/[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 200 by ph10, Wed Aug 1 09:10:40 2007 UTC revision 1273 by ph10, Fri Mar 8 10:25:02 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 37  POSSIBILITY OF SUCH DAMAGE. Line 37  POSSIBILITY OF SUCH DAMAGE.
37  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
38  */  */
39    
   
40  /* This module contains pcre_exec(), the externally visible function that does  /* This module contains pcre_exec(), the externally visible function that does
41  pattern matching using an NFA algorithm, trying to mimic Perl as closely as  pattern matching using an NFA algorithm, trying to mimic Perl as closely as
42  possible. There are also some static supporting functions. */  possible. There are also some static supporting functions. */
43    
44  #ifdef HAVE_CONFIG_H  #ifdef HAVE_CONFIG_H
45  #include <config.h>  #include "config.h"
46  #endif  #endif
47    
48  #define NLBLOCK md             /* Block containing newline information */  #define NLBLOCK md             /* Block containing newline information */
# Line 57  possible. There are also some static sup Line 56  possible. There are also some static sup
56  #undef min  #undef min
57  #undef max  #undef max
58    
59  /* Flag bits for the match() function */  /* The md->capture_last field uses the lower 16 bits for the last captured
60    substring (which can never be greater than 65535) and a bit in the top half
61    to mean "capture vector overflowed". This odd way of doing things was
62    implemented when it was realized that preserving and restoring the overflow bit
63    whenever the last capture number was saved/restored made for a neater
64    interface, and doing it this way saved on (a) another variable, which would
65    have increased the stack frame size (a big NO-NO in PCRE) and (b) another
66    separate set of save/restore instructions. The following defines are used in
67    implementing this. */
68    
69    #define CAPLMASK    0x0000ffff    /* The bits used for last_capture */
70    #define OVFLMASK    0xffff0000    /* The bits used for the overflow flag */
71    #define OVFLBIT     0x00010000    /* The bit that is set for overflow */
72    
73    /* Values for setting in md->match_function_type to indicate two special types
74    of call to match(). We do it this way to save on using another stack variable,
75    as stack usage is to be discouraged. */
76    
77  #define match_condassert     0x01  /* Called to check a condition assertion */  #define MATCH_CONDASSERT     1  /* Called to check a condition assertion */
78  #define match_cbegroup       0x02  /* Could-be-empty unlimited repeat group */  #define MATCH_CBEGROUP       2  /* Could-be-empty unlimited repeat group */
79    
80  /* Non-error returns from the match() function. Error returns are externally  /* Non-error returns from the match() function. Error returns are externally
81  defined PCRE_ERROR_xxx codes, which are all negative. */  defined PCRE_ERROR_xxx codes, which are all negative. */
# Line 68  defined PCRE_ERROR_xxx codes, which are Line 83  defined PCRE_ERROR_xxx codes, which are
83  #define MATCH_MATCH        1  #define MATCH_MATCH        1
84  #define MATCH_NOMATCH      0  #define MATCH_NOMATCH      0
85    
86    /* Special internal returns from the match() function. Make them sufficiently
87    negative to avoid the external error codes. */
88    
89    #define MATCH_ACCEPT       (-999)
90    #define MATCH_KETRPOS      (-998)
91    #define MATCH_ONCE         (-997)
92    /* The next 5 must be kept together and in sequence so that a test that checks
93    for any one of them can use a range. */
94    #define MATCH_COMMIT       (-996)
95    #define MATCH_PRUNE        (-995)
96    #define MATCH_SKIP         (-994)
97    #define MATCH_SKIP_ARG     (-993)
98    #define MATCH_THEN         (-992)
99    #define MATCH_BACKTRACK_MAX MATCH_THEN
100    #define MATCH_BACKTRACK_MIN MATCH_COMMIT
101    
102  /* Maximum number of ints of offset to save on the stack for recursive calls.  /* Maximum number of ints of offset to save on the stack for recursive calls.
103  If the offset vector is bigger, malloc is used. This should be a multiple of 3,  If the offset vector is bigger, malloc is used. This should be a multiple of 3,
104  because the offset vector is always a multiple of 3 long. */  because the offset vector is always a multiple of 3 long. */
# Line 79  because the offset vector is always a mu Line 110  because the offset vector is always a mu
110  static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };  static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };
111  static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };  static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };
112    
113    #ifdef PCRE_DEBUG
   
 #ifdef DEBUG  
114  /*************************************************  /*************************************************
115  *        Debugging function to print chars       *  *        Debugging function to print chars       *
116  *************************************************/  *************************************************/
# Line 99  Returns:     nothing Line 128  Returns:     nothing
128  */  */
129    
130  static void  static void
131  pchars(const uschar *p, int length, BOOL is_subject, match_data *md)  pchars(const pcre_uchar *p, int length, BOOL is_subject, match_data *md)
132  {  {
133  unsigned int c;  pcre_uint32 c;
134    BOOL utf = md->utf;
135  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;  if (is_subject && length > md->end_subject - p) length = md->end_subject - p;
136  while (length-- > 0)  while (length-- > 0)
137    if (isprint(c = *(p++))) printf("%c", c); else printf("\\x%02x", c);    if (isprint(c = RAWUCHARINCTEST(p))) printf("%c", (char)c); else printf("\\x{%02x}", c);
138  }  }
139  #endif  #endif
140    
# Line 114  while (length-- > 0) Line 144  while (length-- > 0)
144  *          Match a back-reference                *  *          Match a back-reference                *
145  *************************************************/  *************************************************/
146    
147  /* If a back reference hasn't been set, the length that is passed is greater  /* Normally, if a back reference hasn't been set, the length that is passed is
148  than the number of characters left in the string, so the match fails.  negative, so the match always fails. However, in JavaScript compatibility mode,
149    the length passed is zero. Note that in caseless UTF-8 mode, the number of
150    subject bytes matched may be different to the number of reference bytes.
151    
152  Arguments:  Arguments:
153    offset      index into the offset vector    offset      index into the offset vector
154    eptr        points into the subject    eptr        pointer into the subject
155    length      length to be matched    length      length of reference to be matched (number of bytes)
156    md          points to match data block    md          points to match data block
157    ims         the ims flags    caseless    TRUE if caseless
158    
159  Returns:      TRUE if matched  Returns:      >= 0 the number of subject bytes matched
160                  -1 no match
161                  -2 partial match; always given if at end subject
162  */  */
163    
164  static BOOL  static int
165  match_ref(int offset, register USPTR eptr, int length, match_data *md,  match_ref(int offset, register PCRE_PUCHAR eptr, int length, match_data *md,
166    unsigned long int ims)    BOOL caseless)
167  {  {
168  USPTR p = md->start_subject + md->offset_vector[offset];  PCRE_PUCHAR eptr_start = eptr;
169    register PCRE_PUCHAR p = md->start_subject + md->offset_vector[offset];
170    #ifdef SUPPORT_UTF
171    BOOL utf = md->utf;
172    #endif
173    
174  #ifdef DEBUG  #ifdef PCRE_DEBUG
175  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
176    printf("matching subject <null>");    printf("matching subject <null>");
177  else  else
# Line 146  pchars(p, length, FALSE, md); Line 184  pchars(p, length, FALSE, md);
184  printf("\n");  printf("\n");
185  #endif  #endif
186    
187  /* Always fail if not enough characters left */  /* Always fail if reference not set (and not JavaScript compatible - in that
188    case the length is passed as zero). */
189    
190  if (length > md->end_subject - eptr) return FALSE;  if (length < 0) return -1;
191    
192  /* Separate the caselesss case for speed */  /* Separate the caseless case for speed. In UTF-8 mode we can only do this
193    properly if Unicode properties are supported. Otherwise, we can check only
194    ASCII characters. */
195    
196  if ((ims & PCRE_CASELESS) != 0)  if (caseless)
197    {    {
198    while (length-- > 0)  #ifdef SUPPORT_UTF
199      if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;  #ifdef SUPPORT_UCP
200      if (utf)
201        {
202        /* Match characters up to the end of the reference. NOTE: the number of
203        data units matched may differ, because in UTF-8 there are some characters
204        whose upper and lower case versions code have different numbers of bytes.
205        For example, U+023A (2 bytes in UTF-8) is the upper case version of U+2C65
206        (3 bytes in UTF-8); a sequence of 3 of the former uses 6 bytes, as does a
207        sequence of two of the latter. It is important, therefore, to check the
208        length along the reference, not along the subject (earlier code did this
209        wrong). */
210    
211        PCRE_PUCHAR endptr = p + length;
212        while (p < endptr)
213          {
214          pcre_uint32 c, d;
215          const ucd_record *ur;
216          if (eptr >= md->end_subject) return -2;   /* Partial match */
217          GETCHARINC(c, eptr);
218          GETCHARINC(d, p);
219          ur = GET_UCD(d);
220          if (c != d && c != d + ur->other_case)
221            {
222            const pcre_uint32 *pp = PRIV(ucd_caseless_sets) + ur->caseset;
223            for (;;)
224              {
225              if (c < *pp) return -1;
226              if (c == *pp++) break;
227              }
228            }
229          }
230        }
231      else
232    #endif
233    #endif
234    
235      /* The same code works when not in UTF-8 mode and in UTF-8 mode when there
236      is no UCP support. */
237        {
238        while (length-- > 0)
239          {
240          pcre_uint32 cc, cp;
241          if (eptr >= md->end_subject) return -2;   /* Partial match */
242          cc = RAWUCHARTEST(eptr);
243          cp = RAWUCHARTEST(p);
244          if (TABLE_GET(cp, md->lcc, cp) != TABLE_GET(cc, md->lcc, cc)) return -1;
245          p++;
246          eptr++;
247          }
248        }
249    }    }
250    
251    /* In the caseful case, we can just compare the bytes, whether or not we
252    are in UTF-8 mode. */
253    
254  else  else
255    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    {
256      while (length-- > 0)
257        {
258        if (eptr >= md->end_subject) return -2;   /* Partial match */
259        if (RAWUCHARINCTEST(p) != RAWUCHARINCTEST(eptr)) return -1;
260        }
261      }
262    
263  return TRUE;  return (int)(eptr - eptr_start);
264  }  }
265    
266    
# Line 203  variable instead of being passed in the Line 303  variable instead of being passed in the
303  ****************************************************************************  ****************************************************************************
304  ***************************************************************************/  ***************************************************************************/
305    
306    /* Numbers for RMATCH calls. When this list is changed, the code at HEAP_RETURN
307  /* Numbers for RMATCH calls */  below must be updated in sync.  */
308    
309  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,  enum { RM1=1, RM2,  RM3,  RM4,  RM5,  RM6,  RM7,  RM8,  RM9,  RM10,
310         RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,         RM11,  RM12, RM13, RM14, RM15, RM16, RM17, RM18, RM19, RM20,
311         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,         RM21,  RM22, RM23, RM24, RM25, RM26, RM27, RM28, RM29, RM30,
312         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,         RM31,  RM32, RM33, RM34, RM35, RM36, RM37, RM38, RM39, RM40,
313         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50 };         RM41,  RM42, RM43, RM44, RM45, RM46, RM47, RM48, RM49, RM50,
314           RM51,  RM52, RM53, RM54, RM55, RM56, RM57, RM58, RM59, RM60,
315           RM61,  RM62, RM63, RM64, RM65, RM66, RM67, RM68 };
316    
317  /* These versions of the macros use the stack, as normal. There are debugging  /* These versions of the macros use the stack, as normal. There are debugging
318  versions and production versions. Note that the "rw" argument of RMATCH isn't  versions and production versions. Note that the "rw" argument of RMATCH isn't
319  actuall used in this definition. */  actually used in this definition. */
320    
321  #ifndef NO_RECURSE  #ifndef NO_RECURSE
322  #define REGISTER register  #define REGISTER register
323    
324  #ifdef DEBUG  #ifdef PCRE_DEBUG
325  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
326    { \    { \
327    printf("match() called in line %d\n", __LINE__); \    printf("match() called in line %d\n", __LINE__); \
328    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1); \    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1); \
329    printf("to line %d\n", __LINE__); \    printf("to line %d\n", __LINE__); \
330    }    }
331  #define RRETURN(ra) \  #define RRETURN(ra) \
332    { \    { \
333    printf("match() returned %d from line %d ", ra, __LINE__); \    printf("match() returned %d from line %d\n", ra, __LINE__); \
334    return ra; \    return ra; \
335    }    }
336  #else  #else
337  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw) \  #define RMATCH(ra,rb,rc,rd,re,rw) \
338    rrc = match(ra,rb,mstart,rc,rd,re,rf,rg,rdepth+1)    rrc = match(ra,rb,mstart,rc,rd,re,rdepth+1)
339  #define RRETURN(ra) return ra  #define RRETURN(ra) return ra
340  #endif  #endif
341    
# Line 247  argument of match(), which never changes Line 348  argument of match(), which never changes
348    
349  #define REGISTER  #define REGISTER
350    
351  #define RMATCH(ra,rb,rc,rd,re,rf,rg,rw)\  #define RMATCH(ra,rb,rc,rd,re,rw)\
352    {\    {\
353    heapframe *newframe = (pcre_stack_malloc)(sizeof(heapframe));\    heapframe *newframe = frame->Xnextframe;\
354    frame->Xwhere = rw; \    if (newframe == NULL)\
355        {\
356        newframe = (heapframe *)(PUBL(stack_malloc))(sizeof(heapframe));\
357        if (newframe == NULL) RRETURN(PCRE_ERROR_NOMEMORY);\
358        newframe->Xnextframe = NULL;\
359        frame->Xnextframe = newframe;\
360        }\
361      frame->Xwhere = rw;\
362    newframe->Xeptr = ra;\    newframe->Xeptr = ra;\
363    newframe->Xecode = rb;\    newframe->Xecode = rb;\
364    newframe->Xmstart = mstart;\    newframe->Xmstart = mstart;\
365    newframe->Xoffset_top = rc;\    newframe->Xoffset_top = rc;\
366    newframe->Xims = re;\    newframe->Xeptrb = re;\
   newframe->Xeptrb = rf;\  
   newframe->Xflags = rg;\  
367    newframe->Xrdepth = frame->Xrdepth + 1;\    newframe->Xrdepth = frame->Xrdepth + 1;\
368    newframe->Xprevframe = frame;\    newframe->Xprevframe = frame;\
369    frame = newframe;\    frame = newframe;\
# Line 269  argument of match(), which never changes Line 375  argument of match(), which never changes
375    
376  #define RRETURN(ra)\  #define RRETURN(ra)\
377    {\    {\
378    heapframe *newframe = frame;\    heapframe *oldframe = frame;\
379    frame = newframe->Xprevframe;\    frame = oldframe->Xprevframe;\
   (pcre_stack_free)(newframe);\  
380    if (frame != NULL)\    if (frame != NULL)\
381      {\      {\
382      rrc = ra;\      rrc = ra;\
# Line 285  argument of match(), which never changes Line 390  argument of match(), which never changes
390    
391  typedef struct heapframe {  typedef struct heapframe {
392    struct heapframe *Xprevframe;    struct heapframe *Xprevframe;
393      struct heapframe *Xnextframe;
394    
395    /* Function arguments that may change */    /* Function arguments that may change */
396    
397    const uschar *Xeptr;    PCRE_PUCHAR Xeptr;
398    const uschar *Xecode;    const pcre_uchar *Xecode;
399    const uschar *Xmstart;    PCRE_PUCHAR Xmstart;
400    int Xoffset_top;    int Xoffset_top;
   long int Xims;  
401    eptrblock *Xeptrb;    eptrblock *Xeptrb;
   int Xflags;  
402    unsigned int Xrdepth;    unsigned int Xrdepth;
403    
404    /* Function local variables */    /* Function local variables */
405    
406    const uschar *Xcallpat;    PCRE_PUCHAR Xcallpat;
407    const uschar *Xcharptr;  #ifdef SUPPORT_UTF
408    const uschar *Xdata;    PCRE_PUCHAR Xcharptr;
409    const uschar *Xnext;  #endif
410    const uschar *Xpp;    PCRE_PUCHAR Xdata;
411    const uschar *Xprev;    PCRE_PUCHAR Xnext;
412    const uschar *Xsaved_eptr;    PCRE_PUCHAR Xpp;
413      PCRE_PUCHAR Xprev;
414      PCRE_PUCHAR Xsaved_eptr;
415    
416    recursion_info Xnew_recursive;    recursion_info Xnew_recursive;
417    
# Line 313  typedef struct heapframe { Line 419  typedef struct heapframe {
419    BOOL Xcondition;    BOOL Xcondition;
420    BOOL Xprev_is_word;    BOOL Xprev_is_word;
421    
   unsigned long int Xoriginal_ims;  
   
