE;
1523    
1524        /* Loop for each branch */
1525    
1526        do
1527          {
1528          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1529    
1530          /* A match means that the assertion is true; break out of the loop
1531          that matches its alternatives. */
1532    
1533          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1534            {
1535            mstart = md->start_match_ptr;   /* In case \K reset it */
1536            break;
1537            }
1538    
1539          /* If not matched, restore the previous mark setting. */
1540    
1541          md->mark = save_mark;
1542    
1543          /* See comment in the code for capturing groups above about handling
1544          THEN. */
1545    
1546          if (rrc == MATCH_THEN)
1547            {
1548            next = ecode + GET(ecode,1);
1549            if (md->start_match_ptr < next &&
1550                (*ecode == OP_ALT || *next == OP_ALT))
1551              rrc = MATCH_NOMATCH;
1552            }
1553    
1554          /* Anything other than NOMATCH causes the entire assertion to fail,
1555          passing back the return code. This includes COMMIT, SKIP, PRUNE and an
1556          uncaptured THEN, which means they take their normal effect. This
1557          consistent approach does not always have exactly the same effect as in
1558          Perl. */
1559    
1560          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1561          ecode += GET(ecode, 1);
1562          }
1563        while (*ecode == OP_ALT);   /* Continue for next alternative */
1564    
1565        /* If we have tried all the alternative branches, the assertion has
1566        failed. If not, we broke out after a match. */
1567    
1568        if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1569    
1570        /* If checking an assertion for a condition, return MATCH_MATCH. */
1571    
1572        if (condassert) RRETURN(MATCH_MATCH);
1573    
1574        /* Continue from after a successful assertion, updating the offsets high
1575        water mark, since extracts may have been taken during the assertion. */
1576    
1577        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1578        ecode += 1 + LINK_SIZE;
1579        offset_top = md->end_offset_top;
1580        continue;
1581    
1582        /* Negative assertion: all branches must fail to match for the assertion to
1583        succeed. */
1584    
1585        case OP_ASSERT_NOT:
1586        case OP_ASSERTBACK_NOT:
1587        save_mark = md->mark;
1588        if (md->match_function_type == MATCH_CONDASSERT)
1589          {
1590          condassert = TRUE;
1591          md->match_function_type = 0;
1592          }
1593        else condassert = FALSE;
1594    
1595        /* Loop for each alternative branch. */
1596    
1597        do
1598          {
1599          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1600          md->mark = save_mark;   /* Always restore the mark setting */
1601    
1602          switch(rrc)
1603            {
1604            case MATCH_MATCH:            /* A successful match means */
1605            case MATCH_ACCEPT:           /* the assertion has failed. */
1606            RRETURN(MATCH_NOMATCH);
1607    
1608            case MATCH_NOMATCH:          /* Carry on with next branch */
1609            break;
1610    
1611            /* See comment in the code for capturing groups above about handling
1612            THEN. */
1613    
1614            case MATCH_THEN:
1615            next = ecode + GET(ecode,1);
1616            if (md->start_match_ptr < next &&
1617                (*ecode == OP_ALT || *next == OP_ALT))
1618              {
1619              rrc = MATCH_NOMATCH;
1620              break;
1621              }
1622            /* Otherwise fall through. */
1623    
1624            /* COMMIT, SKIP, PRUNE, and an uncaptured THEN cause the whole
1625            assertion to fail to match, without considering any more alternatives.
1626            Failing to match means the assertion is true. This is a consistent
1627            approach, but does not always have the same effect as in Perl. */
1628    
1629            case MATCH_COMMIT:
1630            case MATCH_SKIP:
1631            case MATCH_SKIP_ARG:
1632            case MATCH_PRUNE:
1633            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1634            goto NEG_ASSERT_TRUE;   /* Break out of alternation loop */
1635    
1636            /* Anything else is an error */
1637    
1638            default:
1639            RRETURN(rrc);
1640            }
1641    
1642          /* Continue with next branch */
1643    
1644          ecode += GET(ecode,1);
1645          }
1646        while (*ecode == OP_ALT);
1647    
1648        /* All branches in the assertion failed to match. */
1649    
1650        NEG_ASSERT_TRUE:
1651        if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1652        ecode += 1 + LINK_SIZE;                /* Continue with current branch */
1653        continue;
1654    
1655        /* Move the subject pointer back. This occurs only at the start of
1656        each branch of a lookbehind assertion. If we are too close to the start to
1657        move back, this match function fails. When working with UTF-8 we move
1658        back a number of characters, not bytes. */
1659    
1660        case OP_REVERSE:
1661    #ifdef SUPPORT_UTF
1662        if (utf)
1663          {
1664          i = GET(ecode, 1);
1665          while (i-- > 0)
1666            {
1667            eptr--;
1668            if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1669            BACKCHAR(eptr);
1670            }
1671          }
1672        else
1673    #endif
1674    
1675        /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1676    
1677          {
1678          eptr -= GET(ecode, 1);
1679          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1680          }
1681    
1682        /* Save the earliest consulted character, then skip to next op code */
1683    
1684        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1685        ecode += 1 + LINK_SIZE;
1686        break;
1687    
1688      /* The callout item calls an external function, if one is provided, passing      /* The callout item calls an external function, if one is provided, passing
1689      details of the match so far. This is mainly for debugging, though the      details of the match so far. This is mainly for debugging, though the
1690      function is able to force a failure. */      function is able to force a failure. */
1691    
1692      case OP_CALLOUT:      case OP_CALLOUT:
1693      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1694        {        {
1695        pcre_callout_block cb;        PUBL(callout_block) cb;
1696        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1697        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1698        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1699    #if defined COMPILE_PCRE8
1700        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1701        cb.subject_length   = md->end_subject - md->start_subject;  #elif defined COMPILE_PCRE16
1702        cb.start_match      = md->start_match - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1703        cb.current_position = eptr - md->start_subject;  #elif defined COMPILE_PCRE32
1704          cb.subject          = (PCRE_SPTR32)md->start_subject;
1705    #endif
1706          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1707          cb.start_match      = (int)(mstart - md->start_subject);
1708          cb.current_position = (int)(eptr - md->start_subject);
1709        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1710        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1711        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1712        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last & CAPLMASK;
1713          /* Internal change requires this for API compatibility. */
1714          if (cb.capture_last == 0) cb.capture_last = -1;
1715        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1716        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1717          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1718        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1719        }        }
1720      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 847  for (;;) Line 1724  for (;;)
1724      offset data is the offset to the starting bracket from the start of the      offset data is the offset to the starting bracket from the start of the
1725      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1726    
1727      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1728      save their starting points and reinstate them after the recursion. However,      re-instated afterwards. We don't know how many are started and not yet
1729      we don't know how many such there are (offset_top records the completed      finished (offset_top records the completed total) so we just have to save
1730      total) so we just have to save all the potential data. There may be up to      all the potential data. There may be up to 65535 such values, which is too
1731      65535 such values, which is too large to put on the stack, but using malloc      large to put on the stack, but using malloc for small numbers seems
1732      for small numbers seems expensive. As a compromise, the stack is used when      expensive. As a compromise, the stack is used when there are no more than
1733      there are no more than REC_STACK_SAVE_MAX values to store; otherwise malloc      REC_STACK_SAVE_MAX values to store; otherwise malloc is used.
     is used. A problem is what to do if the malloc fails ... there is no way of  
     returning to the top level with an error. Save the top REC_STACK_SAVE_MAX  
     values on the stack, and accept that the rest may be wrong.  
