/[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 1313 by ph10, Wed Apr 24 12:07:09 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 preceed 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        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,      When we get to the final alternative within the brackets, we used to return
1042          match_isgroup);      the result of a recursive call to match() whatever happened so it was
1043        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      possible to reduce stack usage by turning this into a tail recursion,
1044        ecode += GET(ecode, 1);      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      while (*ecode == OP_ALT);      optimization is no longer always possible.
1047      DPRINTF(("bracket 0 failed\n"));  
1048      RRETURN(MATCH_NOMATCH);      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      /* Conditional group: compilation checked that there are no more than      case OP_ONCE:
1059      two branches. If the condition is false, skipping the first branch takes us      case OP_BRA:
1060      past the end if there is only one branch, but that's OK because that is      case OP_SBRA:
1061      exactly what going to the ket would do. */      DPRINTF(("start non-capturing bracket\n"));
1062    
1063      case OP_COND:      for (;;)
     if (ecode[LINK_SIZE+1] == OP_CREF) /* Condition extract or recurse test */  
1064        {        {
1065        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */        if (op >= OP_SBRA || op == OP_ONCE)
1066        condition = (offset == CREF_RECURSE * 2)?          md->match_function_type = MATCH_CBEGROUP;
         (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);  
       }  
1067    
1068      /* The condition is an assertion. Call match() to evaluate it - setting        /* If this is not a possibly empty group, and there are no (*THEN)s in
1069      the final argument TRUE causes it to stop at the end of an assertion. */        the pattern, and this is the final alternative, optimize as described
1070          above. */
1071    
1072      else        else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
       {  
       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,  
           match_condassert | match_isgroup);  
       if (rrc == MATCH_MATCH)  
1073          {          {
1074          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE+2);          ecode += PRIV(OP_lengths)[*ecode];
1075          while (*ecode == OP_ALT) ecode += GET(ecode, 1);          goto TAIL_RECURSE;
1076          }          }
1077        else if (rrc != MATCH_NOMATCH)  
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          RRETURN(rrc);         /* Need braces because of following else */          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        else ecode += GET(ecode, 1);  
1096        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,        if (rrc != MATCH_NOMATCH)
1097          match_isgroup);          {
1098        RRETURN(rrc);          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);
1111          md->mark = save_mark;
1112          if (*ecode != OP_ALT) break;
1113          md->capture_last = save_capture_last;
1114        }        }
     /* Control never reaches here */  
1115    
1116      /* Skip over conditional reference or large extraction number data if      RRETURN(MATCH_NOMATCH);
     encountered. */  
1117    
1118      case OP_CREF:      /* Handle possessive capturing brackets with an unlimited repeat. We come
1119      case OP_BRANUMBER:      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1120      ecode += 3;      handled similarly to the normal case above. However, the matching is
1121      break;      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      /* End of the pattern. If we are in a recursion, we should restore the      POSSESSIVE_CAPTURE:
1131      offsets appropriately and continue from after the call. */      number = GET2(ecode, 1+LINK_SIZE);
1132        offset = number << 1;
1133    
1134      case OP_END:  #ifdef PCRE_DEBUG
1135      if (md->recursive != NULL && md->recursive->group_num == 0)      printf("start possessive bracket %d\n", number);
1136        printf("subject=");
1137        pchars(eptr, 16, TRUE, md);
1138        printf("\n");
1139    #endif
1140    
1141        if (offset < md->offset_max)
1142        {        {
1143        recursion_info *rec = md->recursive;        matched_once = FALSE;
1144        DPRINTF(("End of pattern in a (?0) recursion\n"));        code_offset = (int)(ecode - md->start_code);
       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;  
       }  
1145    
1146      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty        save_offset1 = md->offset_vector[offset];
1147      string - backtracking will then try other alternatives, if any. */        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      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
     md->end_match_ptr = eptr;          /* Record where we ended */  
     md->end_offset_top = offset_top;   /* and how many extracts were taken */  
     RRETURN(MATCH_MATCH);  
   