422  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
423    int Xprop_type;    int Xprop_type;
424    int Xprop_value;    unsigned int Xprop_value;
425    int Xprop_fail_result;    int Xprop_fail_result;
   int Xprop_category;  
   int Xprop_chartype;  
   int Xprop_script;  
426    int Xoclength;    int Xoclength;
427    uschar Xocchars[8];    pcre_uchar Xocchars[6];
428  #endif  #endif
429    
430      int Xcodelink;
431    int Xctype;    int Xctype;
432    unsigned int Xfc;    unsigned int Xfc;
433    int Xfi;    int Xfi;
434    int Xlength;    int Xlength;
435    int Xmax;    int Xmax;
436    int Xmin;    int Xmin;
437    int Xnumber;    unsigned int Xnumber;
438    int Xoffset;    int Xoffset;
439    int Xop;    unsigned int Xop;
440    int Xsave_capture_last;    pcre_int32 Xsave_capture_last;
441    int Xsave_offset1, Xsave_offset2, Xsave_offset3;    int Xsave_offset1, Xsave_offset2, Xsave_offset3;
442    int Xstacksave[REC_STACK_SAVE_MAX];    int Xstacksave[REC_STACK_SAVE_MAX];
443    
# Line 361  typedef struct heapframe { Line 463  typedef struct heapframe {
463    
464  /* This function is called recursively in many circumstances. Whenever it  /* This function is called recursively in many circumstances. Whenever it
465  returns a negative (error) response, the outer incarnation must also return the  returns a negative (error) response, the outer incarnation must also return the
466  same response.  same response. */
467    
468    /* These macros pack up tests that are used for partial matching, and which
469    appear several times in the code. We set the "hit end" flag if the pointer is
470    at the end of the subject and also past the start of the subject (i.e.
471    something has been matched). For hard partial matching, we then return
472    immediately. The second one is used when we already know we are past the end of
473    the subject. */
474    
475    #define CHECK_PARTIAL()\
476      if (md->partial != 0 && eptr >= md->end_subject && \
477          eptr > md->start_used_ptr) \
478        { \
479        md->hitend = TRUE; \
480        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
481        }
482    
483    #define SCHECK_PARTIAL()\
484      if (md->partial != 0 && eptr > md->start_used_ptr) \
485        { \
486        md->hitend = TRUE; \
487        if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL); \
488        }
489    
490  Performance note: It might be tempting to extract commonly used fields from the  
491  md structure (e.g. utf8, end_subject) into individual variables to improve  /* Performance note: It might be tempting to extract commonly used fields from
492    the md structure (e.g. utf, end_subject) into individual variables to improve
493  performance. Tests using gcc on a SPARC disproved this; in the first case, it  performance. Tests using gcc on a SPARC disproved this; in the first case, it
494  made performance worse.  made performance worse.
495    
# Line 375  Arguments: Line 500  Arguments:
500                   by encountering \K)                   by encountering \K)
501     offset_top  current top pointer     offset_top  current top pointer
502     md          pointer to "static" info for the match     md          pointer to "static" info for the match
    ims         current /i, /m, and /s options  
503     eptrb       pointer to chain of blocks containing eptr at start of     eptrb       pointer to chain of blocks containing eptr at start of
504                   brackets - for testing for empty matches                   brackets - for testing for empty matches
    flags       can contain  
                  match_condassert - this is an assertion condition  
                  match_cbegroup - this is the start of an unlimited repeat  
                    group that can match an empty string  
505     rdepth      the recursion depth     rdepth      the recursion depth
506    
507  Returns:       MATCH_MATCH if matched            )  these values are >= 0  Returns:       MATCH_MATCH if matched            )  these values are >= 0
508                 MATCH_NOMATCH if failed to match  )                 MATCH_NOMATCH if failed to match  )
509                   a negative MATCH_xxx value for PRUNE, SKIP, etc
510                 a negative PCRE_ERROR_xxx value if aborted by an error condition                 a negative PCRE_ERROR_xxx value if aborted by an error condition
511                   (e.g. stopped by repeated call or recursion limit)                   (e.g. stopped by repeated call or recursion limit)
512  */  */
513    
514  static int  static int
515  match(REGISTER USPTR eptr, REGISTER const uschar *ecode, const uschar *mstart,  match(REGISTER PCRE_PUCHAR eptr, REGISTER const pcre_uchar *ecode,
516    int offset_top, match_data *md, unsigned long int ims, eptrblock *eptrb,    PCRE_PUCHAR mstart, int offset_top, match_data *md, eptrblock *eptrb,
517    int flags, unsigned int rdepth)    unsigned int rdepth)
518  {  {
519  /* These variables do not need to be preserved over recursion in this function,  /* These variables do not need to be preserved over recursion in this function,
520  so they can be ordinary variables in all cases. Mark some of them with  so they can be ordinary variables in all cases. Mark some of them with
# Line 401  so they can be ordinary variables in all Line 522  so they can be ordinary variables in all
522    
523  register int  rrc;         /* Returns from recursive calls */  register int  rrc;         /* Returns from recursive calls */
524  register int  i;           /* Used for loops not involving calls to RMATCH() */  register int  i;           /* Used for loops not involving calls to RMATCH() */
525  register unsigned int c;   /* Character values not kept over RMATCH() calls */  register pcre_uint32 c;    /* Character values not kept over RMATCH() calls */
526  register BOOL utf8;        /* Local copy of UTF-8 flag for speed */  register BOOL utf;         /* Local copy of UTF flag for speed */
527    
528  BOOL minimize, possessive; /* Quantifier options */  BOOL minimize, possessive; /* Quantifier options */
529    BOOL caseless;
530    int condcode;
531    
532  /* When recursion is not being used, all "local" variables that have to be  /* When recursion is not being used, all "local" variables that have to be
533  preserved over calls to RMATCH() are part of a "frame" which is obtained from  preserved over calls to RMATCH() are part of a "frame". We set up the top-level
534  heap storage. Set up the top-level frame here; others are obtained from the  frame on the stack here; subsequent instantiations are obtained from the heap
535  heap whenever RMATCH() does a "recursion". See the macro definitions above. */  whenever RMATCH() does a "recursion". See the macro definitions above. Putting
536    the top-level on the stack rather than malloc-ing them all gives a performance
537    boost in many cases where there is not much "recursion". */
538    
539  #ifdef NO_RECURSE  #ifdef NO_RECURSE
540  heapframe *frame = (pcre_stack_malloc)(sizeof(heapframe));  heapframe *frame = (heapframe *)md->match_frames_base;
 frame->Xprevframe = NULL;            /* Marks the top level */  
541    
542  /* Copy in the original argument variables */  /* Copy in the original argument variables */
543    
# Line 421  frame->Xeptr = eptr; Line 545  frame->Xeptr = eptr;
545  frame->Xecode = ecode;  frame->Xecode = ecode;
546  frame->Xmstart = mstart;  frame->Xmstart = mstart;
547  frame->Xoffset_top = offset_top;  frame->Xoffset_top = offset_top;
 frame->Xims = ims;  
548  frame->Xeptrb = eptrb;  frame->Xeptrb = eptrb;
 frame->Xflags = flags;  
549  frame->Xrdepth = rdepth;  frame->Xrdepth = rdepth;
550    
551  /* This is where control jumps back to to effect "recursion" */  /* This is where control jumps back to to effect "recursion" */
# Line 436  HEAP_RECURSE: Line 558  HEAP_RECURSE:
558  #define ecode              frame->Xecode  #define ecode              frame->Xecode
559  #define mstart             frame->Xmstart  #define mstart             frame->Xmstart
560  #define offset_top         frame->Xoffset_top  #define offset_top         frame->Xoffset_top
 #define ims                frame->Xims  
561  #define eptrb              frame->Xeptrb  #define eptrb              frame->Xeptrb
 #define flags              frame->Xflags  
562  #define rdepth             frame->Xrdepth  #define rdepth             frame->Xrdepth
563    
564  /* Ditto for the local variables */  /* Ditto for the local variables */
565    
566  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
567  #define charptr            frame->Xcharptr  #define charptr            frame->Xcharptr
568  #endif  #endif
569  #define callpat            frame->Xcallpat  #define callpat            frame->Xcallpat
570    #define codelink           frame->Xcodelink
571  #define data               frame->Xdata  #define data               frame->Xdata
572  #define next               frame->Xnext  #define next               frame->Xnext
573  #define pp                 frame->Xpp  #define pp                 frame->Xpp
# Line 459  HEAP_RECURSE: Line 580  HEAP_RECURSE:
580  #define condition          frame->Xcondition  #define condition          frame->Xcondition
581  #define prev_is_word       frame->Xprev_is_word  #define prev_is_word       frame->Xprev_is_word
582    
 #define original_ims       frame->Xoriginal_ims  
   
583  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
584  #define prop_type          frame->Xprop_type  #define prop_type          frame->Xprop_type
585  #define prop_value         frame->Xprop_value  #define prop_value         frame->Xprop_value
586  #define prop_fail_result   frame->Xprop_fail_result  #define prop_fail_result   frame->Xprop_fail_result
 #define prop_category      frame->Xprop_category  
 #define prop_chartype      frame->Xprop_chartype  
 #define prop_script        frame->Xprop_script  
587  #define oclength           frame->Xoclength  #define oclength           frame->Xoclength
588  #define occhars            frame->Xocchars  #define occhars            frame->Xocchars
589  #endif  #endif
# Line 497  i, and fc and c, can be the same variabl Line 613  i, and fc and c, can be the same variabl
613  #define fi i  #define fi i
614  #define fc c  #define fc c
615    
616    /* Many of the following variables are used only in small blocks of the code.
617    My normal style of coding would have declared them within each of those blocks.
618    However, in order to accommodate the version of this code that uses an external
619    "stack" implemented on the heap, it is easier to declare them all here, so the
620    declarations can be cut out in a block. The only declarations within blocks
621    below are for variables that do not have to be preserved over a recursive call
622    to RMATCH(). */
623    
624    #ifdef SUPPORT_UTF
625    const pcre_uchar *charptr;
626    #endif
627    const pcre_uchar *callpat;
628    const pcre_uchar *data;
629    const pcre_uchar *next;
630    PCRE_PUCHAR       pp;
631    const pcre_uchar *prev;
632    PCRE_PUCHAR       saved_eptr;
633    
634  #ifdef SUPPORT_UTF8                /* Many of these variables are used only  */  recursion_info new_recursive;
635  const uschar *charptr;             /* in small blocks of the code. My normal */  
636  #endif                             /* style of coding would have declared    */  BOOL cur_is_word;
 const uschar *callpat;             /* them within each of those blocks.      */  
 const uschar *data;                /* However, in order to accommodate the   */  
 const uschar *next;                /* version of this code that uses an      */  
 USPTR         pp;                  /* external "stack" implemented on the    */  
 const uschar *prev;                /* heap, it is easier to declare them all */  
 USPTR         saved_eptr;          /* here, so the declarations can be cut   */  
                                    /* out in a block. The only declarations  */  
 recursion_info new_recursive;      /* within blocks below are for variables  */  
                                    /* that do not have to be preserved over  */  
 BOOL cur_is_word;                  /* a recursive call to RMATCH().          */  
637  BOOL condition;  BOOL condition;
638  BOOL prev_is_word;  BOOL prev_is_word;
639    
 unsigned long int original_ims;  
   
640  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
641  int prop_type;  int prop_type;
642  int prop_value;  unsigned int prop_value;
643  int prop_fail_result;  int prop_fail_result;
 int prop_category;  
 int prop_chartype;  
 int prop_script;  
644  int oclength;  int oclength;
645  uschar occhars[8];  pcre_uchar occhars[6];
646  #endif  #endif
647    
648    int codelink;
649  int ctype;  int ctype;
650  int length;  int length;
651  int max;  int max;
652  int min;  int min;
653  int number;  unsigned int number;
654  int offset;  int offset;
655  int op;  unsigned int op;
656  int save_capture_last;  pcre_int32 save_capture_last;
657  int save_offset1, save_offset2, save_offset3;  int save_offset1, save_offset2, save_offset3;
658  int stacksave[REC_STACK_SAVE_MAX];  int stacksave[REC_STACK_SAVE_MAX];
659    
660  eptrblock newptrb;  eptrblock newptrb;
661    
662    /* There is a special fudge for calling match() in a way that causes it to
663    measure the size of its basic stack frame when the stack is being used for
664    recursion. The second argument (ecode) being NULL triggers this behaviour. It
665    cannot normally ever be NULL. The return is the negated value of the frame
666    size. */
667    
668    if (ecode == NULL)
669      {
670      if (rdepth == 0)
671        return match((PCRE_PUCHAR)&rdepth, NULL, NULL, 0, NULL, NULL, 1);
672      else
673        {
674        int len = (char *)&rdepth - (char *)eptr;
675        return (len > 0)? -len : len;
676        }
677      }
678  #endif     /* NO_RECURSE */  #endif     /* NO_RECURSE */
679    
680    /* To save space on the stack and in the heap frame, I have doubled up on some
681    of the local variables that are used only in localised parts of the code, but
682    still need to be preserved over recursive calls of match(). These macros define
683    the alternative names that are used. */
684    
685    #define allow_zero    cur_is_word
686    #define cbegroup      condition
687    #define code_offset   codelink
688    #define condassert    condition
689    #define matched_once  prev_is_word
690    #define foc           number
691    #define save_mark     data
692    
693  /* These statements are here to stop the compiler complaining about unitialized  /* These statements are here to stop the compiler complaining about unitialized
694  variables. */  variables. */
695    
# Line 560  TAIL_RECURSE: Line 709  TAIL_RECURSE:
709  /* OK, now we can get on with the real code of the function. Recursive calls  /* OK, now we can get on with the real code of the function. Recursive calls
710  are specified by the macro RMATCH and RRETURN is used to return. When  are specified by the macro RMATCH and RRETURN is used to return. When
711  NO_RECURSE is *not* defined, these just turn into a recursive call to match()  NO_RECURSE is *not* defined, these just turn into a recursive call to match()
712  and a "return", respectively (possibly with some debugging if DEBUG is  and a "return", respectively (possibly with some debugging if PCRE_DEBUG is
713  defined). However, RMATCH isn't like a function call because it's quite a  defined). However, RMATCH isn't like a function call because it's quite a
714  complicated macro. It has to be used in one particular way. This shouldn't,  complicated macro. It has to be used in one particular way. This shouldn't,
715  however, impact performance when true recursion is being used. */  however, impact performance when true recursion is being used. */
716    
717  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
718  utf8 = md->utf8;       /* Local copy of the flag */  utf = md->utf;       /* Local copy of the flag */
719  #else  #else
720  utf8 = FALSE;  utf = FALSE;
721  #endif  #endif
722    
723  /* First check that we haven't called match() too many times, or that we  /* First check that we haven't called match() too many times, or that we
# Line 577  haven't exceeded the recursive call limi Line 726  haven't exceeded the recursive call limi
726  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);  if (md->match_call_count++ >= md->match_limit) RRETURN(PCRE_ERROR_MATCHLIMIT);
727  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);  if (rdepth >= md->match_limit_recursion) RRETURN(PCRE_ERROR_RECURSIONLIMIT);
728    
 original_ims = ims;    /* Save for resetting on ')' */  
   