1734    
1735      There are also other values that have to be saved. We use a chained      There are also other values that have to be saved. We use a chained
1736      sequence of blocks that actually live on the stack. Thanks to Robin Houston      sequence of blocks that actually live on the stack. Thanks to Robin Houston
1737      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1738        a lot, so he is not to blame for the current way it works. */
1739    
1740      case OP_RECURSE:      case OP_RECURSE:
1741        {        {
1742        callpat = md->start_code + GET(ecode, 1);        recursion_info *ri;
1743        new_recursive.group_num = *callpat - OP_BRA;        unsigned int recno;
1744    
1745        /* For extended extraction brackets (large number), we have to fish out        callpat = md->start_code + GET(ecode, 1);
1746        the number from a dummy opcode at the start. */        recno = (callpat == md->start_code)? 0 :
1747            GET2(callpat, 1 + LINK_SIZE);
1748    
1749        if (new_recursive.group_num > EXTRACT_BASIC_MAX)        /* Check for repeating a recursion without advancing the subject pointer.
1750          new_recursive.group_num = GET2(callpat, 2+LINK_SIZE);        This should catch convoluted mutual recursions. (Some simple cases are
1751          caught at compile time.) */
1752    
1753          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1754            if (recno == ri->group_num && eptr == ri->subject_position)
1755              RRETURN(PCRE_ERROR_RECURSELOOP);
1756    
1757        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1758    
1759          new_recursive.group_num = recno;
1760          new_recursive.saved_capture_last = md->capture_last;
1761          new_recursive.subject_position = eptr;
1762        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1763        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1764    
1765        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1766    
1767        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1768    
1769        /* Now save the offset data. */        /* Now save the offset data */
1770    
1771        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1772        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 891  for (;;) Line 1774  for (;;)
1774        else        else
1775          {          {
1776          new_recursive.offset_save =          new_recursive.offset_save =
1777            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1778          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1779          }          }
   
1780        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1781              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = md->start_match;  
       md->start_match = eptr;  
1782    
1783        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1784        restore the offset and recursion data. */        restore the offset data and the last captured value. If there were nested
1785          recursions, md->recursive might be changed, so reset it before looping.
1786          */
1787    
1788        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1789          cbegroup = (*callpat >= OP_SBRA);
1790        do        do
1791          {          {
1792          RMATCH(rrc, eptr, callpat + 1 + LINK_SIZE, offset_top, md, ims,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1793              eptrb, match_isgroup);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1794          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1795            memcpy(md->offset_vector, new_recursive.offset_save,
1796                new_recursive.saved_max * sizeof(int));
1797            md->capture_last = new_recursive.saved_capture_last;
1798            md->recursive = new_recursive.prevrec;
1799            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1800            {            {
1801            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1802            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1803              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1804            RRETURN(MATCH_MATCH);  
1805              /* Set where we got to in the subject, and reset the start in case
1806              it was changed by \K. This *is* propagated back out of a recursion,
1807              for Perl compatibility. */
1808    
1809              eptr = md->end_match_ptr;
1810              mstart = md->start_match_ptr;
1811              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1812            }            }
1813          else if (rrc != MATCH_NOMATCH)  
1814            /* PCRE does not allow THEN, SKIP, PRUNE or COMMIT to escape beyond a
1815            recursion; they cause a NOMATCH for the entire recursion. These codes
1816            are defined in a range that can be tested for. */
1817    
1818            if (rrc >= MATCH_BACKTRACK_MIN && rrc <= MATCH_BACKTRACK_MAX)
1819              RRETURN(MATCH_NOMATCH);
1820    
1821            /* Any return code other than NOMATCH is an error. */
1822    
1823            if (rrc != MATCH_NOMATCH)
1824            {            {
1825            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1826              if (new_recursive.offset_save != stacksave)
1827                (PUBL(free))(new_recursive.offset_save);
1828            RRETURN(rrc);            RRETURN(rrc);
1829            }            }
1830    
1831          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1832          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1833          }          }
1834        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 932  for (;;) Line 1836  for (;;)
1836        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1837        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1838        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1839          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1840        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1841        }        }
     /* 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(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,  
           eptrb, match_isgroup);  
         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. */  
1842    
1843        do ecode += GET(ecode,1); while (*ecode == OP_ALT);      RECURSION_MATCHED:
1844        break;
       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. We need to reset any options  
       that changed within the bracket before re-running it, so check the next  
       opcode. */  
   
       if (ecode[1+LINK_SIZE] == OP_OPT)  
         {  
         ims = (ims & ~PCRE_IMS) | ecode[4];  
         DPRINTF(("ims set to %02lx at group repeat\n", ims));  
         }  
   
       if (*ecode == OP_KETRMIN)  
         {  
         RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0);  
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);  
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         }  
       else  /* OP_KETRMAX */  
         {  
         RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);  
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);  
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         }  
       }  
     RRETURN(MATCH_NOMATCH);  
1845    
1846      /* 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
1847      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1018  for (;;) Line 1850  for (;;)
1850      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1851      break;      break;
1852    
1853      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1854      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
1855      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
1856      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
1857      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1858    
1859      case OP_BRAZERO:      case OP_BRAZERO:
1860        {      next = ecode + 1;
1861        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1862        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, match_isgroup);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1863        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1864        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1+LINK_SIZE;  
       }  
1865      break;      break;
1866    
1867      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1868        {      next = ecode + 1;
1869        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1870        do next += GET(next,1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1871        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb,      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1872          match_isgroup);      ecode++;
1873        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      break;
1874        ecode++;  
1875        }      case OP_SKIPZERO:
1876        next = ecode+1;
1877        do next += GET(next,1); while (*next == OP_ALT);
1878        ecode = next + 1 + LINK_SIZE;
1879      break;      break;
1880    
1881      /* End of a group, repeated or non-repeating. If we are at the end of      /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1882      an assertion "group", stop matching and return MATCH_MATCH, but record the      here; just jump to the group, with allow_zero set TRUE. */
1883      current high water mark for use by positive assertions. Do this also  
1884      for the "once" (not-backup up) groups. */      case OP_BRAPOSZERO:
1885        op = *(++ecode);
1886        allow_zero = TRUE;
1887        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1888          goto POSSESSIVE_NON_CAPTURE;
1889    
1890        /* End of a group, repeated or non-repeating. */
1891    
1892      case OP_KET:      case OP_KET:
1893      case OP_KETRMIN:      case OP_KETRMIN:
1894      case OP_KETRMAX:      case OP_KETRMAX:
1895        {      case OP_KETRPOS:
1896        prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
       saved_eptr = eptrb->epb_saved_eptr;  
1897    
