/[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 91 by nigel, Sat Feb 24 21:41:34 2007 UTC revision 1100 by chpe, Tue Oct 16 15:56:26 2012 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2006 University of Cambridge             Copyright (c) 1997-2012 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 37  POSSIBILITY OF SUCH DAMAGE. Line 37  POSSIBILITY OF SUCH DAMAGE.
37  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
38  */  */
39    
   
40  /* This module contains pcre_exec(), the externally visible function that does  /* This module contains pcre_exec(), the externally visible function that does
41  pattern matching using an NFA algorithm, trying to mimic Perl as closely as  pattern matching using an NFA algorithm, trying to mimic Perl as closely as
42  possible. There are also some static supporting functions. */  possible. There are also some static supporting functions. */
43    
44  #define NLBLOCK md           /* The block containing newline information */  #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 */  /* Values for setting in md->match_function_type to indicate two special types
60    of call to match(). We do it this way to save on using another stack variable,
61    as stack usage is to be discouraged. */
62    
63  #define match_condassert   0x01    /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
64  #define match_isgroup      0x02    /* Set if start of bracketed group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
65    
66  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
67  defined PCRE_ERROR_xxx codes, which are all negative. */  defined PCRE_ERROR_xxx codes, which are all negative. */
# Line 68  defined PCRE_ERROR_xxx codes, which are Line 69  defined PCRE_ERROR_xxx codes, which are
69  #define MATCH_MATCH        1  #define MATCH_MATCH        1
70  #define MATCH_NOMATCH      0  #define MATCH_NOMATCH      0
71    
72    /* Special internal returns from the match() function. Make them sufficiently
73    negative to avoid the external error codes. */
74    
75    #define MATCH_ACCEPT       (-999)
76    #define MATCH_COMMIT       (-998)
77    #define MATCH_KETRPOS      (-997)
78    #define MATCH_ONCE         (-996)
79    #define MATCH_PRUNE        (-995)
80    #define MATCH_SKIP         (-994)
81    #define MATCH_SKIP_ARG     (-993)
82    #define MATCH_THEN         (-992)
83    
84  /* Maximum number of ints of offset to save on the stack for recursive calls.  /* Maximum number of ints of offset to save on the stack for recursive calls.
85  If the offset vector is bigger, malloc is used. This should be a multiple of 3,  If the offset vector is bigger, malloc is used. This should be a multiple of 3,
86  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 92  because the offset vector is always a mu
92  static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };  static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };
93  static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };  static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };
94    
95    #ifdef PCRE_DEBUG
   
 #ifdef DEBUG  
96  /*************************************************  /*************************************************
97  *        Debugging function to print chars       *  *        Debugging function to print chars       *
98  *************************************************/  *************************************************/
# Line 99  Returns:     nothing Line 110  Returns:     nothing
110  */  */
111    
112  static void  static void
113  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)
114  {  {
115  int c;  pcre_uint32 c;
116    BOOL utf = md->utf;
117  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;
118  while (length-- > 0)  while (length-- > 0)
119    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);
120  }  }
121  #endif  #endif
122    
# Line 114  while (length-- > 0) Line 126  while (length-- > 0)
126  *          Match a back-reference                *  *          Match a back-reference                *
127  *************************************************/  *************************************************/
128    
129  /* 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
130  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
131    the length passed is zero. Note that in caseless UTF-8 mode, the number of
132    subject bytes matched may be different to the number of reference bytes.
133    
134  Arguments:  Arguments:
135    offset      index into the offset vector    offset      index into the offset vector
136    eptr        points into the subject    eptr        pointer into the subject
137    length      length to be matched    length      length of reference to be matched (number of bytes)
138    md          points to match data block    md          points to match data block
139    ims         the ims flags    caseless    TRUE if caseless
140    
141  Returns:      TRUE if matched  Returns:      >= 0 the number of subject bytes matched
142                  -1 no match
143                  -2 partial match; always given if at end subject
144  */  */
145    
146  static BOOL  static int
147  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,
148    unsigned long int ims)    BOOL caseless)
149  {  {
150  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
151    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
152    BOOL utf = md->utf;
153    
154  #ifdef DEBUG  #ifdef PCRE_DEBUG
155  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
156    printf("matching subject <null>");    printf("matching subject <null>");
157  else  else
# Line 146  pchars(p, length, FALSE, md); Line 164  pchars(p, length, FALSE, md);
164  printf("\n");  printf("\n");
165  #endif  #endif
166    
167  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible - in that
168    case the length is passed as zero). */
169    
170  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
171    
172  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
173    properly if Unicode properties are supported. Otherwise, we can check only
174    ASCII characters. */
175    
176  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
177    {    {
178    while (length-- > 0)  #ifdef SUPPORT_UTF
179      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
180      if (utf)
181        {
182        /* Match characters up to the end of the reference. NOTE: the number of
183        data units matched may differ, because in UTF-8 there are some characters
184        whose upper and lower case versions code have different numbers of bytes.
185        For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65
186        (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a
187        sequence of two of the latter. It is important, therefore, to check the
188        length along the reference, not along the subject (earlier code did this
189        wrong). */
190    
191        PCRE_PUCHAR endptr = p + length;
192        while (p < endptr)
193          {
194          pcre_uint32 c, d;
195          const ucd_record *ur;
196          if (eptr >= md->end_subject) return -2;   /* Partial match */
197          GETCHARINC(c, eptr);
198          GETCHARINC(d, p);
199          ur = GET_UCD(d);
200          if (c != d && c != d + ur->other_case)
201            {
202            const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
203            for (;;)
204              {
205              if (c < *pp) return -1;
206              if (c == *pp++) break;
207              }
208            }
209          }
210        }
211      else
212    #endif
213    #endif
214    
215      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
216      is no UCP support. */
217        {
218        while (length-- > 0)
219          {
220          pcre_uchar cc, cp;
221          if (eptr >= md->end_subject) return -2;   /* Partial match */
222          cc = RAWUCHARTEST(eptr);
223          cp = RAWUCHARTEST(p);
224          if (TABLE_GET(cp, md->lcc, cp) != TABLE_GET(cc, md->lcc, cc)) return -1;
225          p++;
226          eptr++;
227          }
228        }
229    }    }
230    
231    /* In the caseful case, we can just compare the bytes, whether or not we
232    are in UTF-8 mode. */
233    
234  else  else
235    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
236      while (length-- > 0)
237        {
238        if (eptr >= md->end_subject) return -2;   /* Partial match */
239        if (RAWUCHARINCTEST(p) != RAWUCHARINCTEST(eptr)) return -1;
240        }
241      }
242    
243  return TRUE;  return (int)(eptr - eptr_start);
244  }  }
245    
246    
# Line 186  calls by keeping local variables that ne Line 266  calls by keeping local variables that ne
266  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
267  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
268  always used to.  always used to.
269    
270    The original heap-recursive code used longjmp(). However, it seems that this
271    can be very slow on some operating systems. Following a suggestion from Stan
272    Switzer, the use of longjmp() has been abolished, at the cost of having to
273    provide a unique number for each call to RMATCH. There is no way of generating
274    a sequence of numbers at compile time in C. I have given them names, to make
275    them stand out more clearly.
276    
277    Crude tests on x86 Linux show a small speedup of around 5-8%. However, on
278    FreeBSD, avoiding longjmp() more than halves the time taken to run the standard
279    tests. Furthermore, not using longjmp() means that local dynamic variables
280    don't have indeterminate values; this has meant that the frame size can be
281    reduced because the result can be "passed back" by straight setting of the
282    variable instead of being passed in the frame.
283  ****************************************************************************  ****************************************************************************
284  ***************************************************************************/  ***************************************************************************/
285    
286    /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
287    below must be updated in sync.  */
288    
289    enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
290           RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
291           RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
292           RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
293           RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
294           RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
295           RM61,  RM62, RM63, RM64, RM65, RM66 };
296    
297  /* 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
298  versions and production versions. */  versions and production versions. Note that the "rw" argument of RMATCH isn't
299    actually used in this definition. */
300    
301  #ifndef NO_RECURSE  #ifndef NO_RECURSE
302  #define REGISTER register  #define REGISTER register
303  #ifdef DEBUG  
304  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #ifdef PCRE_DEBUG
305    #define RMATCH(ra,rb,rc,rd,re,rw) \
306    { \    { \
307    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
308    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
309    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
310    }    }
311  #define RRETURN(ra) \  #define RRETURN(ra) \
312    { \    { \
313    printf("match() returned %d from line %d ", ra, __LINE__); \    printf("match() returned %d from line %d\n", ra, __LINE__); \
314    return ra; \    return ra; \
315    }    }
316  #else  #else
317  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
318    rx = match(ra,rb,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
319  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
320  #endif  #endif
321    
322  #else  #else
323    
324    
325  /* 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
326  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
327  match(), which never changes. */  argument of match(), which never changes. */
328    
329  #define REGISTER  #define REGISTER
330    
331  #define RMATCH(rx,ra,rb,rc,rd,re,rf,rg)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
332    {\    {\
333    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = frame->Xnextframe;\
334    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\  
335      {\      {\
336      DPRINTF(("longjumped back to line %d\n", __LINE__));\      newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
337      frame = md->thisframe;\      if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
338      rx = frame->Xresult;\      newframe->Xnextframe = NULL;\
339        frame->Xnextframe = newframe;\
340      }\      }\
341      frame->Xwhere = rw;\
342      newframe->Xeptr = ra;\
343      newframe->Xecode = rb;\
344      newframe->Xmstart = mstart;\
345      newframe->Xoffset_top = rc;\
346      newframe->Xeptrb = re;\
347      newframe->Xrdepth = frame->Xrdepth + 1;\
348      newframe->Xprevframe = frame;\
349      frame = newframe;\
350      DPRINTF(("restarting from line %d\n", __LINE__));\
351      goto HEAP_RECURSE;\
352      L_##rw:\
353      DPRINTF(("jumped back to line %d\n", __LINE__));\
354    }    }
355    
356  #define RRETURN(ra)\  #define RRETURN(ra)\
357    {\    {\
358    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
359    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
   (pcre_stack_free)(newframe);\  
360    if (frame != NULL)\    if (frame != NULL)\
361      {\      {\
362      frame->Xresult = ra;\      rrc = ra;\
363      md->thisframe = frame;\      goto HEAP_RETURN;\
     longjmp(frame->Xwhere, 1);\  
364      }\      }\
365    return ra;\    return ra;\
366    }    }
# Line 266  match(), which never changes. */ Line 370  match(), which never changes. */
370    
371  typedef struct heapframe {  typedef struct heapframe {
372    struct heapframe *Xprevframe;    struct heapframe *Xprevframe;
373      struct heapframe *Xnextframe;
374    
375    /* Function arguments that may change */    /* Function arguments that may change */
376    
377    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
378    const uschar *Xecode;    const pcre_uchar *Xecode;
379      PCRE_PUCHAR Xmstart;
380    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
381    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
382    unsigned int Xrdepth;    unsigned int Xrdepth;
383    
384    /* Function local variables */    /* Function local variables */
385    
386    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
387    const uschar *Xcharptr;  #ifdef SUPPORT_UTF
388    const uschar *Xdata;    PCRE_PUCHAR Xcharptr;
389    const uschar *Xnext;  #endif
390    const uschar *Xpp;    PCRE_PUCHAR Xdata;
391    const uschar *Xprev;    PCRE_PUCHAR Xnext;
392    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xpp;
393      PCRE_PUCHAR Xprev;
394      PCRE_PUCHAR Xsaved_eptr;
395    
396    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
397    
398    BOOL Xcur_is_word;    BOOL Xcur_is_word;
399    BOOL Xcondition;    BOOL Xcondition;
   BOOL Xminimize;  
400    BOOL Xprev_is_word;    BOOL Xprev_is_word;
401    
   unsigned long int Xoriginal_ims;  
   