729  /* At the start of a group with an unlimited repeat that may match an empty  /* At the start of a group with an unlimited repeat that may match an empty
730  string, the match_cbegroup flag is set. When this is the case, add the current  string, the variable md->match_function_type is set to MATCH_CBEGROUP. It is
731  subject pointer to the chain of such remembered pointers, to be checked when we  done this way to save having to use another function argument, which would take
732  hit the closing ket, in order to break infinite loops that match no characters.  up space on the stack. See also MATCH_CONDASSERT below.
733  When match() is called in other circumstances, don't add to the chain. The  
734  match_cbegroup flag must NOT be used with tail recursion, because the memory  When MATCH_CBEGROUP is set, add the current subject pointer to the chain of
735  block that is used is on the stack, so a new one may be required for each  such remembered pointers, to be checked when we hit the closing ket, in order
736  match(). */  to break infinite loops that match no characters. When match() is called in
737    other circumstances, don't add to the chain. The MATCH_CBEGROUP feature must
738    NOT be used with tail recursion, because the memory block that is used is on
739    the stack, so a new one may be required for each match(). */
740    
741  if ((flags & match_cbegroup) != 0)  if (md->match_function_type == MATCH_CBEGROUP)
742    {    {
743    newptrb.epb_saved_eptr = eptr;    newptrb.epb_saved_eptr = eptr;
744    newptrb.epb_prev = eptrb;    newptrb.epb_prev = eptrb;
745    eptrb = &newptrb;    eptrb = &newptrb;
746      md->match_function_type = 0;
747    }    }
748    
749  /* Now start processing the opcodes. */  /* Now start processing the opcodes. */
# Line 602  for (;;) Line 753  for (;;)
753    minimize = possessive = FALSE;    minimize = possessive = FALSE;
754    op = *ecode;    op = *ecode;
755    
   /* For partial matching, remember if we ever hit the end of the subject after  
   matching at least one subject character. */  
   
   if (md->partial &&  
       eptr >= md->end_subject &&  
       eptr > mstart)  
     md->hitend = TRUE;  
   
756    switch(op)    switch(op)
757      {      {
758      /* Handle a capturing bracket. If there is space in the offset vector, save      case OP_MARK:
759      the current subject position in the working slot at the top of the vector.      md->nomatch_mark = ecode + 2;
760      We mustn't change the current values of the data slot, because they may be      md->mark = NULL;    /* In case previously set by assertion */
761      set from a previous iteration of this group, and be referred to by a      RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
762      reference inside the group.        eptrb, RM55);
763        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
764      If the bracket fails to match, we need to restore this value and also the           md->mark == NULL) md->mark = ecode + 2;
765      values of the final offsets, in case they were set by a previous iteration  
766      of the same bracket.      /* A return of MATCH_SKIP_ARG means that matching failed at SKIP with an
767        argument, and we must check whether that argument matches this MARK's
768        argument. It is passed back in md->start_match_ptr (an overloading of that
769        variable). If it does match, we reset that variable to the current subject
770        position and return MATCH_SKIP. Otherwise, pass back the return code
771        unaltered. */
772    
773        else if (rrc == MATCH_SKIP_ARG &&
774            STRCMP_UC_UC_TEST(ecode + 2, md->start_match_ptr) == 0)
775          {
776          md->start_match_ptr = eptr;
777          RRETURN(MATCH_SKIP);
778          }
779        RRETURN(rrc);
780    
781        case OP_FAIL:
782        RRETURN(MATCH_NOMATCH);
783    
784        /* COMMIT overrides PRUNE, SKIP, and THEN */
785    
786        case OP_COMMIT:
787        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
788          eptrb, RM52);
789        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE &&
790            rrc != MATCH_SKIP && rrc != MATCH_SKIP_ARG &&
791            rrc != MATCH_THEN)
792          RRETURN(rrc);
793        RRETURN(MATCH_COMMIT);
794    
795        /* PRUNE overrides THEN */
796    
797        case OP_PRUNE:
798        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
799          eptrb, RM51);
800        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
801        RRETURN(MATCH_PRUNE);
802    
803        case OP_PRUNE_ARG:
804        md->nomatch_mark = ecode + 2;
805        md->mark = NULL;    /* In case previously set by assertion */
806        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
807          eptrb, RM56);
808        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
809             md->mark == NULL) md->mark = ecode + 2;
810        if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
811        RRETURN(MATCH_PRUNE);
812    
813        /* SKIP overrides PRUNE and THEN */
814    
815        case OP_SKIP:
816        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
817          eptrb, RM53);
818        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
819          RRETURN(rrc);
820        md->start_match_ptr = eptr;   /* Pass back current position */
821        RRETURN(MATCH_SKIP);
822    
823        /* Note that, for Perl compatibility, SKIP with an argument does NOT set
824        nomatch_mark. There is a flag that disables this opcode when re-matching a
825        pattern that ended with a SKIP for which there was not a matching MARK. */
826    
827        case OP_SKIP_ARG:
828        if (md->ignore_skip_arg)
829          {
830          ecode += PRIV(OP_lengths)[*ecode] + ecode[1];
831          break;
832          }
833        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top, md,
834          eptrb, RM57);
835        if (rrc != MATCH_NOMATCH && rrc != MATCH_PRUNE && rrc != MATCH_THEN)
836          RRETURN(rrc);
837    
838        /* Pass back the current skip name by overloading md->start_match_ptr and
839        returning the special MATCH_SKIP_ARG return code. This will either be
840        caught by a matching MARK, or get to the top, where it causes a rematch
841        with the md->ignore_skip_arg flag set. */
842    
843        md->start_match_ptr = ecode + 2;
844        RRETURN(MATCH_SKIP_ARG);
845    
846        /* For THEN (and THEN_ARG) we pass back the address of the opcode, so that
847        the branch in which it occurs can be determined. Overload the start of
848        match pointer to do this. */
849    
850        case OP_THEN:
851        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
852          eptrb, RM54);
853        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
854        md->start_match_ptr = ecode;
855        RRETURN(MATCH_THEN);
856    
857        case OP_THEN_ARG:
858        md->nomatch_mark = ecode + 2;
859        md->mark = NULL;    /* In case previously set by assertion */
860        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode] + ecode[1], offset_top,
861          md, eptrb, RM58);
862        if ((rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) &&
863             md->mark == NULL) md->mark = ecode + 2;
864        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
865        md->start_match_ptr = ecode;
866        RRETURN(MATCH_THEN);
867    
868        /* Handle an atomic group that does not contain any capturing parentheses.
869        This can be handled like an assertion. Prior to 8.13, all atomic groups
870        were handled this way. In 8.13, the code was changed as below for ONCE, so
871        that backups pass through the group and thereby reset captured values.
872        However, this uses a lot more stack, so in 8.20, atomic groups that do not
873        contain any captures generate OP_ONCE_NC, which can be handled in the old,
874        less stack intensive way.
875    
876        Check the alternative branches in turn - the matching won't pass the KET
877        for this kind of subpattern. If any one branch matches, we carry on as at
878        the end of a normal bracket, leaving the subject pointer, but resetting
879        the start-of-match value in case it was changed by \K. */
880    
881        case OP_ONCE_NC:
882        prev = ecode;
883        saved_eptr = eptr;
884        save_mark = md->mark;
885        do
886          {
887          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM64);
888          if (rrc == MATCH_MATCH)  /* Note: _not_ MATCH_ACCEPT */
889            {
890            mstart = md->start_match_ptr;
891            break;
892            }
893          if (rrc == MATCH_THEN)
894            {
895            next = ecode + GET(ecode,1);
896            if (md->start_match_ptr < next &&
897                (*ecode == OP_ALT || *next == OP_ALT))
898              rrc = MATCH_NOMATCH;
899            }
900    
901          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
902          ecode += GET(ecode,1);
903          md->mark = save_mark;
904          }
905        while (*ecode == OP_ALT);
906    
907        /* If hit the end of the group (which could be repeated), fail */
908    
909        if (*ecode != OP_ONCE_NC && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);
910    
911        /* Continue as from after the group, updating the offsets high water
912        mark, since extracts may have been taken. */
913    
914        do ecode += GET(ecode, 1); while (*ecode == OP_ALT);
915    
916        offset_top = md->end_offset_top;
917        eptr = md->end_match_ptr;
918    
919        /* For a non-repeating ket, just continue at this level. This also
920        happens for a repeating ket if no characters were matched in the group.
921        This is the forcible breaking of infinite loops as implemented in Perl
922        5.005. */
923    
924        if (*ecode == OP_KET || eptr == saved_eptr)
925          {
926          ecode += 1+LINK_SIZE;
927          break;
928          }
929    
930        /* The repeating kets try the rest of the pattern or restart from the
931        preceding bracket, in the appropriate order. The second "call" of match()
932        uses tail recursion, to avoid using another stack frame. */
933    
934        if (*ecode == OP_KETRMIN)
935          {
936          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM65);
937          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
938          ecode = prev;
939          goto TAIL_RECURSE;
940          }
941        else  /* OP_KETRMAX */
942          {
943          RMATCH(eptr, prev, offset_top, md, eptrb, RM66);
944          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
945          ecode += 1 + LINK_SIZE;
946          goto TAIL_RECURSE;
947          }
948        /* Control never gets here */
949    
950        /* Handle a capturing bracket, other than those that are possessive with an
951        unlimited repeat. If there is space in the offset vector, save the current
952        subject position in the working slot at the top of the vector. We mustn't
953        change the current values of the data slot, because they may be set from a
954        previous iteration of this group, and be referred to by a reference inside
955        the group. A failure to match might occur after the group has succeeded,
956        if something later on doesn't match. For this reason, we need to restore
957        the working value and also the values of the final offsets, in case they
958        were set by a previous iteration of the same bracket.
959    
960      If there isn't enough space in the offset vector, treat this as if it were      If there isn't enough space in the offset vector, treat this as if it were
961      a non-capturing bracket. Don't worry about setting the flag for the error      a non-capturing bracket. Don't worry about setting the flag for the error
# Line 631  for (;;) Line 966  for (;;)
966      number = GET2(ecode, 1+LINK_SIZE);      number = GET2(ecode, 1+LINK_SIZE);
967      offset = number << 1;      offset = number << 1;
968    
969  #ifdef DEBUG  #ifdef PCRE_DEBUG
970      printf("start bracket %d\n", number);      printf("start bracket %d\n", number);
971      printf("subject=");      printf("subject=");
972      pchars(eptr, 16, TRUE, md);      pchars(eptr, 16, TRUE, md);
# Line 644  for (;;) Line 979  for (;;)
979        save_offset2 = md->offset_vector[offset+1];        save_offset2 = md->offset_vector[offset+1];
980        save_offset3 = md->offset_vector[md->offset_end - number];        save_offset3 = md->offset_vector[md->offset_end - number];
981        save_capture_last = md->capture_last;        save_capture_last = md->capture_last;
982          save_mark = md->mark;
983    
984        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
985        md->offset_vector[md->offset_end - number] = eptr - md->start_subject;        md->offset_vector[md->offset_end - number] =
986            (int)(eptr - md->start_subject);
987    
988        flags = (op == OP_SCBRA)? match_cbegroup : 0;        for (;;)
       do  
989          {          {
990          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md,          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
991            ims, eptrb, flags, RM1);          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
992              eptrb, RM1);
993            if (rrc == MATCH_ONCE) break;  /* Backing up through an atomic group */
994    
995            /* If we backed up to a THEN, check whether it is within the current
996            branch by comparing the address of the THEN that is passed back with
997            the end of the branch. If it is within the current branch, and the
998            branch is one of two or more alternatives (it either starts or ends
999            with OP_ALT), we have reached the limit of THEN's action, so convert
1000            the return code to NOMATCH, which will cause normal backtracking to
1001            happen from now on. Otherwise, THEN is passed back to an outer
1002            alternative. This implements Perl's treatment of parenthesized groups,
1003            where a group not containing | does not affect the current alternative,
1004            that is, (X) is NOT the same as (X|(*F)). */
1005    
1006            if (rrc == MATCH_THEN)
1007              {
1008              next = ecode + GET(ecode,1);
1009              if (md->start_match_ptr < next &&
1010                  (*ecode == OP_ALT || *next == OP_ALT))
1011                rrc = MATCH_NOMATCH;
1012              }
1013    
1014            /* Anything other than NOMATCH is passed back. */
1015    
1016          if (rrc != MATCH_NOMATCH) RRETURN(rrc);          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1017          md->capture_last = save_capture_last;          md->capture_last = save_capture_last;
1018          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
1019            md->mark = save_mark;
1020            if (*ecode != OP_ALT) break;
1021          }          }
       while (*ecode == OP_ALT);  
1022    
1023        DPRINTF(("bracket %d failed\n", number));        DPRINTF(("bracket %d failed\n", number));
   