     /* Change option settings */  
   
     case OP_OPT:  
     ims = ecode[1];  
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1152    
1153      /* Assertion brackets. Check the alternative branches in turn - the        /* Each time round the loop, save the current subject position for use
1154      matching won't pass the KET for an assertion. If any one branch matches,        when the group matches. For MATCH_MATCH, the group has matched, so we
1155      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the        restart it with a new subject starting position, remembering that we had
1156      start of each branch to move the current point backwards, so the code at        at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1157      this level is identical to the lookahead case. */        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      case OP_ASSERT:          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1191      case OP_ASSERTBACK:          md->capture_last = save_capture_last;
1192      do          ecode += GET(ecode, 1);
1193        {          if (*ecode != OP_ALT) break;
1194        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,          }
         match_isgroup);  
       if (rrc == MATCH_MATCH) break;  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += GET(ecode, 1);  
       }  
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1195    
1196      /* If checking an assertion for a condition, return MATCH_MATCH. */        if (!matched_once)
1197            {
1198            md->offset_vector[offset] = save_offset1;
1199            md->offset_vector[offset+1] = save_offset2;
1200            md->offset_vector[md->offset_end - number] = save_offset3;
1201            }
1202    
1203      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);        if (allow_zero || matched_once)
1204            {
1205            ecode += 1 + LINK_SIZE;
1206            break;
1207            }
1208    
1209      /* Continue from after the assertion, updating the offsets high water        RRETURN(MATCH_NOMATCH);
1210      mark, since extracts may have been taken during the assertion. */        }
1211    
1212      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1213      ecode += 1 + LINK_SIZE;      as a non-capturing bracket. */
     offset_top = md->end_offset_top;  
     continue;  
1214    
1215      /* Negative assertion: all branches must fail to match */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1216        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1217    
1218      case OP_ASSERT_NOT:      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
     case OP_ASSERTBACK_NOT:  
     do  
       {  
       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,  
         match_isgroup);  
       if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
1219    
1220      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1221        /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1222    
1223      ecode += 1 + LINK_SIZE;      /* Non-capturing possessive bracket with unlimited repeat. We come here
1224      continue;      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      /* Move the subject pointer back. This occurs only at the start of      case OP_BRAPOS:
1229      each branch of a lookbehind assertion. If we are too close to the start to      case OP_SBRAPOS:
1230      move back, this match function fails. When working with UTF-8 we move      allow_zero = FALSE;
     back a number of characters, not bytes. */  
1231    
1232      case OP_REVERSE:      POSSESSIVE_NON_CAPTURE:
1233  #ifdef SUPPORT_UTF8      matched_once = FALSE;
1234      if (utf8)      code_offset = (int)(ecode - md->start_code);
1235        save_capture_last = md->capture_last;
1236    
1237        for (;;)
1238        {        {
1239        c = GET(ecode,1);        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1240        for (i = 0; i < c; i++)        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1241            eptrb, RM48);
1242          if (rrc == MATCH_KETRPOS)
1243          {          {
1244          eptr--;          offset_top = md->end_offset_top;
1245          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          eptr = md->end_match_ptr;
1246          BACKCHAR(eptr)          ecode = md->start_code + code_offset;
1247            matched_once = TRUE;
1248            continue;
1249          }          }
       }  
     else  
 #endif  
1250    
1251      /* No UTF-8 support, or not in UTF-8 mode: count is byte count */        /* 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);
1263          ecode += GET(ecode, 1);
1264          if (*ecode != OP_ALT) break;
1265          md->capture_last = save_capture_last;
1266          }
1267    
1268        if (matched_once || allow_zero)
1269        {        {
1270        eptr -= GET(ecode,1);        ecode += 1 + LINK_SIZE;
1271        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        break;
1272        }        }
1273        RRETURN(MATCH_NOMATCH);
1274    
1275      /* Skip to next op code */      /* Control never reaches here. */
1276    
1277      ecode += 1 + LINK_SIZE;      /* Conditional group: compilation checked that there are no more than
1278      break;      two branches. If the condition is false, skipping the first branch takes us
1279        past the end if there is only one branch, but that's OK because that is
1280        exactly what going to the ket would do. */
1281    
1282      /* The callout item calls an external function, if one is provided, passing      case OP_COND:
1283      details of the match so far. This is mainly for debugging, though the      case OP_SCOND:
1284      function is able to force a failure. */      codelink = GET(ecode, 1);
1285    
1286      case OP_CALLOUT:      /* Because of the way auto-callout works during compile, a callout item is
1287      if (pcre_callout != NULL)      inserted between OP_COND and an assertion condition. */
1288    
1289        if (ecode[LINK_SIZE+1] == OP_CALLOUT)
1290        {        {
1291        pcre_callout_block cb;        if (PUBL(callout) != NULL)
1292        cb.version          = 1;   /* Version 1 of the callout block */          {
1293        cb.callout_number   = ecode[1];          PUBL(callout_block) cb;
1294        cb.offset_vector    = md->offset_vector;          cb.version          = 2;   /* Version 1 of the callout block */
1295        cb.subject          = (PCRE_SPTR)md->start_subject;          cb.callout_number   = ecode[LINK_SIZE+2];
1296        cb.subject_length   = md->end_subject - md->start_subject;          cb.offset_vector    = md->offset_vector;
1297        cb.start_match      = md->start_match - md->start_subject;  #if defined COMPILE_PCRE8
1298        cb.current_position = eptr - md->start_subject;          cb.subject          = (PCRE_SPTR)md->start_subject;
1299    #elif defined COMPILE_PCRE16
1300            cb.subject          = (PCRE_SPTR16)md->start_subject;
1301    #elif defined COMPILE_PCRE32
1302            cb.subject          = (PCRE_SPTR32)md->start_subject;
1303    #endif
1304            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1305            cb.start_match      = (int)(mstart - md->start_subject);
1306            cb.current_position = (int)(eptr - md->start_subject);
1307            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1308            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1309            cb.capture_top      = offset_top/2;
1310            cb.capture_last     = md->capture_last & CAPLMASK;
1311            /* Internal change requires this for API compatibility. */
1312            if (cb.capture_last == 0) cb.capture_last = -1;
1313            cb.callout_data     = md->callout_data;
1314            cb.mark             = md->nomatch_mark;
1315            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1316            if (rrc < 0) RRETURN(rrc);
1317            }
1318          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1319          codelink -= PRIV(OP_lengths)[OP_CALLOUT];
1320          }
1321    
1322        condcode = ecode[LINK_SIZE+1];
1323    
1324        /* Now see what the actual condition is */
1325    
1326        if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1327          {
1328          if (md->recursive == NULL)                /* Not recursing => FALSE */
1329            {
1330            condition = FALSE;
1331            ecode += GET(ecode, 1);
1332            }
1333          else
1334            {
1335            unsigned int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1336            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1337    
1338            /* If the test is for recursion into a specific subpattern, and it is
1339            false, but the test was set up by name, scan the table to see if the
1340            name refers to any other numbers, and test them. The condition is true
1341            if any one is set. */
1342    
1343            if (!condition && condcode == OP_NRREF)
1344              {
1345              pcre_uchar *slotA = md->name_table;
1346              for (i = 0; i < md->name_count; i++)
1347                {
1348                if (GET2(slotA, 0) == recno) break;
1349                slotA += md->name_entry_size;
1350                }
1351    
1352              /* Found a name for the number - there can be only one; duplicate
1353              names for different numbers are allowed, but not vice versa. First
1354              scan down for duplicates. */
1355    
1356              if (i < md->name_count)
1357                {
1358                pcre_uchar *slotB = slotA;
1359                while (slotB > md->name_table)
1360                  {
1361                  slotB -= md->name_entry_size;
1362                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1363                    {
1364                    condition = GET2(slotB, 0) == md->recursive->group_num;
1365                    if (condition) break;
1366                    }
1367                  else break;
1368                  }
1369    
1370                /* Scan up for duplicates */
1371    
1372                if (!condition)
1373                  {
1374                  slotB = slotA;
1375                  for (i++; i < md->name_count; i++)
1376                    {
1377                    slotB += md->name_entry_size;
1378                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1379                      {
1380                      condition = GET2(slotB, 0) == md->recursive->group_num;
1381                      if (condition) break;
1382                      }
1383                    else break;
1384                    }
1385                  }
1386                }
1387              }
1388    
1389            /* Chose branch according to the condition */
1390    
1391            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1392            }
1393          }
1394    
1395        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1396          {
1397          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1398          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1399    
1400          /* If the numbered capture is unset, but the reference was by name,
1401          scan the table to see if the name refers to any other numbers, and test
1402          them. The condition is true if any one is set. This is tediously similar
1403          to the code above, but not close enough to try to amalgamate. */
1404    
1405          if (!condition && condcode == OP_NCREF)
1406            {
1407            unsigned int refno = offset >> 1;
1408            pcre_uchar *slotA = md->name_table;
1409    
1410            for (i = 0; i < md->name_count; i++)
1411              {
1412              if (GET2(slotA, 0) == refno) break;
1413              slotA += md->name_entry_size;
1414              }
1415    
1416            /* Found a name for the number - there can be only one; duplicate names
1417            for different numbers are allowed, but not vice versa. First scan down
1418            for duplicates. */
1419    
1420            if (i < md->name_count)
1421              {
1422              pcre_uchar *slotB = slotA;
1423              while (slotB > md->name_table)
1424                {
1425                slotB -= md->name_entry_size;
1426                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1427                  {
1428                  offset = GET2(slotB, 0) << 1;
1429                  condition = offset < offset_top &&
1430                    md->offset_vector[offset] >= 0;
1431                  if (condition) break;
1432                  }
1433                else break;
1434                }
1435    
1436              /* Scan up for duplicates */
1437    
1438              if (!condition)
1439                {
1440                slotB = slotA;
1441                for (i++; i < md->name_count; i++)
1442                  {
1443                  slotB += md->name_entry_size;
1444                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1445                    {
1446                    offset = GET2(slotB, 0) << 1;
1447                    condition = offset < offset_top &&
1448                      md->offset_vector[offset] >= 0;
1449                    if (condition) break;
1450                    }
1451                  else break;
1452                  }
1453                }
1454              }
1455            }
1456    
1457          /* Chose branch according to the condition */
1458    
1459          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1460          }
1461    
1462        else if (condcode == OP_DEF)     /* DEFINE - always false */
1463          {
1464          condition = FALSE;
1465          ecode += GET(ecode, 1);
1466          }
1467    
1468        /* The condition is an assertion. Call match() to evaluate it - setting
1469        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1470        an assertion. */
1471    
1472        else
1473          {
1474          md->match_function_type = MATCH_CONDASSERT;
1475          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1476          if (rrc == MATCH_MATCH)
1477            {
1478            if (md->end_offset_top > offset_top)
1479              offset_top = md->end_offset_top;  /* Captures may have happened */
1480            condition = TRUE;
1481            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1482            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1483            }
1484    
1485          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1486          assertion; it is therefore treated as NOMATCH. */
1487    
1488          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1489            {
1490            RRETURN(rrc);         /* Need braces because of following else */
1491            }
1492          else
1493            {
1494            condition = FALSE;
1495            ecode += codelink;
1496            }
1497          }
1498    
1499        /* We are now at the branch that is to be obeyed. As there is only one, can
1500        use tail recursion to avoid using another stack frame, except when there is
1501        unlimited repeat of a possibly empty group. In the latter case, a recursive
1502        call to match() is always required, unless the second alternative doesn't
1503        exist, in which case we can just plough on. Note that, for compatibility
1504        with Perl, the | in a conditional group is NOT treated as creating two
1505        alternatives. If a THEN is encountered in the branch, it propagates out to
1506        the enclosing alternative (unless nested in a deeper set of alternatives,
1507        of course). */
1508    
1509        if (condition || *ecode == OP_ALT)
1510          {
1511          if (op != OP_SCOND)
1512            {
1513            ecode += 1 + LINK_SIZE;
1514            goto TAIL_RECURSE;
1515            }
1516    
1517          md->match_function_type = MATCH_CBEGROUP;
1518          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1519          RRETURN(rrc);
1520          }
1521    
1522         /* Condition false & no alternative; continue after the group. */
1523    
1524        else
1525          {
1526          ecode += 1 + LINK_SIZE;
1527          }
1528        break;
1529    
1530    
1531        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1532        to close any currently open capturing brackets. */
1533    
1534        case OP_CLOSE:
1535        number = GET2(ecode, 1);   /* Must be less than 65536 */
1536        offset = number << 1;
1537    
1538    #ifdef PCRE_DEBUG
1539          printf("end bracket %d at *ACCEPT", number);
1540          printf("\n");
1541    #endif
1542    
1543        md->capture_last = (md->capture_last & OVFLMASK) | number;
1544        if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1545          {
1546          md->offset_vector[offset] =
1547            md->offset_vector[md->offset_end - number];
1548          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1549          if (offset_top <= offset) offset_top = offset + 2;
1550          }
1551        ecode += 1 + IMM2_SIZE;
1552        break;
1553    
1554    
1555        /* End of the pattern, either real or forced. */
1556    
1557        case OP_END:
1558        case OP_ACCEPT:
1559        case OP_ASSERT_ACCEPT:
1560    
1561        /* If we have matched an empty string, fail if not in an assertion and not
1562        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1563        is set and we have matched at the start of the subject. In both cases,
1564        backtracking will then try other alternatives, if any. */
1565    
1566        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1567             md->recursive == NULL &&
1568             (md->notempty ||
1569               (md->notempty_atstart &&
1570                 mstart == md->start_subject + md->start_offset)))
1571          RRETURN(MATCH_NOMATCH);
1572    
1573        /* Otherwise, we have a match. */
1574    
1575        md->end_match_ptr = eptr;           /* Record where we ended */
1576        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1577        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1578    
1579        /* For some reason, the macros don't work properly if an expression is
1580        given as the argument to RRETURN when the heap is in use. */
1581    
1582        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1583        RRETURN(rrc);
1584    
1585        /* Assertion brackets. Check the alternative branches in turn - the
1586        matching won't pass the KET for an assertion. If any one branch matches,
1587        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1588        start of each branch to move the current point backwards, so the code at
1589        this level is identical to the lookahead case. When the assertion is part
1590        of a condition, we want to return immediately afterwards. The caller of
1591        this incarnation of the match() function will have set MATCH_CONDASSERT in
1592        md->match_function type, and one of these opcodes will be the first opcode
1593        that is processed. We use a local variable that is preserved over calls to
1594        match() to remember this case. */
1595    
1596        case OP_ASSERT:
1597        case OP_ASSERTBACK:
1598        save_mark = md->mark;
1599        if (md->match_function_type == MATCH_CONDASSERT)
1600          {
1601          condassert = TRUE;
1602          md->match_function_type = 0;
1603          }
1604        else condassert = FALSE;
1605    
1606        /* Loop for each branch */
1607    
1608        do
1609          {
1610          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1611    
1612          /* A match means that the assertion is true; break out of the loop
1613          that matches its alternatives. */
1614    
1615          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1616            {
1617            mstart = md->start_match_ptr;   /* In case \K reset it */
1618            break;
1619            }
1620    
1621          /* If not matched, restore the previous mark setting. */
1622    
1623          md->mark = save_mark;
1624    
1625          /* See comment in the code for capturing groups above about handling
1626          THEN. */
1627    
1628          if (rrc == MATCH_THEN)
1629            {
1630            next = ecode + GET(ecode,1);
1631            if (md->start_match_ptr < next &&
1632                (*ecode == OP_ALT || *next == OP_ALT))
1633              rrc = MATCH_NOMATCH;
1634            }
1635    
1636          /* Anything other than NOMATCH causes the entire assertion to fail,
1637          passing back the return code. This includes COMMIT, SKIP, PRUNE and an
1638          uncaptured THEN, which means they take their normal effect. This
1639          consistent approach does not always have exactly the same effect as in
1640          Perl. */
1641    
1642          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1643          ecode += GET(ecode, 1);
1644          }
1645        while (*ecode == OP_ALT);   /* Continue for next alternative */
1646    
1647        /* If we have tried all the alternative branches, the assertion has
1648        failed. If not, we broke out after a match. */
1649    
1650        if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1651    
1652        /* If checking an assertion for a condition, return MATCH_MATCH. */
1653    
1654        if (condassert) RRETURN(MATCH_MATCH);
1655    
1656        /* Continue from after a successful assertion, updating the offsets high
1657        water mark, since extracts may have been taken during the assertion. */
1658    
1659        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1660        ecode += 1 + LINK_SIZE;
1661        offset_top = md->end_offset_top;
1662        continue;
1663    
1664        /* Negative assertion: all branches must fail to match for the assertion to
1665        succeed. */
1666    
1667        case OP_ASSERT_NOT:
1668        case OP_ASSERTBACK_NOT:
1669        save_mark = md->mark;
1670        if (md->match_function_type == MATCH_CONDASSERT)
1671          {
1672          condassert = TRUE;
1673          md->match_function_type = 0;
1674          }
1675        else condassert = FALSE;
1676    
1677        /* Loop for each alternative branch. */
1678    
1679        do
1680          {
1681          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1682          md->mark = save_mark;   /* Always restore the mark setting */
1683    
1684          switch(rrc)
1685            {
1686            case MATCH_MATCH:            /* A successful match means */
1687            case MATCH_ACCEPT:           /* the assertion has failed. */
1688            RRETURN(MATCH_NOMATCH);
1689    
1690            case MATCH_NOMATCH:          /* Carry on with next branch */
1691            break;
1692    
1693            /* See comment in the code for capturing groups above about handling
1694            THEN. */
1695    
1696            case MATCH_THEN:
1697            next = ecode + GET(ecode,1);
1698            if (md->start_match_ptr < next &&
1699                (*ecode == OP_ALT || *next == OP_ALT))
1700              {
1701              rrc = MATCH_NOMATCH;
1702              break;
1703              }
1704            /* Otherwise fall through. */
1705    
1706            /* COMMIT, SKIP, PRUNE, and an uncaptured THEN cause the whole
1707            assertion to fail to match, without considering any more alternatives.
1708            Failing to match means the assertion is true. This is a consistent
1709            approach, but does not always have the same effect as in Perl. */
1710    
1711            case MATCH_COMMIT:
1712            case MATCH_SKIP:
1713            case MATCH_SKIP_ARG:
1714            case MATCH_PRUNE:
1715            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1716            goto NEG_ASSERT_TRUE;   /* Break out of alternation loop */
1717    
1718            /* Anything else is an error */
1719    
1720            default:
1721            RRETURN(rrc);
1722            }
1723    
1724          /* Continue with next branch */
1725    
1726          ecode += GET(ecode,1);
1727          }
1728        while (*ecode == OP_ALT);
1729    
1730        /* All branches in the assertion failed to match. */
1731    
1732        NEG_ASSERT_TRUE:
1733        if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1734        ecode += 1 + LINK_SIZE;                /* Continue with current branch */
1735        continue;
1736    
1737        /* Move the subject pointer back. This occurs only at the start of
1738        each branch of a lookbehind assertion. If we are too close to the start to
1739        move back, this match function fails. When working with UTF-8 we move
1740        back a number of characters, not bytes. */
1741    
1742        case OP_REVERSE:
1743    #ifdef SUPPORT_UTF
1744        if (utf)
1745          {
1746          i = GET(ecode, 1);
1747          while (i-- > 0)
1748            {
1749            eptr--;
1750            if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1751            BACKCHAR(eptr);
1752            }
1753          }
1754        else
1755    #endif
1756    
1757        /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1758    
1759          {
1760          eptr -= GET(ecode, 1);
1761          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1762          }
1763    
1764        /* Save the earliest consulted character, then skip to next op code */
1765    
1766        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1767        ecode += 1 + LINK_SIZE;
1768        break;
1769    
1770        /* The callout item calls an external function, if one is provided, passing
1771        details of the match so far. This is mainly for debugging, though the
1772        function is able to force a failure. */
1773    
1774        case OP_CALLOUT:
1775        if (PUBL(callout) != NULL)
1776          {
1777          PUBL(callout_block) cb;
1778          cb.version          = 2;   /* Version 1 of the callout block */
1779          cb.callout_number   = ecode[1];
1780          cb.offset_vector    = md->offset_vector;
1781    #if defined COMPILE_PCRE8
1782          cb.subject          = (PCRE_SPTR)md->start_subject;
1783    #elif defined COMPILE_PCRE16
1784          cb.subject          = (PCRE_SPTR16)md->start_subject;
1785    #elif defined COMPILE_PCRE32
1786          cb.subject          = (PCRE_SPTR32)md->start_subject;
1787    #endif
1788          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1789          cb.start_match      = (int)(mstart - md->start_subject);
1790          cb.current_position = (int)(eptr - md->start_subject);
1791        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1792        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1793        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1794        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last & CAPLMASK;
1795          /* Internal change requires this for API compatibility. */
1796          if (cb.capture_last == 0) cb.capture_last = -1;
1797        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1798        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1799          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1800        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1801        }        }
1802      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 847  for (;;) Line 1806  for (;;)
1806      offset data is the offset to the starting bracket from the start of the      offset data is the offset to the starting bracket from the start of the
1807      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1808    
1809      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1810      save their starting points and reinstate them after the recursion. However,      re-instated afterwards. We don't know how many are started and not yet
1811      we don't know how many such there are (offset_top records the completed      finished (offset_top records the completed total) so we just have to save
1812      total) so we just have to save all the potential data. There may be up to      all the potential data. There may be up to 65535 such values, which is too
1813      65535 such values, which is too large to put on the stack, but using malloc      large to put on the stack, but using malloc for small numbers seems
1814      for small numbers seems expensive. As a compromise, the stack is used when      expensive. As a compromise, the stack is used when there are no more than
1815      there are no more than REC_STACK_SAVE_MAX values to store; otherwise malloc      REC_STACK_SAVE_MAX values to store; otherwise malloc is used.
     is used. A problem is what to do if the malloc fails ... there is no way of  
     returning to the top level with an error. Save the top REC_STACK_SAVE_MAX  
     values on the stack, and accept that the rest may be wrong.  
1816    
1817      There are also other values that have to be saved. We use a chained      There are also other values that have to be saved. We use a chained
1818      sequence of blocks that actually live on the stack. Thanks to Robin Houston      sequence of blocks that actually live on the stack. Thanks to Robin Houston
1819      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1820        a lot, so he is not to blame for the current way it works. */
1821    
1822      case OP_RECURSE:      case OP_RECURSE:
1823        {        {
1824        callpat = md->start_code + GET(ecode, 1);        recursion_info *ri;
1825        new_recursive.group_num = *callpat - OP_BRA;        unsigned int recno;
1826    
1827        /* For extended extraction brackets (large number), we have to fish out        callpat = md->start_code + GET(ecode, 1);
1828        the number from a dummy opcode at the start. */        recno = (callpat == md->start_code)? 0 :
1829            GET2(callpat, 1 + LINK_SIZE);
1830    
1831        if (new_recursive.group_num > EXTRACT_BASIC_MAX)        /* Check for repeating a recursion without advancing the subject pointer.
1832          new_recursive.group_num = GET2(callpat, 2+LINK_SIZE);        This should catch convoluted mutual recursions. (Some simple cases are
1833          caught at compile time.) */
1834    
1835          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1836            if (recno == ri->group_num && eptr == ri->subject_position)
1837              RRETURN(PCRE_ERROR_RECURSELOOP);
1838    
1839        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1840    
1841          new_recursive.group_num = recno;
1842          new_recursive.saved_capture_last = md->capture_last;
1843          new_recursive.subject_position = eptr;
1844        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1845        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1846    
1847        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1848    
1849        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1850    
1851        /* Now save the offset data. */        /* Now save the offset data */
1852    
1853        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1854        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 891  for (;;) Line 1856  for (;;)
1856        else        else
1857          {          {
1858          new_recursive.offset_save =          new_recursive.offset_save =
1859            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1860          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1861          }          }
   