1898        /* Back up the stack of bracket start pointers. */      /* If this was a group that remembered the subject start, in order to break
1899        infinite repeats of empty string matches, retrieve the subject start from
1900        the chain. Otherwise, set it NULL. */
1901    
1902        eptrb = eptrb->epb_prev;      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1903          {
1904        if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1905            *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1906            *prev == OP_ONCE)        }
1907          {      else saved_eptr = NULL;
         md->end_match_ptr = eptr;      /* For ONCE */  
         md->end_offset_top = offset_top;  
         RRETURN(MATCH_MATCH);  
         }  
1908    
1909        /* In all other cases except a conditional group we have to check the      /* If we are at the end of an assertion group or a non-capturing atomic
1910        group number back at the start and if necessary complete handling an      group, stop matching and return MATCH_MATCH, but record the current high
1911        extraction by setting the offsets and bumping the high water mark. */      water mark for use by positive assertions. We also need to record the match
1912        start in case it was changed by \K. */
1913    
1914        if (*prev != OP_COND)      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1915          {           *prev == OP_ONCE_NC)
1916          number = *prev - OP_BRA;        {
1917          md->end_match_ptr = eptr;      /* For ONCE_NC */
1918          md->end_offset_top = offset_top;
1919          md->start_match_ptr = mstart;
1920          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1921          }
1922    
1923          /* For extended extraction brackets (large number), we have to fish out      /* For capturing groups we have to check the group number back at the start
1924          the number from a dummy opcode at the start. */      and if necessary complete handling an extraction by setting the offsets and
1925        bumping the high water mark. Whole-pattern recursion is coded as a recurse
1926        into group 0, so it won't be picked up here. Instead, we catch it when the
1927        OP_END is reached. Other recursion is handled here. We just have to record
1928        the current subject position and start match pointer and give a MATCH
1929        return. */
1930    
1931          if (number > EXTRACT_BASIC_MAX) number = GET2(prev, 2+LINK_SIZE);      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1932          offset = number << 1;          *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1933          {
1934          number = GET2(prev, 1+LINK_SIZE);
1935          offset = number << 1;
1936    
1937  #ifdef DEBUG  #ifdef PCRE_DEBUG
1938          printf("end bracket %d", number);        printf("end bracket %d", number);
1939          printf("\n");        printf("\n");
1940  #endif  #endif
1941    
1942          /* Test for a numbered group. This includes groups called as a result        /* Handle a recursively called group. */
         of recursion. Note that whole-pattern recursion is coded as a recurse  
         into group 0, so it won't be picked up here. Instead, we catch it when  
         the OP_END is reached. */  
1943    
1944          if (number > 0)        if (md->recursive != NULL && md->recursive->group_num == number)
1945            {          {
1946            md->capture_last = number;          md->end_match_ptr = eptr;
1947            if (offset >= md->offset_max) md->offset_overflow = TRUE; else          md->start_match_ptr = mstart;
1948              {          RRETURN(MATCH_MATCH);
1949              md->offset_vector[offset] =          }
               md->offset_vector[md->offset_end - number];  
             md->offset_vector[offset+1] = eptr - md->start_subject;  
             if (offset_top <= offset) offset_top = offset + 2;  
             }  
1950    
1951            /* Handle a recursively called group. Restore the offsets        /* Deal with capturing */
           appropriately and continue from after the call. */  
1952    
1953            if (md->recursive != NULL && md->recursive->group_num == number)        md->capture_last = (md->capture_last & OVFLMASK) | number;
1954              {        if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1955              recursion_info *rec = md->recursive;          {
1956              DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          /* If offset is greater than offset_top, it means that we are
1957              md->recursive = rec->prevrec;          "skipping" a capturing group, and that group's offsets must be marked
1958              md->start_match = rec->save_start;          unset. In earlier versions of PCRE, all the offsets were unset at the
1959              memcpy(md->offset_vector, rec->offset_save,          start of matching, but this doesn't work because atomic groups and
1960                rec->saved_max * sizeof(int));          assertions can cause a value to be set that should later be unset.
1961              ecode = rec->after_call;          Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1962              ims = original_ims;          part of the atomic group, but this is not on the final matching path,
1963              break;          so must be unset when 2 is set. (If there is no group 2, there is no
1964              }          problem, because offset_top will then be 2, indicating no capture.) */
1965    
1966            if (offset > offset_top)
1967              {
1968              register int *iptr = md->offset_vector + offset_top;
1969              register int *iend = md->offset_vector + offset;
1970              while (iptr < iend) *iptr++ = -1;
1971            }            }
         }  
1972    
1973        /* Reset the value of the ims flags, in case they got changed during          /* Now make the extraction */
       the group. */  
1974    
1975        ims = original_ims;          md->offset_vector[offset] =
1976        DPRINTF(("ims reset to %02lx\n", ims));            md->offset_vector[md->offset_end - number];
1977            md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1978            if (offset_top <= offset) offset_top = offset + 2;
1979            }
1980          }
1981    
1982        /* For a non-repeating ket, just continue at this level. This also      /* For an ordinary non-repeating ket, just continue at this level. This
1983        happens for a repeating ket if no characters were matched in the group.      also happens for a repeating ket if no characters were matched in the
1984        This is the forcible breaking of infinite loops as implemented in Perl      group. This is the forcible breaking of infinite loops as implemented in
1985        5.005. If there is an options reset, it will get obeyed in the normal      Perl 5.005. For a non-repeating atomic group that includes captures,
1986        course of events. */      establish a backup point by processing the rest of the pattern at a lower
1987        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1988        original OP_ONCE level, thereby bypassing intermediate backup points, but
1989        resetting any captures that happened along the way. */
1990    
1991        if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1992          {
1993          if (*prev == OP_ONCE)
1994          {          {
1995          ecode += 1 + LINK_SIZE;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1996          break;          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1997            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1998            RRETURN(MATCH_ONCE);
1999          }          }
2000          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
2001          break;
2002          }
2003    
2004        /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
2005        and return the MATCH_KETRPOS. This makes it possible to do the repeats one
2006        at a time from the outer level, thus saving stack. */
2007    
2008        if (*ecode == OP_KETRPOS)
2009          {
2010          md->end_match_ptr = eptr;
2011          md->end_offset_top = offset_top;
2012          RRETURN(MATCH_KETRPOS);
2013          }
2014    
2015        /* The repeating kets try the rest of the pattern or restart from the      /* The normal repeating kets try the rest of the pattern or restart from
2016        preceding bracket, in the appropriate order. */      the preceding bracket, in the appropriate order. In the second case, we can
2017        use tail recursion to avoid using another stack frame, unless we have an
2018        an atomic group or an unlimited repeat of a group that can match an empty
2019        string. */
2020    
2021        if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
2022          {
2023          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2024          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2025          if (*prev == OP_ONCE)
2026          {          {
2027          RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);          RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
         if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
         RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);  
2028          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2029            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2030            RRETURN(MATCH_ONCE);
2031          }          }
2032        else  /* OP_KETRMAX */        if (*prev >= OP_SBRA)    /* Could match an empty string */
2033          {          {
2034          RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2035          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          RRETURN(rrc);
2036          RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);          }
2037          ecode = prev;
2038          goto TAIL_RECURSE;
2039          }
2040        else  /* OP_KETRMAX */
2041          {