1024        md->offset_vector[offset] = save_offset1;        md->offset_vector[offset] = save_offset1;
1025        md->offset_vector[offset+1] = save_offset2;        md->offset_vector[offset+1] = save_offset2;
1026        md->offset_vector[md->offset_end - number] = save_offset3;        md->offset_vector[md->offset_end - number] = save_offset3;
1027    
1028        RRETURN(MATCH_NOMATCH);        /* At this point, rrc will be one of MATCH_ONCE or MATCH_NOMATCH. */
1029    
1030          RRETURN(rrc);
1031        }        }
1032    
1033      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
# Line 679  for (;;) Line 1041  for (;;)
1041      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1042      /* VVVVVVVVVVVVVVVVVVVVVVVVV */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1043    
1044      /* Non-capturing bracket. Loop for all the alternatives. When we get to the      /* Non-capturing or atomic group, except for possessive with unlimited
1045      final alternative within the brackets, we would return the result of a      repeat and ONCE group with no captures. Loop for all the alternatives.
     recursive call to match() whatever happened. We can reduce stack usage by  
     turning this into a tail recursion, except in the case when match_cbegroup  
     is set.*/  
1046    
1047        When we get to the final alternative within the brackets, we used to return
1048        the result of a recursive call to match() whatever happened so it was
1049        possible to reduce stack usage by turning this into a tail recursion,
1050        except in the case of a possibly empty group. However, now that there is
1051        the possiblity of (*THEN) occurring in the final alternative, this
1052        optimization is no longer always possible.
1053    
1054        We can optimize if we know there are no (*THEN)s in the pattern; at present
1055        this is the best that can be done.
1056    
1057        MATCH_ONCE is returned when the end of an atomic group is successfully
1058        reached, but subsequent matching fails. It passes back up the tree (causing
1059        captured values to be reset) until the original atomic group level is
1060        reached. This is tested by comparing md->once_target with the start of the
1061        group. At this point, the return is converted into MATCH_NOMATCH so that
1062        previous backup points can be taken. */
1063    
1064        case OP_ONCE:
1065      case OP_BRA:      case OP_BRA:
1066      case OP_SBRA:      case OP_SBRA:
1067      DPRINTF(("start non-capturing bracket\n"));      DPRINTF(("start non-capturing bracket\n"));
1068      flags = (op >= OP_SBRA)? match_cbegroup : 0;  
1069      for (;;)      for (;;)
1070        {        {
1071        if (ecode[GET(ecode, 1)] != OP_ALT)   /* Final alternative */        if (op >= OP_SBRA || op == OP_ONCE)
1072          {          md->match_function_type = MATCH_CBEGROUP;
         if (flags == 0)    /* Not a possibly empty group */  
           {  
           ecode += _pcre_OP_lengths[*ecode];  
           DPRINTF(("bracket 0 tail recursion\n"));  
           goto TAIL_RECURSE;  
           }  
1073    
1074          /* Possibly empty group; can't use tail recursion. */        /* If this is not a possibly empty group, and there are no (*THEN)s in
1075          the pattern, and this is the final alternative, optimize as described
1076          above. */
1077    
1078          RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        else if (!md->hasthen && ecode[GET(ecode, 1)] != OP_ALT)
1079            eptrb, flags, RM48);          {
1080          RRETURN(rrc);          ecode += PRIV(OP_lengths)[*ecode];
1081            goto TAIL_RECURSE;
1082          }          }
1083    
1084        /* For non-final alternatives, continue the loop for a NOMATCH result;        /* In all other cases, we have to make another call to match(). */
       otherwise return. */  
1085    
1086        RMATCH(eptr, ecode + _pcre_OP_lengths[*ecode], offset_top, md, ims,        save_mark = md->mark;
1087          eptrb, flags, RM2);        save_capture_last = md->capture_last;
1088        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md, eptrb,
1089            RM2);
1090    
1091          /* See comment in the code for capturing groups above about handling
1092          THEN. */
1093    
1094          if (rrc == MATCH_THEN)
1095            {
1096            next = ecode + GET(ecode,1);
1097            if (md->start_match_ptr < next &&
1098                (*ecode == OP_ALT || *next == OP_ALT))
1099              rrc = MATCH_NOMATCH;
1100            }
1101    
1102          if (rrc != MATCH_NOMATCH)
1103            {
1104            if (rrc == MATCH_ONCE)
1105              {
1106              const pcre_uchar *scode = ecode;
1107              if (*scode != OP_ONCE)           /* If not at start, find it */
1108                {
1109                while (*scode == OP_ALT) scode += GET(scode, 1);
1110                scode -= GET(scode, 1);
1111                }
1112              if (md->once_target == scode) rrc = MATCH_NOMATCH;
1113              }
1114            RRETURN(rrc);
1115            }
1116        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1117        }        md->mark = save_mark;
1118      /* Control never reaches here. */        if (*ecode != OP_ALT) break;
1119          md->capture_last = save_capture_last;
1120          }
1121    
1122        RRETURN(MATCH_NOMATCH);
1123    
1124        /* Handle possessive capturing brackets with an unlimited repeat. We come
1125        here from BRAZERO with allow_zero set TRUE. The offset_vector values are
1126        handled similarly to the normal case above. However, the matching is
1127        different. The end of these brackets will always be OP_KETRPOS, which
1128        returns MATCH_KETRPOS without going further in the pattern. By this means
1129        we can handle the group by iteration rather than recursion, thereby
1130        reducing the amount of stack needed. */
1131    
1132        case OP_CBRAPOS:
1133        case OP_SCBRAPOS:
1134        allow_zero = FALSE;
1135    
1136      /* Conditional group: compilation checked that there are no more than      POSSESSIVE_CAPTURE:
1137      two branches. If the condition is false, skipping the first branch takes us      number = GET2(ecode, 1+LINK_SIZE);
1138      past the end if there is only one branch, but that's OK because that is      offset = number << 1;
     exactly what going to the ket would do. As there is only one branch to be  
     obeyed, we can use tail recursion to avoid using another stack frame. */  
1139    
1140      case OP_COND:  #ifdef PCRE_DEBUG
1141      case OP_SCOND:      printf("start possessive bracket %d\n", number);
1142      if (ecode[LINK_SIZE+1] == OP_RREF)         /* Recursion test */      printf("subject=");
1143        {      pchars(eptr, 16, TRUE, md);
1144        offset = GET2(ecode, LINK_SIZE + 2);     /* Recursion group number*/      printf("\n");
1145        condition = md->recursive != NULL &&  #endif
         (offset == RREF_ANY || offset == md->recursive->group_num);  
       ecode += condition? 3 : GET(ecode, 1);  
       }  
1146    
1147      else if (ecode[LINK_SIZE+1] == OP_CREF)    /* Group used test */      if (offset < md->offset_max)
1148        {        {
1149        offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */        matched_once = FALSE;
1150        condition = offset < offset_top && md->offset_vector[offset] >= 0;        code_offset = (int)(ecode - md->start_code);
       ecode += condition? 3 : GET(ecode, 1);  
       }  
1151    
1152      else if (ecode[LINK_SIZE+1] == OP_DEF)     /* DEFINE - always false */        save_offset1 = md->offset_vector[offset];
1153        {        save_offset2 = md->offset_vector[offset+1];
1154        condition = FALSE;        save_offset3 = md->offset_vector[md->offset_end - number];
1155        ecode += GET(ecode, 1);        save_capture_last = md->capture_last;
       }  
1156    
1157      /* The condition is an assertion. Call match() to evaluate it - setting        DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
     the final argument match_condassert causes it to stop at the end of an  
     assertion. */  
1158    
1159      else        /* Each time round the loop, save the current subject position for use
1160        {        when the group matches. For MATCH_MATCH, the group has matched, so we
1161        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL,        restart it with a new subject starting position, remembering that we had
1162            match_condassert, RM3);        at least one match. For MATCH_NOMATCH, carry on with the alternatives, as
1163        if (rrc == MATCH_MATCH)        usual. If we haven't matched any alternatives in any iteration, check to
1164          {        see if a previous iteration matched. If so, the group has matched;
1165          condition = TRUE;        continue from afterwards. Otherwise it has failed; restore the previous
1166          ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);        capture values before returning NOMATCH. */
1167          while (*ecode == OP_ALT) ecode += GET(ecode, 1);  
1168          }        for (;;)
1169        else if (rrc != MATCH_NOMATCH)          {
1170          {          md->offset_vector[md->offset_end - number] =
1171          RRETURN(rrc);         /* Need braces because of following else */            (int)(eptr - md->start_subject);
1172          }          if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1173        else          RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1174          {            eptrb, RM63);
1175          condition = FALSE;          if (rrc == MATCH_KETRPOS)
1176              {
1177              offset_top = md->end_offset_top;
1178              eptr = md->end_match_ptr;
1179              ecode = md->start_code + code_offset;
1180              save_capture_last = md->capture_last;
1181              matched_once = TRUE;
1182              continue;
1183              }
1184    
1185            /* See comment in the code for capturing groups above about handling
1186            THEN. */
1187    
1188            if (rrc == MATCH_THEN)
1189              {
1190              next = ecode + GET(ecode,1);
1191              if (md->start_match_ptr < next &&
1192                  (*ecode == OP_ALT || *next == OP_ALT))
1193                rrc = MATCH_NOMATCH;
1194              }
1195    
1196            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1197            md->capture_last = save_capture_last;
1198          ecode += GET(ecode, 1);          ecode += GET(ecode, 1);
1199            if (*ecode != OP_ALT) break;
1200          }          }
       }  
   
     /* We are now at the branch that is to be obeyed. As there is only one,  
     we can use tail recursion to avoid using another stack frame, except when  
     match_cbegroup is required for an unlimited repeat of a possibly empty  
     group. If the second alternative doesn't exist, we can just plough on. */  
1201    
1202      if (condition || *ecode == OP_ALT)        if (!matched_once)
       {  
       ecode += 1 + LINK_SIZE;  
       if (op == OP_SCOND)        /* Possibly empty group */  
1203          {          {
1204          RMATCH(eptr, ecode, offset_top, md, ims, eptrb, match_cbegroup, RM49);          md->offset_vector[offset] = save_offset1;
1205          RRETURN(rrc);          md->offset_vector[offset+1] = save_offset2;
1206            md->offset_vector[md->offset_end - number] = save_offset3;
1207          }          }
1208        else                       /* Group must match something */  
1209          if (allow_zero || matched_once)
1210          {          {
1211          flags = 0;          ecode += 1 + LINK_SIZE;
1212          goto TAIL_RECURSE;          break;
1213          }          }
       }  
     else                         /* Condition false & no 2nd alternative */  
       {  
       ecode += 1 + LINK_SIZE;  
       }  
     break;  
   