1862        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1863              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = md->start_match;  
       md->start_match = eptr;  
1864    
1865        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1866        restore the offset and recursion data. */        restore the offset data and the last captured value. If there were nested
1867          recursions, md->recursive might be changed, so reset it before looping.
1868          */
1869    
1870        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1871          cbegroup = (*callpat >= OP_SBRA);
1872        do        do
1873          {          {
1874          RMATCH(rrc, eptr, callpat + 1 + LINK_SIZE, offset_top, md, ims,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1875              eptrb, match_isgroup);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1876          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1877            memcpy(md->offset_vector, new_recursive.offset_save,
1878                new_recursive.saved_max * sizeof(int));
1879            md->capture_last = new_recursive.saved_capture_last;
1880            md->recursive = new_recursive.prevrec;
1881            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1882            {            {
1883            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1884            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1885              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1886            RRETURN(MATCH_MATCH);  
1887            }            /* Set where we got to in the subject, and reset the start in case
1888          else if (rrc != MATCH_NOMATCH)            it was changed by \K. This *is* propagated back out of a recursion,
1889              for Perl compatibility. */
1890    
1891              eptr = md->end_match_ptr;
1892              mstart = md->start_match_ptr;
1893              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1894              }
1895    
1896            /* PCRE does not allow THEN, SKIP, PRUNE or COMMIT to escape beyond a
1897            recursion; they cause a NOMATCH for the entire recursion. These codes
1898            are defined in a range that can be tested for. */
1899    
1900            if (rrc >= MATCH_BACKTRACK_MIN && rrc <= MATCH_BACKTRACK_MAX)
1901              RRETURN(MATCH_NOMATCH);
1902    
1903            /* Any return code other than NOMATCH is an error. */
1904    
1905            if (rrc != MATCH_NOMATCH)
1906            {            {
1907            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1908              if (new_recursive.offset_save != stacksave)
1909                (PUBL(free))(new_recursive.offset_save);
1910            RRETURN(rrc);            RRETURN(rrc);
1911            }            }
1912    
1913          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1914          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1915          }          }
1916        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 932  for (;;) Line 1918  for (;;)
1918        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1919        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1920        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1921          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1922        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1923        }        }
     /* Control never reaches here */  
   