2042          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2043          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2044          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2045          if (*prev == OP_ONCE)
2046            {
2047            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2048          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2049            md->once_target = prev;
2050            RRETURN(MATCH_ONCE);
2051          }          }
2052          ecode += 1 + LINK_SIZE;
2053          goto TAIL_RECURSE;
2054        }        }
2055        /* Control never gets here */
2056    
2057      RRETURN(MATCH_NOMATCH);      /* Not multiline mode: start of subject assertion, unless notbol. */
   
     /* Start of subject unless notbol, or after internal newline if multiline */  
2058    
2059      case OP_CIRC:      case OP_CIRC:
2060      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[-1] != NEWLINE)  
         RRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
2061    
2062      /* Start of subject assertion */      /* Start of subject assertion */
2063    
# Line 1182  for (;;) Line 2066  for (;;)
2066      ecode++;      ecode++;
2067      break;      break;
2068    
2069        /* Multiline mode: start of subject unless notbol, or after any newline. */
2070    
2071        case OP_CIRCM:
2072        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2073        if (eptr != md->start_subject &&
2074            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2075          RRETURN(MATCH_NOMATCH);
2076        ecode++;
2077        break;
2078    
2079      /* Start of match assertion */      /* Start of match assertion */
2080    
2081      case OP_SOM:      case OP_SOM:
# Line 1189  for (;;) Line 2083  for (;;)
2083      ecode++;      ecode++;
2084      break;      break;
2085    
2086      /* Assert before internal newline if multiline, or before a terminating      /* Reset the start of match point */
     newline unless endonly is set, else end of subject unless noteol is set. */  
2087    
2088      case OP_DOLL:      case OP_SET_SOM:
2089      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
2090        ecode++;
2091        break;
2092    
2093        /* Multiline mode: assert before any newline, or before end of subject
2094        unless noteol is set. */
2095    
2096        case OP_DOLLM:
2097        if (eptr < md->end_subject)
2098        {        {
2099        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2100          { if (*eptr != NEWLINE) RRETURN(MATCH_NOMATCH); }          {
2101        else          if (md->partial != 0 &&
2102          { if (md->noteol) RRETURN(MATCH_NOMATCH); }              eptr + 1 >= md->end_subject &&
2103        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2104        break;              NLBLOCK->nllen == 2 &&
2105                RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2106              {
2107              md->hitend = TRUE;
2108              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2109              }
2110            RRETURN(MATCH_NOMATCH);
2111            }
2112        }        }
2113      else      else
2114        {        {
2115        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2116        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr < md->end_subject - 1 ||  
            (eptr == md->end_subject - 1 && *eptr != NEWLINE))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2117        }        }
2118      /* ... else fall through */      ecode++;
2119        break;
2120    
2121        /* Not multiline mode: assert before a terminating newline or before end of
2122        subject unless noteol is set. */
2123    
2124        case OP_DOLL:
2125        if (md->noteol) RRETURN(MATCH_NOMATCH);
2126        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2127    
2128        /* ... else fall through for endonly */
2129    
2130      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2131    
2132      case OP_EOD:      case OP_EOD:
2133      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2134        SCHECK_PARTIAL();
2135      ecode++;      ecode++;
2136      break;      break;
2137    
2138      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2139    
2140      case OP_EODN:      case OP_EODN:
2141      if (eptr < md->end_subject - 1 ||      ASSERT_NL_OR_EOS:
2142         (eptr == md->end_subject - 1 && *eptr != NEWLINE)) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject &&
2143            (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2144          {
2145          if (md->partial != 0 &&
2146              eptr + 1 >= md->end_subject &&
2147              NLBLOCK->nltype == NLTYPE_FIXED &&
2148              NLBLOCK->nllen == 2 &&
2149              RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2150            {
2151            md->hitend = TRUE;
2152            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2153            }
2154          RRETURN(MATCH_NOMATCH);
2155          }
2156    
2157        /* Either at end of string or \n before end. */
2158    
2159        SCHECK_PARTIAL();
2160      ecode++;      ecode++;
2161      break;      break;
2162    
# Line 1239  for (;;) Line 2168  for (;;)
2168    
2169        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2170        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
2171        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2172          partial matching. */
2173    
2174  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2175        if (utf8)        if (utf)
2176          {          {
2177            /* Get status of previous character */
2178    
2179          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2180            {            {
2181            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2182            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2183              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2184            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2185    #ifdef SUPPORT_UCP
2186              if (md->use_ucp)
2187                {
2188                if (c == '_') prev_is_word = TRUE; else
2189                  {
2190                  int cat = UCD_CATEGORY(c);
2191                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2192                  }
2193                }
2194              else
2195    #endif
2196            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2197            }            }
2198          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2199            /* Get status of next character */
2200    
2201            if (eptr >= md->end_subject)
2202              {
2203              SCHECK_PARTIAL();
2204              cur_is_word = FALSE;
2205              }
2206            else
2207            {            {
2208            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2209    #ifdef SUPPORT_UCP
2210              if (md->use_ucp)
2211                {
2212                if (c == '_') cur_is_word = TRUE; else
2213                  {
2214                  int cat = UCD_CATEGORY(c);
2215                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2216                  }
2217                }
2218              else
2219    #endif
2220            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2221            }            }
2222          }          }
2223        else        else
2224  #endif  #endif
2225    
2226        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2227          consistency with the behaviour of \w we do use it in this case. */
2228    
2229          {          {
2230          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2231            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2232          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2233            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2234              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2235    #ifdef SUPPORT_UCP
2236              if (md->use_ucp)
2237                {
2238                c = eptr[-1];
2239                if (c == '_') prev_is_word = TRUE; else
2240                  {
2241                  int cat = UCD_CATEGORY(c);
2242                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2243                  }
2244                }
2245              else
2246    #endif
2247              prev_is_word = MAX_255(eptr[-1])
2248                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2249              }
2250    
2251            /* Get status of next character */
2252    
2253            if (eptr >= md->end_subject)
2254              {
2255              SCHECK_PARTIAL();
2256              cur_is_word = FALSE;
2257              }
2258            else
2259    #ifdef SUPPORT_UCP
2260            if (md->use_ucp)
2261              {
2262              c = *eptr;
2263              if (c == '_') cur_is_word = TRUE; else
2264                {
2265                int cat = UCD_CATEGORY(c);
2266                cur_is_word = (cat == ucp_L || cat == ucp_N);
2267                }
2268              }
2269            else
2270    #endif
2271            cur_is_word = MAX_255(*eptr)
2272              && ((md->ctypes[*eptr] & ctype_word) != 0);
2273          }          }
2274    
2275        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1277  for (;;) Line 2280  for (;;)
2280        }        }
2281      break;      break;
2282    
2283      /* Match a single character type; inline for speed */      /* Match any single character type except newline; have to take care with
2284        CRLF newlines and partial matching. */
2285    
2286      case OP_ANY:      case OP_ANY:
2287      if ((ims & PCRE_DOTALL) == 0 && eptr < md->end_subject && *eptr == NEWLINE)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2288        if (md->partial != 0 &&
2289            eptr + 1 >= md->end_subject &&
2290            NLBLOCK->nltype == NLTYPE_FIXED &&
2291            NLBLOCK->nllen == 2 &&
2292            RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2293          {
2294          md->hitend = TRUE;