   
     /* End of the pattern. If we are in a top-level recursion, we should  
     restore the offsets appropriately and continue from after the call. */  
1214    
1215      case OP_END:        RRETURN(MATCH_NOMATCH);
     if (md->recursive != NULL && md->recursive->group_num == 0)  
       {  
       recursion_info *rec = md->recursive;  
       DPRINTF(("End of pattern in a (?0) recursion\n"));  
       md->recursive = rec->prevrec;  
       memmove(md->offset_vector, rec->offset_save,  
         rec->saved_max * sizeof(int));  
       mstart = rec->save_start;  
       ims = original_ims;  
       ecode = rec->after_call;  
       break;  
1216        }        }
1217    
1218      /* Otherwise, if PCRE_NOTEMPTY is set, fail if we have matched an empty      /* FALL THROUGH ... Insufficient room for saving captured contents. Treat
1219      string - backtracking will then try other alternatives, if any. */      as a non-capturing bracket. */
1220    
1221      if (md->notempty && eptr == mstart) RRETURN(MATCH_NOMATCH);      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1222      md->end_match_ptr = eptr;           /* Record where we ended */      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     md->end_offset_top = offset_top;    /* and how many extracts were taken */  
     md->start_match_ptr = mstart;  /* and the start (\K can modify) */  
     RRETURN(MATCH_MATCH);  
1223    
1224      /* Change option settings */      DPRINTF(("insufficient capture room: treat as non-capturing\n"));
1225    
1226      case OP_OPT:      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
1227      ims = ecode[1];      /* VVVVVVVVVVVVVVVVVVVVVVVVV */
     ecode += 2;  
     DPRINTF(("ims set to %02lx\n", ims));  
     break;  
1228    
1229      /* Assertion brackets. Check the alternative branches in turn - the      /* Non-capturing possessive bracket with unlimited repeat. We come here
1230      matching won't pass the KET for an assertion. If any one branch matches,      from BRAZERO with allow_zero = TRUE. The code is similar to the above,
1231      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the      without the capturing complication. It is written out separately for speed
1232      start of each branch to move the current point backwards, so the code at      and cleanliness. */
1233      this level is identical to the lookahead case. */  
1234        case OP_BRAPOS:
1235        case OP_SBRAPOS:
1236        allow_zero = FALSE;
1237    
1238        POSSESSIVE_NON_CAPTURE:
1239        matched_once = FALSE;
1240        code_offset = (int)(ecode - md->start_code);
1241        save_capture_last = md->capture_last;
1242    
1243      case OP_ASSERT:      for (;;)
     case OP_ASSERTBACK:  
     do  
1244        {        {
1245        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        if (op >= OP_SBRA) md->match_function_type = MATCH_CBEGROUP;
1246          RM4);        RMATCH(eptr, ecode + PRIV(OP_lengths)[*ecode], offset_top, md,
1247        if (rrc == MATCH_MATCH) break;          eptrb, RM48);
1248          if (rrc == MATCH_KETRPOS)
1249            {
1250            offset_top = md->end_offset_top;
1251            eptr = md->end_match_ptr;
1252            ecode = md->start_code + code_offset;
1253            matched_once = TRUE;
1254            continue;
1255            }
1256    
1257          /* See comment in the code for capturing groups above about handling
1258          THEN. */
1259    
1260          if (rrc == MATCH_THEN)
1261            {
1262            next = ecode + GET(ecode,1);
1263            if (md->start_match_ptr < next &&
1264                (*ecode == OP_ALT || *next == OP_ALT))
1265              rrc = MATCH_NOMATCH;
1266            }
1267    
1268        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1269        ecode += GET(ecode, 1);        ecode += GET(ecode, 1);
1270          if (*ecode != OP_ALT) break;
1271          md->capture_last = save_capture_last;
1272        }        }
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);  
1273    
1274      /* If checking an assertion for a condition, return MATCH_MATCH. */      if (matched_once || allow_zero)
1275          {
1276          ecode += 1 + LINK_SIZE;
1277          break;
1278          }
1279        RRETURN(MATCH_NOMATCH);
1280    
1281      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      /* Control never reaches here. */
1282    
1283      /* Continue from after the assertion, updating the offsets high water      /* Conditional group: compilation checked that there are no more than
1284      mark, since extracts may have been taken during the assertion. */      two branches. If the condition is false, skipping the first branch takes us
1285        past the end if there is only one branch, but that's OK because that is
1286        exactly what going to the ket would do. */
1287    
1288      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      case OP_COND:
1289      ecode += 1 + LINK_SIZE;      case OP_SCOND:
1290      offset_top = md->end_offset_top;      codelink = GET(ecode, 1);
     continue;  
1291    
1292      /* Negative assertion: all branches must fail to match */      /* Because of the way auto-callout works during compile, a callout item is
1293        inserted between OP_COND and an assertion condition. */
1294    
1295      case OP_ASSERT_NOT:      if (ecode[LINK_SIZE+1] == OP_CALLOUT)
     case OP_ASSERTBACK_NOT:  
     do  
1296        {        {
1297        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, NULL, 0,        if (PUBL(callout) != NULL)
1298          RM5);          {
1299        if (rrc == MATCH_MATCH) RRETURN(MATCH_NOMATCH);          PUBL(callout_block) cb;
1300        if (rrc != MATCH_NOMATCH) RRETURN(rrc);          cb.version          = 2;   /* Version 1 of the callout block */
1301        ecode += GET(ecode,1);          cb.callout_number   = ecode[LINK_SIZE+2];
1302            cb.offset_vector    = md->offset_vector;
1303    #if defined COMPILE_PCRE8
1304            cb.subject          = (PCRE_SPTR)md->start_subject;
1305    #elif defined COMPILE_PCRE16
1306            cb.subject          = (PCRE_SPTR16)md->start_subject;
1307    #elif defined COMPILE_PCRE32
1308            cb.subject          = (PCRE_SPTR32)md->start_subject;
1309    #endif
1310            cb.subject_length   = (int)(md->end_subject - md->start_subject);
1311            cb.start_match      = (int)(mstart - md->start_subject);
1312            cb.current_position = (int)(eptr - md->start_subject);
1313            cb.pattern_position = GET(ecode, LINK_SIZE + 3);
1314            cb.next_item_length = GET(ecode, 3 + 2*LINK_SIZE);
1315            cb.capture_top      = offset_top/2;
1316            cb.capture_last     = md->capture_last & CAPLMASK;
1317            /* Internal change requires this for API compatibility. */
1318            if (cb.capture_last == 0) cb.capture_last = -1;
1319            cb.callout_data     = md->callout_data;
1320            cb.mark             = md->nomatch_mark;
1321            if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1322            if (rrc < 0) RRETURN(rrc);
1323            }
1324          ecode += PRIV(OP_lengths)[OP_CALLOUT];
1325          codelink -= PRIV(OP_lengths)[OP_CALLOUT];
1326          }
1327    
1328        condcode = ecode[LINK_SIZE+1];
1329    
1330        /* Now see what the actual condition is */
1331    
1332        if (condcode == OP_RREF || condcode == OP_NRREF)    /* Recursion test */
1333          {
1334          if (md->recursive == NULL)                /* Not recursing => FALSE */
1335            {
1336            condition = FALSE;
1337            ecode += GET(ecode, 1);
1338            }
1339          else
1340            {
1341            unsigned int recno = GET2(ecode, LINK_SIZE + 2);   /* Recursion group number*/
1342            condition = (recno == RREF_ANY || recno == md->recursive->group_num);
1343    
1344            /* If the test is for recursion into a specific subpattern, and it is
1345            false, but the test was set up by name, scan the table to see if the
1346            name refers to any other numbers, and test them. The condition is true
1347            if any one is set. */
1348    
1349            if (!condition && condcode == OP_NRREF)
1350              {
1351              pcre_uchar *slotA = md->name_table;
1352              for (i = 0; i < md->name_count; i++)
1353                {
1354                if (GET2(slotA, 0) == recno) break;
1355                slotA += md->name_entry_size;
1356                }
1357    
1358              /* Found a name for the number - there can be only one; duplicate
1359              names for different numbers are allowed, but not vice versa. First
1360              scan down for duplicates. */
1361    
1362              if (i < md->name_count)
1363                {
1364                pcre_uchar *slotB = slotA;
1365                while (slotB > md->name_table)
1366                  {
1367                  slotB -= md->name_entry_size;
1368                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1369                    {
1370                    condition = GET2(slotB, 0) == md->recursive->group_num;
1371                    if (condition) break;
1372                    }
1373                  else break;
1374                  }
1375    
1376                /* Scan up for duplicates */
1377    
1378                if (!condition)
1379                  {
1380                  slotB = slotA;
1381                  for (i++; i < md->name_count; i++)
1382                    {
1383                    slotB += md->name_entry_size;
1384                    if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1385                      {
1386                      condition = GET2(slotB, 0) == md->recursive->group_num;
1387                      if (condition) break;
1388                      }
1389                    else break;
1390                    }
1391                  }
1392                }
1393              }
1394    
1395            /* Chose branch according to the condition */
1396    
1397            ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1398            }
1399          }
1400    
1401        else if (condcode == OP_CREF || condcode == OP_NCREF)  /* Group used test */
1402          {
1403          offset = GET2(ecode, LINK_SIZE+2) << 1;  /* Doubled ref number */
1404          condition = offset < offset_top && md->offset_vector[offset] >= 0;
1405    
1406          /* If the numbered capture is unset, but the reference was by name,
1407          scan the table to see if the name refers to any other numbers, and test
1408          them. The condition is true if any one is set. This is tediously similar
1409          to the code above, but not close enough to try to amalgamate. */
1410    
1411          if (!condition && condcode == OP_NCREF)
1412            {
1413            unsigned int refno = offset >> 1;
1414            pcre_uchar *slotA = md->name_table;
1415    
1416            for (i = 0; i < md->name_count; i++)
1417              {
1418              if (GET2(slotA, 0) == refno) break;
1419              slotA += md->name_entry_size;
1420              }
1421    
1422            /* Found a name for the number - there can be only one; duplicate names
1423            for different numbers are allowed, but not vice versa. First scan down
1424            for duplicates. */
1425    
1426            if (i < md->name_count)
1427              {
1428              pcre_uchar *slotB = slotA;
1429              while (slotB > md->name_table)
1430                {
1431                slotB -= md->name_entry_size;
1432                if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1433                  {
1434                  offset = GET2(slotB, 0) << 1;
1435                  condition = offset < offset_top &&
1436                    md->offset_vector[offset] >= 0;
1437                  if (condition) break;
1438                  }
1439                else break;
1440                }
1441    
1442              /* Scan up for duplicates */
1443    
1444              if (!condition)
1445                {
1446                slotB = slotA;
1447                for (i++; i < md->name_count; i++)
1448                  {
1449                  slotB += md->name_entry_size;
1450                  if (STRCMP_UC_UC(slotA + IMM2_SIZE, slotB + IMM2_SIZE) == 0)
1451                    {
1452                    offset = GET2(slotB, 0) << 1;
1453                    condition = offset < offset_top &&
1454                      md->offset_vector[offset] >= 0;
1455                    if (condition) break;
1456                    }
1457                  else break;
1458                  }
1459                }
1460              }
1461            }
1462    
1463          /* Chose branch according to the condition */
1464    
1465          ecode += condition? 1 + IMM2_SIZE : GET(ecode, 1);
1466          }
1467    
1468        else if (condcode == OP_DEF)     /* DEFINE - always false */
1469          {
1470          condition = FALSE;
1471          ecode += GET(ecode, 1);
1472          }
1473    
1474        /* The condition is an assertion. Call match() to evaluate it - setting
1475        md->match_function_type to MATCH_CONDASSERT causes it to stop at the end of
1476        an assertion. */
1477    
1478        else
1479          {
1480          md->match_function_type = MATCH_CONDASSERT;
1481          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM3);
1482          if (rrc == MATCH_MATCH)
1483            {
1484            if (md->end_offset_top > offset_top)
1485              offset_top = md->end_offset_top;  /* Captures may have happened */
1486            condition = TRUE;
1487            ecode += 1 + LINK_SIZE + GET(ecode, LINK_SIZE + 2);
1488            while (*ecode == OP_ALT) ecode += GET(ecode, 1);
1489            }
1490    
1491          /* PCRE doesn't allow the effect of (*THEN) to escape beyond an
1492          assertion; it is therefore treated as NOMATCH. */
1493    
1494          else if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN)
1495            {
1496            RRETURN(rrc);         /* Need braces because of following else */
1497            }
1498          else
1499            {
1500            condition = FALSE;
1501            ecode += codelink;
1502            }
1503          }
1504    
1505        /* We are now at the branch that is to be obeyed. As there is only one, can
1506        use tail recursion to avoid using another stack frame, except when there is
1507        unlimited repeat of a possibly empty group. In the latter case, a recursive
1508        call to match() is always required, unless the second alternative doesn't
1509        exist, in which case we can just plough on. Note that, for compatibility
1510        with Perl, the | in a conditional group is NOT treated as creating two
1511        alternatives. If a THEN is encountered in the branch, it propagates out to
1512        the enclosing alternative (unless nested in a deeper set of alternatives,
1513        of course). */
1514    
1515        if (condition || *ecode == OP_ALT)
1516          {
1517          if (op != OP_SCOND)
1518            {
1519            ecode += 1 + LINK_SIZE;
1520            goto TAIL_RECURSE;
1521            }
1522    
1523          md->match_function_type = MATCH_CBEGROUP;
1524          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM49);
1525          RRETURN(rrc);
1526          }
1527    
1528         /* Condition false & no alternative; continue after the group. */
1529    
1530        else
1531          {
1532          ecode += 1 + LINK_SIZE;
1533          }
1534        break;
1535    
1536    
1537        /* Before OP_ACCEPT there may be any number of OP_CLOSE opcodes,
1538        to close any currently open capturing brackets. */
1539    
1540        case OP_CLOSE:
1541        number = GET2(ecode, 1);   /* Must be less than 65536 */
1542        offset = number << 1;
1543    
1544    #ifdef PCRE_DEBUG
1545          printf("end bracket %d at *ACCEPT", number);
1546          printf("\n");
1547    #endif
1548    
1549        md->capture_last = (md->capture_last & OVFLMASK) | number;
1550        if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1551          {
1552          md->offset_vector[offset] =
1553            md->offset_vector[md->offset_end - number];
1554          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
1555          if (offset_top <= offset) offset_top = offset + 2;
1556          }
1557        ecode += 1 + IMM2_SIZE;
1558        break;
1559    
1560    
1561        /* End of the pattern, either real or forced. */
1562    
1563        case OP_END:
1564        case OP_ACCEPT:
1565        case OP_ASSERT_ACCEPT:
1566    
1567        /* If we have matched an empty string, fail if not in an assertion and not
1568        in a recursion if either PCRE_NOTEMPTY is set, or if PCRE_NOTEMPTY_ATSTART
1569        is set and we have matched at the start of the subject. In both cases,
1570        backtracking will then try other alternatives, if any. */
1571    
1572        if (eptr == mstart && op != OP_ASSERT_ACCEPT &&
1573             md->recursive == NULL &&
1574             (md->notempty ||
1575               (md->notempty_atstart &&
1576                 mstart == md->start_subject + md->start_offset)))
1577          RRETURN(MATCH_NOMATCH);
1578    
1579        /* Otherwise, we have a match. */
1580    
1581        md->end_match_ptr = eptr;           /* Record where we ended */
1582        md->end_offset_top = offset_top;    /* and how many extracts were taken */
1583        md->start_match_ptr = mstart;       /* and the start (\K can modify) */
1584    
1585        /* For some reason, the macros don't work properly if an expression is
1586        given as the argument to RRETURN when the heap is in use. */
1587    
1588        rrc = (op == OP_END)? MATCH_MATCH : MATCH_ACCEPT;
1589        RRETURN(rrc);
1590    
1591        /* Assertion brackets. Check the alternative branches in turn - the
1592        matching won't pass the KET for an assertion. If any one branch matches,
1593        the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
1594        start of each branch to move the current point backwards, so the code at
1595        this level is identical to the lookahead case. When the assertion is part
1596        of a condition, we want to return immediately afterwards. The caller of
1597        this incarnation of the match() function will have set MATCH_CONDASSERT in
1598        md->match_function type, and one of these opcodes will be the first opcode
1599        that is processed. We use a local variable that is preserved over calls to
1600        match() to remember this case. */
1601    
1602        case OP_ASSERT:
1603        case OP_ASSERTBACK:
1604        save_mark = md->mark;
1605        if (md->match_function_type == MATCH_CONDASSERT)
1606          {
1607          condassert = TRUE;
1608          md->match_function_type = 0;
1609          }
1610        else condassert = FALSE;
1611    
1612        do
1613          {
1614          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM4);
1615          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1616            {
1617            mstart = md->start_match_ptr;   /* In case \K reset it */
1618            break;
1619            }
1620          md->mark = save_mark;
1621    
1622          /* A COMMIT failure must fail the entire assertion, without trying any
1623          subsequent branches. */
1624    
1625          if (rrc == MATCH_COMMIT) RRETURN(MATCH_NOMATCH);
1626    
1627          /* PCRE does not allow THEN to escape beyond an assertion; it
1628          is treated as NOMATCH. */
1629    
1630          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1631          ecode += GET(ecode, 1);
1632          }
1633        while (*ecode == OP_ALT);
1634    
1635        if (*ecode == OP_KET) RRETURN(MATCH_NOMATCH);
1636    
1637        /* If checking an assertion for a condition, return MATCH_MATCH. */
1638    
1639        if (condassert) RRETURN(MATCH_MATCH);
1640    
1641        /* Continue from after the assertion, updating the offsets high water
1642        mark, since extracts may have been taken during the assertion. */
1643    
1644        do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1645        ecode += 1 + LINK_SIZE;
1646        offset_top = md->end_offset_top;
1647        continue;
1648    
1649        /* Negative assertion: all branches must fail to match. Encountering SKIP,
1650        PRUNE, or COMMIT means we must assume failure without checking subsequent
1651        branches. */
1652    
1653        case OP_ASSERT_NOT:
1654        case OP_ASSERTBACK_NOT:
1655        save_mark = md->mark;
1656        if (md->match_function_type == MATCH_CONDASSERT)
1657          {
1658          condassert = TRUE;
1659          md->match_function_type = 0;
1660          }
1661        else condassert = FALSE;
1662    
1663        do
1664          {
1665          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, NULL, RM5);
1666          md->mark = save_mark;
1667          if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT) RRETURN(MATCH_NOMATCH);
1668          if (rrc == MATCH_SKIP || rrc == MATCH_PRUNE || rrc == MATCH_COMMIT)
1669            {
1670            do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1671            break;
1672            }
1673    
1674          /* PCRE does not allow THEN to escape beyond an assertion; it is treated
1675          as NOMATCH. */
1676    
1677          if (rrc != MATCH_NOMATCH && rrc != MATCH_THEN) RRETURN(rrc);
1678          ecode += GET(ecode,1);
1679        }        }
1680      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
1681    
1682      if ((flags & match_condassert) != 0) RRETURN(MATCH_MATCH);      if (condassert) RRETURN(MATCH_MATCH);  /* Condition assertion */
1683    
1684      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1685      continue;      continue;
# Line 887  for (;;) Line 1690  for (;;)
1690      back a number of characters, not bytes. */      back a number of characters, not bytes. */
1691    
1692      case OP_REVERSE:      case OP_REVERSE:
1693  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1694      if (utf8)      if (utf)
1695        {        {
1696        i = GET(ecode, 1);        i = GET(ecode, 1);
1697        while (i-- > 0)        while (i-- > 0)
1698          {          {
1699          eptr--;          eptr--;
1700          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);          if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1701          BACKCHAR(eptr)          BACKCHAR(eptr);
1702          }          }
1703        }        }
1704      else      else
# Line 908  for (;;) Line 1711  for (;;)
1711        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);        if (eptr < md->start_subject) RRETURN(MATCH_NOMATCH);
1712        }        }
1713    
1714      /* Skip to next op code */      /* Save the earliest consulted character, then skip to next op code */
1715    
1716        if (eptr < md->start_used_ptr) md->start_used_ptr = eptr;
1717      ecode += 1 + LINK_SIZE;      ecode += 1 + LINK_SIZE;
1718      break;      break;
1719    
# Line 918  for (;;) Line 1722  for (;;)
1722      function is able to force a failure. */      function is able to force a failure. */
1723    
1724      case OP_CALLOUT:      case OP_CALLOUT:
1725      if (pcre_callout != NULL)      if (PUBL(callout) != NULL)
1726        {        {
1727        pcre_callout_block cb;        PUBL(callout_block) cb;
1728        cb.version          = 1;   /* Version 1 of the callout block */        cb.version          = 2;   /* Version 1 of the callout block */
1729        cb.callout_number   = ecode[1];        cb.callout_number   = ecode[1];
1730        cb.offset_vector    = md->offset_vector;        cb.offset_vector    = md->offset_vector;
1731    #if defined COMPILE_PCRE8
1732        cb.subject          = (PCRE_SPTR)md->start_subject;        cb.subject          = (PCRE_SPTR)md->start_subject;
1733        cb.subject_length   = md->end_subject - md->start_subject;  #elif defined COMPILE_PCRE16
1734        cb.start_match      = mstart - md->start_subject;        cb.subject          = (PCRE_SPTR16)md->start_subject;
1735        cb.current_position = eptr - md->start_subject;  #elif defined COMPILE_PCRE32
1736          cb.subject          = (PCRE_SPTR32)md->start_subject;
1737    #endif
1738          cb.subject_length   = (int)(md->end_subject - md->start_subject);
1739          cb.start_match      = (int)(mstart - md->start_subject);
1740          cb.current_position = (int)(eptr - md->start_subject);
1741        cb.pattern_position = GET(ecode, 2);        cb.pattern_position = GET(ecode, 2);
1742        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);        cb.next_item_length = GET(ecode, 2 + LINK_SIZE);
1743        cb.capture_top      = offset_top/2;        cb.capture_top      = offset_top/2;
1744        cb.capture_last     = md->capture_last;        cb.capture_last     = md->capture_last & CAPLMASK;
1745          /* Internal change requires this for API compatibility. */
1746          if (cb.capture_last == 0) cb.capture_last = -1;
1747        cb.callout_data     = md->callout_data;        cb.callout_data     = md->callout_data;
1748        if ((rrc = (*pcre_callout)(&cb)) > 0) RRETURN(MATCH_NOMATCH);        cb.mark             = md->nomatch_mark;
1749          if ((rrc = (*PUBL(callout))(&cb)) > 0) RRETURN(MATCH_NOMATCH);
1750        if (rrc < 0) RRETURN(rrc);        if (rrc < 0) RRETURN(rrc);
1751        }        }
1752      ecode += 2 + 2*LINK_SIZE;      ecode += 2 + 2*LINK_SIZE;
# Line 943  for (;;) Line 1756  for (;;)
1756      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
1757      whole pattern. (This is so that it works from duplicated subpatterns.)      whole pattern. (This is so that it works from duplicated subpatterns.)
1758    
1759      If there are any capturing brackets started but not finished, we have to      The state of the capturing groups is preserved over recursion, and
1760      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
1761      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
1762      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
1763      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
1764      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
1765      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.  
1766    
1767      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
1768      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
1769      for the original version of this logic. */      for the original version of this logic. It has, however, been hacked around
1770        a lot, so he is not to blame for the current way it works. */
1771    
1772      case OP_RECURSE:      case OP_RECURSE:
1773        {        {
1774          recursion_info *ri;
1775          unsigned int recno;
1776    
1777        callpat = md->start_code + GET(ecode, 1);        callpat = md->start_code + GET(ecode, 1);
1778        new_recursive.group_num = (callpat == md->start_code)? 0 :        recno = (callpat == md->start_code)? 0 :
1779          GET2(callpat, 1 + LINK_SIZE);          GET2(callpat, 1 + LINK_SIZE);
1780    
1781          /* Check for repeating a recursion without advancing the subject pointer.
1782          This should catch convoluted mutual recursions. (Some simple cases are
1783          caught at compile time.) */
1784    
1785          for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
1786            if (recno == ri->group_num && eptr == ri->subject_position)
1787              RRETURN(PCRE_ERROR_RECURSELOOP);
1788    
1789        /* Add to "recursing stack" */        /* Add to "recursing stack" */
1790    
1791          new_recursive.group_num = recno;
1792          new_recursive.saved_capture_last = md->capture_last;
1793          new_recursive.subject_position = eptr;
1794        new_recursive.prevrec = md->recursive;        new_recursive.prevrec = md->recursive;
1795        md->recursive = &new_recursive;        md->recursive = &new_recursive;
1796    
1797        /* Find where to continue from afterwards */        /* Where to continue from afterwards */
1798    
1799        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       new_recursive.after_call = ecode;  
1800    
1801        /* Now save the offset data. */        /* Now save the offset data */
1802    
1803        new_recursive.saved_max = md->offset_end;        new_recursive.saved_max = md->offset_end;
1804        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)        if (new_recursive.saved_max <= REC_STACK_SAVE_MAX)
# Line 982  for (;;) Line 1806  for (;;)
1806        else        else
1807          {          {
1808          new_recursive.offset_save =          new_recursive.offset_save =
1809            (int *)(pcre_malloc)(new_recursive.saved_max * sizeof(int));            (int *)(PUBL(malloc))(new_recursive.saved_max * sizeof(int));
1810          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);          if (new_recursive.offset_save == NULL) RRETURN(PCRE_ERROR_NOMEMORY);
1811          }          }
   