     /* "Once" brackets are like assertion brackets except that after a match,  
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
   
     case OP_ONCE:  
       {  
       prev = ecode;  
       saved_eptr = eptr;  
   
       do  
         {  
         RMATCH(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);  
1924    
1925        /* If hit the end of the group (which could be repeated), fail */      RECURSION_MATCHED:
1926        break;
       if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
   
       /* Continue as from after the assertion, updating the offsets high water  
       mark, since extracts may have been taken. */  
   
       do ecode += GET(ecode,1); while (*ecode == OP_ALT);  
   
       offset_top = md->end_offset_top;  
       eptr = md->end_match_ptr;  
   
       /* For a non-repeating ket, just continue at this level. This also  
       happens for a repeating ket if no characters were matched in the group.  
       This is the forcible breaking of infinite loops as implemented in Perl  
       5.005. If there is an options reset, it will get obeyed in the normal  
       course of events. */  
   
       if (*ecode == OP_KET || eptr == saved_eptr)  
         {  
         ecode += 1+LINK_SIZE;  
         break;  
         }  
   
       /* The repeating kets try the rest of the pattern or restart from the  
       preceding bracket, in the appropriate order. 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);  
1927    
1928      /* An alternation is the end of a branch; scan along to find the end of the      /* An alternation is the end of a branch; scan along to find the end of the
1929      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1018  for (;;) Line 1932  for (;;)
1932      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1933      break;      break;
1934    
1935      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1936      that it may occur zero times. It may repeat infinitely, or not at all -      indicating that it may occur zero times. It may repeat infinitely, or not
1937      i.e. it could be ()* or ()? in the pattern. Brackets with fixed upper      at all - i.e. it could be ()* or ()? or even (){0} in the pattern. Brackets
1938      repeat limits are compiled as a number of copies, with the optional ones      with fixed upper repeat limits are compiled as a number of copies, with the
1939      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1940    
1941      case OP_BRAZERO:      case OP_BRAZERO:
1942        {      next = ecode + 1;
1943        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1944        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, match_isgroup);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1945        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1946        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1+LINK_SIZE;  
       }  
1947      break;      break;
1948    
1949      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1950        {      next = ecode + 1;
1951        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1952        do next += GET(next,1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1953        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb,      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1954          match_isgroup);      ecode++;
1955        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      break;
1956        ecode++;  
1957        }      case OP_SKIPZERO:
1958        next = ecode+1;
1959        do next += GET(next,1); while (*next == OP_ALT);
1960        ecode = next + 1 + LINK_SIZE;
1961      break;      break;
1962    
1963      /* 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
1964      an assertion "group", stop matching and return MATCH_MATCH, but record the      here; just jump to the group, with allow_zero set TRUE. */
1965      current high water mark for use by positive assertions. Do this also  
1966      for the "once" (not-backup up) groups. */      case OP_BRAPOSZERO:
1967        op = *(++ecode);
1968        allow_zero = TRUE;
1969        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1970          goto POSSESSIVE_NON_CAPTURE;
1971    
1972        /* End of a group, repeated or non-repeating. */
1973    
1974      case OP_KET:      case OP_KET:
1975      case OP_KETRMIN:      case OP_KETRMIN:
1976      case OP_KETRMAX:      case OP_KETRMAX:
1977        {      case OP_KETRPOS:
1978        prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
       saved_eptr = eptrb->epb_saved_eptr;  
1979    
1980        /* Back up the stack of bracket start pointers. */      /* If this was a group that remembered the subject start, in order to break
1981        infinite repeats of empty string matches, retrieve the subject start from
1982        the chain. Otherwise, set it NULL. */
1983    
1984        eptrb = eptrb->epb_prev;      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1985          {
1986        if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1987            *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1988            *prev == OP_ONCE)        }
1989          {      else saved_eptr = NULL;
         md->end_match_ptr = eptr;      /* For ONCE */  
         md->end_offset_top = offset_top;  
         RRETURN(MATCH_MATCH);  
         }  
1990    
1991        /* 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
1992        group number back at the start and if necessary complete handling an      group, stop matching and return MATCH_MATCH, but record the current high
1993        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
1994        start in case it was changed by \K. */
1995    
1996        if (*prev != OP_COND)      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1997          {           *prev == OP_ONCE_NC)
1998          number = *prev - OP_BRA;        {
1999          md->end_match_ptr = eptr;      /* For ONCE_NC */
2000          md->end_offset_top = offset_top;
2001          md->start_match_ptr = mstart;
2002          RRETURN(MATCH_MATCH);         /* Sets md->mark */
2003          }
2004    
2005          /* 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
2006          the number from a dummy opcode at the start. */      and if necessary complete handling an extraction by setting the offsets and
2007        bumping the high water mark. Whole-pattern recursion is coded as a recurse
2008        into group 0, so it won't be picked up here. Instead, we catch it when the
2009        OP_END is reached. Other recursion is handled here. We just have to record
2010        the current subject position and start match pointer and give a MATCH
2011        return. */
2012    
2013          if (number > EXTRACT_BASIC_MAX) number = GET2(prev, 2+LINK_SIZE);      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
2014          offset = number << 1;          *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
2015          {
2016          number = GET2(prev, 1+LINK_SIZE);
2017          offset = number << 1;
2018    
2019  #ifdef DEBUG  #ifdef PCRE_DEBUG
2020          printf("end bracket %d", number);        printf("end bracket %d", number);
2021          printf("\n");        printf("\n");
2022  #endif  #endif
2023    
2024          /* 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. */  
2025    
2026          if (number > 0)        if (md->recursive != NULL && md->recursive->group_num == number)
2027            {          {
2028            md->capture_last = number;          md->end_match_ptr = eptr;
2029            if (offset >= md->offset_max) md->offset_overflow = TRUE; else          md->start_match_ptr = mstart;
2030              {          RRETURN(MATCH_MATCH);
2031              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;  
             }  
2032    
2033            /* Handle a recursively called group. Restore the offsets        /* Deal with capturing */
           appropriately and continue from after the call. */  
2034    
2035            if (md->recursive != NULL && md->recursive->group_num == number)        md->capture_last = (md->capture_last & OVFLMASK) | number;
2036              {        if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
2037              recursion_info *rec = md->recursive;          {
2038              DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          /* If offset is greater than offset_top, it means that we are
2039              md->recursive = rec->prevrec;          "skipping" a capturing group, and that group's offsets must be marked
2040              md->start_match = rec->save_start;          unset. In earlier versions of PCRE, all the offsets were unset at the
2041              memcpy(md->offset_vector, rec->offset_save,          start of matching, but this doesn't work because atomic groups and
2042                rec->saved_max * sizeof(int));          assertions can cause a value to be set that should later be unset.
2043              ecode = rec->after_call;          Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
2044              ims = original_ims;          part of the atomic group, but this is not on the final matching path,
2045              break;          so must be unset when 2 is set. (If there is no group 2, there is no
2046              }          problem, because offset_top will then be 2, indicating no capture.) */
2047    
2048            if (offset > offset_top)
2049              {
2050              register int *iptr = md->offset_vector + offset_top;
2051              register int *iend = md->offset_vector + offset;
2052              while (iptr < iend) *iptr++ = -1;
2053            }            }
         }  
2054    
2055        /* Reset the value of the ims flags, in case they got changed during          /* Now make the extraction */
       the group. */  
2056    
2057        ims = original_ims;          md->offset_vector[offset] =
2058        DPRINTF(("ims reset to %02lx\n", ims));            md->offset_vector[md->offset_end - number];
2059            md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
2060            if (offset_top <= offset) offset_top = offset + 2;
2061            }
2062          }
2063    
2064        /* For a non-repeating ket, just continue at this level. This also      /* For an ordinary non-repeating ket, just continue at this level. This
2065        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
2066        This is the forcible breaking of infinite loops as implemented in Perl      group. This is the forcible breaking of infinite loops as implemented in
2067        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,
2068        course of events. */      establish a backup point by processing the rest of the pattern at a lower
2069        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
2070        original OP_ONCE level, thereby bypassing intermediate backup points, but
2071        resetting any captures that happened along the way. */
2072    
2073        if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
2074          {
2075          if (*prev == OP_ONCE)
2076          {          {
2077          ecode += 1 + LINK_SIZE;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
2078          break;          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2079            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2080            RRETURN(MATCH_ONCE);
2081          }          }
2082          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
2083          break;
2084          }
2085    
2086        /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
2087        preceding bracket, in the appropriate order. */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
2088        at a time from the outer level, thus saving stack. */
2089    
2090        if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRPOS)
2091          {
2092          md->end_match_ptr = eptr;
2093          md->end_offset_top = offset_top;
2094          RRETURN(MATCH_KETRPOS);
2095          }
2096    
2097        /* The normal repeating kets try the rest of the pattern or restart from
2098        the preceding bracket, in the appropriate order. In the second case, we can
2099        use tail recursion to avoid using another stack frame, unless we have an
2100        an atomic group or an unlimited repeat of a group that can match an empty
2101        string. */
2102    
2103        if (*ecode == OP_KETRMIN)
2104          {
2105          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2106          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2107          if (*prev == OP_ONCE)
2108          {          {
2109          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);  
2110          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2111            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2112            RRETURN(MATCH_ONCE);
2113          }          }
2114        else  /* OP_KETRMAX */        if (*prev >= OP_SBRA)    /* Could match an empty string */
2115          {          {
2116          RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2117          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          RRETURN(rrc);
2118          RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);          }
2119          ecode = prev;
2120          goto TAIL_RECURSE;
2121          }
2122        else  /* OP_KETRMAX */
2123          {
2124          RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2125          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2126          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2127          if (*prev == OP_ONCE)
2128            {
2129            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2130          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2131            md->once_target = prev;
2132            RRETURN(MATCH_ONCE);
2133          }          }
2134          ecode += 1 + LINK_SIZE;
2135          goto TAIL_RECURSE;
2136        }        }
2137        /* Control never gets here */
2138    
2139      RRETURN(MATCH_NOMATCH);      /* Not multiline mode: start of subject assertion, unless notbol. */
   