2295          if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2296          }
2297    
2298        /* Fall through */
2299    
2300        /* Match any single character whatsoever. */
2301    
2302        case OP_ALLANY:
2303        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2304          {                            /* not be updated before SCHECK_PARTIAL. */
2305          SCHECK_PARTIAL();
2306        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2307      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);        }
2308  #ifdef SUPPORT_UTF8      eptr++;
2309      if (utf8)  #ifdef SUPPORT_UTF
2310        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2311  #endif  #endif
2312      ecode++;      ecode++;
2313      break;      break;
# Line 1294  for (;;) Line 2316  for (;;)
2316      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2317    
2318      case OP_ANYBYTE:      case OP_ANYBYTE:
2319      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2320          {                            /* not be updated before SCHECK_PARTIAL. */
2321          SCHECK_PARTIAL();
2322          RRETURN(MATCH_NOMATCH);
2323          }
2324        eptr++;
2325      ecode++;      ecode++;
2326      break;      break;
2327    
2328      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2329      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2330          {
2331          SCHECK_PARTIAL();
2332          RRETURN(MATCH_NOMATCH);
2333          }
2334      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2335      if (      if (
2336  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2337         c < 256 &&         c < 256 &&
2338  #endif  #endif
2339         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1312  for (;;) Line 2343  for (;;)
2343      break;      break;
2344    
2345      case OP_DIGIT:      case OP_DIGIT:
2346      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2347          {
2348          SCHECK_PARTIAL();
2349          RRETURN(MATCH_NOMATCH);
2350          }
2351      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2352      if (      if (
2353  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2354         c >= 256 ||         c > 255 ||
2355  #endif  #endif
2356         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2357         )         )
# Line 1325  for (;;) Line 2360  for (;;)
2360      break;      break;
2361    
2362      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2363      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2364          {
2365          SCHECK_PARTIAL();
2366          RRETURN(MATCH_NOMATCH);
2367          }
2368      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2369      if (      if (
2370  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2371         c < 256 &&         c < 256 &&
2372  #endif  #endif
2373         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1338  for (;;) Line 2377  for (;;)
2377      break;      break;
2378    
2379      case OP_WHITESPACE:      case OP_WHITESPACE:
2380      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2381          {
2382          SCHECK_PARTIAL();
2383          RRETURN(MATCH_NOMATCH);
2384          }
2385      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2386      if (      if (
2387  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2388         c >= 256 ||         c > 255 ||
2389  #endif  #endif
2390         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2391         )         )
# Line 1351  for (;;) Line 2394  for (;;)
2394      break;      break;
2395    
2396      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2397      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2398          {
2399          SCHECK_PARTIAL();
2400          RRETURN(MATCH_NOMATCH);
2401          }
2402      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2403      if (      if (
2404  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2405         c < 256 &&         c < 256 &&
2406  #endif  #endif
2407         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1364  for (;;) Line 2411  for (;;)
2411      break;      break;
2412    
2413      case OP_WORDCHAR:      case OP_WORDCHAR:
2414      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2415          {
2416          SCHECK_PARTIAL();
2417          RRETURN(MATCH_NOMATCH);
2418          }
2419      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2420      if (      if (
2421  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2422         c >= 256 ||         c > 255 ||
2423  #endif  #endif
2424         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2425         )         )
# Line 1376  for (;;) Line 2427  for (;;)
2427      ecode++;      ecode++;
2428      break;      break;
2429    
2430        case OP_ANYNL:
2431        if (eptr >= md->end_subject)
2432          {
2433          SCHECK_PARTIAL();
2434          RRETURN(MATCH_NOMATCH);
2435          }
2436        GETCHARINCTEST(c, eptr);
2437        switch(c)
2438          {
2439          default: RRETURN(MATCH_NOMATCH);
2440    
2441          case CHAR_CR:
2442          if (eptr >= md->end_subject)
2443            {
2444            SCHECK_PARTIAL();
2445            }
2446          else if (RAWUCHARTEST(eptr) == CHAR_LF) eptr++;
2447          break;
2448    
2449          case CHAR_LF:
2450          break;
2451    
2452          case CHAR_VT:
2453          case CHAR_FF:
2454          case CHAR_NEL:
2455    #ifndef EBCDIC
2456          case 0x2028:
2457          case 0x2029:
2458    #endif  /* Not EBCDIC */
2459          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2460          break;
2461          }
2462        ecode++;
2463        break;
2464    
2465        case OP_NOT_HSPACE:
2466        if (eptr >= md->end_subject)
2467          {
2468          SCHECK_PARTIAL();
2469          RRETURN(MATCH_NOMATCH);
2470          }
2471        GETCHARINCTEST(c, eptr);
2472        switch(c)
2473          {
2474          HSPACE_CASES: RRETURN(MATCH_NOMATCH);  /* Byte and multibyte cases */
2475          default: break;
2476          }
2477        ecode++;
2478        break;
2479    
2480        case OP_HSPACE:
2481        if (eptr >= md->end_subject)
2482          {
2483          SCHECK_PARTIAL();
2484          RRETURN(MATCH_NOMATCH);
2485          }
2486        GETCHARINCTEST(c, eptr);
2487        switch(c)
2488          {
2489          HSPACE_CASES: break;  /* Byte and multibyte cases */
2490          default: RRETURN(MATCH_NOMATCH);
2491          }
2492        ecode++;
2493        break;
2494    
2495        case OP_NOT_VSPACE:
2496        if (eptr >= md->end_subject)
2497          {
2498          SCHECK_PARTIAL();
2499          RRETURN(MATCH_NOMATCH);
2500          }
2501        GETCHARINCTEST(c, eptr);
2502        switch(c)
2503          {
2504          VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2505          default: break;
2506          }
2507        ecode++;
2508        break;
2509    
2510        case OP_VSPACE:
2511        if (eptr >= md->end_subject)
2512          {
2513          SCHECK_PARTIAL();
2514          RRETURN(MATCH_NOMATCH);
2515          }
2516        GETCHARINCTEST(c, eptr);
2517        switch(c)
2518          {
2519          VSPACE_CASES: break;
2520          default: RRETURN(MATCH_NOMATCH);
2521          }
2522        ecode++;
2523        break;
2524    
2525  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2526      /* Check the next character by Unicode property. We will get here only      /* Check the next character by Unicode property. We will get here only
2527      if the support is in the binary; otherwise a compile-time error occurs. */      if the support is in the binary; otherwise a compile-time error occurs. */
2528    
2529      case OP_PROP:      case OP_PROP:
2530      case OP_NOTPROP:      case OP_NOTPROP:
2531      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2532          {
2533          SCHECK_PARTIAL();
2534          RRETURN(MATCH_NOMATCH);
2535          }
2536      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2537        {        {
2538        int chartype, script;        const pcre_uint32 *cp;
2539        int category = _pcre_ucp_findprop(c, &chartype, &script);        const ucd_record *prop = GET_UCD(c);
2540    
2541        switch(ecode[1])        switch(ecode[1])
2542          {          {
# Line 1395  for (;;) Line 2545  for (;;)
2545          break;          break;
2546    
2547          case PT_LAMP:          case PT_LAMP:
2548          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2549               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2550               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2551            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2552           break;          break;
2553    
2554          case PT_GC:          case PT_GC:
2555          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2556            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2557          break;          break;
2558    
2559          case PT_PC:          case PT_PC:
2560          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2561            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2562          break;          break;
2563    
2564          case PT_SC:          case PT_SC:
2565          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2566              RRETURN(MATCH_NOMATCH);
2567            break;
2568    
2569            /* These are specials */
2570    
2571            case PT_ALNUM:
2572            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2573                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2574              RRETURN(MATCH_NOMATCH);
2575            break;
2576    
2577            /* Perl space used to exclude VT, but from Perl 5.18 it is included,
2578            which means that Perl space and POSIX space are now identical. PCRE
2579            was changed at release 8.34. */
2580    
2581            case PT_SPACE:    /* Perl space */
2582            case PT_PXSPACE:  /* POSIX space */
2583            switch(c)