1812        memcpy(new_recursive.offset_save, md->offset_vector,        memcpy(new_recursive.offset_save, md->offset_vector,
1813              new_recursive.saved_max * sizeof(int));              new_recursive.saved_max * sizeof(int));
       new_recursive.save_start = mstart;  
       mstart = eptr;  
1814    
1815        /* OK, now we can do the recursion. For each top-level alternative we        /* OK, now we can do the recursion. After processing each alternative,
1816        restore the offset and recursion data. */        restore the offset data and the last captured value. If there were nested
1817          recursions, md->recursive might be changed, so reset it before looping.
1818          */
1819    
1820        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));        DPRINTF(("Recursing into group %d\n", new_recursive.group_num));
1821        flags = (*callpat >= OP_SBRA)? match_cbegroup : 0;        cbegroup = (*callpat >= OP_SBRA);
1822        do        do
1823          {          {
1824          RMATCH(eptr, callpat + _pcre_OP_lengths[*callpat], offset_top,          if (cbegroup) md->match_function_type = MATCH_CBEGROUP;
1825            md, ims, eptrb, flags, RM6);          RMATCH(eptr, callpat + PRIV(OP_lengths)[*callpat], offset_top,
1826          if (rrc == MATCH_MATCH)            md, eptrb, RM6);
1827            memcpy(md->offset_vector, new_recursive.offset_save,
1828                new_recursive.saved_max * sizeof(int));
1829            md->capture_last = new_recursive.saved_capture_last;
1830            md->recursive = new_recursive.prevrec;
1831            if (rrc == MATCH_MATCH || rrc == MATCH_ACCEPT)
1832            {            {
1833            DPRINTF(("Recursion matched\n"));            DPRINTF(("Recursion matched\n"));
           md->recursive = new_recursive.prevrec;  
1834            if (new_recursive.offset_save != stacksave)            if (new_recursive.offset_save != stacksave)
1835              (pcre_free)(new_recursive.offset_save);              (PUBL(free))(new_recursive.offset_save);
1836            RRETURN(MATCH_MATCH);  
1837              /* Set where we got to in the subject, and reset the start in case
1838              it was changed by \K. This *is* propagated back out of a recursion,
1839              for Perl compatibility. */
1840    
1841              eptr = md->end_match_ptr;
1842              mstart = md->start_match_ptr;
1843              goto RECURSION_MATCHED;        /* Exit loop; end processing */
1844            }            }
1845          else if (rrc != MATCH_NOMATCH)  
1846            /* PCRE does not allow THEN, SKIP, PRUNE or COMMIT to escape beyond a
1847            recursion; they are treated as NOMATCH. These codes are defined in a
1848            range that can be tested for. Any other return code is an error. */
1849    
1850            else if (rrc != MATCH_NOMATCH &&
1851                     (rrc < MATCH_BACKTRACK_MIN || rrc > MATCH_BACKTRACK_MAX))
1852            {            {
1853            DPRINTF(("Recursion gave error %d\n", rrc));            DPRINTF(("Recursion gave error %d\n", rrc));
1854              if (new_recursive.offset_save != stacksave)
1855                (PUBL(free))(new_recursive.offset_save);
1856            RRETURN(rrc);            RRETURN(rrc);
1857            }            }
1858    
1859          md->recursive = &new_recursive;          md->recursive = &new_recursive;
         memcpy(md->offset_vector, new_recursive.offset_save,  
             new_recursive.saved_max * sizeof(int));  
1860          callpat += GET(callpat, 1);          callpat += GET(callpat, 1);
1861          }          }
1862        while (*callpat == OP_ALT);        while (*callpat == OP_ALT);
# Line 1024  for (;;) Line 1864  for (;;)
1864        DPRINTF(("Recursion didn't match\n"));        DPRINTF(("Recursion didn't match\n"));
1865        md->recursive = new_recursive.prevrec;        md->recursive = new_recursive.prevrec;
1866        if (new_recursive.offset_save != stacksave)        if (new_recursive.offset_save != stacksave)
1867          (pcre_free)(new_recursive.offset_save);          (PUBL(free))(new_recursive.offset_save);
1868        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
1869        }        }
     /* Control never reaches here */  
   
     /* "Once" brackets are like assertion brackets except that after a match,  
     the point in the subject string is not moved back. Thus there can never be  
     a move back into the brackets. Friedl calls these "atomic" subpatterns.  
     Check the alternative branches in turn - the matching won't pass the KET  
     for this kind of subpattern. If any one branch matches, we carry on as at  
     the end of a normal bracket, leaving the subject pointer. */  
   
     case OP_ONCE:  
     prev = ecode;  
     saved_eptr = eptr;  
   
     do  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM7);  
       if (rrc == MATCH_MATCH) break;  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += GET(ecode,1);  
       }  
     while (*ecode == OP_ALT);  
   
     /* If hit the end of the group (which could be repeated), fail */  
   
     if (*ecode != OP_ONCE && *ecode != OP_ALT) RRETURN(MATCH_NOMATCH);  
   
     /* Continue as from after the assertion, updating the offsets high water  
     mark, since extracts may have been taken. */  
   
     do ecode += GET(ecode, 1); while (*ecode == OP_ALT);  
   
     offset_top = md->end_offset_top;  
     eptr = md->end_match_ptr;  
   
     /* For a non-repeating ket, just continue at this level. This also  
     happens for a repeating ket if no characters were matched in the group.  
     This is the forcible breaking of infinite loops as implemented in Perl  
     5.005. If there is an options reset, it will get obeyed in the normal  
     course of events. */  
   
     if (*ecode == OP_KET || eptr == saved_eptr)  
       {  
       ecode += 1+LINK_SIZE;  
       break;  
       }  
1870    
1871      /* The repeating kets try the rest of the pattern or restart from the      RECURSION_MATCHED:
1872      preceding bracket, in the appropriate order. The second "call" of match()      break;
     uses tail recursion, to avoid using another stack frame. We need to reset  
     any options that changed within the bracket before re-running it, so  
     check the next opcode. */  
   
     if (ecode[1+LINK_SIZE] == OP_OPT)  
       {  
       ims = (ims & ~PCRE_IMS) | ecode[4];  
       DPRINTF(("ims set to %02lx at group repeat\n", ims));  
       }  
   