     /* Start of subject unless notbol, or after internal newline if multiline */  
2140    
2141      case OP_CIRC:      case OP_CIRC:
2142      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
     if ((ims & PCRE_MULTILINE) != 0)  
       {  
       if (eptr != md->start_subject && eptr[-1] != NEWLINE)  
         RRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
2143    
2144      /* Start of subject assertion */      /* Start of subject assertion */
2145    
# Line 1182  for (;;) Line 2148  for (;;)
2148      ecode++;      ecode++;
2149      break;      break;
2150    
2151        /* Multiline mode: start of subject unless notbol, or after any newline. */
2152    
2153        case OP_CIRCM:
2154        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2155        if (eptr != md->start_subject &&
2156            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2157          RRETURN(MATCH_NOMATCH);
2158        ecode++;
2159        break;
2160    
2161      /* Start of match assertion */      /* Start of match assertion */
2162    
2163      case OP_SOM:      case OP_SOM:
# Line 1189  for (;;) Line 2165  for (;;)
2165      ecode++;      ecode++;
2166      break;      break;
2167    
2168      /* 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. */  
2169    
2170      case OP_DOLL:      case OP_SET_SOM:
2171      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
2172        ecode++;
2173        break;
2174    
2175        /* Multiline mode: assert before any newline, or before end of subject
2176        unless noteol is set. */
2177    
2178        case OP_DOLLM:
2179        if (eptr < md->end_subject)
2180        {        {
2181        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2182          { if (*eptr != NEWLINE) RRETURN(MATCH_NOMATCH); }          {
2183        else          if (md->partial != 0 &&
2184          { if (md->noteol) RRETURN(MATCH_NOMATCH); }              eptr + 1 >= md->end_subject &&
2185        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2186        break;              NLBLOCK->nllen == 2 &&
2187                RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2188              {
2189              md->hitend = TRUE;
2190              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2191              }
2192            RRETURN(MATCH_NOMATCH);
2193            }
2194        }        }
2195      else      else
2196        {        {
2197        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2198        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr < md->end_subject - 1 ||  
            (eptr == md->end_subject - 1 && *eptr != NEWLINE))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2199        }        }
2200      /* ... else fall through */      ecode++;
2201        break;
2202    
2203        /* Not multiline mode: assert before a terminating newline or before end of
2204        subject unless noteol is set. */
2205    
2206        case OP_DOLL:
2207        if (md->noteol) RRETURN(MATCH_NOMATCH);
2208        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2209    
2210        /* ... else fall through for endonly */
2211    
2212      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2213    
2214      case OP_EOD:      case OP_EOD:
2215      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2216        SCHECK_PARTIAL();
2217      ecode++;      ecode++;
2218      break;      break;
2219    
2220      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2221    
2222      case OP_EODN:      case OP_EODN:
2223      if (eptr < md->end_subject - 1 ||      ASSERT_NL_OR_EOS:
2224         (eptr == md->end_subject - 1 && *eptr != NEWLINE)) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject &&
2225            (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2226          {
2227          if (md->partial != 0 &&
2228              eptr + 1 >= md->end_subject &&
2229              NLBLOCK->nltype == NLTYPE_FIXED &&
2230              NLBLOCK->nllen == 2 &&
2231              RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2232            {
2233            md->hitend = TRUE;
2234            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2235            }
2236          RRETURN(MATCH_NOMATCH);
2237          }
2238    
2239        /* Either at end of string or \n before end. */
2240    
2241        SCHECK_PARTIAL();
2242      ecode++;      ecode++;
2243      break;      break;
2244    
# Line 1237  for (;;) Line 2248  for (;;)
2248      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
2249        {        {
2250    
2251        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2252        It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to        It takes a bit more work in UTF-8 mode. Characters > 255 are assumed to
2253        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2254          partial matching. */
2255    
2256    #ifdef SUPPORT_UTF
2257          if (utf)
2258            {
2259            /* Get status of previous character */
2260    
2261            if (eptr == md->start_subject) prev_is_word = FALSE; else
2262              {
2263              PCRE_PUCHAR lastptr = eptr - 1;
2264              BACKCHAR(lastptr);
2265              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2266              GETCHAR(c, lastptr);
2267    #ifdef SUPPORT_UCP
2268              if (md->use_ucp)
2269                {
2270                if (c == '_') prev_is_word = TRUE; else
2271                  {
2272                  int cat = UCD_CATEGORY(c);
2273                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2274                  }
2275                }
2276              else
2277    #endif
2278              prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2279              }
2280    
2281            /* Get status of next character */
2282    
2283            if (eptr >= md->end_subject)
2284              {
2285              SCHECK_PARTIAL();
2286              cur_is_word = FALSE;
2287              }
2288            else
2289              {
2290              GETCHAR(c, eptr);
2291    #ifdef SUPPORT_UCP
2292              if (md->use_ucp)
2293                {
2294                if (c == '_') cur_is_word = TRUE; else
2295                  {
2296                  int cat = UCD_CATEGORY(c);
2297                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2298                  }
2299                }
2300              else
2301    #endif
2302              cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2303              }
2304            }
2305          else
2306    #endif
2307    
2308          /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2309          consistency with the behaviour of \w we do use it in this case. */
2310    
2311            {
2312            /* Get status of previous character */
2313    
2314            if (eptr == md->start_subject) prev_is_word = FALSE; else
2315              {
2316              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2317    #ifdef SUPPORT_UCP
2318              if (md->use_ucp)
2319                {
2320                c = eptr[-1];
2321                if (c == '_') prev_is_word = TRUE; else
2322                  {
2323                  int cat = UCD_CATEGORY(c);
2324                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2325                  }
2326                }
2327              else
2328    #endif
2329              prev_is_word = MAX_255(eptr[-1])
2330                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2331              }
2332    
2333            /* Get status of next character */
2334    
2335  #ifdef SUPPORT_UTF8          if (eptr >= md->end_subject)
       if (utf8)  
         {  
         if (eptr == md->start_subject) prev_is_word = FALSE; else  
2336            {            {
2337            const uschar *lastptr = eptr - 1;            SCHECK_PARTIAL();
2338            while((*lastptr & 0xc0) == 0x80) lastptr--;            cur_is_word = FALSE;
           GETCHAR(c, lastptr);  
           prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;  
2339            }            }
2340          if (eptr >= md->end_subject) cur_is_word = FALSE; else          else
2341    #ifdef SUPPORT_UCP
2342            if (md->use_ucp)
2343            {            {
2344            GETCHAR(c, eptr);            c = *eptr;
2345            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            if (c == '_') cur_is_word = TRUE; else
2346                {
2347                int cat = UCD_CATEGORY(c);
2348                cur_is_word = (cat == ucp_L || cat == ucp_N);
2349                }
2350            }            }
2351          }          else
       else  
2352  #endif  #endif
2353            cur_is_word = MAX_255(*eptr)
2354        /* More streamlined when not in UTF-8 mode */            && ((md->ctypes[*eptr] & ctype_word) != 0);
   