402  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
403    int Xprop_type;    int Xprop_type;
404    int Xprop_value;    int Xprop_value;
405    int Xprop_fail_result;    int Xprop_fail_result;
406    int Xprop_category;    int Xoclength;
407    int Xprop_chartype;    pcre_uchar Xocchars[6];
   int Xprop_script;  
   int *Xprop_test_variable;  
408  #endif  #endif
409    
410      int Xcodelink;
411    int Xctype;    int Xctype;
412    int Xfc;    unsigned int Xfc;
413    int Xfi;    int Xfi;
414    int Xlength;    int Xlength;
415    int Xmax;    int Xmax;
# Line 321  typedef struct heapframe { Line 423  typedef struct heapframe {
423    
424    eptrblock Xnewptrb;    eptrblock Xnewptrb;
425    
426    /* Place to pass back result, and where to jump back to */    /* Where to jump back to */
427    
428    int  Xresult;    int Xwhere;
   jmp_buf Xwhere;  
429    
430  } heapframe;  } heapframe;
431    
# Line 340  typedef struct heapframe { Line 441  typedef struct heapframe {
441  *         Match from current position            *  *         Match from current position            *
442  *************************************************/  *************************************************/
443    
444  /* 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  
445  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
446  same response.  same response. */
447    
448    /* These macros pack up tests that are used for partial matching, and which
449    appear several times in the code. We set the "hit end" flag if the pointer is
450    at the end of the subject and also past the start of the subject (i.e.
451    something has been matched). For hard partial matching, we then return
452    immediately. The second one is used when we already know we are past the end of
453    the subject. */
454    
455    #define CHECK_PARTIAL()\
456      if (md->partial != 0 && eptr >= md->end_subject && \
457          eptr > md->start_used_ptr) \
458        { \
459        md->hitend = TRUE; \
460        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
461        }
462    
463    #define SCHECK_PARTIAL()\
464      if (md->partial != 0 && eptr > md->start_used_ptr) \
465        { \
466        md->hitend = TRUE; \
467        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
468        }
469    
470    
471  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
472  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
473  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
474  made performance worse.  made performance worse.
475    
476  Arguments:  Arguments:
477     eptr        pointer in subject     eptr        pointer to current character in subject
478     ecode       position in code     ecode       pointer to current position in compiled code
479       mstart      pointer to the current match start position (can be modified
480                     by encountering \K)
481     offset_top  current top pointer     offset_top  current top pointer
482     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
483     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
484                   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  
485     rdepth      the recursion depth     rdepth      the recursion depth
486    
487  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
488                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
489                   a negative MATCH_xxx value for PRUNE, SKIP, etc
490                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
491                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
492  */  */
493    
494  static int  static int
495  match(REGISTER USPTR eptr, REGISTER const uschar *ecode,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
496    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
497    int flags, unsigned int rdepth)    unsigned int rdepth)
498  {  {
499  /* 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,
500  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
501  because they are used a lot in loops. */  "register" because they are used a lot in loops. */
502    
503  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
504  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
505  register unsigned int  c;  /* Character values not kept over RMATCH() calls */  register pcre_uint32 c;    /* Character values not kept over RMATCH() calls */
506  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
507    
508    BOOL minimize, possessive; /* Quantifier options */
509    BOOL caseless;
510    int condcode;
511    
512  /* 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
513  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
514  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
515  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
516    the top-level on the stack rather than malloc-ing them all gives a performance
517    boost in many cases where there is not much "recursion". */
518    
519  #ifdef NO_RECURSE  #ifdef NO_RECURSE
520  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)md->match_frames_base;
 frame->Xprevframe = NULL;            /* Marks the top level */  
521    
522  /* Copy in the original argument variables */  /* Copy in the original argument variables */
523    
524  frame->Xeptr = eptr;  frame->Xeptr = eptr;
525  frame->Xecode = ecode;  frame->Xecode = ecode;
526    frame->Xmstart = mstart;
527  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
528  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
529  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
530    
531  /* 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 536  HEAP_RECURSE:
536    
537  #define eptr               frame->Xeptr  #define eptr               frame->Xeptr
538  #define ecode              frame->Xecode  #define ecode              frame->Xecode
539    #define mstart             frame->Xmstart
540  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
541  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
542  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
543    
544  /* Ditto for the local variables */  /* Ditto for the local variables */
545    
546  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
547  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
548  #endif  #endif
549  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
550    #define codelink           frame->Xcodelink
551  #define data               frame->Xdata  #define data               frame->Xdata
552  #define next               frame->Xnext  #define next               frame->Xnext
553  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 434  HEAP_RECURSE: Line 558  HEAP_RECURSE:
558    
559  #define cur_is_word        frame->Xcur_is_word  #define cur_is_word        frame->Xcur_is_word
560  #define condition          frame->Xcondition  #define condition          frame->Xcondition
 #define minimize           frame->Xminimize  
561  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
562    
 #define original_ims       frame->Xoriginal_ims  
   
563  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
564  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
565  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
566  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
567  #define prop_category      frame->Xprop_category  #define oclength           frame->Xoclength
568  #define prop_chartype      frame->Xprop_chartype  #define occhars            frame->Xocchars
 #define prop_script        frame->Xprop_script  
 #define prop_test_variable frame->Xprop_test_variable  
569  #endif  #endif
570    
571  #define ctype              frame->Xctype  #define ctype              frame->Xctype
# Line 470  HEAP_RECURSE: Line 589  HEAP_RECURSE:
589  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
590  i, and fc and c, can be the same variables. */  i, and fc and c, can be the same variables. */
591    
592  #else  #else         /* NO_RECURSE not defined */
593  #define fi i  #define fi i
594  #define fc c  #define fc c
595    
596    /* Many of the following variables are used only in small blocks of the code.
597    My normal style of coding would have declared them within each of those blocks.
598    However, in order to accommodate the version of this code that uses an external
599    "stack" implemented on the heap, it is easier to declare them all here, so the
600    declarations can be cut out in a block. The only declarations within blocks
601    below are for variables that do not have to be preserved over a recursive call
602    to RMATCH(). */
603    
604    #ifdef SUPPORT_UTF
605    const pcre_uchar *charptr;
606    #endif
607    const pcre_uchar *callpat;
608    const pcre_uchar *data;
609    const pcre_uchar *next;
610    PCRE_PUCHAR       pp;
611    const pcre_uchar *prev;
612    PCRE_PUCHAR       saved_eptr;
613    
614    recursion_info new_recursive;
615    
616  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  BOOL cur_is_word;
 const uschar *charptr;             /* in small blocks of the code. My normal */  
 #endif                             /* style of coding would have declared    */  
 const uschar *callpat;             /* them within each of those blocks.      */  
 const uschar *data;                /* However, in order to accommodate the   */  
 const uschar *next;                /* version of this code that uses an      */  
 USPTR         pp;                  /* external "stack" implemented on the    */  
 const uschar *prev;                /* heap, it is easier to declare them all */  
 USPTR         saved_eptr;          /* here, so the declarations can be cut   */  
                                    /* out in a block. The only declarations  */  
 recursion_info new_recursive;      /* within blocks below are for variables  */  
                                    /* that do not have to be preserved over  */  
 BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  
617  BOOL condition;  BOOL condition;
 BOOL minimize;  
618  BOOL prev_is_word;  BOOL prev_is_word;
619    
 unsigned long int original_ims;  
   
620  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
621  int prop_type;  int prop_type;
622  int prop_value;  int prop_value;
623  int prop_fail_result;  int prop_fail_result;
624  int prop_category;  int oclength;
625  int prop_chartype;  pcre_uchar occhars[6];
 int prop_script;  
 int *prop_test_variable;  
626  #endif  #endif
627    
628    int codelink;
629  int ctype;  int ctype;
630  int length;  int length;
631  int max;  int max;
632  int min;  int min;
633  int number;  int number;
634  int offset;  int offset;
635  int op;  pcre_uchar op;
636  int save_capture_last;  int save_capture_last;
637  int save_offset1, save_offset2, save_offset3;  int save_offset1, save_offset2, save_offset3;
638  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
639    
640  eptrblock newptrb;  eptrblock newptrb;
641  #endif  
642    /* There is a special fudge for calling match() in a way that causes it to
643    measure the size of its basic stack frame when the stack is being used for
644    recursion. The second argument (ecode) being NULL triggers this behaviour. It
645    cannot normally ever be NULL. The return is the negated value of the frame
646    size. */
647    
648    if (ecode == NULL)
649      {
650      if (rdepth == 0)
651        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
652      else
653        {
654        int len = (char *)&rdepth - (char *)eptr;
655        return (len > 0)? -len : len;
656        }
657      }
658    #endif     /* NO_RECURSE */
659    
660    /* To save space on the stack and in the heap frame, I have doubled up on some
661    of the local variables that are used only in localised parts of the code, but
662    still need to be preserved over recursive calls of match(). These macros define
663    the alternative names that are used. */
664    
665    #define allow_zero    cur_is_word
666    #define cbegroup      condition
667    #define code_offset   codelink
668    #define condassert    condition
669    #define matched_once  prev_is_word
670    #define foc           number
671    #define save_mark     data
672    
673  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
674  variables. */  variables. */
# Line 524  variables. */ Line 676  variables. */
676  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
677  prop_value = 0;  prop_value = 0;
678  prop_fail_result = 0;  prop_fail_result = 0;
 prop_test_variable = NULL;  
679  #endif  #endif
680    
681    
682  /* This label is used for tail recursion, which is used in a few cases even  /* This label is used for tail recursion, which is used in a few cases even
683  when NO_RECURSE is not defined, in order to reduce the amount of stack that is  when NO_RECURSE is not defined, in order to reduce the amount of stack that is
684  used. Thanks to Ian Taylor for noticing this possibility and sending the  used. Thanks to Ian Taylor for noticing this possibility and sending the
# Line 537  TAIL_RECURSE: Line 689  TAIL_RECURSE:
689  /* 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
690  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
691  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()
692  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
693  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
694  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,
695  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
696    
697    #ifdef SUPPORT_UTF
698    utf = md->utf;       /* Local copy of the flag */
699    #else
700    utf = FALSE;
701    #endif
702    
703  /* 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
704  haven't exceeded the recursive call limit. */  haven't exceeded the recursive call limit. */
705    
706  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
707  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
708    
709  original_ims = ims;    /* Save for resetting on ')' */  /* At the start of a group with an unlimited repeat that may match an empty
710    string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
711    done this way to save having to use another function argument, which would take
712    up space on the stack. See also MATCH_CONDASSERT below.
713    
714    When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
715    such remembered pointers, to be checked when we hit the closing ket, in order
716    to break infinite loops that match no characters. When match() is called in
717    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
718    NOT be used with tail recursion, because the memory block that is used is on
719    the stack, so a new one may be required for each match(). */
720    
721  #ifdef SUPPORT_UTF8  if (md->match_function_type == MATCH_CBEGROUP)
 utf8 = md->utf8;       /* Local copy of the flag */  
 #else  
 utf8 = FALSE;  
 #endif  
   
 /* At the start of a bracketed group, add the current subject pointer to the  
 stack of such pointers, to be re-instated at the end of the group when we hit  
 the closing ket. When match() is called in other circumstances, we don't add to  
 this stack. */  
   
 if ((flags & match_isgroup) != 0)  
722    {    {
   newptrb.epb_prev = eptrb;  
723    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
724      newptrb.epb_prev = eptrb;
725    eptrb = &newptrb;    eptrb = &newptrb;
726      md->match_function_type = 0;
727    }    }
728    
729  /* Now start processing the operations. */  /* Now start processing the opcodes. */
730    
731  for (;;)  for (;;)
732    {    {
733      minimize = possessive = FALSE;
734    op = *ecode;    op = *ecode;
   minimize = FALSE;  
735    
736    /* For partial matching, remember if we ever hit the end of the subject after    switch(op)
737    matching at least one subject character. */      {
738        case OP_MARK:
739        md->nomatch_mark = ecode + 2;
740        md->mark = NULL;    /* In case previously set by assertion */
741        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
742          eptrb, RM55);
743        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
744             md->mark == NULL) md->mark = ecode + 2;
745    
746        /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
747        argument, and we must check whether that argument matches this MARK's
748        argument. It is passed back in md->start_match_ptr (an overloading of that
749        variable). If it does match, we reset that variable to the current subject
750        position and return MATCH_SKIP. Otherwise, pass back the return code
751        unaltered. */
752    
753        else if (rrc == MATCH_SKIP_ARG &&
754            STRCMP_UC_UC_TEST(ecode + 2, md->start_match_ptr) == 0)
755          {
756          md->start_match_ptr = eptr;
757          RRETURN(MATCH_SKIP);
758          }
759        RRETURN(rrc);
760    
761        case OP_FAIL:
762        RRETURN(MATCH_NOMATCH);
763    
764        /* COMMIT overrides PRUNE, SKIP, and THEN */
765    
766        case OP_COMMIT:
767        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
768          eptrb, RM52);
769        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
770            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
771            rrc != MATCH_THEN)
772          RRETURN(rrc);
773        RRETURN(MATCH_COMMIT);
774    
775        /* PRUNE overrides THEN */
776    
777        case OP_PRUNE:
778        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
779          eptrb, RM51);
780        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
781        RRETURN(MATCH_PRUNE);
782    
783        case OP_PRUNE_ARG:
784        md->nomatch_mark = ecode + 2;
785        md->mark = NULL;    /* In case previously set by assertion */
786        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
787          eptrb, RM56);
788        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
789             md->mark == NULL) md->mark = ecode + 2;
790        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
791        RRETURN(MATCH_PRUNE);
792    
793        /* SKIP overrides PRUNE and THEN */
794    
795        case OP_SKIP:
796        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
797          eptrb, RM53);
798        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
799          RRETURN(rrc);
800        md->start_match_ptr = eptr;   /* Pass back current position */
801        RRETURN(MATCH_SKIP);
802    
803        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
804        nomatch_mark. There is a flag that disables this opcode when re-matching a
805        pattern that ended with a SKIP for which there was not a matching MARK. */
806    
807        case OP_SKIP_ARG:
808        if (md->ignore_skip_arg)
809          {
810          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
811          break;
812          }
813        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
814          eptrb, RM57);
815        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
816          RRETURN(rrc);
817    
818        /* Pass back the current skip name by overloading md->start_match_ptr and
819        returning the special MATCH_SKIP_ARG return code. This will either be
820        caught by a matching MARK, or get to the top, where it causes a rematch
821        with the md->ignore_skip_arg flag set. */
822    
823        md->start_match_ptr = ecode + 2;
824        RRETURN(MATCH_SKIP_ARG);
825    
826        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
827        the branch in which it occurs can be determined. Overload the start of
828        match pointer to do this. */
829    
830        case OP_THEN:
831        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
832          eptrb, RM54);
833        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
834        md->start_match_ptr = ecode;
835        RRETURN(MATCH_THEN);
836    
837        case OP_THEN_ARG:
838        md->nomatch_mark = ecode + 2;
839        md->mark = NULL;    /* In case previously set by assertion */
840        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
841          md, eptrb, RM58);
842        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
843             md->mark == NULL) md->mark = ecode + 2;
844        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
845        md->start_match_ptr = ecode;
846        RRETURN(MATCH_THEN);
847    
848        /* Handle an atomic group that does not contain any capturing parentheses.
849        This can be handled like an assertion. Prior to 8.13, all atomic groups
850        were handled this way. In 8.13, the code was changed as below for ONCE, so
851        that backups pass through the group and thereby reset captured values.
852        However, this uses a lot more stack, so in 8.20, atomic groups that do not
853        contain any captures generate OP_ONCE_NC, which can be handled in the old,
854        less stack intensive way.
855    
856    if (md->partial &&      Check the alternative branches in turn - the matching won't pass the KET
857        eptr >= md->end_subject &&      for this kind of subpattern. If any one branch matches, we carry on as at
858        eptr > md->start_match)      the end of a normal bracket, leaving the subject pointer, but resetting
859      md->hitend = TRUE;      the start-of-match value in case it was changed by \K. */
   
   /* Opening capturing bracket. If there is space in the offset vector, save  
   the current subject position in the working slot at the top of the vector. We  
   mustn't change the current values of the data slot, because they may be set  
   from a previous iteration of this group, and be referred to by a reference  
   inside the group.  
   