2584              {
2585              HSPACE_CASES:
2586              VSPACE_CASES:
2587              if (op == OP_NOTPROP) RRETURN(MATCH_NOMATCH);
2588              break;
2589    
2590              default:
2591              if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z) ==
2592                (op == OP_NOTPROP)) RRETURN(MATCH_NOMATCH);
2593              break;
2594              }
2595            break;
2596    
2597            case PT_WORD:
2598            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2599                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2600                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2601            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2602          break;          break;
2603    
2604            case PT_CLIST:
2605            cp = PRIV(ucd_caseless_sets) + ecode[2];
2606            for (;;)
2607              {
2608              if (c < *cp)
2609                { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; }
2610              if (c == *cp++)
2611                { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } }
2612              }
2613            break;
2614    
2615            case PT_UCNC:
2616            if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
2617                 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
2618                 c >= 0xe000) == (op == OP_NOTPROP))
2619              RRETURN(MATCH_NOMATCH);
2620            break;
2621    
2622            /* This should never occur */
2623    
2624          default:          default:
2625          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
         break;  
2626          }          }
2627    
2628        ecode += 3;        ecode += 3;
# Line 1429  for (;;) Line 2633  for (;;)
2633      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2634    
2635      case OP_EXTUNI:      case OP_EXTUNI:
2636      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2637      GETCHARINCTEST(c, eptr);        {
2638          SCHECK_PARTIAL();
2639          RRETURN(MATCH_NOMATCH);
2640          }
2641        else
2642        {        {
2643        int chartype, script;        int lgb, rgb;
2644        int category = _pcre_ucp_findprop(c, &chartype, &script);        GETCHARINCTEST(c, eptr);
2645        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        lgb = UCD_GRAPHBREAK(c);
2646        while (eptr < md->end_subject)        while (eptr < md->end_subject)
2647          {          {
2648          int len = 1;          int len = 1;
2649          if (!utf8) c = *eptr; else          if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2650            {          rgb = UCD_GRAPHBREAK(c);
2651            GETCHARLEN(c, eptr, len);          if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2652            }          lgb = rgb;
         category = _pcre_ucp_findprop(c, &chartype, &script);  
         if (category != ucp_M) break;  
2653          eptr += len;          eptr += len;
2654          }          }
2655        }        }
2656        CHECK_PARTIAL();
2657      ecode++;      ecode++;
2658      break;      break;
2659  #endif  #endif  /* SUPPORT_UCP */
2660    
2661    
2662      /* 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 1458  for (;;) Line 2665  for (;;)
2665      similar code to character type repeats - written out again for speed.      similar code to character type repeats - written out again for speed.
2666      However, if the referenced string is the empty string, always treat      However, if the referenced string is the empty string, always treat
2667      it as matched, any number of times (otherwise there could be infinite      it as matched, any number of times (otherwise there could be infinite
2668      loops). */      loops). If the reference is unset, there are two possibilities:
2669    
2670      case OP_REF:      (a) In the default, Perl-compatible state, set the length negative;
2671        {      this ensures that every attempt at a match fails. We can't just fail
2672        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      here, because of the possibility of quantifiers with zero minima.
       ecode += 3;                                 /* Advance past item */  
2673    
2674        /* If the reference is unset, set the length to be longer than the amount      (b) If the JavaScript compatibility flag is set, set the length to zero
2675        of subject left; this ensures that every attempt at a match fails. We      so that the back reference matches an empty string.
       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];  
2676    
2677        /* Set up for repetition, or handle the non-repeated case */      Otherwise, set the length to the length of what was matched by the
2678        referenced subpattern.
2679    
2680        switch (*ecode)      The OP_REF and OP_REFI opcodes are used for a reference to a numbered group
2681        or to a non-duplicated named group. For a duplicated named group, OP_DNREF
2682        and OP_DNREFI are used. In this case we must scan the list of groups to
2683        which the name refers, and use the first one that is set. */
2684    
2685        case OP_DNREF:
2686        case OP_DNREFI:
2687        caseless = op == OP_DNREFI;
2688          {
2689          int count = GET2(ecode, 1+IMM2_SIZE);
2690          pcre_uchar *slot = md->name_table + GET2(ecode, 1) * md->name_entry_size;
2691          ecode += 1 + 2*IMM2_SIZE;
2692    
2693          while (count-- > 0)
2694          {          {
2695          case OP_CRSTAR:          offset = GET2(slot, 0) << 1;
2696          case OP_CRMINSTAR:          if (offset < offset_top && md->offset_vector[offset] >= 0) break;
2697          case OP_CRPLUS:          slot += md->name_entry_size;
2698          case OP_CRMINPLUS:          }
2699          case OP_CRQUERY:        if (count < 0)
2700          case OP_CRMINQUERY:          length = (md->jscript_compat)? 0 : -1;
2701          c = *ecode++ - OP_CRSTAR;        else
2702          minimize = (c & 1) != 0;          length = md->offset_vector[offset+1] - md->offset_vector[offset];
2703          min = rep_min[c];                 /* Pick up values from tables; */        }
2704          max = rep_max[c];                 /* zero for max => infinity */      goto REF_REPEAT;
         if (max == 0) max = INT_MAX;  
         break;  
2705    
2706          case OP_CRRANGE:      case OP_REF:
2707          case OP_CRMINRANGE:      case OP_REFI:
2708          minimize = (*ecode == OP_CRMINRANGE);      caseless = op == OP_REFI;
2709          min = GET2(ecode, 1);      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2710          max = GET2(ecode, 3);      ecode += 1 + IMM2_SIZE;
2711          if (max == 0) max = INT_MAX;      if (offset >= offset_top || md->offset_vector[offset] < 0)
2712          ecode += 5;        length = (md->jscript_compat)? 0 : -1;
2713          break;      else
2714          length = md->offset_vector[offset+1] - md->offset_vector[offset];
2715    
2716          default:               /* No repeat follows */      /* Set up for repetition, or handle the non-repeated case */
2717          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);  
2718          eptr += length;      REF_REPEAT:
2719          continue;              /* With the main loop */      switch (*ecode)
2720          {
2721          case OP_CRSTAR:
2722          case OP_CRMINSTAR:
2723          case OP_CRPLUS:
2724          case OP_CRMINPLUS:
2725          case OP_CRQUERY:
2726          case OP_CRMINQUERY:
2727          c = *ecode++ - OP_CRSTAR;
2728          minimize = (c & 1) != 0;
2729          min = rep_min[c];                 /* Pick up values from tables; */
2730          max = rep_max[c];                 /* zero for max => infinity */
2731          if (max == 0) max = INT_MAX;
2732          break;
2733    
2734          case OP_CRRANGE:
2735          case OP_CRMINRANGE:
2736          minimize = (*ecode == OP_CRMINRANGE);
2737          min = GET2(ecode, 1);
2738          max = GET2(ecode, 1 + IMM2_SIZE);
2739          if (max == 0) max = INT_MAX;
2740          ecode += 1 + 2 * IMM2_SIZE;
2741          break;
2742    
2743          default:               /* No repeat follows */
2744          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2745            {
2746            if (length == -2) eptr = md->end_subject;   /* Partial match */
2747            CHECK_PARTIAL();
2748            RRETURN(MATCH_NOMATCH);
2749          }          }
2750          eptr += length;
2751          continue;              /* With the main loop */
2752          }
2753    
2754        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2755        main loop. */      zero, just continue with the main loop. If the length is negative, it
2756        means the reference is unset in non-Java-compatible mode. If the minimum is
2757        zero, we can continue at the same level without recursion. For any other
2758        minimum, carrying on will result in NOMATCH. */
2759    
2760        if (length == 0) continue;      if (length == 0) continue;
2761        if (length < 0 && min == 0) continue;
2762    
2763        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2764        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2765        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2766    
2767        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2768          {
2769          int slength;
2770          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2771          {          {
2772          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          if (slength == -2) eptr = md->end_subject;   /* Partial match */
2773          eptr += length;          CHECK_PARTIAL();
2774            RRETURN(MATCH_NOMATCH);
2775          }          }
2776          eptr += slength;
2777          }
2778    
2779        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2780        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2781    
2782        if (min == max) continue;      if (min == max) continue;
2783    