         {  
         prev_is_word = (eptr != md->start_subject) &&  
           ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
         cur_is_word = (eptr < md->end_subject) &&  
           ((md->ctypes[*eptr] & ctype_word) != 0);  
2355          }          }
2356    
2357        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1277  for (;;) Line 2362  for (;;)
2362        }        }
2363      break;      break;
2364    
2365      /* Match a single character type; inline for speed */      /* Match any single character type except newline; have to take care with
2366        CRLF newlines and partial matching. */
2367    
2368      case OP_ANY:      case OP_ANY:
2369      if ((ims & PCRE_DOTALL) == 0 && eptr < md->end_subject && *eptr == NEWLINE)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2370        if (md->partial != 0 &&
2371            eptr + 1 >= md->end_subject &&
2372            NLBLOCK->nltype == NLTYPE_FIXED &&
2373            NLBLOCK->nllen == 2 &&
2374            RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2375          {
2376          md->hitend = TRUE;
2377          if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2378          }
2379    
2380        /* Fall through */
2381    
2382        /* Match any single character whatsoever. */
2383    
2384        case OP_ALLANY:
2385        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2386          {                            /* not be updated before SCHECK_PARTIAL. */
2387          SCHECK_PARTIAL();
2388        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2389      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);        }
2390  #ifdef SUPPORT_UTF8      eptr++;
2391      if (utf8)  #ifdef SUPPORT_UTF
2392        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2393  #endif  #endif
2394      ecode++;      ecode++;
2395      break;      break;
# Line 1294  for (;;) Line 2398  for (;;)
2398      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2399    
2400      case OP_ANYBYTE:      case OP_ANYBYTE:
2401      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2402          {                            /* not be updated before SCHECK_PARTIAL. */
2403          SCHECK_PARTIAL();
2404          RRETURN(MATCH_NOMATCH);
2405          }
2406        eptr++;
2407      ecode++;      ecode++;
2408      break;      break;
2409    
2410      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2411      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2412          {
2413          SCHECK_PARTIAL();
2414          RRETURN(MATCH_NOMATCH);
2415          }
2416      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2417      if (      if (
2418  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2419         c < 256 &&         c < 256 &&
2420  #endif  #endif
2421         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1312  for (;;) Line 2425  for (;;)
2425      break;      break;
2426    
2427      case OP_DIGIT:      case OP_DIGIT:
2428      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2429          {
2430          SCHECK_PARTIAL();
2431          RRETURN(MATCH_NOMATCH);
2432          }
2433      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2434      if (      if (
2435  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2436         c >= 256 ||         c > 255 ||
2437  #endif  #endif
2438         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2439         )         )
# Line 1325  for (;;) Line 2442  for (;;)
2442      break;      break;
2443    
2444      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2445      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2446          {
2447          SCHECK_PARTIAL();
2448          RRETURN(MATCH_NOMATCH);
2449          }
2450      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2451      if (      if (
2452  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2453         c < 256 &&         c < 256 &&
2454  #endif  #endif
2455         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1338  for (;;) Line 2459  for (;;)
2459      break;      break;
2460    
2461      case OP_WHITESPACE:      case OP_WHITESPACE:
2462      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2463          {
2464          SCHECK_PARTIAL();
2465          RRETURN(MATCH_NOMATCH);
2466          }
2467      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2468      if (      if (
2469  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2470         c >= 256 ||         c > 255 ||
2471  #endif  #endif
2472         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2473         )         )
# Line 1351  for (;;) Line 2476  for (;;)
2476      break;      break;
2477    
2478      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2479      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2480          {
2481          SCHECK_PARTIAL();
2482          RRETURN(MATCH_NOMATCH);
2483          }
2484      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2485      if (      if (
2486  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2487         c < 256 &&         c < 256 &&
2488  #endif  #endif
2489         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1364  for (;;) Line 2493  for (;;)
2493      break;      break;
2494    
2495      case OP_WORDCHAR:      case OP_WORDCHAR:
2496      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2497          {
2498          SCHECK_PARTIAL();
2499          RRETURN(MATCH_NOMATCH);
2500          }
2501      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2502      if (      if (
2503  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2504         c >= 256 ||         c > 255 ||
2505  #endif  #endif
2506         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2507         )         )
# Line 1376  for (;;) Line 2509  for (;;)
2509      ecode++;      ecode++;
2510      break;      break;
2511    
2512        case OP_ANYNL:
2513        if (eptr >= md->end_subject)
2514          {
2515          SCHECK_PARTIAL();
2516          RRETURN(MATCH_NOMATCH);
2517          }
2518        GETCHARINCTEST(c, eptr);
2519        switch(c)
2520          {
2521          default: RRETURN(MATCH_NOMATCH);
2522    
2523          case CHAR_CR:
2524          if (eptr >= md->end_subject)
2525            {
2526            SCHECK_PARTIAL();
2527            }
2528          else if (RAWUCHARTEST(eptr) == CHAR_LF) eptr++;
2529          break;
2530    
2531          case CHAR_LF:
2532          break;
2533    
2534          case CHAR_VT:
2535          case CHAR_FF:
2536          case CHAR_NEL:
2537    #ifndef EBCDIC
2538          case 0x2028:
2539          case 0x2029:
2540    #endif  /* Not EBCDIC */
2541          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2542          break;
2543          }
2544        ecode++;
2545        break;
2546    
2547        case OP_NOT_HSPACE:
2548        if (eptr >= md->end_subject)
2549          {
2550          SCHECK_PARTIAL();
2551          RRETURN(MATCH_NOMATCH);
2552          }
2553        GETCHARINCTEST(c, eptr);
2554        switch(c)
2555          {
2556          HSPACE_CASES: RRETURN(MATCH_NOMATCH);  /* Byte and multibyte cases */
2557          default: break;
2558          }
2559        ecode++;
2560        break;
2561    
2562        case OP_HSPACE:
2563        if (eptr >= md->end_subject)
2564          {
2565          SCHECK_PARTIAL();
2566          RRETURN(MATCH_NOMATCH);
2567          }
2568        GETCHARINCTEST(c, eptr);
2569        switch(c)
2570          {
2571          HSPACE_CASES: break;  /* Byte and multibyte cases */
2572          default: RRETURN(MATCH_NOMATCH);
2573          }
2574        ecode++;
2575        break;
2576    
2577        case OP_NOT_VSPACE:
2578        if (eptr >= md->end_subject)
2579          {
2580          SCHECK_PARTIAL();
2581          RRETURN(MATCH_NOMATCH);
2582          }
2583        GETCHARINCTEST(c, eptr);
2584        switch(c)
2585          {
2586          VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2587          default: break;
2588          }
2589        ecode++;
2590        break;
2591    
2592        case OP_VSPACE:
2593        if (eptr >= md->end_subject)
2594          {
2595          SCHECK_PARTIAL();
2596          RRETURN(MATCH_NOMATCH);
2597          }
2598        GETCHARINCTEST(c, eptr);
2599        switch(c)
2600          {
2601          VSPACE_CASES: break;
2602          default: RRETURN(MATCH_NOMATCH);
2603          }
2604        ecode++;
2605        break;
2606    
2607  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2608      /* Check the next character by Unicode property. We will get here only      /* Check the next character by Unicode property. We will get here only
2609      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. */
2610    
2611      case OP_PROP:      case OP_PROP:
2612      case OP_NOTPROP:      case OP_NOTPROP:
2613      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2614          {
2615          SCHECK_PARTIAL();
2616          RRETURN(MATCH_NOMATCH);
2617          }
2618      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2619        {        {
2620        int chartype, script;        const pcre_uint32 *cp;
2621        int category = _pcre_ucp_findprop(c, &chartype, &script);        const ucd_record *prop = GET_UCD(c);
2622    
2623        switch(ecode[1])        switch(ecode[1])
2624          {          {
# Line 1395  for (;;) Line 2627  for (;;)
2627          break;          break;
2628    
2629          case PT_LAMP:          case PT_LAMP:
2630          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2631               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2632               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2633            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2634           break;          break;
2635    
2636          case PT_GC:          case PT_GC:
2637          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2638            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2639          break;          break;
2640    
2641          case PT_PC:          case PT_PC:
2642          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2643            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2644          break;          break;
2645    
2646          case PT_SC:          case PT_SC:
2647          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2648              RRETURN(MATCH_NOMATCH);
2649            break;
2650    
2651            /* These are specials */
2652    
2653            case PT_ALNUM:
2654            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2655                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2656              RRETURN(MATCH_NOMATCH);
2657            break;
2658    
2659            case PT_SPACE:    /* Perl space */
2660            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2661                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2662                   == (op == OP_NOTPROP))
2663              RRETURN(MATCH_NOMATCH);
2664            break;
2665    
2666            case PT_PXSPACE:  /* POSIX space */
2667            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2668                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2669                 c == CHAR_FF || c == CHAR_CR)
2670                   == (op == OP_NOTPROP))
2671              RRETURN(MATCH_NOMATCH);
2672            break;
2673    
2674            case PT_WORD:
2675            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2676                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2677                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2678              RRETURN(MATCH_NOMATCH);
2679            break;
2680    
2681            case PT_CLIST:
2682            cp = PRIV(ucd_caseless_sets) + ecode[2];
2683            for (;;)
2684              {
2685              if (c < *cp)
2686                { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; }
2687              if (c == *cp++)
2688                { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } }
2689              }
2690            break;
2691    
2692            case PT_UCNC:
2693            if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
2694                 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
2695                 c >= 0xe000) == (op == OP_NOTPROP))
2696            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2697          break;          break;
2698    
2699            /* This should never occur */
2700    
2701          default:          default:
2702          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
         break;  
2703          }          }
2704    
2705        ecode += 3;        ecode += 3;
# Line 1429  for (;;) Line 2710  for (;;)
2710      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2711    
2712      case OP_EXTUNI:      case OP_EXTUNI:
2713      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2714      GETCHARINCTEST(c, eptr);        {
2715          SCHECK_PARTIAL();
2716          RRETURN(MATCH_NOMATCH);
2717          }
2718        else
2719        {        {
2720        int chartype, script;        int lgb, rgb;
2721        int category = _pcre_ucp_findprop(c, &chartype, &script);        GETCHARINCTEST(c, eptr);
2722        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        lgb = UCD_GRAPHBREAK(c);
2723        while (eptr < md->end_subject)        while (eptr < md->end_subject)
2724          {          {
2725          int len = 1;          int len = 1;
2726          if (!utf8) c = *eptr; else          if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2727            {          rgb = UCD_GRAPHBREAK(c);
2728            GETCHARLEN(c, eptr, len);          if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2729            }          lgb = rgb;
         category = _pcre_ucp_findprop(c, &chartype, &script);  
         if (category != ucp_M) break;  
2730          eptr += len;          eptr += len;
2731          }          }
2732        }        }
2733        CHECK_PARTIAL();
2734      ecode++;      ecode++;
2735      break;      break;
2736  #endif  #endif  /* SUPPORT_UCP */
2737    
2738    
2739      /* Match a back reference, possibly repeatedly. Look past the end of the      /* Match a back reference, possibly repeatedly. Look past the end of the
# Line 1461  for (;;) Line 2745  for (;;)
2745      loops). */      loops). */
2746    
2747      case OP_REF:      case OP_REF:
2748        {      case OP_REFI:
2749        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2750        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2751        ecode += 1 + IMM2_SIZE;
2752    
2753        /* If the reference is unset, set the length to be longer than the amount      /* If the reference is unset, there are two possibilities:
       of subject left; this ensures that every attempt at a match fails. We  
       can't just fail here, because of the possibility of quantifiers with zero  
       minima. */  
   