   If the bracket fails to match, we need to restore this value and also the  
   values of the final offsets, in case they were set by a previous iteration of  
   the same bracket.  
   
   If there isn't enough space in the offset vector, treat this as if it were a  
   non-capturing bracket. Don't worry about setting the flag for the error case  
   here; that is handled in the code for KET. */  
860    
861    if (op > OP_BRA)      case OP_ONCE_NC:
862      {      prev = ecode;
863      number = op - OP_BRA;      saved_eptr = eptr;
864        save_mark = md->mark;
865        do
866          {
867          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
868          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
869            {
870            mstart = md->start_match_ptr;
871            break;
872            }
873          if (rrc == MATCH_THEN)
874            {
875            next = ecode + GET(ecode,1);
876            if (md->start_match_ptr < next &&
877                (*ecode == OP_ALT || *next == OP_ALT))
878              rrc = MATCH_NOMATCH;
879            }
880    
881          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
882          ecode += GET(ecode,1);
883          md->mark = save_mark;
884          }
885        while (*ecode == OP_ALT);
886    
887        /* If hit the end of the group (which could be repeated), fail */
888    
889        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
890    
891      /* For extended extraction brackets (large number), we have to fish out the      /* Continue as from after the group, updating the offsets high water
892      number from a dummy opcode at the start. */      mark, since extracts may have been taken. */
893    
894        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
895    
896        offset_top = md->end_offset_top;
897        eptr = md->end_match_ptr;
898    
899        /* For a non-repeating ket, just continue at this level. This also
900        happens for a repeating ket if no characters were matched in the group.
901        This is the forcible breaking of infinite loops as implemented in Perl
902        5.005. */
903    
904        if (*ecode == OP_KET || eptr == saved_eptr)
905          {
906          ecode += 1+LINK_SIZE;
907          break;
908          }
909    
910        /* The repeating kets try the rest of the pattern or restart from the
911        preceding bracket, in the appropriate order. The second "call" of match()
912        uses tail recursion, to avoid using another stack frame. */
913    
914        if (*ecode == OP_KETRMIN)
915          {
916          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
917          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
918          ecode = prev;
919          goto TAIL_RECURSE;
920          }
921        else  /* OP_KETRMAX */
922          {
923          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
924          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
925          ecode += 1 + LINK_SIZE;
926          goto TAIL_RECURSE;
927          }
928        /* Control never gets here */
929    
930      if (number > EXTRACT_BASIC_MAX)      /* Handle a capturing bracket, other than those that are possessive with an
931        number = GET2(ecode, 2+LINK_SIZE);      unlimited repeat. If there is space in the offset vector, save the current
932        subject position in the working slot at the top of the vector. We mustn't
933        change the current values of the data slot, because they may be set from a
934        previous iteration of this group, and be referred to by a reference inside
935        the group. A failure to match might occur after the group has succeeded,
936        if something later on doesn't match. For this reason, we need to restore
937        the working value and also the values of the final offsets, in case they
938        were set by a previous iteration of the same bracket.
939    
940        If there isn't enough space in the offset vector, treat this as if it were
941        a non-capturing bracket. Don't worry about setting the flag for the error
942        case here; that is handled in the code for KET. */
943    
944        case OP_CBRA:
945        case OP_SCBRA:
946        number = GET2(ecode, 1+LINK_SIZE);
947      offset = number << 1;      offset = number << 1;
948    
949  #ifdef DEBUG  #ifdef PCRE_DEBUG
950      printf("start bracket %d subject=", number);      printf("start bracket %d\n", number);
951        printf("subject=");
952      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
953      printf("\n");      printf("\n");
954  #endif  #endif
# Line 620  for (;;) Line 959  for (;;)
959        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
960        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
961        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
962          save_mark = md->mark;
963    
964        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
965        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
966            (int)(eptr - md->start_subject);
967    
968        do        for (;;)
969          {          {
970          RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
971            match_isgroup);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
972              eptrb, RM1);
973            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
974    
975            /* If we backed up to a THEN, check whether it is within the current
976            branch by comparing the address of the THEN that is passed back with
977            the end of the branch. If it is within the current branch, and the
978            branch is one of two or more alternatives (it either starts or ends
979            with OP_ALT), we have reached the limit of THEN's action, so convert
980            the return code to NOMATCH, which will cause normal backtracking to
981            happen from now on. Otherwise, THEN is passed back to an outer
982            alternative. This implements Perl's treatment of parenthesized groups,
983            where a group not containing | does not affect the current alternative,
984            that is, (X) is NOT the same as (X|(*F)). */
985    
986            if (rrc == MATCH_THEN)
987              {
988              next = ecode + GET(ecode,1);
989              if (md->start_match_ptr < next &&
990                  (*ecode == OP_ALT || *next == OP_ALT))
991                rrc = MATCH_NOMATCH;
992              }
993    
994            /* Anything other than NOMATCH is passed back. */
995    
996          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
997          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
998          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
999            md->mark = save_mark;
1000            if (*ecode != OP_ALT) break;
1001          }          }
       while (*ecode == OP_ALT);  
1002    
1003        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
1004        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
1005        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
1006        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
1007    
1008        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1009    
1010          RRETURN(rrc);
1011        }        }
1012    
1013      /* Insufficient room for saving captured contents */      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1014        as a non-capturing bracket. */
1015    
1016      else op = OP_BRA;      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1017      }      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1018    
1019    /* Other types of node can be handled by a switch */      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1020    
1021    switch(op)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1022      {      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1023      case OP_BRA:     /* Non-capturing bracket: optimized */  
1024      DPRINTF(("start bracket 0\n"));      /* Non-capturing or atomic group, except for possessive with unlimited
1025        repeat and ONCE group with no captures. Loop for all the alternatives.
1026    
1027        When we get to the final alternative within the brackets, we used to return
1028        the result of a recursive call to match() whatever happened so it was
1029        possible to reduce stack usage by turning this into a tail recursion,
1030        except in the case of a possibly empty group. However, now that there is
1031        the possiblity of (*THEN) occurring in the final alternative, this
1032        optimization is no longer always possible.
1033    
1034      /* Loop for all the alternatives */      We can optimize if we know there are no (*THEN)s in the pattern; at present
1035        this is the best that can be done.
1036    
1037        MATCH_ONCE is returned when the end of an atomic group is successfully
1038        reached, but subsequent matching fails. It passes back up the tree (causing
1039        captured values to be reset) until the original atomic group level is
1040        reached. This is tested by comparing md->once_target with the start of the
1041        group. At this point, the return is converted into MATCH_NOMATCH so that
1042        previous backup points can be taken. */
1043    
1044        case OP_ONCE:
1045        case OP_BRA:
1046        case OP_SBRA:
1047        DPRINTF(("start non-capturing bracket\n"));
1048    
1049      for (;;)      for (;;)
1050        {        {
1051        /* When we get to the final alternative within the brackets, we would        if (op >= OP_SBRA || op == OP_ONCE)
1052        return the result of a recursive call to match() whatever happened. We          md->match_function_type = MATCH_CBEGROUP;
       can reduce stack usage by turning this into a tail recursion. */  
   
       if (ecode[GET(ecode, 1)] != OP_ALT)  
        {  
        ecode += 1 + LINK_SIZE;  
        flags = match_isgroup;  
        DPRINTF(("bracket 0 tail recursion\n"));  
        goto TAIL_RECURSE;  
        }  
1053    
1054        /* For non-final alternatives, continue the loop for a NOMATCH result;        /* If this is not a possibly empty group, and there are no (*THEN)s in
1055        otherwise return. */        the pattern, and this is the final alternative, optimize as described
1056          above. */
1057    
1058        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb,        else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1059          match_isgroup);          {
1060        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          ecode += PRIV(OP_lengths)[*ecode];
1061            goto TAIL_RECURSE;
1062            }
1063    
1064          /* In all other cases, we have to make another call to match(). */
1065    
1066          save_mark = md->mark;
1067          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1068            RM2);
1069    
1070          /* See comment in the code for capturing groups above about handling
1071          THEN. */
1072    
1073          if (rrc == MATCH_THEN)
1074            {
1075            next = ecode + GET(ecode,1);
1076            if (md->start_match_ptr < next &&
1077                (*ecode == OP_ALT || *next == OP_ALT))
1078              rrc = MATCH_NOMATCH;
1079            }
1080    
1081          if (rrc != MATCH_NOMATCH)
1082            {
1083            if (rrc == MATCH_ONCE)
1084              {
1085              const pcre_uchar *scode = ecode;
1086              if (*scode != OP_ONCE)           /* If not at start, find it */
1087                {
1088                while (*scode == OP_ALT) scode += GET(scode, 1);
1089                scode -= GET(scode, 1);
1090                }
1091              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1092              }
1093            RRETURN(rrc);
1094            }
1095        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1096          md->mark = save_mark;
1097          if (*ecode != OP_ALT) break;
1098        }        }
     /* Control never reaches here. */  
1099    
1100      /* Conditional group: compilation checked that there are no more than      RRETURN(MATCH_NOMATCH);
     two branches. If the condition is false, skipping the first branch takes us  
     past the end if there is only one branch, but that's OK because that is  
     exactly what going to the ket would do. As there is only one branch to be  
     obeyed, we can use tail recursion to avoid using another stack frame. */  
1101    
1102      case OP_COND:      /* Handle possessive capturing brackets with an unlimited repeat. We come
1103      if (ecode[LINK_SIZE+1] == OP_CREF) /* Condition extract or recurse test */      here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1104        {      handled similarly to the normal case above. However, the matching is
1105        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */      different. The end of these brackets will always be OP_KETRPOS, which
1106        condition = (offset == CREF_RECURSE * 2)?      returns MATCH_KETRPOS without going further in the pattern. By this means
1107          (md->recursive != NULL) :      we can handle the group by iteration rather than recursion, thereby
1108          (offset < offset_top && md->offset_vector[offset] >= 0);      reducing the amount of stack needed. */
1109        ecode += condition? (LINK_SIZE + 4) : (LINK_SIZE + 1 + GET(ecode, 1));  
1110        flags = match_isgroup;      case OP_CBRAPOS:
1111        goto TAIL_RECURSE;      case OP_SCBRAPOS:
1112        }      allow_zero = FALSE;
1113    
1114      /* The condition is an assertion. Call match() to evaluate it - setting      POSSESSIVE_CAPTURE:
1115      the final argument TRUE causes it to stop at the end of an assertion. */      number = GET2(ecode, 1+LINK_SIZE);
1116        offset = number << 1;
1117    
1118      else  #ifdef PCRE_DEBUG
1119        printf("start possessive bracket %d\n", number);
1120        printf("subject=");
1121        pchars(eptr, 16, TRUE, md);
1122        printf("\n");
1123    #endif
1124    
1125        if (offset < md->offset_max)
1126        {        {
1127        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        matched_once = FALSE;
1128            match_condassert | match_isgroup);        code_offset = (int)(ecode - md->start_code);
1129        if (rrc == MATCH_MATCH)  
1130          {        save_offset1 = md->offset_vector[offset];
1131          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE+2);        save_offset2 = md->offset_vector[offset+1];
1132          while (*ecode == OP_ALT) ecode += GET(ecode, 1);        save_offset3 = md->offset_vector[md->offset_end - number];
1133          save_capture_last = md->capture_last;
1134    
1135          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
1136    
1137          /* Each time round the loop, save the current subject position for use
1138          when the group matches. For MATCH_MATCH, the group has matched, so we
1139          restart it with a new subject starting position, remembering that we had
1140          at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1141          usual. If we haven't matched any alternatives in any iteration, check to
1142          see if a previous iteration matched. If so, the group has matched;
1143          continue from afterwards. Otherwise it has failed; restore the previous
1144          capture values before returning NOMATCH. */
1145    
1146          for (;;)
1147            {
1148            md->offset_vector[md->offset_end - number] =
1149              (int)(eptr - md->start_subject);
1150            if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1151            RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1152              eptrb, RM63);
1153            if (rrc == MATCH_KETRPOS)
1154              {
1155              offset_top = md->end_offset_top;
1156              eptr = md->end_match_ptr;
1157              ecode = md->start_code + code_offset;
1158              save_capture_last = md->capture_last;
1159              matched_once = TRUE;
1160              continue;
1161              }
1162    
1163            /* See comment in the code for capturing groups above about handling
1164            THEN. */
1165    
1166            if (rrc == MATCH_THEN)
1167              {
1168              next = ecode + GET(ecode,1);
1169              if (md->start_match_ptr < next &&
1170                  (*ecode == OP_ALT || *next == OP_ALT))
1171                rrc = MATCH_NOMATCH;
1172              }
1173    
1174            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1175            md->capture_last = save_capture_last;
1176            ecode += GET(ecode, 1);
1177            if (*ecode != OP_ALT) break;
1178          }          }
1179        else if (rrc != MATCH_NOMATCH)  
1180          if (!matched_once)
1181          {          {
1182          RRETURN(rrc);         /* Need braces because of following else */          md->offset_vector[offset] = save_offset1;
1183            md->offset_vector[offset+1] = save_offset2;
1184            md->offset_vector[md->offset_end - number] = save_offset3;
1185          }          }
       else ecode += GET(ecode, 1);  
1186    
1187        /* We are now at the branch that is to be obeyed. As there is only one,        if (allow_zero || matched_once)
1188        we can use tail recursion to avoid using another stack frame. */          {
1189            ecode += 1 + LINK_SIZE;
1190            break;
1191            }
1192    
1193        ecode += 1 + LINK_SIZE;        RRETURN(MATCH_NOMATCH);
       flags = match_isgroup;  
       goto TAIL_RECURSE;  
1194        }        }
     /* Control never reaches here */  
1195    
1196      /* Skip over conditional reference or large extraction number data if      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1197      encountered. */      as a non-capturing bracket. */
1198    
1199      case OP_CREF:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1200      case OP_BRANUMBER:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     ecode += 3;  
     break;  
1201    
1202      /* End of the pattern. If we are in a recursion, we should restore the      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
     offsets appropriately and continue from after the call. */  
1203    
1204      case OP_END:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1205      if (md->recursive != NULL && md->recursive->group_num == 0)      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
       {  
       recursion_info *rec = md->recursive;  
       DPRINTF(("End of pattern in a (?0) recursion\n"));  
       md->recursive = rec->prevrec;  
       memmove(md->offset_vector, rec->offset_save,  
         rec->saved_max * sizeof(int));  
       md->start_match = rec->save_start;  
       ims = original_ims;  
       ecode = rec->after_call;  
       break;  
       }  
1206    
1207      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      /* Non-capturing possessive bracket with unlimited repeat. We come here
1208      string - backtracking will then try other alternatives, if any. */      from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1209        without the capturing complication. It is written out separately for speed
1210        and cleanliness. */
1211    
1212      if (md->notempty && eptr == md->start_match) RRETURN(MATCH_NOMATCH);      case OP_BRAPOS:
1213      md->end_match_ptr = eptr;          /* Record where we ended */      case OP_SBRAPOS:
1214      md->end_offset_top = offset_top;   /* and how many extracts were taken */      allow_zero = FALSE;
     RRETURN(MATCH_MATCH);  
   