2784        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2785    
2786        if (minimize)      if (minimize)
2787          {
2788          for (fi = min;; fi++)
2789          {          {
2790          for (fi = min;; fi++)          int slength;
2791            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2792            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2793            if (fi >= max) RRETURN(MATCH_NOMATCH);
2794            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2795            {            {
2796            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2797            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2798            if (fi >= max || !match_ref(offset, eptr, length, md, ims))            RRETURN(MATCH_NOMATCH);
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2799            }            }
2800          /* Control never gets here */          eptr += slength;
2801          }          }
2802          /* Control never gets here */
2803          }
2804    
2805        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2806    
2807        else      else
2808          {
2809          pp = eptr;
2810          for (i = min; i < max; i++)
2811          {          {
2812          pp = eptr;          int slength;
2813          for (i = min; i < max; i++)          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
           {  
           if (!match_ref(offset, eptr, length, md, ims)) break;  
           eptr += length;  
           }  
         while (eptr >= pp)  
2814            {            {
2815            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2816            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            the soft partial matching case. */
2817            eptr -= length;  
2818              if (slength == -2 && md->partial != 0 &&
2819                  md->end_subject > md->start_used_ptr)
2820                {
2821                md->hitend = TRUE;
2822                if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2823                }
2824              break;
2825            }            }
2826          RRETURN(MATCH_NOMATCH);          eptr += slength;
2827            }
2828    
2829          while (eptr >= pp)
2830            {
2831            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2832            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2833            eptr -= length;
2834          }          }
2835          RRETURN(MATCH_NOMATCH);
2836        }        }
2837      /* Control never gets here */      /* Control never gets here */
2838    
   
   
2839      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2840      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,
2841      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 1578  for (;;) Line 2850  for (;;)
2850      case OP_NCLASS:      case OP_NCLASS:
2851      case OP_CLASS:      case OP_CLASS:
2852        {        {
2853          /* The data variable is saved across frames, so the byte map needs to
2854          be stored there. */
2855    #define BYTE_MAP ((pcre_uint8 *)data)
2856        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2857        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2858    
2859        switch (*ecode)        switch (*ecode)
2860          {          {
# Line 1600  for (;;) Line 2875  for (;;)
2875          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2876          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2877          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2878          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2879          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2880          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2881          break;          break;
2882    
2883          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1612  for (;;) Line 2887  for (;;)
2887    
2888        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2889    
2890  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2891        /* UTF-8 mode */        if (utf)
       if (utf8)  
2892          {          {
2893          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2894            {            {
2895            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2896                {
2897                SCHECK_PARTIAL();
2898                RRETURN(MATCH_NOMATCH);
2899                }
2900            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2901            if (c > 255)            if (c > 255)
2902              {              {
2903              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2904              }              }
2905            else            else
2906              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2907            }            }
2908          }          }
2909        else        else
2910  #endif  #endif
2911        /* Not UTF-8 mode */        /* Not UTF mode */
2912          {          {
2913          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2914            {            {
2915            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2916                {
2917                SCHECK_PARTIAL();
2918                RRETURN(MATCH_NOMATCH);
2919                }
2920            c = *eptr++;            c = *eptr++;
2921            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2922              if (c > 255)
2923                {
2924                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2925                }
2926              else
2927    #endif
2928                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2929            }            }
2930          }          }
2931    
# Line 1652  for (;;) Line 2939  for (;;)
2939    
2940        if (minimize)        if (minimize)
2941          {          {
2942  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2943          /* UTF-8 mode */          if (utf)
         if (utf8)  
2944            {            {
2945            for (fi = min;; fi++)            for (fi = min;; fi++)
2946              {              {
2947              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2948              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2949              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2950                if (eptr >= md->end_subject)
2951                  {
2952                  SCHECK_PARTIAL();
2953                  RRETURN(MATCH_NOMATCH);
2954                  }
2955              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2956              if (c > 255)              if (c > 255)
2957                {                {
2958                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2959                }                }
2960              else              else
2961                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2962              }              }
2963            }            }
2964          else          else
2965  #endif  #endif
2966          /* Not UTF-8 mode */          /* Not UTF mode */
2967            {            {
2968            for (fi = min;; fi++)            for (fi = min;; fi++)
2969              {              {
2970              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2971              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2972              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2973                if (eptr >= md->end_subject)
2974                  {
2975                  SCHECK_PARTIAL();
2976                  RRETURN(MATCH_NOMATCH);
2977                  }
2978              c = *eptr++;              c = *eptr++;
2979              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2980                if (c > 255)
2981                  {
2982                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2983                  }
2984                else
2985    #endif
2986                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2987              }              }
2988            }            }
2989          /* Control never gets here */          /* Control never gets here */
# Line 1694  for (;;) Line 2995  for (;;)
2995          {          {
2996          pp = eptr;          pp = eptr;
2997    
2998  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2999          /* UTF-8 mode */          if (utf)
         if (utf8)  
3000            {            {
3001            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3002              {              {
3003              int len = 1;              int len = 1;
3004              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3005                  {
3006                  SCHECK_PARTIAL();
3007                  break;
3008                  }
3009              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
3010              if (c > 255)              if (c > 255)
3011                {                {
3012                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
3013                }                }
3014              else              else
3015                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
3016              eptr += len;              eptr += len;
3017              }              }
3018            for (;;)            for (;;)
3019              {              {
3020              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
3021              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3022              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3023              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1723  for (;;) Line 3025  for (;;)
3025            }            }
3026          else          else
3027  #endif  #endif
3028            /* Not UTF-8 mode */            /* Not UTF mode */
3029            {            {
3030            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3031              {              {
3032              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3033                  {
3034                  SCHECK_PARTIAL();