       length = (offset >= offset_top || md->offset_vector[offset] < 0)?  
         md->end_subject - eptr + 1 :  
         md->offset_vector[offset+1] - md->offset_vector[offset];  
2754    
2755        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2756        this ensures that every attempt at a match fails. We can't just fail
2757        here, because of the possibility of quantifiers with zero minima.
2758    
2759        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2760          {      so that the back reference matches an empty string.
         case OP_CRSTAR:  
         case OP_CRMINSTAR:  
         case OP_CRPLUS:  
         case OP_CRMINPLUS:  
         case OP_CRQUERY:  
         case OP_CRMINQUERY:  
         c = *ecode++ - OP_CRSTAR;  
         minimize = (c & 1) != 0;  
         min = rep_min[c];                 /* Pick up values from tables; */  
         max = rep_max[c];                 /* zero for max => infinity */  
         if (max == 0) max = INT_MAX;  
         break;  
2761    
2762          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2763          case OP_CRMINRANGE:      referenced subpattern. */
         minimize = (*ecode == OP_CRMINRANGE);  
         min = GET2(ecode, 1);  
         max = GET2(ecode, 3);  
         if (max == 0) max = INT_MAX;  
         ecode += 5;  
         break;  
2764    
2765          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2766          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2767          eptr += length;      else
2768          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2769    
2770        /* Set up for repetition, or handle the non-repeated case */
2771    
2772        switch (*ecode)
2773          {
2774          case OP_CRSTAR:
2775          case OP_CRMINSTAR:
2776          case OP_CRPLUS:
2777          case OP_CRMINPLUS:
2778          case OP_CRQUERY:
2779          case OP_CRMINQUERY:
2780          c = *ecode++ - OP_CRSTAR;
2781          minimize = (c & 1) != 0;
2782          min = rep_min[c];                 /* Pick up values from tables; */
2783          max = rep_max[c];                 /* zero for max => infinity */
2784          if (max == 0) max = INT_MAX;
2785          break;
2786    
2787          case OP_CRRANGE:
2788          case OP_CRMINRANGE:
2789          minimize = (*ecode == OP_CRMINRANGE);
2790          min = GET2(ecode, 1);
2791          max = GET2(ecode, 1 + IMM2_SIZE);
2792          if (max == 0) max = INT_MAX;
2793          ecode += 1 + 2 * IMM2_SIZE;
2794          break;
2795    
2796          default:               /* No repeat follows */
2797          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2798            {
2799            if (length == -2) eptr = md->end_subject;   /* Partial match */
2800            CHECK_PARTIAL();
2801            RRETURN(MATCH_NOMATCH);
2802          }          }
2803          eptr += length;
2804          continue;              /* With the main loop */
2805          }
2806    
2807        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2808        main loop. */      zero, just continue with the main loop. If the length is negative, it
2809        means the reference is unset in non-Java-compatible mode. If the minimum is
2810        zero, we can continue at the same level without recursion. For any other
2811        minimum, carrying on will result in NOMATCH. */
2812    
2813        if (length == 0) continue;      if (length == 0) continue;
2814        if (length < 0 && min == 0) continue;
2815    
2816        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2817        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2818        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2819    
2820        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2821          {
2822          int slength;
2823          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2824          {          {
2825          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          if (slength == -2) eptr = md->end_subject;   /* Partial match */
2826          eptr += length;          CHECK_PARTIAL();
2827            RRETURN(MATCH_NOMATCH);
2828          }          }
2829          eptr += slength;
2830          }
2831    
2832        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2833        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2834    
2835        if (min == max) continue;      if (min == max) continue;
2836    
2837        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2838    
2839        if (minimize)      if (minimize)
2840          {
2841          for (fi = min;; fi++)
2842          {          {
2843          for (fi = min;; fi++)          int slength;
2844            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2845            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2846            if (fi >= max) RRETURN(MATCH_NOMATCH);
2847            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2848            {            {
2849            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2850            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2851            if (fi >= max || !match_ref(offset, eptr, length, md, ims))            RRETURN(MATCH_NOMATCH);
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2852            }            }
2853          /* Control never gets here */          eptr += slength;
2854          }          }
2855          /* Control never gets here */
2856          }
2857    
2858        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2859    
2860        else      else
2861          {
2862          pp = eptr;
2863          for (i = min; i < max; i++)
2864          {          {
2865          pp = eptr;          int slength;
2866          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)  
2867            {            {
2868            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2869            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            the soft partial matching case. */
2870            eptr -= length;  
2871              if (slength == -2 && md->partial != 0 &&
2872                  md->end_subject > md->start_used_ptr)
2873                {
2874                md->hitend = TRUE;
2875                if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2876                }
2877              break;
2878            }            }
2879          RRETURN(MATCH_NOMATCH);          eptr += slength;
2880            }
2881    
2882          while (eptr >= pp)
2883            {
2884            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2885            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2886            eptr -= length;
2887          }          }
2888          RRETURN(MATCH_NOMATCH);
2889        }        }
2890      /* Control never gets here */      /* Control never gets here */
2891    
   