     /* Change option settings */  
   
     case OP_OPT:  
     ims = ecode[1];  
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1215    
1216      /* Assertion brackets. Check the alternative branches in turn - the      POSSESSIVE_NON_CAPTURE:
1217      matching won't pass the KET for an assertion. If any one branch matches,      matched_once = FALSE;
1218      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      code_offset = (int)(ecode - md->start_code);
     start of each branch to move the current point backwards, so the code at  
     this level is identical to the lookahead case. */  
1219    
1220      case OP_ASSERT:      for (;;)
     case OP_ASSERTBACK:  
     do  
1221        {        {
1222        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1223          match_isgroup);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1224        if (rrc == MATCH_MATCH) break;          eptrb, RM48);
1225          if (rrc == MATCH_KETRPOS)
1226            {
1227            offset_top = md->end_offset_top;
1228            eptr = md->end_match_ptr;
1229            ecode = md->start_code + code_offset;
1230            matched_once = TRUE;
1231            continue;
1232            }
1233    
1234          /* See comment in the code for capturing groups above about handling
1235          THEN. */
1236    
1237          if (rrc == MATCH_THEN)
1238            {
1239            next = ecode + GET(ecode,1);
1240            if (md->start_match_ptr < next &&
1241                (*ecode == OP_ALT || *next == OP_ALT))
1242              rrc = MATCH_NOMATCH;
1243            }
1244    
1245        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1246        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1247          if (*ecode != OP_ALT) break;
1248        }        }
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1249    
1250      /* If checking an assertion for a condition, return MATCH_MATCH. */      if (matched_once || allow_zero)
1251          {
1252          ecode += 1 + LINK_SIZE;
1253          break;
1254          }
1255        RRETURN(MATCH_NOMATCH);
1256    
1257      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* Control never reaches here. */
1258    
1259      /* Continue from after the assertion, updating the offsets high water      /* Conditional group: compilation checked that there are no more than
1260      mark, since extracts may have been taken during the assertion. */      two branches. If the condition is false, skipping the first branch takes us
1261        past the end if there is only one branch, but that's OK because that is
1262        exactly what going to the ket would do. */
1263    
1264      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_COND:
1265      ecode += 1 + LINK_SIZE;      case OP_SCOND:
1266      offset_top = md->end_offset_top;      codelink = GET(ecode, 1);
     continue;  
1267    
1268      /* Negative assertion: all branches must fail to match */      /* Because of the way auto-callout works during compile, a callout item is
1269        inserted between OP_COND and an assertion condition. */
1270    
1271      case OP_ASSERT_NOT:      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
     case OP_ASSERTBACK_NOT:  
     do  
1272        {        {
1273        RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        if (PUBL(callout) != NULL)
1274          match_isgroup);          {
1275        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);          PUBL(callout_block) cb;
1276        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          cb.version          = 2;   /* Version 1 of the callout block */
1277        ecode += GET(ecode,1);          cb.callout_number   = ecode[LINK_SIZE+2];
1278            cb.offset_vector    = md->offset_vector;
1279    #if defined COMPILE_PCRE8
1280            cb.subject          = (PCRE_SPTR)md->start_subject;
1281    #elif defined COMPILE_PCRE16
1282            cb.subject          = (PCRE_SPTR16)md->start_subject;
1283    #elif defined COMPILE_PCRE32
1284            cb.subject          = (PCRE_SPTR32)md->start_subject;
1285    #endif
1286            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1287            cb.start_match      = (int)(mstart - md->start_subject);
1288            cb.current_position = (int)(eptr - md->start_subject);
1289            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1290            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1291            cb.capture_top      = offset_top/2;
1292            cb.capture_last     = md->capture_last;
1293            cb.callout_data     = md->callout_data;
1294            cb.mark             = md->nomatch_mark;
1295            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1296            if (rrc < 0) RRETURN(rrc);
1297            }
1298          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1299        }        }
     while (*ecode == OP_ALT);  
   