     if (*ecode == OP_KETRMIN)  
       {  
       RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM8);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode = prev;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     else  /* OP_KETRMAX */  
       {  
       RMATCH(eptr, prev, offset_top, md, ims, eptrb, match_cbegroup, RM9);  
       if (rrc != MATCH_NOMATCH) RRETURN(rrc);  
       ecode += 1 + LINK_SIZE;  
       flags = 0;  
       goto TAIL_RECURSE;  
       }  
     /* Control never gets here */  
1873    
1874      /* 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
1875      bracketed group and go to there. */      bracketed group and go to there. */
# Line 1110  for (;;) Line 1878  for (;;)
1878      do ecode += GET(ecode,1); while (*ecode == OP_ALT);      do ecode += GET(ecode,1); while (*ecode == OP_ALT);
1879      break;      break;
1880    
1881      /* BRAZERO and BRAMINZERO occur just before a bracket group, indicating      /* BRAZERO, BRAMINZERO and SKIPZERO occur just before a bracket group,
1882      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
1883      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
1884      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
1885      preceded by BRAZERO or BRAMINZERO. */      optional ones preceded by BRAZERO or BRAMINZERO. */
1886    
1887      case OP_BRAZERO:      case OP_BRAZERO:
1888        {      next = ecode + 1;
1889        next = ecode+1;      RMATCH(eptr, next, offset_top, md, eptrb, RM10);
1890        RMATCH(eptr, next, offset_top, md, ims, eptrb, 0, RM10);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1891        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      do next += GET(next, 1); while (*next == OP_ALT);
1892        do next += GET(next,1); while (*next == OP_ALT);      ecode = next + 1 + LINK_SIZE;
       ecode = next + 1 + LINK_SIZE;  
       }  
1893      break;      break;
1894    
1895      case OP_BRAMINZERO:      case OP_BRAMINZERO:
1896        {      next = ecode + 1;
1897        next = ecode+1;      do next += GET(next, 1); while (*next == OP_ALT);
1898        do next += GET(next, 1); while (*next == OP_ALT);      RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, eptrb, RM11);
1899        RMATCH(eptr, next + 1+LINK_SIZE, offset_top, md, ims, eptrb, 0, RM11);      if (rrc != MATCH_NOMATCH) RRETURN(rrc);
1900        if (rrc != MATCH_NOMATCH) RRETURN(rrc);      ecode++;
1901        ecode++;      break;
1902        }  
1903        case OP_SKIPZERO:
1904        next = ecode+1;
1905        do next += GET(next,1); while (*next == OP_ALT);
1906        ecode = next + 1 + LINK_SIZE;
1907      break;      break;
1908    
1909        /* BRAPOSZERO occurs before a possessive bracket group. Don't do anything
1910        here; just jump to the group, with allow_zero set TRUE. */
1911    
1912        case OP_BRAPOSZERO:
1913        op = *(++ecode);
1914        allow_zero = TRUE;
1915        if (op == OP_CBRAPOS || op == OP_SCBRAPOS) goto POSSESSIVE_CAPTURE;
1916          goto POSSESSIVE_NON_CAPTURE;
1917    
1918      /* End of a group, repeated or non-repeating. */      /* End of a group, repeated or non-repeating. */
1919    
1920      case OP_KET:      case OP_KET:
1921      case OP_KETRMIN:      case OP_KETRMIN:
1922      case OP_KETRMAX:      case OP_KETRMAX:
1923        case OP_KETRPOS:
1924      prev = ecode - GET(ecode, 1);      prev = ecode - GET(ecode, 1);
1925    
1926      /* If this was a group that remembered the subject start, in order to break      /* If this was a group that remembered the subject start, in order to break
1927      infinite repeats of empty string matches, retrieve the subject start from      infinite repeats of empty string matches, retrieve the subject start from
1928      the chain. Otherwise, set it NULL. */      the chain. Otherwise, set it NULL. */
1929    
1930      if (*prev >= OP_SBRA)      if (*prev >= OP_SBRA || *prev == OP_ONCE)
1931        {        {
1932        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */        saved_eptr = eptrb->epb_saved_eptr;   /* Value at start of group */
1933        eptrb = eptrb->epb_prev;              /* Backup to previous group */        eptrb = eptrb->epb_prev;              /* Backup to previous group */
1934        }        }
1935      else saved_eptr = NULL;      else saved_eptr = NULL;
1936    
1937      /* If we are at the end of an assertion group, stop matching and return      /* If we are at the end of an assertion group or a non-capturing atomic
1938      MATCH_MATCH, but record the current high water mark for use by positive      group, stop matching and return MATCH_MATCH, but record the current high
1939      assertions. Do this also for the "once" (atomic) groups. */      water mark for use by positive assertions. We also need to record the match
1940        start in case it was changed by \K. */
1941      if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||  
1942          *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||      if ((*prev >= OP_ASSERT && *prev <= OP_ASSERTBACK_NOT) ||
1943          *prev == OP_ONCE)           *prev == OP_ONCE_NC)
1944        {        {
1945        md->end_match_ptr = eptr;      /* For ONCE */        md->end_match_ptr = eptr;      /* For ONCE_NC */
1946        md->end_offset_top = offset_top;        md->end_offset_top = offset_top;
1947        RRETURN(MATCH_MATCH);        md->start_match_ptr = mstart;
1948          RRETURN(MATCH_MATCH);         /* Sets md->mark */
1949        }        }
1950    
1951      /* For capturing groups we have to check the group number back at the start      /* For capturing groups we have to check the group number back at the start
1952      and if necessary complete handling an extraction by setting the offsets and      and if necessary complete handling an extraction by setting the offsets and
1953      bumping the high water mark. Note that whole-pattern recursion is coded as      bumping the high water mark. Whole-pattern recursion is coded as a recurse
1954      a recurse into group 0, so it won't be picked up here. Instead, we catch it      into group 0, so it won't be picked up here. Instead, we catch it when the
1955      when the OP_END is reached. Other recursion is handled here. */      OP_END is reached. Other recursion is handled here. We just have to record
1956        the current subject position and start match pointer and give a MATCH
1957        return. */
1958    
1959      if (*prev == OP_CBRA || *prev == OP_SCBRA)      if (*prev == OP_CBRA || *prev == OP_SCBRA ||
1960            *prev == OP_CBRAPOS || *prev == OP_SCBRAPOS)
1961        {        {
1962        number = GET2(prev, 1+LINK_SIZE);        number = GET2(prev, 1+LINK_SIZE);
1963        offset = number << 1;        offset = number << 1;
1964    
1965  #ifdef DEBUG  #ifdef PCRE_DEBUG
1966        printf("end bracket %d", number);        printf("end bracket %d", number);
1967        printf("\n");        printf("\n");
1968  #endif  #endif
1969    
1970          /* Handle a recursively called group. */
1971    
1972          if (md->recursive != NULL && md->recursive->group_num == number)
1973            {
1974            md->end_match_ptr = eptr;
1975            md->start_match_ptr = mstart;
1976            RRETURN(MATCH_MATCH);
1977            }
1978    
1979          /* Deal with capturing */
1980    
1981        md->capture_last = number;        md->capture_last = (md->capture_last & OVFLMASK) | number;
1982        if (offset >= md->offset_max) md->offset_overflow = TRUE; else        if (offset >= md->offset_max) md->capture_last |= OVFLBIT; else
1983          {          {
1984            /* If offset is greater than offset_top, it means that we are
1985            "skipping" a capturing group, and that group's offsets must be marked
1986            unset. In earlier versions of PCRE, all the offsets were unset at the
1987            start of matching, but this doesn't work because atomic groups and
1988            assertions can cause a value to be set that should later be unset.
1989            Example: matching /(?>(a))b|(a)c/ against "ac". This sets group 1 as
1990            part of the atomic group, but this is not on the final matching path,
1991            so must be unset when 2 is set. (If there is no group 2, there is no
1992            problem, because offset_top will then be 2, indicating no capture.) */
1993    
1994            if (offset > offset_top)
1995              {
1996              register int *iptr = md->offset_vector + offset_top;
1997              register int *iend = md->offset_vector + offset;
1998              while (iptr < iend) *iptr++ = -1;
1999              }
2000    
2001            /* Now make the extraction */
2002    
2003          md->offset_vector[offset] =          md->offset_vector[offset] =
2004            md->offset_vector[md->offset_end - number];            md->offset_vector[md->offset_end - number];
2005          md->offset_vector[offset+1] = eptr - md->start_subject;          md->offset_vector[offset+1] = (int)(eptr - md->start_subject);
2006          if (offset_top <= offset) offset_top = offset + 2;          if (offset_top <= offset) offset_top = offset + 2;
2007          }          }
2008          }
2009    
2010        /* Handle a recursively called group. Restore the offsets      /* For an ordinary non-repeating ket, just continue at this level. This
2011        appropriately and continue from after the call. */      also happens for a repeating ket if no characters were matched in the
2012        group. This is the forcible breaking of infinite loops as implemented in
2013        Perl 5.005. For a non-repeating atomic group that includes captures,
2014        establish a backup point by processing the rest of the pattern at a lower
2015        level. If this results in a NOMATCH return, pass MATCH_ONCE back to the
2016        original OP_ONCE level, thereby bypassing intermediate backup points, but
2017        resetting any captures that happened along the way. */
2018    
2019        if (md->recursive != NULL && md->recursive->group_num == number)      if (*ecode == OP_KET || eptr == saved_eptr)
2020          {
2021          if (*prev == OP_ONCE)
2022          {          {
2023          recursion_info *rec = md->recursive;          RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM12);
2024          DPRINTF(("Recursion (%d) succeeded - continuing\n", number));          if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2025          md->recursive = rec->prevrec;          md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2026          mstart = rec->save_start;          RRETURN(MATCH_ONCE);
         memcpy(md->offset_vector, rec->offset_save,  
           rec->saved_max * sizeof(int));  
         ecode = rec->after_call;  
         ims = original_ims;  
         break;  
2027          }          }
2028          ecode += 1 + LINK_SIZE;    /* Carry on at this level */
2029          break;
2030        }        }
2031    
2032      /* For both capturing and non-capturing groups, reset the value of the ims      /* OP_KETRPOS is a possessive repeating ket. Remember the current position,
2033      flags, in case they got changed during the group. */      and return the MATCH_KETRPOS. This makes it possible to do the repeats one
2034        at a time from the outer level, thus saving stack. */
     ims = original_ims;  
     DPRINTF(("ims reset to %02lx\n", ims));  
2035    
2036      /* For a non-repeating ket, just continue at this level. This also      if (*ecode == OP_KETRPOS)
     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)  
2037        {        {
2038        ecode += 1 + LINK_SIZE;        md->end_match_ptr = eptr;
2039        break;        md->end_offset_top = offset_top;
2040          RRETURN(MATCH_KETRPOS);
2041        }        }
2042    
2043      /* The repeating kets try the rest of the pattern or restart from the      /* The normal repeating kets try the rest of the pattern or restart from
2044      preceding bracket, in the appropriate order. In the second case, we can use      the preceding bracket, in the appropriate order. In the second case, we can
2045      tail recursion to avoid using another stack frame, unless we have an      use tail recursion to avoid using another stack frame, unless we have an
2046      unlimited repeat of a group that can match an empty string. */      an atomic group or an unlimited repeat of a group that can match an empty
2047        string. */
     flags = (*prev >= OP_SBRA)? match_cbegroup : 0;  
2048    
2049      if (*ecode == OP_KETRMIN)      if (*ecode == OP_KETRMIN)
2050        {        {
2051        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, ims, eptrb, 0, RM12);        RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM7);
2052        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2053        if (flags != 0)    /* Could match an empty string */        if (*prev == OP_ONCE)
2054            {
2055            RMATCH(eptr, prev, offset_top, md, eptrb, RM8);
2056            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2057            md->once_target = prev;  /* Level at which to change to MATCH_NOMATCH */
2058            RRETURN(MATCH_ONCE);
2059            }
2060          if (*prev >= OP_SBRA)    /* Could match an empty string */
2061          {          {
2062          RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM50);          RMATCH(eptr, prev, offset_top, md, eptrb, RM50);
2063          RRETURN(rrc);          RRETURN(rrc);
2064          }          }
2065        ecode = prev;        ecode = prev;
# Line 1248  for (;;) Line 2067  for (;;)
2067        }        }
2068      else  /* OP_KETRMAX */      else  /* OP_KETRMAX */
2069        {        {
2070        RMATCH(eptr, prev, offset_top, md, ims, eptrb, flags, RM13);        RMATCH(eptr, prev, offset_top, md, eptrb, RM13);
2071          if (rrc == MATCH_ONCE && md->once_target == prev) rrc = MATCH_NOMATCH;
2072        if (rrc != MATCH_NOMATCH) RRETURN(rrc);        if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2073          if (*prev == OP_ONCE)
2074            {
2075            RMATCH(eptr, ecode + 1 + LINK_SIZE, offset_top, md, eptrb, RM9);
2076            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2077            md->once_target = prev;
2078            RRETURN(MATCH_ONCE);
2079            }
2080        ecode += 1 + LINK_SIZE;        ecode += 1 + LINK_SIZE;
       flags = 0;  
2081        goto TAIL_RECURSE;        goto TAIL_RECURSE;
2082        }        }
2083      /* Control never gets here */      /* Control never gets here */
2084    
2085      /* Start of subject unless notbol, or after internal newline if multiline */      /* Not multiline mode: start of subject assertion, unless notbol. */
2086    
2087      case OP_CIRC:      case OP_CIRC:
2088      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);      if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
     if ((ims & PCRE_MULTILINE) != 0)  
       {  
       if (eptr != md->start_subject &&  
           (eptr == md->end_subject || !WAS_NEWLINE(eptr)))  
         RRETURN(MATCH_NOMATCH);  
       ecode++;  
       break;  
       }  
     /* ... else fall through */  
2089    
2090      /* Start of subject assertion */      /* Start of subject assertion */
2091    
# Line 1277  for (;;) Line 2094  for (;;)
2094      ecode++;      ecode++;
2095      break;      break;
2096    
2097        /* Multiline mode: start of subject unless notbol, or after any newline. */
2098    
2099        case OP_CIRCM:
2100        if (md->notbol && eptr == md->start_subject) RRETURN(MATCH_NOMATCH);
2101        if (eptr != md->start_subject &&
2102            (eptr == md->end_subject || !WAS_NEWLINE(eptr)))
2103          RRETURN(MATCH_NOMATCH);
2104        ecode++;
2105        break;
2106    
2107      /* Start of match assertion */      /* Start of match assertion */
2108    
2109      case OP_SOM:      case OP_SOM:
# Line 1291  for (;;) Line 2118  for (;;)
2118      ecode++;      ecode++;
2119      break;      break;
2120    
2121      /* Assert before internal newline if multiline, or before a terminating      /* Multiline mode: assert before any newline, or before end of subject
2122      newline unless endonly is set, else end of subject unless noteol is set. */      unless noteol is set. */
2123    
2124      case OP_DOLL:      case OP_DOLLM:
2125      if ((ims & PCRE_MULTILINE) != 0)      if (eptr < md->end_subject)
2126        {        {
2127        if (eptr < md->end_subject)        if (!IS_NEWLINE(eptr))
2128          { if (!IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH); }          {
2129        else          if (md->partial != 0 &&
2130          { if (md->noteol) RRETURN(MATCH_NOMATCH); }              eptr + 1 >= md->end_subject &&
2131        ecode++;              NLBLOCK->nltype == NLTYPE_FIXED &&
2132        break;              NLBLOCK->nllen == 2 &&
2133                RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2134              {
2135              md->hitend = TRUE;
2136              if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2137              }
2138            RRETURN(MATCH_NOMATCH);
2139            }
2140        }        }
2141      else      else
2142        {        {
2143        if (md->noteol) RRETURN(MATCH_NOMATCH);        if (md->noteol) RRETURN(MATCH_NOMATCH);
2144        if (!md->endonly)        SCHECK_PARTIAL();
         {  
         if (eptr != md->end_subject &&  
             (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))  
           RRETURN(MATCH_NOMATCH);  
         ecode++;  
         break;  
         }  
2145        }        }
2146        ecode++;
2147        break;
2148    
2149        /* Not multiline mode: assert before a terminating newline or before end of
2150        subject unless noteol is set. */
2151    
2152        case OP_DOLL:
2153        if (md->noteol) RRETURN(MATCH_NOMATCH);
2154        if (!md->endonly) goto ASSERT_NL_OR_EOS;
2155    
2156      /* ... else fall through for endonly */      /* ... else fall through for endonly */
2157    
2158      /* End of subject assertion (\z) */      /* End of subject assertion (\z) */
2159    