3035                  break;
3036                  }
3037              c = *eptr;              c = *eptr;
3038              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
3039                if (c > 255)
3040                  {
3041                  if (op == OP_CLASS) break;
3042                  }
3043                else
3044    #endif
3045                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3046              eptr++;              eptr++;
3047              }              }
3048            while (eptr >= pp)            while (eptr >= pp)
3049              {              {
3050              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3051              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3052              eptr--;              eptr--;
3053              }              }
# Line 1742  for (;;) Line 3055  for (;;)
3055    
3056          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3057          }          }
3058    #undef BYTE_MAP
3059        }        }
3060      /* Control never gets here */      /* Control never gets here */
3061    
3062    
3063      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
3064      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
3065        mode, because Unicode properties are supported in non-UTF-8 mode. */
3066    
3067  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3068      case OP_XCLASS:      case OP_XCLASS:
3069        {        {
3070        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 1774  for (;;) Line 3089  for (;;)
3089          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3090          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3091          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3092          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3093          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3094          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3095          break;          break;
3096    
3097          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1788  for (;;) Line 3103  for (;;)
3103    
3104        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3105          {          {
3106          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3107          GETCHARINC(c, eptr);            {
3108          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
3109              RRETURN(MATCH_NOMATCH);
3110              }
3111            GETCHARINCTEST(c, eptr);
3112            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3113          }          }
3114    
3115        /* 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 1805  for (;;) Line 3124  for (;;)
3124          {          {
3125          for (fi = min;; fi++)          for (fi = min;; fi++)
3126            {            {
3127            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3128            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3129            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3130            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3131            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3132                SCHECK_PARTIAL();
3133                RRETURN(MATCH_NOMATCH);
3134                }
3135              GETCHARINCTEST(c, eptr);
3136              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3137            }            }
3138          /* Control never gets here */          /* Control never gets here */
3139          }          }
# Line 1822  for (;;) Line 3146  for (;;)
3146          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3147            {            {
3148            int len = 1;            int len = 1;
3149            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3150            GETCHARLEN(c, eptr, len);              {
3151            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3152                break;
3153                }
3154    #ifdef SUPPORT_UTF
3155              GETCHARLENTEST(c, eptr, len);
3156    #else
3157              c = *eptr;
3158    #endif
3159              if (!PRIV(xclass)(c, data, utf)) break;
3160            eptr += len;            eptr += len;
3161            }            }
3162          for(;;)          for(;;)
3163            {            {
3164            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3165            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3166            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3167            BACKCHAR(eptr)  #ifdef SUPPORT_UTF
3168              if (utf) BACKCHAR(eptr);
3169    #endif
3170            }            }
3171          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3172          }          }
# Line 1844  for (;;) Line 3178  for (;;)
3178      /* Match a single character, casefully */      /* Match a single character, casefully */
3179    
3180      case OP_CHAR:      case OP_CHAR:
3181  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3182      if (utf8)      if (utf)
3183        {        {
3184        length = 1;        length = 1;
3185        ecode++;        ecode++;
3186        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3187        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3188        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);          {
3189            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3190            RRETURN(MATCH_NOMATCH);
3191            }
3192          while (length-- > 0) if (*ecode++ != RAWUCHARINC(eptr)) RRETURN(MATCH_NOMATCH);
3193        }        }
3194      else      else
3195  #endif  #endif
3196        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3197        {        {
3198        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3199            {
3200            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3201            RRETURN(MATCH_NOMATCH);
3202            }
3203        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3204        ecode += 2;        ecode += 2;
3205        }        }
3206      break;      break;
3207    
3208      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3209        subject, give up immediately. */
3210    
3211        case OP_CHARI:
3212        if (eptr >= md->end_subject)
3213          {
3214          SCHECK_PARTIAL();
3215          RRETURN(MATCH_NOMATCH);
3216          }
3217    
3218      case OP_CHARNC:  #ifdef SUPPORT_UTF
3219  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3220        {        {
3221        length = 1;        length = 1;
3222        ecode++;        ecode++;
3223        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3224    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3225        /* 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
3226        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3227          fast lookup table. We know that there is at least one byte left in the
3228          subject. */
3229    
3230        if (fc < 128)        if (fc < 128)
3231          {          {
3232          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          pcre_uint32 cc = RAWUCHAR(eptr);
3233            if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH);
3234            ecode++;
3235            eptr++;
3236          }          }
3237    
3238        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3239          use the value of "length" to check for sufficient bytes left, because the
3240          other case of the character may have more or fewer bytes.  */
3241    
3242        else        else
3243          {          {
3244          int dc;          pcre_uint32 dc;
3245          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
3246          ecode += length;          ecode += length;
3247    
# Line 1898  for (;;) Line 3251  for (;;)
3251          if (fc != dc)          if (fc != dc)
3252            {            {
3253  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3254            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3255  #endif  #endif
3256              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3257            }            }
3258          }          }
3259        }        }
3260      else      else
3261  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3262    
3263      /* Non-UTF-8 mode */      /* Not UTF mode */
3264        {        {
3265        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3266        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3267          eptr++;
3268        ecode += 2;        ecode += 2;
3269        }        }
3270      break;      break;
3271    
3272      /* Match a single character repeatedly; different opcodes share code. */      /* Match a single character repeatedly. */
3273    
3274      case OP_EXACT:      case OP_EXACT:
3275        case OP_EXACTI:
3276      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3277      ecode += 3;      ecode += 1 + IMM2_SIZE;
3278      goto REPEATCHAR;      goto REPEATCHAR;
3279    
3280        case OP_POSUPTO:
3281        case OP_POSUPTOI:
3282        possessive = TRUE;
3283        /* Fall through */
3284    
3285      case OP_UPTO:      case OP_UPTO:
3286        case OP_UPTOI:
3287      case OP_MINUPTO:      case OP_MINUPTO:
3288        case OP_MINUPTOI:
3289      min = 0;      min = 0;
3290      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3291      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3292      ecode += 3;      ecode += 1 + IMM2_SIZE;
3293        goto REPEATCHAR;
3294    
3295        case OP_POSSTAR:
3296        case OP_POSSTARI:
3297        possessive = TRUE;
3298        min = 0;
3299        max = INT_MAX;
3300        ecode++;
3301        goto REPEATCHAR;
3302    
3303        case OP_POSPLUS:
3304        case OP_POSPLUSI:
3305        possessive = TRUE;
3306        min = 1;
3307        max = INT_MAX;
3308        ecode++;
3309        goto REPEATCHAR;
3310    
3311        case OP_POSQUERY:
3312        case OP_POSQUERYI:
3313        possessive = TRUE;
3314        min = 0;
3315        max = 1;
3316        ecode++;
3317      goto REPEATCHAR;      goto REPEATCHAR;
3318