     if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);  
1300    
1301      ecode += 1 + LINK_SIZE;      condcode = ecode[LINK_SIZE+1];
     continue;  
1302    
1303      /* Move the subject pointer back. This occurs only at the start of      /* Now see what the actual condition is */
     each branch of a lookbehind assertion. If we are too close to the start to  
     move back, this match function fails. When working with UTF-8 we move  
     back a number of characters, not bytes. */  
1304    
1305      case OP_REVERSE:      if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
 #ifdef SUPPORT_UTF8  
     if (utf8)  
1306        {        {
1307        c = GET(ecode,1);        if (md->recursive == NULL)                /* Not recursing => FALSE */
       for (i = 0; i < c; i++)  
1308          {          {
1309          eptr--;          condition = FALSE;
1310          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          ecode += GET(ecode, 1);
         BACKCHAR(eptr)  
1311          }          }
1312        }        else
1313      else          {
1314            int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1315            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1316    
1317            /* If the test is for recursion into a specific subpattern, and it is
1318            false, but the test was set up by name, scan the table to see if the
1319            name refers to any other numbers, and test them. The condition is true
1320            if any one is set. */
1321    
1322            if (!condition && condcode == OP_NRREF)
1323              {
1324              pcre_uchar *slotA = md->name_table;
1325              for (i = 0; i < md->name_count; i++)
1326                {
1327                if (GET2(slotA, 0) == recno) break;
1328                slotA += md->name_entry_size;
1329                }
1330    
1331              /* Found a name for the number - there can be only one; duplicate
1332              names for different numbers are allowed, but not vice versa. First
1333              scan down for duplicates. */
1334    
1335              if (i < md->name_count)
1336                {
1337                pcre_uchar *slotB = slotA;
1338                while (slotB > md->name_table)
1339                  {
1340                  slotB -= md->name_entry_size;
1341                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1342                    {
1343                    condition = GET2(slotB, 0) == md->recursive->group_num;
1344                    if (condition) break;
1345                    }
1346                  else break;
1347                  }
1348    
1349                /* Scan up for duplicates */
1350    
1351                if (!condition)
1352                  {
1353                  slotB = slotA;
1354                  for (i++; i < md->name_count; i++)
1355                    {
1356                    slotB += md->name_entry_size;
1357                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1358                      {
1359                      condition = GET2(slotB, 0) == md->recursive->group_num;
1360                      if (condition) break;
1361                      }
1362                    else break;
1363                    }
1364                  }
1365                }
1366              }
1367    
1368            /* Chose branch according to the condition */
1369    
1370            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1371            }
1372          }
1373    
1374        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1375          {
1376          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1377          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1378    
1379          /* If the numbered capture is unset, but the reference was by name,
1380          scan the table to see if the name refers to any other numbers, and test
1381          them. The condition is true if any one is set. This is tediously similar
1382          to the code above, but not close enough to try to amalgamate. */
1383    
1384          if (!condition && condcode == OP_NCREF)
1385            {
1386            int refno = offset >> 1;
1387            pcre_uchar *slotA = md->name_table;
1388    
1389            for (i = 0; i < md->name_count; i++)
1390              {
1391              if (GET2(slotA, 0) == refno) break;
1392              slotA += md->name_entry_size;
1393              }
1394    
1395            /* Found a name for the number - there can be only one; duplicate names
1396            for different numbers are allowed, but not vice versa. First scan down
1397            for duplicates. */
1398    
1399            if (i < md->name_count)
1400              {
1401              pcre_uchar *slotB = slotA;
1402              while (slotB > md->name_table)
1403                {
1404                slotB -= md->name_entry_size;
1405                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1406                  {
1407                  offset = GET2(slotB, 0) << 1;
1408                  condition = offset < offset_top &&
1409                    md->offset_vector[offset] >= 0;
1410                  if (condition) break;
1411                  }
1412                else break;
1413                }
1414    
1415              /* Scan up for duplicates */
1416    
1417              if (!condition)
1418                {
1419                slotB = slotA;
1420                for (i++; i < md->name_count; i++)
1421                  {
1422                  slotB += md->name_entry_size;
1423                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1424                    {
1425                    offset = GET2(slotB, 0) << 1;
1426                    condition = offset < offset_top &&
1427                      md->offset_vector[offset] >= 0;
1428                    if (condition) break;
1429                    }
1430                  else break;
1431                  }
1432                }
1433              }
1434            }
1435    
1436          /* Chose branch according to the condition */
1437    
1438          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1439          }
1440    
1441        else if (condcode == OP_DEF)     /* DEFINE - always false */
1442          {
1443          condition = FALSE;
1444          ecode += GET(ecode, 1);
1445          }
1446    
1447        /* The condition is an assertion. Call match() to evaluate it - setting
1448        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1449        an assertion. */
1450    
1451        else
1452          {
1453          md->match_function_type = MATCH_CONDASSERT;
1454          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1455          if (rrc == MATCH_MATCH)
1456            {
1457            if (md->end_offset_top > offset_top)
1458              offset_top = md->end_offset_top;  /* Captures may have happened */
1459            condition = TRUE;
1460            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1461            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1462            }
1463    
1464          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1465          assertion; it is therefore treated as NOMATCH. */
1466    
1467          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1468            {
1469            RRETURN(rrc);         /* Need braces because of following else */
1470            }
1471          else
1472            {
1473            condition = FALSE;
1474            ecode += codelink;
1475            }
1476          }
1477    
1478        /* We are now at the branch that is to be obeyed. As there is only one, can
1479        use tail recursion to avoid using another stack frame, except when there is
1480        unlimited repeat of a possibly empty group. In the latter case, a recursive
1481        call to match() is always required, unless the second alternative doesn't
1482        exist, in which case we can just plough on. Note that, for compatibility
1483        with Perl, the | in a conditional group is NOT treated as creating two
1484        alternatives. If a THEN is encountered in the branch, it propagates out to
1485        the enclosing alternative (unless nested in a deeper set of alternatives,
1486        of course). */
1487    
1488        if (condition || *ecode == OP_ALT)
1489          {
1490          if (op != OP_SCOND)
1491            {
1492            ecode += 1 + LINK_SIZE;
1493            goto TAIL_RECURSE;
1494            }
1495    
1496          md->match_function_type = MATCH_CBEGROUP;
1497          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1498          RRETURN(rrc);
1499          }
1500    
1501         /* Condition false & no alternative; continue after the group. */
1502    
1503        else
1504          {
1505          ecode += 1 + LINK_SIZE;
1506          }
1507        break;
1508    
1509    
1510        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1511        to close any currently open capturing brackets. */
1512    
1513        case OP_CLOSE:
1514        number = GET2(ecode, 1);
1515        offset = number << 1;
1516    
1517    #ifdef PCRE_DEBUG
1518          printf("end bracket %d at *ACCEPT", number);
1519          printf("\n");
1520    #endif
1521    
1522        md->capture_last = number;
1523        if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1524          {
1525          md->offset_vector[offset] =
1526            md->offset_vector[md->offset_end - number];
1527          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1528          if (offset_top <= offset) offset_top = offset + 2;
1529          }
1530        ecode += 1 + IMM2_SIZE;
1531        break;
1532    
1533    
1534        /* End of the pattern, either real or forced. */
1535    
1536        case OP_END:
1537        case OP_ACCEPT:
1538        case OP_ASSERT_ACCEPT:
1539    
1540        /* If we have matched an empty string, fail if not in an assertion and not
1541        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1542        is set and we have matched at the start of the subject. In both cases,
1543        backtracking will then try other alternatives, if any. */
1544    
1545        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1546             md->recursive == NULL &&
1547             (md->notempty ||
1548               (md->notempty_atstart &&
1549                 mstart == md->start_subject + md->start_offset)))
1550          RRETURN(MATCH_NOMATCH);
1551    
1552        /* Otherwise, we have a match. */
1553    
1554        md->end_match_ptr = eptr;           /* Record where we ended */
1555        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1556        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1557    
1558        /* For some reason, the macros don't work properly if an expression is
1559        given as the argument to RRETURN when the heap is in use. */
1560    
1561        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1562        RRETURN(rrc);
1563    
1564        /* Assertion brackets. Check the alternative branches in turn - the
1565        matching won't pass the KET for an assertion. If any one branch matches,
1566        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1567        start of each branch to move the current point backwards, so the code at
1568        this level is identical to the lookahead case. When the assertion is part
1569        of a condition, we want to return immediately afterwards. The caller of
1570        this incarnation of the match() function will have set MATCH_CONDASSERT in
1571        md->match_function type, and one of these opcodes will be the first opcode
1572        that is processed. We use a local variable that is preserved over calls to
1573        match() to remember this case. */
1574    
1575        case OP_ASSERT:
1576        case OP_ASSERTBACK:
1577        save_mark = md->mark;
1578        if (md->match_function_type == MATCH_CONDASSERT)
1579          {
1580          condassert = TRUE;
1581          md->match_function_type = 0;
1582          }
1583        else condassert = FALSE;
1584    
1585        do
1586          {
1587          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1588          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1589            {
1590            mstart = md->start_match_ptr;   /* In case \K reset it */
1591            break;
1592            }
1593          md->mark = save_mark;
1594    
1595          /* A COMMIT failure must fail the entire assertion, without trying any
1596          subsequent branches. */
1597    
1598          if (rrc == MATCH_COMMIT) RRETURN(MATCH_NOMATCH);
1599    
1600          /* PCRE does not allow THEN to escape beyond an assertion; it
1601          is treated as NOMATCH. */
1602    
1603          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1604          ecode += GET(ecode, 1);
1605          }
1606        while (*ecode == OP_ALT);
1607    
1608        if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1609    
1610        /* If checking an assertion for a condition, return MATCH_MATCH. */
1611    
1612        if (condassert) RRETURN(MATCH_MATCH);
1613    
1614        /* Continue from after the assertion, updating the offsets high water
1615        mark, since extracts may have been taken during the assertion. */
1616    
1617        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1618        ecode += 1 + LINK_SIZE;
1619        offset_top = md->end_offset_top;
1620        continue;
1621    
1622        /* Negative assertion: all branches must fail to match. Encountering SKIP,
1623        PRUNE, or COMMIT means we must assume failure without checking subsequent
1624        branches. */
1625    
1626        case OP_ASSERT_NOT:
1627        case OP_ASSERTBACK_NOT:
1628        save_mark = md->mark;
1629        if (md->match_function_type == MATCH_CONDASSERT)
1630          {
1631          condassert = TRUE;
1632          md->match_function_type = 0;
1633          }
1634        else condassert = FALSE;
1635    
1636        do
1637          {
1638          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1639          md->mark = save_mark;
1640          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1641          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1642            {
1643            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1644            break;
1645            }
1646    
1647          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1648          as NOMATCH. */
1649    
1650          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1651          ecode += GET(ecode,1);
1652          }
1653        while (*ecode == OP_ALT);
1654    
1655        if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1656    
1657        ecode += 1 + LINK_SIZE;
1658        continue;
1659    
1660        /* Move the subject pointer back. This occurs only at the start of
1661        each branch of a lookbehind assertion. If we are too close to the start to
1662        move back, this match function fails. When working with UTF-8 we move
1663        back a number of characters, not bytes. */
1664    
1665        case OP_REVERSE:
1666    #ifdef SUPPORT_UTF
1667        if (utf)
1668          {
1669          i = GET(ecode, 1);
1670          while (i-- > 0)
1671            {
1672            eptr--;
1673            if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1674            BACKCHAR(eptr);
1675            }
1676          }
1677        else
1678  #endif  #endif
1679    
1680      /* No UTF-8 support, or not in UTF-8 mode: count is byte count */      /* No UTF-8 support, or not in UTF-8 mode: count is byte count */
1681    
1682        {        {
1683        eptr -= GET(ecode,1);        eptr -= GET(ecode, 1);
1684        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1685        }        }
1686    
1687      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1688    
1689        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1690      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1691      break;      break;
1692    
# Line 854  for (;;) Line 1695  for (;;)
1695      function is able to force a failure. */      function is able to force a failure. */
1696    
1697      case OP_CALLOUT:      case OP_CALLOUT:
1698      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1699        {        {
1700        pcre_callout_block cb;        PUBL(callout_block) cb;
1701        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1702        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1703        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1704    #if defined COMPILE_PCRE8
1705        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1706        cb.subject_length   = md->end_subject - md->start_subject;  #elif defined COMPILE_PCRE16
1707        cb.start_match      = md->start_match - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1708        cb.current_position = eptr - md->start_subject;  #elif defined COMPILE_PCRE32
1709          cb.subject          = (PCRE_SPTR32)md->start_subject;
1710    #endif
1711          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1712          cb.start_match      = (int)(mstart - md->start_subject);
1713          cb.current_position = (int)(eptr - md->start_subject);
1714        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1715        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1716        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1717        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last;
1718        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1719        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1720          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1721        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1722        }        }
1723      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 879  for (;;) Line 1727  for (;;)
1727      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
1728      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1729    
1730      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1731      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
1732      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
1733      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
1734      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
1735      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
1736      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.  
1737    
1738      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
1739      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
1740      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1741        a lot, so he is not to blame for the current way it works. */
1742    
1743      case OP_RECURSE:      case OP_RECURSE:
1744        {        {
1745        callpat = md->start_code + GET(ecode, 1);        recursion_info *ri;
1746        new_recursive.group_num = *callpat - OP_BRA;        int recno;
1747    
1748        /* For extended extraction brackets (large number), we have to fish out        callpat = md->start_code + GET(ecode, 1);
1749        the number from a dummy opcode at the start. */        recno = (callpat == md->start_code)? 0 :
1750            GET2(callpat, 1 + LINK_SIZE);
1751    
1752        if (new_recursive.group_num > EXTRACT_BASIC_MAX)        /* Check for repeating a recursion without advancing the subject pointer.
1753          new_recursive.group_num = GET2(callpat, 2+LINK_SIZE);        This should catch convoluted mutual recursions. (Some simple cases are
1754          caught at compile time.) */
1755    
1756          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1757            if (recno == ri->group_num && eptr == ri->subject_position)
1758              RRETURN(PCRE_ERROR_RECURSELOOP);
1759    
1760        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1761    
1762          new_recursive.group_num = recno;
1763          new_recursive.subject_position = eptr;
1764        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1765        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1766    
1767        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1768    
1769        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1770    
1771        /* Now save the offset data. */        /* Now save the offset data */
1772    
1773        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1774        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 923  for (;;) Line 1776  for (;;)
1776        else        else
1777          {          {
1778          new_recursive.offset_save =          new_recursive.offset_save =
1779            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1780          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1781          }          }
   
1782        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1783              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = md->start_match;  
       md->start_match = eptr;  
1784    
1785        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1786        restore the offset and recursion data. */        restore the offset data. If there were nested recursions, md->recursive
1787          might be changed, so reset it before looping. */
1788    
1789        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1790          cbegroup = (*callpat >= OP_SBRA);
1791        do        do
1792          {          {
1793          RMATCH(rrc, eptr, callpat + 1 + LINK_SIZE, offset_top, md, ims,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1794              eptrb, match_isgroup);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1795          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1796            memcpy(md->offset_vector, new_recursive.offset_save,
1797                new_recursive.saved_max * sizeof(int));
1798            md->recursive = new_recursive.prevrec;
1799            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1800            {            {
1801            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1802            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1803              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1804            RRETURN(MATCH_MATCH);  
1805              /* Set where we got to in the subject, and reset the start in case
1806              it was changed by \K. This *is* propagated back out of a recursion,
1807              for Perl compatibility. */
1808    
1809              eptr = md->end_match_ptr;
1810              mstart = md->start_match_ptr;
1811              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1812            }            }
1813          else if (rrc != MATCH_NOMATCH)  
1814            /* PCRE does not allow THEN or COMMIT to escape beyond a recursion; it
1815            is treated as NOMATCH. */
1816    
1817            else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN &&
1818                     rrc != MATCH_COMMIT)
1819            {            {
1820            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1821              if (new_recursive.offset_save != stacksave)
1822                (PUBL(free))(new_recursive.offset_save);
1823            RRETURN(rrc);            RRETURN(rrc);
1824            }            }
1825    
1826          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1827          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1828          }          }
1829        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 964  for (;;) Line 1831  for (;;)
1831        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1832        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1833        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1834          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1835        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1836        }        }
     /* Control never reaches here */  
1837    
1838      /* "Once" brackets are like assertion brackets except that after a match,      RECURSION_MATCHED:
1839      the point in the subject string is not moved back. Thus there can never be      break;
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims,  
         eptrb, match_isgroup);  
       if (rrc == MATCH_MATCH) break;  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
   
     /* If hit the end of the group (which could be repeated), fail */  
   
     if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
   
     /* Continue as from after the assertion, updating the offsets high water  
     mark, since extracts may have been taken. */  
   
     do ecode += GET(ecode,1); while (*ecode == OP_ALT);  
   
     offset_top = md->end_offset_top;  
     eptr = md->end_match_ptr;  
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
       {  
       ecode += 1+LINK_SIZE;  
       break;  
       }  
   
     /* The repeating kets try the rest of the pattern or restart from the  
     preceding bracket, in the appropriate order. The second "call" of match()  
     uses tail recursion, to avoid using another stack frame. We need to reset  
     any options that changed within the bracket before re-running it, so  
     check the next opcode. */  
   
     if (ecode[1+LINK_SIZE] == OP_OPT)  
       {  
       ims = (ims & ~PCRE_IMS) | ecode[4];  
       DPRINTF(("ims set to %02lx at group repeat\n", ims));  
       }  
   
     if (*ecode == OP_KETRMIN)  
       {  
       RMATCH(rrc, eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       flags = match_isgroup;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1840    
1841      /* 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
1842      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1051  for (;;) Line 1845  for (;;)
1845      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1846      break;      break;
1847    
1848      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1849      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
1850      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
1851      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
1852      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1853    
1854      case OP_BRAZERO:      case OP_BRAZERO:
1855        {      next = ecode + 1;
1856        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1857        RMATCH(rrc, eptr, next, offset_top, md, ims, eptrb, match_isgroup);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1858        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1859        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1+LINK_SIZE;  
       }  
1860      break;      break;
1861    
1862      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1863        {      next = ecode + 1;
1864        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1865        do next += GET(next,1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1866        RMATCH(rrc, eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb,      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1867          match_isgroup);      ecode++;
1868        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      break;
1869        ecode++;  
1870        }      case OP_SKIPZERO:
1871        next = ecode+1;
1872        do next += GET(next,1); while (*next == OP_ALT);
1873        ecode = next + 1 + LINK_SIZE;
1874      break;      break;
1875    
1876      /* 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
1877      an assertion "group", stop matching and return MATCH_MATCH, but record the      here; just jump to the group, with allow_zero set TRUE. */
1878      current high water mark for use by positive assertions. Do this also  
1879      for the "once" (not-backup up) groups. */      case OP_BRAPOSZERO:
1880        op = *(++ecode);
1881        allow_zero = TRUE;
1882        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1883          goto POSSESSIVE_NON_CAPTURE;
1884    
1885        /* End of a group, repeated or non-repeating. */
1886    
1887      case OP_KET:      case OP_KET:
1888      case OP_KETRMIN:      case OP_KETRMIN:
1889      case OP_KETRMAX:      case OP_KETRMAX:
1890        case OP_KETRPOS:
1891      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
     saved_eptr = eptrb->epb_saved_eptr;  
1892    
1893      /* Back up the stack of bracket start pointers. */      /* If this was a group that remembered the subject start, in order to break
1894        infinite repeats of empty string matches, retrieve the subject start from
1895        the chain. Otherwise, set it NULL. */
1896    
1897      eptrb = eptrb->epb_prev;      if (*prev >= OP_SBRA || *prev == OP_ONCE)
   