   
2892      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2893      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,
2894      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 2903  for (;;)
2903      case OP_NCLASS:      case OP_NCLASS:
2904      case OP_CLASS:      case OP_CLASS:
2905        {        {
2906          /* The data variable is saved across frames, so the byte map needs to
2907          be stored there. */
2908    #define BYTE_MAP ((pcre_uint8 *)data)
2909        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2910        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2911    
2912        switch (*ecode)        switch (*ecode)
2913          {          {
# Line 1600  for (;;) Line 2928  for (;;)
2928          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2929          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2930          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2931          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2932          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2933          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2934          break;          break;
2935    
2936          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1612  for (;;) Line 2940  for (;;)
2940    
2941        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2942    
2943  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2944        /* UTF-8 mode */        if (utf)
       if (utf8)  
2945          {          {
2946          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2947            {            {
2948            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2949                {
2950                SCHECK_PARTIAL();
2951                RRETURN(MATCH_NOMATCH);
2952                }
2953            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2954            if (c > 255)            if (c > 255)
2955              {              {
2956              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2957              }              }
2958            else            else
2959              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2960            }            }
2961          }          }
2962        else        else
2963  #endif  #endif
2964        /* Not UTF-8 mode */        /* Not UTF mode */
2965          {          {
2966          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2967            {            {
2968            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2969                {
2970                SCHECK_PARTIAL();
2971                RRETURN(MATCH_NOMATCH);
2972                }
2973            c = *eptr++;            c = *eptr++;
2974            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2975              if (c > 255)
2976                {
2977                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2978                }
2979              else
2980    #endif
2981                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2982            }            }
2983          }          }
2984    
# Line 1652  for (;;) Line 2992  for (;;)
2992    
2993        if (minimize)        if (minimize)
2994          {          {
2995  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2996          /* UTF-8 mode */          if (utf)
         if (utf8)  
2997            {            {
2998            for (fi = min;; fi++)            for (fi = min;; fi++)
2999              {              {
3000              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
3001              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3002              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3003                if (eptr >= md->end_subject)
3004                  {
3005                  SCHECK_PARTIAL();
3006                  RRETURN(MATCH_NOMATCH);
3007                  }
3008              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
3009              if (c > 255)              if (c > 255)
3010                {                {
3011                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
3012                }                }
3013              else              else
3014                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
3015              }              }
3016            }            }
3017          else          else
3018  #endif  #endif
3019          /* Not UTF-8 mode */          /* Not UTF mode */
3020            {            {
3021            for (fi = min;; fi++)            for (fi = min;; fi++)
3022              {              {
3023              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
3024              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3025              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3026                if (eptr >= md->end_subject)
3027                  {
3028                  SCHECK_PARTIAL();
3029                  RRETURN(MATCH_NOMATCH);
3030                  }
3031              c = *eptr++;              c = *eptr++;
3032              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
3033                if (c > 255)
3034                  {
3035                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
3036                  }
3037                else
3038    #endif
3039                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
3040              }              }
3041            }            }
3042          /* Control never gets here */          /* Control never gets here */
# Line 1694  for (;;) Line 3048  for (;;)
3048          {          {
3049          pp = eptr;          pp = eptr;
3050    
3051  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3052          /* UTF-8 mode */          if (utf)
         if (utf8)  
3053            {            {
3054            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3055              {              {
3056              int len = 1;              int len = 1;
3057              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3058                  {
3059                  SCHECK_PARTIAL();
3060                  break;
3061                  }
3062              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
3063              if (c > 255)              if (c > 255)
3064                {                {
3065                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
3066                }                }
3067              else              else
3068                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
3069              eptr += len;              eptr += len;
3070              }              }
3071            for (;;)            for (;;)
3072              {              {
3073              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
3074              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3075              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3076              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1723  for (;;) Line 3078  for (;;)
3078            }            }
3079          else          else
3080  #endif  #endif
3081            /* Not UTF-8 mode */            /* Not UTF mode */
3082            {            {
3083            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3084              {              {
3085              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3086                  {
3087                  SCHECK_PARTIAL();
3088                  break;
3089                  }
3090              c = *eptr;              c = *eptr;
3091              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
3092                if (c > 255)
3093                  {
3094                  if (op == OP_CLASS) break;
3095                  }
3096                else
3097    #endif
3098                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3099              eptr++;              eptr++;
3100              }              }
3101            while (eptr >= pp)            while (eptr >= pp)
3102              {              {
3103              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3104              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3105              eptr--;              eptr--;
3106              }              }
# Line 1742  for (;;) Line 3108  for (;;)
3108    
3109          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3110          }          }
3111    #undef BYTE_MAP
3112        }        }
3113      /* Control never gets here */      /* Control never gets here */
3114    
3115    
3116      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
3117      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
3118        mode, because Unicode properties are supported in non-UTF-8 mode. */
3119    
3120  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3121      case OP_XCLASS:      case OP_XCLASS:
3122        {        {
3123        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 1774  for (;;) Line 3142  for (;;)
3142          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3143          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3144          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3145          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3146          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3147          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3148          break;          break;
3149    
3150          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1788  for (;;) Line 3156  for (;;)
3156    
3157        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3158          {          {
3159          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3160          GETCHARINC(c, eptr);            {
3161          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
3162              RRETURN(MATCH_NOMATCH);
3163              }
3164            GETCHARINCTEST(c, eptr);
3165            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3166          }          }
3167    
3168        /* 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 3177  for (;;)
3177          {          {
3178          for (fi = min;; fi++)          for (fi = min;; fi++)
3179            {            {
3180            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3181            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3182            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3183            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3184            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3185                SCHECK_PARTIAL();
3186                RRETURN(MATCH_NOMATCH);
3187                }
3188              GETCHARINCTEST(c, eptr);
3189              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3190            }            }
3191          /* Control never gets here */          /* Control never gets here */
3192          }          }
# Line 1822  for (;;) Line 3199  for (;;)
3199          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3200            {            {
3201            int len = 1;            int len = 1;
3202            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3203            GETCHARLEN(c, eptr, len);              {
3204            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3205                break;
3206                }
3207    #ifdef SUPPORT_UTF
3208              GETCHARLENTEST(c, eptr, len);
3209    #else
3210              c = *eptr;
3211    #endif
3212              if (!PRIV(xclass)(c, data, utf)) break;
3213            eptr += len;            eptr += len;
3214            }            }
3215          for(;;)          for(;;)
3216            {            {
3217            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3218            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3219            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3220            BACKCHAR(eptr)  #ifdef SUPPORT_UTF
3221              if (utf) BACKCHAR(eptr);
3222    #endif
3223            }            }
3224          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3225          }          }
# Line 1844  for (;;) Line 3231  for (;;)
3231      /* Match a single character, casefully */      /* Match a single character, casefully */
3232    
3233      case OP_CHAR:      case OP_CHAR:
3234  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3235      if (utf8)      if (utf)
3236        {        {
3237        length = 1;        length = 1;
3238        ecode++;        ecode++;
3239        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3240        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3241        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);          {
3242            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3243            RRETURN(MATCH_NOMATCH);
3244            }
3245          while (length-- > 0) if (*ecode++ != RAWUCHARINC(eptr)) RRETURN(MATCH_NOMATCH);
3246        }        }
3247      else      else
3248  #endif  #endif
3249        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3250        {        {
3251        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3252            {
3253            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3254            RRETURN(MATCH_NOMATCH);
3255            }
3256        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3257        ecode += 2;        ecode += 2;
3258        }        }
3259      break;      break;
3260    
3261      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3262        subject, give up immediately. */
3263    
3264        case OP_CHARI:
3265        if (eptr >= md->end_subject)
3266          {
3267          SCHECK_PARTIAL();
3268          RRETURN(MATCH_NOMATCH);
3269          }
3270    
3271      case OP_CHARNC:  #ifdef SUPPORT_UTF
3272  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3273        {        {
3274        length = 1;        length = 1;
3275        ecode++;        ecode++;
3276        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3277    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3278        /* 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
3279        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3280          fast lookup table. We know that there is at least one byte left in the
3281          subject. */
3282    
3283        if (fc < 128)        if (fc < 128)
3284          {          {
3285          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          pcre_uint32 cc = RAWUCHAR(eptr);
3286            if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH);
3287            ecode++;
3288            eptr++;
3289          }          }
3290    
3291        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3292          use the value of "length" to check for sufficient bytes left, because the
3293          other case of the character may have more or fewer bytes.  */
3294    
3295        else        else
3296          {          {
3297          int dc;          pcre_uint32 dc;
3298          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
3299          ecode += length;          ecode += length;
3300    
# Line 1898  for (;;) Line 3304  for (;;)
3304          if (fc != dc)          if (fc != dc)
3305            {            {
3306  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3307            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3308  #endif  #endif
3309              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3310            }            }
3311          }          }
3312        }        }
3313      else      else
3314  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3315    
3316      /* Non-UTF-8 mode */      /* Not UTF mode */
3317        {        {
3318        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3319        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3320          eptr++;
3321        ecode += 2;        ecode += 2;
3322        }        }
3323      break;      break;
3324    
3325      /* Match a single character repeatedly; different opcodes share code. */      /* Match a single character repeatedly. */
3326    
3327      case OP_EXACT:      case OP_EXACT:
3328        case OP_EXACTI:
3329      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3330      ecode += 3;      ecode += 1 + IMM2_SIZE;
3331      goto REPEATCHAR;      goto REPEATCHAR;
3332    
3333        case OP_POSUPTO:
3334        case OP_POSUPTOI:
3335        possessive = TRUE;
3336        /* Fall through */
3337    
3338      case OP_UPTO:      case OP_UPTO:
3339        case OP_UPTOI:
3340      case OP_MINUPTO:      case OP_MINUPTO:
3341        case OP_MINUPTOI:
3342      min = 0;      min = 0;
3343      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3344      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3345      ecode += 3;      ecode += 1 + IMM2_SIZE;
3346        goto REPEATCHAR;
3347    
3348        case OP_POSSTAR:
3349        case OP_POSSTARI:
3350        possessive = TRUE;
3351        min = 0;
3352        max = INT_MAX;
3353        ecode++;
3354        goto REPEATCHAR;
3355    
3356        case OP_POSPLUS:
3357        case OP_POSPLUSI:
3358        possessive = TRUE;
3359        min = 1;
3360        max = INT_MAX;
3361        ecode++;
3362        goto REPEATCHAR;
3363    
3364        case OP_POSQUERY:
3365        case OP_POSQUERYI:
3366        possessive = TRUE;
3367        min = 0;
3368        max = 1;
3369        ecode++;
3370      goto REPEATCHAR;      goto REPEATCHAR;
3371    
3372      case OP_STAR:      case OP_STAR:
3373        case OP_STARI:
3374      case OP_MINSTAR:      case OP_MINSTAR:
3375        case OP_MINSTARI:
3376      case OP_PLUS:      case OP_PLUS:
3377        case OP_PLUSI:
3378      case OP_MINPLUS:      case OP_MINPLUS:
3379        case OP_MINPLUSI:
3380      case OP_QUERY:      case OP_QUERY:
3381        case OP_QUERYI:
3382      case OP_MINQUERY:      case OP_MINQUERY:
3383      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3384        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3385      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3386      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3387      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3388      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3389    
3390      /* Common code for all repeated single-character matches. We can give      /* Common code for all repeated single-character matches. We first check
3391      up quickly if there are fewer than the minimum number of characters left in      for the minimum number of characters. If the minimum equals the maximum, we
3392      the subject. */      are done. Otherwise, if minimizing, check the rest of the pattern for a
3393        match; if there isn't one, advance up to the maximum, one character at a
3394        time.
3395    
3396        If maximizing, advance up to the maximum number of matching characters,
3397        until eptr is past the end of the maximum run. If possessive, we are
3398        then done (no backing up). Otherwise, match at this position; anything
3399        other than no match is immediately returned. For nomatch, back up one
3400        character, unless we are matching \R and the last thing matched was
3401        \r\n, in which case, back up two bytes. When we reach the first optional
3402        character position, we can save stack by doing a tail recurse.
3403    
3404        The various UTF/non-UTF and caseful/caseless cases are handled separately,
3405        for speed. */
3406    
3407      REPEATCHAR:      REPEATCHAR:
3408  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3409      if (utf8)      if (utf)
3410        {        {
3411        length = 1;        length = 1;
3412        charptr = ecode;        charptr = ecode;
3413        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3414        ecode += length;        ecode += length;
3415    
3416        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 1961  for (;;) Line 3418  for (;;)
3418    
3419        if (length > 1)        if (length > 1)
3420          {          {
         int oclength = 0;  
         uschar occhars[8];  
   
3421  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3422          int othercase;          pcre_uint32 othercase;
3423          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3424              (othercase = _pcre_ucp_othercase(fc)) >= 0 &&              (othercase = UCD_OTHERCASE(fc)) != fc)
3425               othercase >= 0)            oclength = PRIV(ord2utf)(othercase, occhars);
3426            oclength = _pcre_ord2utf8(othercase, occhars);          else oclength = 0;
3427  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3428    
3429          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3430            {            {
3431            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3432            /* Need braces because of following else */              memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3433            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3434              else if (oclength > 0 &&
3435                       eptr <= md->end_subject - oclength &&
3436                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3437    #endif  /* SUPPORT_UCP */
3438            else            else
3439              {              {
3440              if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);              CHECK_PARTIAL();
3441              eptr += oclength;              RRETURN(MATCH_NOMATCH);
3442              }              }
3443            }            }
3444    
# Line 1990  for (;;) Line 3448  for (;;)
3448            {            {
3449            for (fi = min;; fi++)            for (fi = min;; fi++)
3450              {              {
3451              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM22);
3452              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3453              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
3454              if (memcmp(eptr, charptr, length) == 0) eptr += length;              if (eptr <= md->end_subject - length &&
3455              /* Need braces because of following else */                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3456              else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }  #ifdef SUPPORT_UCP
3457                else if (oclength > 0 &&
3458                         eptr <= md->end_subject - oclength &&
3459                         memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3460    #endif  /* SUPPORT_UCP */
3461              else              else
3462                {                {
3463                if (memcmp(eptr, occhars, oclength) != 0) RRETURN(MATCH_NOMATCH);                CHECK_PARTIAL();
3464                eptr += oclength;                RRETURN(MATCH_NOMATCH);
3465                }                }
3466              }              }
3467            /* Control never gets here */            /* Control never gets here */
3468            }            }
3469          else  
3470            else  /* Maximize */
3471            {            {
3472            pp = eptr;            pp = eptr;
3473            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3474              {              {
3475              if (eptr > md->end_subject - length) break;              if (eptr <= md->end_subject - length &&
3476              if (memcmp(eptr, charptr, length) == 0) eptr += length;                  memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3477              else if (oclength == 0) break;  #ifdef SUPP