     if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
         *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||  
         *prev == OP_ONCE)  
1898        {        {
1899        md->end_match_ptr = eptr;      /* For ONCE */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1900        md->end_offset_top = offset_top;        eptrb = eptrb->epb_prev;              /* Backup to previous group */
       RRETURN(MATCH_MATCH);  
1901        }        }
1902        else saved_eptr = NULL;
1903    
1904      /* 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
1905      group number back at the start and if necessary complete handling an      group, stop matching and return MATCH_MATCH, but record the current high
1906      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
1907        start in case it was changed by \K. */
1908    
1909      if (*prev != OP_COND)      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1910             *prev == OP_ONCE_NC)
1911        {        {
1912        number = *prev - OP_BRA;        md->end_match_ptr = eptr;      /* For ONCE_NC */
1913          md->end_offset_top = offset_top;
1914          md->start_match_ptr = mstart;
1915          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1916          }
1917    
1918        /* 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
1919        the number from a dummy opcode at the start. */      and if necessary complete handling an extraction by setting the offsets and
1920        bumping the high water mark. Whole-pattern recursion is coded as a recurse
1921        into group 0, so it won't be picked up here. Instead, we catch it when the
1922        OP_END is reached. Other recursion is handled here. We just have to record
1923        the current subject position and start match pointer and give a MATCH
1924        return. */
1925    
1926        if (number > EXTRACT_BASIC_MAX) number = GET2(prev, 2+LINK_SIZE);      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1927            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1928          {
1929          number = GET2(prev, 1+LINK_SIZE);
1930        offset = number << 1;        offset = number << 1;
1931    
1932  #ifdef DEBUG  #ifdef PCRE_DEBUG
1933        printf("end bracket %d", number);        printf("end bracket %d", number);
1934        printf("\n");        printf("\n");
1935  #endif  #endif
1936    
1937        /* 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. */  
1938    
1939        if (number > 0)        if (md->recursive != NULL && md->recursive->group_num == number)
1940          {          {
1941          md->capture_last = number;          md->end_match_ptr = eptr;
1942          if (offset >= md->offset_max) md->offset_overflow = TRUE; else          md->start_match_ptr = mstart;
1943            {          RRETURN(MATCH_MATCH);
1944            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;  
           }  
1945    
1946          /* Handle a recursively called group. Restore the offsets        /* Deal with capturing */
1947          appropriately and continue from after the call. */  
1948          md->capture_last = number;
1949          if (offset >= md->offset_max) md->offset_overflow = TRUE; else
1950            {
1951            /* If offset is greater than offset_top, it means that we are
1952            "skipping" a capturing group, and that group's offsets must be marked
1953            unset. In earlier versions of PCRE, all the offsets were unset at the
1954            start of matching, but this doesn't work because atomic groups and
1955            assertions can cause a value to be set that should later be unset.
1956            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1957            part of the atomic group, but this is not on the final matching path,
1958            so must be unset when 2 is set. (If there is no group 2, there is no
1959            problem, because offset_top will then be 2, indicating no capture.) */
1960    
1961          if (md->recursive != NULL && md->recursive->group_num == number)          if (offset > offset_top)
1962            {            {
1963            recursion_info *rec = md->recursive;            register int *iptr = md->offset_vector + offset_top;
1964            DPRINTF(("Recursion (%d) succeeded - continuing\n", number));            register int *iend = md->offset_vector + offset;
1965            md->recursive = rec->prevrec;            while (iptr < iend) *iptr++ = -1;
           md->start_match = rec->save_start;  
           memcpy(md->offset_vector, rec->offset_save,  
             rec->saved_max * sizeof(int));  
           ecode = rec->after_call;  
           ims = original_ims;  
           break;  
1966            }            }
         }  
       }  
1967    
1968      /* Reset the value of the ims flags, in case they got changed during          /* Now make the extraction */
     the group. */  
1969    
1970      ims = original_ims;          md->offset_vector[offset] =
1971      DPRINTF(("ims reset to %02lx\n", ims));            md->offset_vector[md->offset_end - number];
1972            md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1973            if (offset_top <= offset) offset_top = offset + 2;
1974            }
1975          }
1976    
1977      /* For a non-repeating ket, just continue at this level. This also      /* For an ordinary non-repeating ket, just continue at this level. This
1978      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
1979      This is the forcible breaking of infinite loops as implemented in Perl      group. This is the forcible breaking of infinite loops as implemented in
1980      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,
1981      course of events. */      establish a backup point by processing the rest of the pattern at a lower
1982        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
1983        original OP_ONCE level, thereby bypassing intermediate backup points, but
1984        resetting any captures that happened along the way. */
1985    
1986      if (*ecode == OP_KET || eptr == saved_eptr)      if (*ecode == OP_KET || eptr == saved_eptr)
1987        {        {
1988        ecode += 1 + LINK_SIZE;        if (*prev == OP_ONCE)
1989            {
1990            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
1991            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1992            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
1993            RRETURN(MATCH_ONCE);
1994            }
1995          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
1996        break;        break;
1997        }        }
1998    
1999      /* The repeating kets try the rest of the pattern or restart from the      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
2000      preceding bracket, in the appropriate order. In the second case, we can use      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
2001      tail recursion to avoid using another stack frame. */      at a time from the outer level, thus saving stack. */
2002    
2003        if (*ecode == OP_KETRPOS)
2004          {
2005          md->end_match_ptr = eptr;
2006          md->end_offset_top = offset_top;
2007          RRETURN(MATCH_KETRPOS);
2008          }
2009    
2010        /* The normal repeating kets try the rest of the pattern or restart from
2011        the preceding bracket, in the appropriate order. In the second case, we can
2012        use tail recursion to avoid using another stack frame, unless we have an
2013        an atomic group or an unlimited repeat of a group that can match an empty
2014        string. */
2015    
2016      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
2017        {        {
2018        RMATCH(rrc, eptr, ecode + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2019        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2020          if (*prev == OP_ONCE)
2021            {
2022            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2023            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2024            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2025            RRETURN(MATCH_ONCE);
2026            }
2027          if (*prev >= OP_SBRA)    /* Could match an empty string */
2028            {
2029            RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2030            RRETURN(rrc);
2031            }
2032        ecode = prev;        ecode = prev;
       flags = match_isgroup;  
2033        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2034        }        }
2035      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
2036        {        {
2037        RMATCH(rrc, eptr, prev, offset_top, md, ims, eptrb, match_isgroup);        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2038          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2039        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2040          if (*prev == OP_ONCE)
2041            {
2042            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2043            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2044            md->once_target = prev;
2045            RRETURN(MATCH_ONCE);
2046            }
2047        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
2048        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2049        }        }
2050      /* Control never gets here */      /* Control never gets here */
2051    
2052      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2053    
2054      case OP_CIRC:      case OP_CIRC:
2055      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
     if ((ims & PCRE_MULTILINE) != 0)  
       {  
       if (eptr != md->start_subject &&  
           (eptr == md->end_subject ||  
            eptr < md->start_subject + md->nllen ||  
            !IS_NEWLINE(eptr - md->nllen)))  
         RRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
2056    
2057      /* Start of subject assertion */      /* Start of subject assertion */
2058    
# Line 1218  for (;;) Line 2061  for (;;)
2061      ecode++;      ecode++;
2062      break;      break;
2063    
2064        /* Multiline mode: start of subject unless notbol, or after any newline. */
2065    
2066        case OP_CIRCM:
2067        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2068        if (eptr != md->start_subject &&
2069            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2070          RRETURN(MATCH_NOMATCH);
2071        ecode++;
2072        break;
2073    
2074      /* Start of match assertion */      /* Start of match assertion */
2075    
2076      case OP_SOM:      case OP_SOM:
# Line 1225  for (;;) Line 2078  for (;;)
2078      ecode++;      ecode++;
2079      break;      break;
2080    
2081      /* 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. */  
2082    
2083      case OP_DOLL:      case OP_SET_SOM:
2084      if ((ims & PCRE_MULTILINE) != 0)      mstart = eptr;
2085        ecode++;
2086        break;
2087    
2088        /* Multiline mode: assert before any newline, or before end of subject
2089        unless noteol is set. */
2090    
2091        case OP_DOLLM:
2092        if (eptr < md->end_subject)
2093        {        {
2094        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2095          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }          {
2096        else          if (md->partial != 0 &&
2097          { if (md->noteol) RRETURN(MATCH_NOMATCH); }              eptr + 1 >= md->end_subject &&
2098        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2099        break;              NLBLOCK->nllen == 2 &&
2100                RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2101              {
2102              md->hitend = TRUE;
2103              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2104              }
2105            RRETURN(MATCH_NOMATCH);
2106            }
2107        }        }
2108      else      else
2109        {        {
2110        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2111        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (eptr != md->end_subject - md->nllen || !IS_NEWLINE(eptr)))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2112        }        }
2113        ecode++;
2114        break;
2115    
2116        /* Not multiline mode: assert before a terminating newline or before end of
2117        subject unless noteol is set. */
2118    
2119        case OP_DOLL:
2120        if (md->noteol) RRETURN(MATCH_NOMATCH);
2121        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2122    
2123      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2124    
2125      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2126    
2127      case OP_EOD:      case OP_EOD:
2128      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2129        SCHECK_PARTIAL();
2130      ecode++;      ecode++;
2131      break;      break;
2132    
2133      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2134    
2135      case OP_EODN:      case OP_EODN:
2136      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2137          (eptr != md->end_subject - md->nllen || !IS_NEWLINE(eptr)))      if (eptr < md->end_subject &&
2138            (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2139          {
2140          if (md->partial != 0 &&
2141              eptr + 1 >= md->end_subject &&
2142              NLBLOCK->nltype == NLTYPE_FIXED &&
2143              NLBLOCK->nllen == 2 &&
2144              RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2145            {
2146            md->hitend = TRUE;
2147            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2148            }
2149        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2150          }
2151    
2152        /* Either at end of string or \n before end. */
2153    
2154        SCHECK_PARTIAL();
2155      ecode++;      ecode++;
2156      break;      break;
2157    
# Line 1276  for (;;) Line 2163  for (;;)
2163    
2164        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2165        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
2166        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2167          partial matching. */
2168    
2169  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2170        if (utf8)        if (utf)
2171          {          {
2172            /* Get status of previous character */
2173    
2174          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2175            {            {
2176            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2177            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2178              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2179            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2180    #ifdef SUPPORT_UCP
2181              if (md->use_ucp)
2182                {
2183                if (c == '_') prev_is_word = TRUE; else
2184                  {
2185                  int cat = UCD_CATEGORY(c);
2186                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2187                  }
2188                }
2189              else
2190    #endif
2191            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2192            }            }
2193          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2194            /* Get status of next character */
2195    
2196            if (eptr >= md->end_subject)
2197              {
2198              SCHECK_PARTIAL();
2199              cur_is_word = FALSE;
2200              }
2201            else
2202            {            {
2203            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2204    #ifdef SUPPORT_UCP
2205              if (md->use_ucp)
2206                {
2207                if (c == '_') cur_is_word = TRUE; else
2208                  {
2209                  int cat = UCD_CATEGORY(c);
2210                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2211                  }
2212                }
2213              else
2214    #endif
2215            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2216            }            }
2217          }          }
2218        else        else
2219  #endif  #endif
2220    
2221        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2222          consistency with the behaviour of \w we do use it in this case. */
2223    
2224          {          {
2225          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2226            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2227          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2228            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2229              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2230    #ifdef SUPPORT_UCP
2231              if (md->use_ucp)
2232                {
2233                c = eptr[-1];
2234                if (c == '_') prev_is_word = TRUE; else
2235                  {
2236                  int cat = UCD_CATEGORY(c);
2237                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2238                  }
2239                }
2240              else
2241    #endif
2242              prev_is_word = MAX_255(eptr[-1])
2243                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2244              }
2245    
2246            /* Get status of next character */
2247    
2248            if (eptr >= md->end_subject)
2249              {
2250              SCHECK_PARTIAL();
2251              cur_is_word = FALSE;
2252              }
2253            else
2254    #ifdef SUPPORT_UCP
2255            if (md->use_ucp)
2256              {
2257              c = *eptr;
2258              if (c == '_') cur_is_word = TRUE; else
2259                {
2260                int cat = UCD_CATEGORY(c);
2261                cur_is_word = (cat == ucp_L || cat == ucp_N);
2262                }
2263              }
2264            else
2265    #endif
2266            cur_is_word = MAX_255(*eptr)
2267              && ((md->ctypes[*eptr] & ctype_word) != 0);
2268          }          }
2269    
2270        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1314  for (;;) Line 2275  for (;;)
2275        }        }
2276      break;      break;
2277    
2278      /* Match a single character type; inline for speed */      /* Match any single character type except newline; have to take care with
2279        CRLF newlines and partial matching. */
2280    
2281      case OP_ANY:      case OP_ANY:
2282      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2283        if (md->partial != 0 &&
2284            eptr + 1 >= md->end_subject &&
2285            NLBLOCK->nltype == NLTYPE_FIXED &&
2286            NLBLOCK->nllen == 2 &&
2287            RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2288        {        {
2289        if (eptr <= md->end_subject - md->nllen && IS_NEWLINE(eptr))        md->hitend = TRUE;
2290          RRETURN(MATCH_NOMATCH);        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2291          }
2292    
2293        /* Fall through */
2294    
2295        /* Match any single character whatsoever. */
2296    
2297        case OP_ALLANY:
2298        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2299          {                            /* not be updated before SCHECK_PARTIAL. */
2300          SCHECK_PARTIAL();
2301          RRETURN(MATCH_NOMATCH);
2302        }        }
2303      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      eptr++;
2304      if (utf8)  #ifdef SUPPORT_UTF
2305        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;      if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2306    #endif
2307      ecode++;      ecode++;
2308      break;      break;
2309    
# Line 1332  for (;;) Line 2311  for (;;)
2311      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2312    
2313      case OP_ANYBYTE:      case OP_ANYBYTE:
2314      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2315          {                            /* not be updated before SCHECK_PARTIAL. */
2316          SCHECK_PARTIAL();
2317          RRETURN(MATCH_NOMATCH);
2318          }
2319        eptr++;
2320      ecode++;      ecode++;
2321      break;      break;
2322    
2323      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2324      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2325          {
2326          SCHECK_PARTIAL();
2327          RRETURN(MATCH_NOMATCH);
2328          }
2329      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2330      if (      if (
2331  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2332         c < 256 &&         c < 256 &&
2333  #endif  #endif
2334         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1350  for (;;) Line 2338  for (;;)
2338      break;      break;
2339    
2340      case OP_DIGIT:      case OP_DIGIT:
2341      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2342          {
2343          SCHECK_PARTIAL();
2344          RRETURN(MATCH_NOMATCH);
2345          }
2346      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2347      if (      if (
2348  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2349         c >= 256 ||         c > 255 ||
2350  #endif  #endif
2351         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2352         )         )
# Line 1363  for (;;) Line 2355  for (;;)
2355      break;      break;
2356    
2357      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2358      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2359          {
2360          SCHECK_PARTIAL();
2361          RRETURN(MATCH_NOMATCH);
2362          }
2363      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2364      if (      if (
2365  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2366         c < 256 &&         c < 256 &&
2367  #endif  #endif
2368         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1376  for (;;) Line 2372  for (;;)
2372      break;      break;
2373    
2374      case OP_WHITESPACE:      case OP_WHITESPACE:
2375      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2376          {
2377          SCHECK_PARTIAL();
2378          RRETURN(MATCH_NOMATCH);
2379          }
2380      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2381      if (      if (
2382  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2383         c >= 256 ||         c > 255 ||
2384  #endif  #endif
2385         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2386         )         )
# Line 1389  for (;;) Line 2389  for (;;)
2389      break;      break;
2390    
2391      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2392      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2393          {
2394          SCHECK_PARTIAL();
2395          RRETURN(MATCH_NOMATCH);
2396          }
2397      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2398      if (      if (
2399  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2400         c < 256 &&         c < 256 &&
2401  #endif  #endif
2402         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1402  for (;;) Line 2406  for (;;)
2406      break;      break;
2407    
2408      case OP_WORDCHAR:      case OP_WORDCHAR:
2409      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2410          {
2411          SCHECK_PARTIAL();
2412          RRETURN(MATCH_NOMATCH);
2413          }
2414      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2415      if (      if (
2416  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2417         c >= 256 ||         c > 255 ||
2418  #endif  #endif
2419         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2420         )         )
# Line 1414  for (;;) Line 2422  for (;;)
2422      ecode++;      ecode++;
2423      break;      break;
2424    
2425        case OP_ANYNL:
2426        if (eptr >= md->end_subject)
2427          {
2428          SCHECK_PARTIAL();
2429          RRETURN(MATCH_NOMATCH);
2430          }
2431        GETCHARINCTEST(c, eptr);
2432        switch(c)
2433          {
2434          default: RRETURN(MATCH_NOMATCH);
2435    
2436          case CHAR_CR:
2437          if (eptr >= md->end_subject)
2438            {
2439            SCHECK_PARTIAL();
2440            }
2441          else if (RAWUCHARTEST(eptr) == CHAR_LF) eptr++;
2442          break;
2443    
2444          case CHAR_LF:
2445          break;
2446    
2447          case CHAR_VT:
2448          case CHAR_FF:
2449          case CHAR_NEL:
2450    #ifndef EBCDIC
2451          case 0x2028:
2452          case 0x2029:
2453    #endif  /* Not EBCDIC */
2454          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2455          break;
2456          }
2457        ecode++;
2458        break;
2459    
2460        case OP_NOT_HSPACE:
2461        if (eptr >= md->end_subject)
2462          {
2463          SCHECK_PARTIAL();
2464          RRETURN(MATCH_NOMATCH);
2465          }
2466        GETCHARINCTEST(c, eptr);
2467        switch(c)
2468          {
2469          HSPACE_CASES: RRETURN(MATCH_NOMATCH);  /* Byte and multibyte cases */
2470          default: break;
2471          }
2472        ecode++;
2473        break;
2474    
2475        case OP_HSPACE:
2476        if (eptr >= md->end_subject)
2477          {
2478          SCHECK_PARTIAL();
2479          RRETURN(MATCH_NOMATCH);
2480          }
2481        GETCHARINCTEST(c, eptr);
2482        switch(c)
2483          {
2484          HSPACE_CASES: break;  /* Byte and multibyte cases */
2485          default: RRETURN(MATCH_NOMATCH);
2486          }
2487        ecode++;
2488        break;
2489    
2490        case OP_NOT_VSPACE:
2491        if (eptr >= md->end_subject)
2492          {
2493          SCHECK_PARTIAL();
2494          RRETURN(MATCH_NOMATCH);
2495          }
2496        GETCHARINCTEST(c, eptr);
2497        switch(c)
2498          {
2499          VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2500          default: break;
2501          }
2502        ecode++;
2503        break;
2504    
2505        case OP_VSPACE:
2506        if (eptr >= md->end_subject)
2507          {
2508          SCHECK_PARTIAL();
2509          RRETURN(MATCH_NOMATCH);
2510          }
2511        GETCHARINCTEST(c, eptr);
2512        switch(c)
2513          {
2514          VSPACE_CASES: break;
2515          default: RRETURN(MATCH_NOMATCH);
2516          }
2517        ecode++;
2518        break;
2519    
2520  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2521      /* Check the next character by Unicode property. We will get here only      /* Check the next character by Unicode property. We will get here only
2522      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. */
2523    
2524      case OP_PROP:      case OP_PROP:
2525      case OP_NOTPROP:      case OP_NOTPROP:
2526      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2527          {
2528          SCHECK_PARTIAL();
2529          RRETURN(MATCH_NOMATCH);
2530          }
2531      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2532        {        {
2533        int chartype, script;        const pcre_uint32 *cp;
2534        int category = _pcre_ucp_findprop(c, &chartype, &script);        const ucd_record *prop = GET_UCD(c);
2535    
2536        switch(ecode[1])        switch(ecode[1])
2537          {          {
# Line 1433  for (;;) Line 2540  for (;;)
2540          break;          break;
2541    
2542          case PT_LAMP:          case PT_LAMP:
2543          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2544               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2545               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2546            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2547           break;          break;
2548    
2549          case PT_GC:          case PT_GC:
2550          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2551            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2552          break;          break;
2553    
2554          case PT_PC:          case PT_PC:
2555          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2556            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2557          break;          break;
2558    
2559          case PT_SC:          case PT_SC:
2560          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2561              RRETURN(MATCH_NOMATCH);
2562            break;
2563    
2564            /* These are specials */
2565    
2566            case PT_ALNUM:
2567            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2568                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2569              RRETURN(MATCH_NOMATCH);
2570            break;
2571    
2572            case PT_SPACE:    /* Perl space */
2573            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2574                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2575                   == (op == OP_NOTPROP))
2576              RRETURN(MATCH_NOMATCH);
2577            break;
2578    
2579            case PT_PXSPACE:  /* POSIX space */
2580            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2581                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2582                 c == CHAR_FF || c == CHAR_CR)
2583                   == (op == OP_NOTPROP))
2584              RRETURN(MATCH_NOMATCH);
2585            break;
2586    
2587            case PT_WORD:
2588            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2589                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2590                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2591            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2592          break;          break;
2593    
2594            case PT_CLIST:
2595            cp = PRIV(ucd_caseless_sets) + prop->caseset;
2596            for (;;)
2597              {
2598              if (c < *cp)
2599                { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; }
2600              if (c == *cp++)
2601                { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } }
2602              }
2603            break;
2604    
2605            /* This should never occur */
2606    
2607          default:          default:
2608          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
         break;  
2609          }          }
2610    
2611        ecode += 3;        ecode += 3;
# Line 1467  for (;;) Line 2616  for (;;)
2616      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2617    
2618      case OP_EXTUNI:      case OP_EXTUNI:
2619      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
     GETCHARINCTEST(c, eptr);  
2620        {        {
2621        int chartype, script;        SCHECK_PARTIAL();
2622        int category = _pcre_ucp_findprop(c, &chartype, &script);        RRETURN(MATCH_NOMATCH);
2623        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        }
2624        else
2625          {
2626          int lgb, rgb;
2627          GETCHARINCTEST(c, eptr);
2628          lgb = UCD_GRAPHBREAK(c);
2629        while (eptr < md->end_subject)        while (eptr < md->end_subject)
2630          {          {
2631          int len = 1;          int len = 1;
2632          if (!utf8) c = *eptr; else          if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2633            {          rgb = UCD_GRAPHBREAK(c);
2634            GETCHARLEN(c, eptr, len);          if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2635            }          lgb = rgb;
         category = _pcre_ucp_findprop(c, &chartype, &script);  
         if (category != ucp_M) break;  
2636          eptr += len;          eptr += len;
2637          }          }
2638        }        }
2639        CHECK_PARTIAL();
2640      ecode++;      ecode++;
2641      break;      break;
2642  #endif  #endif  /* SUPPORT_UCP */
2643    
2644    
2645      /* 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 1499  for (;;) Line 2651  for (;;)
2651      loops). */      loops). */
2652    
2653      case OP_REF:      case OP_REF:
2654        {      case OP_REFI:
2655        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2656        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2657        ecode += 1 + IMM2_SIZE;
2658    
2659        /* 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];  
2660    
2661        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2662        this ensures that every attempt at a match fails. We can't just fail
2663        here, because of the possibility of quantifiers with zero minima.
2664    
2665        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2666          {      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;  
2667    
2668          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2669          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;  
2670    
2671          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2672          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2673          eptr += length;      else
2674          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2675    
2676        /* Set up for repetition, or handle the non-repeated case */
2677    
2678        switch (*ecode)
2679          {
2680          case OP_CRSTAR:
2681          case OP_CRMINSTAR:
2682          case OP_CRPLUS:
2683          case OP_CRMINPLUS:
2684          case OP_CRQUERY:
2685          case OP_CRMINQUERY:
2686          c = *ecode++ - OP_CRSTAR;
2687          minimize = (c & 1) != 0;
2688          min = rep_min[c];                 /* Pick up values from tables; */
2689          max = rep_max[c];                 /* zero for max => infinity */
2690          if (max == 0) max = INT_MAX;
2691          break;
2692    
2693          case OP_CRRANGE:
2694          case OP_CRMINRANGE:
2695          minimize = (*ecode == OP_CRMINRANGE);
2696          min = GET2(ecode, 1);
2697          max = GET2(ecode, 1 + IMM2_SIZE);
2698          if (max == 0) max = INT_MAX;
2699          ecode += 1 + 2 * IMM2_SIZE;
2700          break;
2701    
2702          default:               /* No repeat follows */
2703          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2704            {
2705            if (length == -2) eptr = md->end_subject;   /* Partial match */
2706            CHECK_PARTIAL();
2707            RRETURN(MATCH_NOMATCH);
2708          }          }
2709          eptr += length;
2710          continue;              /* With the main loop */
2711          }
2712    
2713        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2714        main loop. */      zero, just continue with the main loop. If the length is negative, it
2715        means the reference is unset in non-Java-compatible mode. If the minimum is
2716        zero, we can continue at the same level without recursion. For any other
2717        minimum, carrying on will result in NOMATCH. */
2718    
2719        if (length == 0) continue;      if (length == 0) continue;
2720        if (length < 0 && min == 0) continue;
2721    
2722        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2723        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2724        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2725    
2726        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2727          {
2728          int slength;
2729          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2730          {          {
2731          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          if (slength == -2) eptr = md->end_subject;   /* Partial match */
2732          eptr += length;          CHECK_PARTIAL();
2733            RRETURN(MATCH_NOMATCH);
2734          }          }
2735          eptr += slength;
2736          }
2737    
2738        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2739        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2740    
2741        if (min == max) continue;      if (min == max) continue;
2742    
2743        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2744    
2745        if (minimize)      if (minimize)
2746          {
2747          for (fi = min;; fi++)
2748          {          {
2749          for (fi = min;; fi++)          int slength;
2750            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2751            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2752            if (fi >= max) RRETURN(MATCH_NOMATCH);
2753            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2754            {            {
2755            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2756            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2757            if (fi >= max || !match_ref(offset, eptr, length, md, ims))            RRETURN(MATCH_NOMATCH);
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2758            }            }
2759          /* Control never gets here */          eptr += slength;
2760          }          }
2761          /* Control never gets here */
2762          }
2763    
2764        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2765    
2766        else      else
2767          {
2768          pp = eptr;
2769          for (i = min; i < max; i++)
2770          {          {
2771          pp = eptr;          int slength;
2772          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)  
2773            {            {
2774            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2775            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            the soft partial matching case. */
2776            eptr -= length;  
2777              if (slength == -2 && md->partial != 0 &&
2778                  md->end_subject > md->start_used_ptr)
2779                {
2780                md->hitend = TRUE;
2781                if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2782                }
2783              break;
2784            }            }
2785          RRETURN(MATCH_NOMATCH);          eptr += slength;
2786          }          }
2787    
2788          while (eptr >= pp)
2789            {
2790            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2791            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2792            eptr -= length;
2793            }
2794          RRETURN(MATCH_NOMATCH);
2795        }        }
2796      /* Control never gets here */      /* Control never gets here */
2797    
   
   
2798      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2799      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,
2800      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 1616  for (;;) Line 2809  for (;;)
2809      case OP_NCLASS:      case OP_NCLASS:
2810      case OP_CLASS:      case OP_CLASS:
2811        {        {
2812          /* The data variable is saved across frames, so the byte map needs to
2813          be stored there. */
2814    #define BYTE_MAP ((pcre_uint8 *)data)
2815        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2816        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2817    
2818        switch (*ecode)        switch (*ecode)
2819          {          {
# Line 1638  for (;;) Line 2834  for (;;)
2834          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2835          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2836          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2837          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2838          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2839          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2840          break;          break;
2841    
2842          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1650  for (;;) Line 2846  for (;;)
2846    
2847        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2848    
2849  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2850        /* UTF-8 mode */        if (utf)
       if (utf8)  
2851          {          {
2852          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2853            {            {
2854            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2855                {
2856                SCHECK_PARTIAL();
2857                RRETURN(MATCH_NOMATCH);
2858                }
2859            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2860            if (c > 255)            if (c > 255)
2861              {              {
2862              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2863              }              }
2864            else            else
2865              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2866            }            }
2867          }          }
2868        else        else
2869  #endif  #endif
2870        /* Not UTF-8 mode */        /* Not UTF mode */
2871          {          {
2872          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2873            {            {
2874            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2875                {
2876                SCHECK_PARTIAL();
2877                RRETURN(MATCH_NOMATCH);
2878                }
2879            c = *eptr++;            c = *eptr++;
2880            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2881              if (c > 255)
2882                {
2883                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2884                }
2885              else
2886    #endif
2887                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2888            }            }
2889          }          }
2890    
# Line 1690  for (;;) Line 2898  for (;;)
2898    
2899        if (minimize)        if (minimize)
2900          {          {
2901  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2902          /* UTF-8 mode */          if (utf)
         if (utf8)  
2903            {            {
2904            for (fi = min;; fi++)            for (fi = min;; fi++)
2905              {              {
2906              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2907              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2908              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2909                if (eptr >= md->end_subject)
2910                  {
2911                  SCHECK_PARTIAL();
2912                  RRETURN(MATCH_NOMATCH);
2913                  }
2914              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2915              if (c > 255)              if (c > 255)
2916                {                {
2917                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2918                }                }
2919              else              else
2920                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2921              }              }
2922            }            }
2923          else          else
2924  #endif  #endif
2925          /* Not UTF-8 mode */          /* Not UTF mode */
2926            {            {
2927            for (fi = min;; fi++)            for (fi = min;; fi++)
2928              {              {
2929              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2930              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2931              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2932                if (eptr >= md->end_subject)
2933                  {
2934                  SCHECK_PARTIAL();
2935                  RRETURN(MATCH_NOMATCH);
2936                  }
2937              c = *eptr++;              c = *eptr++;
2938              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2939                if (c > 255)
2940                  {
2941                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2942                  }
2943                else
2944    #endif
2945                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2946              }              }
2947            }            }
2948          /* Control never gets here */          /* Control never gets here */
# Line 1732  for (;;) Line 2954  for (;;)
2954          {          {
2955          pp = eptr;          pp = eptr;
2956    
2957  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2958          /* UTF-8 mode */          if (utf)
         if (utf8)  
2959            {            {
2960            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2961              {              {
2962              int len = 1;              int len = 1;
2963              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2964                  {
2965                  SCHECK_PARTIAL();
2966                  break;
2967                  }
2968              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
2969              if (c > 255)              if (c > 255)
2970                {                {
2971                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
2972                }                }
2973              else              else
2974                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
2975              eptr += len;              eptr += len;
2976              }              }
2977            for (;;)            for (;;)
2978              {              {
2979              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
2980              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2981              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
2982              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1761  for (;;) Line 2984  for (;;)
2984            }            }
2985          else          else
2986  #endif  #endif
2987            /* Not UTF-8 mode */            /* Not UTF mode */
2988            {            {
2989            for (i = min; i < max; i++)            for (i = min; i < max; i++)
2990              {              {
2991              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
2992                  {
2993                  SCHECK_PARTIAL();
2994                  break;
2995                  }
2996              c = *eptr;              c = *eptr;
2997              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
2998                if (c > 255)
2999                  {
3000                  if (op == OP_CLASS) break;
3001                  }
3002                else
3003    #endif
3004                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3005              eptr++;              eptr++;
3006              }              }
3007            while (eptr >= pp)            while (eptr >= pp)
3008              {              {
3009              RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3010              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3011              eptr--;              eptr--;
3012              }              }
# Line 1780  for (;;) Line 3014  for (;;)
3014    
3015          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3016          }          }
3017    #undef BYTE_MAP
3018        }        }
3019      /* Control never gets here */      /* Control never gets here */
3020    
3021    
3022      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
3023      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
3024        mode, because Unicode properties are supported in non-UTF-8 mode. */
3025    
3026  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3027      case OP_XCLASS:      case OP_XCLASS:
3028        {        {
3029        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 1812  for (;;) Line 3048  for (;;)
3048          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3049          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3050          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3051          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3052          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3053          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3054          break;          break;
3055    
3056          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1826  for (;;) Line 3062  for (;;)
3062    
3063        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3064          {          {
3065          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3066          GETCHARINC(c, eptr);            {
3067          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
3068              RRETURN(MATCH_NOMATCH);
3069              }
3070            GETCHARINCTEST(c, eptr);
3071            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3072          }          }
3073    
3074        /* 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 1843  for (;;) Line 3083  for (;;)
3083          {          {
3084          for (fi = min;; fi++)          for (fi = min;; fi++)
3085            {            {
3086            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3087            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3088            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3089            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3090            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3091                SCHECK_PARTIAL();
3092                RRETURN(MATCH_NOMATCH);
3093                }
3094              GETCHARINCTEST(c, eptr);
3095              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3096            }            }
3097          /* Control never gets here */          /* Control never gets here */
3098          }          }
# Line 1860  for (;;) Line 3105  for (;;)
3105          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3106            {            {
3107            int len = 1;            int len = 1;
3108            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3109            GETCHARLEN(c, eptr, len);              {
3110            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3111                break;
3112                }
3113    #ifdef SUPPORT_UTF
3114              GETCHARLENTEST(c, eptr, len);
3115    #else
3116              c = *eptr;
3117    #endif
3118              if (!PRIV(xclass)(c, data, utf)) break;
3119            eptr += len;            eptr += len;
3120            }            }
3121          for(;;)          for(;;)
3122            {            {
3123            RMATCH(rrc, eptr, ecode, offset_top, md, ims, eptrb, 0);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3124            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3125            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3126            BACKCHAR(eptr)  #ifdef SUPPORT_UTF
3127              if (utf) BACKCHAR(eptr);
3128    #endif
3129            }            }
3130          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3131          }          }
# Line 1882  for (;;) Line 3137  for (;;)
3137      /* Match a single character, casefully */      /* Match a single character, casefully */
3138    
3139      case OP_CHAR:      case OP_CHAR:
3140  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3141      if (utf8)      if (utf)
3142        {        {
3143        length = 1;        length = 1;
3144        ecode++;        ecode++;
3145        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3146        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3147        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);          {
3148            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3149            RRETURN(MATCH_NOMATCH);
3150            }
3151          while (length-- > 0) if (*ecode++ != RAWUCHARINC(eptr)) RRETURN(MATCH_NOMATCH);
3152        }        }
3153      else      else
3154  #endif  #endif
3155        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3156        {        {
3157        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3158            {
3159            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3160            RRETURN(MATCH_NOMATCH);
3161            }
3162        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3163        ecode += 2;        ecode += 2;
3164        }        }
3165      break;      break;
3166    
3167      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3168        subject, give up immediately. */
3169    
3170        case OP_CHARI:
3171        if (eptr >= md->end_subject)
3172          {
3173          SCHECK_PARTIAL();
3174          RRETURN(MATCH_NOMATCH);
3175          }
3176    
3177      case OP_CHARNC:  #ifdef SUPPORT_UTF
3178  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3179        {        {
3180        length = 1;        length = 1;
3181        ecode++;        ecode++;
3182        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3183    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3184        /* 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
3185        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3186          fast lookup table. We know that there is at least one byte left in the
3187          subject. */
3188    
3189        if (fc < 128)        if (fc < 128)
3190          {          {
3191          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          pcre_uchar cc = RAWUCHAR(eptr);
3192            if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH);
3193            ecode++;
3194            eptr++;
3195          }          }
3196    
3197        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3198          use the value of "length" to check for sufficient bytes left, because the
3199          other case of the character may have more or fewer bytes.  */
3200    
3201        else        else
3202          {          {
3203          int dc;          pcre_uint32 dc;
3204          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
3205          ecode += length;          ecode += length;
3206    
# Line 1936  for (;;) Line 3210  for (;;)
3210          if (fc != dc)          if (fc != dc)
3211            {            {
3212  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3213            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3214  #endif  #endif
3215              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3216            }            }
3217          }          }
3218        }        }
3219      else      else
3220  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3221    
3222      /* Non-UTF-8 mode */      /* Not UTF mode */
3223        {        {
3224        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3225        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3226          eptr++;
3227        ecode += 2;        ecode += 2;
3228        }        }
3229      break;      break;
3230    
3231      /* Match a single character repeatedly; different opcodes share code. */      /* Match a single character repeatedly. */
3232    
3233      case OP_EXACT:      case OP_EXACT:
3234        case OP_EXACTI:
3235      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3236      ecode += 3;      ecode += 1 + IMM2_SIZE;
3237      goto REPEATCHAR;      goto REPEATCHAR;
3238    
3239        case OP_POSUPTO:
3240        case OP_POSUPTOI:
3241        possessive = TRUE;
3242        /* Fall through */
3243    
3244      case OP_UPTO:      case OP_UPTO:
3245        case OP_UPTOI:
3246      case OP_MINUPTO:      case OP_MINUPTO:
3247        case OP_MINUPTOI:
3248      min = 0;      min = 0;
3249      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3250      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3251      ecode += 3;      ecode += 1 + IMM2_SIZE;
3252        goto REPEATCHAR;
3253    
3254        case OP_POSSTAR:
3255        case OP_POSSTARI:
3256        possessive = TRUE;
3257        min = 0;
3258        max = INT_MAX;
3259        ecode++;
3260        goto REPEATCHAR;
3261    
3262        case OP_POSPLUS:
3263        case OP_POSPLUSI:
3264        possessive = TRUE;
3265        min = 1;
3266        max = INT_MAX;
3267        ecode++;
3268        goto REPEATCHAR;
3269    
3270        case OP_POSQUERY:
3271        case OP_POSQUERYI:
3272        possessive = TRUE;
3273        min = 0;
3274        max = 1;
3275        ecode++;
3276      goto REPEATCHAR;      goto REPEATCHAR;
3277    
3278      case OP_STAR:      case OP_STAR:
3279        case OP_STARI:
3280      case OP_MINSTAR:      case OP_MINSTAR:
3281        case OP_MINSTARI:
3282      case OP_PLUS:      case OP_PLUS:
3283        case OP_PLUSI:
3284      case OP_MINPLUS:      case OP_MINPLUS:
3285        case OP_MINPLUSI:
3286      case OP_QUERY:      case OP_QUERY:
3287        case OP_QUERYI:
3288      case OP_MINQUERY:      case OP_MINQUERY:
3289      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3290        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3291      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3292      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3293      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3294      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3295    
3296      /* Common code for all repeated single-character matches. We can give      /* Common code for all repeated single-character matches. */
     up quickly if there are fewer than the minimum number of characters left in  
     the subject. */  
3297    
3298      REPEATCHAR:      REPEATCHAR:
3299  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3300      if (utf8)      if (utf)
3301        {        {
3302        length = 1;