2160      case OP_EOD:      case OP_EOD:
2161      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr < md->end_subject) RRETURN(MATCH_NOMATCH);
2162        SCHECK_PARTIAL();
2163      ecode++;      ecode++;
2164      break;      break;
2165    
2166      /* End of subject or ending \n assertion (\Z) */      /* End of subject or ending \n assertion (\Z) */
2167    
2168      case OP_EODN:      case OP_EODN:
2169      if (eptr != md->end_subject &&      ASSERT_NL_OR_EOS:
2170        if (eptr < md->end_subject &&
2171          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))          (!IS_NEWLINE(eptr) || eptr != md->end_subject - md->nllen))
2172          {
2173          if (md->partial != 0 &&
2174              eptr + 1 >= md->end_subject &&
2175              NLBLOCK->nltype == NLTYPE_FIXED &&
2176              NLBLOCK->nllen == 2 &&
2177              RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2178            {
2179            md->hitend = TRUE;
2180            if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2181            }
2182        RRETURN(MATCH_NOMATCH);        RRETURN(MATCH_NOMATCH);
2183          }
2184    
2185        /* Either at end of string or \n before end. */
2186    
2187        SCHECK_PARTIAL();
2188      ecode++;      ecode++;
2189      break;      break;
2190    
# Line 1342  for (;;) Line 2196  for (;;)
2196    
2197        /* Find out if the previous and current characters are "word" characters.        /* Find out if the previous and current characters are "word" characters.
2198        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
2199        be "non-word" characters. */        be "non-word" characters. Remember the earliest consulted character for
2200          partial matching. */
2201    
2202  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2203        if (utf8)        if (utf)
2204          {          {
2205            /* Get status of previous character */
2206    
2207          if (eptr == md->start_subject) prev_is_word = FALSE; else          if (eptr == md->start_subject) prev_is_word = FALSE; else
2208            {            {
2209            const uschar *lastptr = eptr - 1;            PCRE_PUCHAR lastptr = eptr - 1;
2210            while((*lastptr & 0xc0) == 0x80) lastptr--;            BACKCHAR(lastptr);
2211              if (lastptr < md->start_used_ptr) md->start_used_ptr = lastptr;
2212            GETCHAR(c, lastptr);            GETCHAR(c, lastptr);
2213    #ifdef SUPPORT_UCP
2214              if (md->use_ucp)
2215                {
2216                if (c == '_') prev_is_word = TRUE; else
2217                  {
2218                  int cat = UCD_CATEGORY(c);
2219                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2220                  }
2221                }
2222              else
2223    #endif
2224            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            prev_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2225            }            }
2226          if (eptr >= md->end_subject) cur_is_word = FALSE; else  
2227            /* Get status of next character */
2228    
2229            if (eptr >= md->end_subject)
2230              {
2231              SCHECK_PARTIAL();
2232              cur_is_word = FALSE;
2233              }
2234            else
2235            {            {
2236            GETCHAR(c, eptr);            GETCHAR(c, eptr);
2237    #ifdef SUPPORT_UCP
2238              if (md->use_ucp)
2239                {
2240                if (c == '_') cur_is_word = TRUE; else
2241                  {
2242                  int cat = UCD_CATEGORY(c);
2243                  cur_is_word = (cat == ucp_L || cat == ucp_N);
2244                  }
2245                }
2246              else
2247    #endif
2248            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;            cur_is_word = c < 256 && (md->ctypes[c] & ctype_word) != 0;
2249            }            }
2250          }          }
2251        else        else
2252  #endif  #endif
2253    
2254        /* More streamlined when not in UTF-8 mode */        /* Not in UTF-8 mode, but we may still have PCRE_UCP set, and for
2255          consistency with the behaviour of \w we do use it in this case. */
2256    
2257          {          {
2258          prev_is_word = (eptr != md->start_subject) &&          /* Get status of previous character */
2259            ((md->ctypes[eptr[-1]] & ctype_word) != 0);  
2260          cur_is_word = (eptr < md->end_subject) &&          if (eptr == md->start_subject) prev_is_word = FALSE; else
2261            ((md->ctypes[*eptr] & ctype_word) != 0);            {
2262              if (eptr <= md->start_used_ptr) md->start_used_ptr = eptr - 1;
2263    #ifdef SUPPORT_UCP
2264              if (md->use_ucp)
2265                {
2266                c = eptr[-1];
2267                if (c == '_') prev_is_word = TRUE; else
2268                  {
2269                  int cat = UCD_CATEGORY(c);
2270                  prev_is_word = (cat == ucp_L || cat == ucp_N);
2271                  }
2272                }
2273              else
2274    #endif
2275              prev_is_word = MAX_255(eptr[-1])
2276                && ((md->ctypes[eptr[-1]] & ctype_word) != 0);
2277              }
2278    
2279            /* Get status of next character */
2280    
2281            if (eptr >= md->end_subject)
2282              {
2283              SCHECK_PARTIAL();
2284              cur_is_word = FALSE;
2285              }
2286            else
2287    #ifdef SUPPORT_UCP
2288            if (md->use_ucp)
2289              {
2290              c = *eptr;
2291              if (c == '_') cur_is_word = TRUE; else
2292                {
2293                int cat = UCD_CATEGORY(c);
2294                cur_is_word = (cat == ucp_L || cat == ucp_N);
2295                }
2296              }
2297            else
2298    #endif
2299            cur_is_word = MAX_255(*eptr)
2300              && ((md->ctypes[*eptr] & ctype_word) != 0);
2301          }          }
2302    
2303        /* Now see if the situation is what we want */        /* Now see if the situation is what we want */
# Line 1380  for (;;) Line 2308  for (;;)
2308        }        }
2309      break;      break;
2310    
2311      /* Match a single character type; inline for speed */      /* Match any single character type except newline; have to take care with
2312        CRLF newlines and partial matching. */
2313    
2314      case OP_ANY:      case OP_ANY:
2315      if ((ims & PCRE_DOTALL) == 0)      if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);
2316        if (md->partial != 0 &&
2317            eptr + 1 >= md->end_subject &&
2318            NLBLOCK->nltype == NLTYPE_FIXED &&
2319            NLBLOCK->nllen == 2 &&
2320            RAWUCHARTEST(eptr) == NLBLOCK->nl[0])
2321        {        {
2322        if (IS_NEWLINE(eptr)) RRETURN(MATCH_NOMATCH);        md->hitend = TRUE;
2323          if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2324        }        }
2325      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);  
2326      if (utf8)      /* Fall through */
2327        while (eptr < md->end_subject && (*eptr & 0xc0) == 0x80) eptr++;  
2328        /* Match any single character whatsoever. */
2329    
2330        case OP_ALLANY:
2331        if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2332          {                            /* not be updated before SCHECK_PARTIAL. */
2333          SCHECK_PARTIAL();
2334          RRETURN(MATCH_NOMATCH);
2335          }
2336        eptr++;
2337    #ifdef SUPPORT_UTF
2338        if (utf) ACROSSCHAR(eptr < md->end_subject, *eptr, eptr++);
2339    #endif
2340      ecode++;      ecode++;
2341      break;      break;
2342    
# Line 1397  for (;;) Line 2344  for (;;)
2344      any byte, even newline, independent of the setting of PCRE_DOTALL. */      any byte, even newline, independent of the setting of PCRE_DOTALL. */
2345    
2346      case OP_ANYBYTE:      case OP_ANYBYTE:
2347      if (eptr++ >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)   /* DO NOT merge the eptr++ here; it must */
2348          {                            /* not be updated before SCHECK_PARTIAL. */
2349          SCHECK_PARTIAL();
2350          RRETURN(MATCH_NOMATCH);
2351          }
2352        eptr++;
2353      ecode++;      ecode++;
2354      break;      break;
2355    
2356      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2357      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2358          {
2359          SCHECK_PARTIAL();
2360          RRETURN(MATCH_NOMATCH);
2361          }
2362      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2363      if (      if (
2364  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2365         c < 256 &&         c < 256 &&
2366  #endif  #endif
2367         (md->ctypes[c] & ctype_digit) != 0         (md->ctypes[c] & ctype_digit) != 0
# Line 1415  for (;;) Line 2371  for (;;)
2371      break;      break;
2372    
2373      case OP_DIGIT:      case OP_DIGIT:
2374      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2375          {
2376          SCHECK_PARTIAL();
2377          RRETURN(MATCH_NOMATCH);
2378          }
2379      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2380      if (      if (
2381  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2382         c >= 256 ||         c > 255 ||
2383  #endif  #endif
2384         (md->ctypes[c] & ctype_digit) == 0         (md->ctypes[c] & ctype_digit) == 0
2385         )         )
# Line 1428  for (;;) Line 2388  for (;;)
2388      break;      break;
2389    
2390      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2391      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2392          {
2393          SCHECK_PARTIAL();
2394          RRETURN(MATCH_NOMATCH);
2395          }
2396      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2397      if (      if (
2398  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2399         c < 256 &&         c < 256 &&
2400  #endif  #endif
2401         (md->ctypes[c] & ctype_space) != 0         (md->ctypes[c] & ctype_space) != 0
# Line 1441  for (;;) Line 2405  for (;;)
2405      break;      break;
2406    
2407      case OP_WHITESPACE:      case OP_WHITESPACE:
2408      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2409          {
2410          SCHECK_PARTIAL();
2411          RRETURN(MATCH_NOMATCH);
2412          }
2413      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2414      if (      if (
2415  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2416         c >= 256 ||         c > 255 ||
2417  #endif  #endif
2418         (md->ctypes[c] & ctype_space) == 0         (md->ctypes[c] & ctype_space) == 0
2419         )         )
# Line 1454  for (;;) Line 2422  for (;;)
2422      break;      break;
2423    
2424      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2425      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2426          {
2427          SCHECK_PARTIAL();
2428          RRETURN(MATCH_NOMATCH);
2429          }
2430      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2431      if (      if (
2432  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2433         c < 256 &&         c < 256 &&
2434  #endif  #endif
2435         (md->ctypes[c] & ctype_word) != 0         (md->ctypes[c] & ctype_word) != 0
# Line 1467  for (;;) Line 2439  for (;;)
2439      break;      break;
2440    
2441      case OP_WORDCHAR:      case OP_WORDCHAR:
2442      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2443          {
2444          SCHECK_PARTIAL();
2445          RRETURN(MATCH_NOMATCH);
2446          }
2447      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2448      if (      if (
2449  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !(defined COMPILE_PCRE8)
2450         c >= 256 ||         c > 255 ||
2451  #endif  #endif
2452         (md->ctypes[c] & ctype_word) == 0         (md->ctypes[c] & ctype_word) == 0
2453         )         )
# Line 1480  for (;;) Line 2456  for (;;)
2456      break;      break;
2457    
2458      case OP_ANYNL:      case OP_ANYNL:
2459      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2460          {
2461          SCHECK_PARTIAL();
2462          RRETURN(MATCH_NOMATCH);
2463          }
2464      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2465      switch(c)      switch(c)
2466        {        {
2467        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
2468        case 0x000d:  
2469        if (eptr < md->end_subject && *eptr == 0x0a) eptr++;        case CHAR_CR:
2470          if (eptr >= md->end_subject)
2471            {
2472            SCHECK_PARTIAL();
2473            }
2474          else if (RAWUCHARTEST(eptr) == CHAR_LF) eptr++;
2475          break;
2476    
2477          case CHAR_LF:
2478        break;        break;
2479        case 0x000a:  
2480        case 0x000b:        case CHAR_VT:
2481        case 0x000c:        case CHAR_FF:
2482        case 0x0085:        case CHAR_NEL:
2483    #ifndef EBCDIC
2484        case 0x2028:        case 0x2028:
2485        case 0x2029:        case 0x2029:
2486    #endif  /* Not EBCDIC */
2487          if (md->bsr_anycrlf) RRETURN(MATCH_NOMATCH);
2488        break;        break;
2489        }        }
2490      ecode++;      ecode++;
2491      break;      break;
2492    
2493      case OP_NOT_HSPACE:      case OP_NOT_HSPACE:
2494      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2495          {
2496          SCHECK_PARTIAL();
2497          RRETURN(MATCH_NOMATCH);
2498          }
2499      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2500      switch(c)      switch(c)
2501        {        {
2502          HSPACE_CASES: RRETURN(MATCH_NOMATCH);  /* Byte and multibyte cases */
2503        default: break;        default: break;
       case 0x09:      /* HT */  
       case 0x20:      /* SPACE */  
       case 0xa0:      /* NBSP */  
       case 0x1680:    /* OGHAM SPACE MARK */  
       case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */  
       case 0x2000:    /* EN QUAD */  
       case 0x2001:    /* EM QUAD */  
       case 0x2002:    /* EN SPACE */  
       case 0x2003:    /* EM SPACE */  
       case 0x2004:    /* THREE-PER-EM SPACE */  
       case 0x2005:    /* FOUR-PER-EM SPACE */  
       case 0x2006:    /* SIX-PER-EM SPACE */  
       case 0x2007:    /* FIGURE SPACE */  
       case 0x2008:    /* PUNCTUATION SPACE */  
       case 0x2009:    /* THIN SPACE */  
       case 0x200A:    /* HAIR SPACE */  
       case 0x202f:    /* NARROW NO-BREAK SPACE */  
       case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */  
       case 0x3000:    /* IDEOGRAPHIC SPACE */  
       RRETURN(MATCH_NOMATCH);  
2504        }        }
2505      ecode++;      ecode++;
2506      break;      break;
2507    
2508      case OP_HSPACE:      case OP_HSPACE:
2509      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2510          {
2511          SCHECK_PARTIAL();
2512          RRETURN(MATCH_NOMATCH);
2513          }
2514      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2515      switch(c)      switch(c)
2516        {        {
2517          HSPACE_CASES: break;  /* Byte and multibyte cases */
2518        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
       case 0x09:      /* HT */  
       case 0x20:      /* SPACE */  
       case 0xa0:      /* NBSP */  
       case 0x1680:    /* OGHAM SPACE MARK */  
       case 0x180e:    /* MONGOLIAN VOWEL SEPARATOR */  
       case 0x2000:    /* EN QUAD */  
       case 0x2001:    /* EM QUAD */  
       case 0x2002:    /* EN SPACE */  
       case 0x2003:    /* EM SPACE */  
       case 0x2004:    /* THREE-PER-EM SPACE */  
       case 0x2005:    /* FOUR-PER-EM SPACE */  
       case 0x2006:    /* SIX-PER-EM SPACE */  
       case 0x2007:    /* FIGURE SPACE */  
       case 0x2008:    /* PUNCTUATION SPACE */  
       case 0x2009:    /* THIN SPACE */  
       case 0x200A:    /* HAIR SPACE */  
       case 0x202f:    /* NARROW NO-BREAK SPACE */  
       case 0x205f:    /* MEDIUM MATHEMATICAL SPACE */  
       case 0x3000:    /* IDEOGRAPHIC SPACE */  
       break;  
2519        }        }
2520      ecode++;      ecode++;
2521      break;      break;
2522    
2523      case OP_NOT_VSPACE:      case OP_NOT_VSPACE:
2524      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2525          {
2526          SCHECK_PARTIAL();
2527          RRETURN(MATCH_NOMATCH);
2528          }
2529      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2530      switch(c)      switch(c)
2531        {        {
2532          VSPACE_CASES: RRETURN(MATCH_NOMATCH);
2533        default: break;        default: break;
       case 0x0a:      /* LF */  
       case 0x0b:      /* VT */  
       case 0x0c:      /* FF */  
       case 0x0d:      /* CR */  
       case 0x85:      /* NEL */  
       case 0x2028:    /* LINE SEPARATOR */  
       case 0x2029:    /* PARAGRAPH SEPARATOR */  
       RRETURN(MATCH_NOMATCH);  
2534        }        }
2535      ecode++;      ecode++;
2536      break;      break;
2537    
2538      case OP_VSPACE:      case OP_VSPACE:
2539      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2540          {
2541          SCHECK_PARTIAL();
2542          RRETURN(MATCH_NOMATCH);
2543          }
2544      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2545      switch(c)      switch(c)
2546        {        {
2547          VSPACE_CASES: break;
2548        default: RRETURN(MATCH_NOMATCH);        default: RRETURN(MATCH_NOMATCH);
       case 0x0a:      /* LF */  
       case 0x0b:      /* VT */  
       case 0x0c:      /* FF */  
       case 0x0d:      /* CR */  
       case 0x85:      /* NEL */  
       case 0x2028:    /* LINE SEPARATOR */  
       case 0x2029:    /* PARAGRAPH SEPARATOR */  
       break;  
2549        }        }
2550      ecode++;      ecode++;
2551      break;      break;
# Line 1601  for (;;) Line 2556  for (;;)
2556    
2557      case OP_PROP:      case OP_PROP:
2558      case OP_NOTPROP:      case OP_NOTPROP:
2559      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2560          {
2561          SCHECK_PARTIAL();
2562          RRETURN(MATCH_NOMATCH);
2563          }
2564      GETCHARINCTEST(c, eptr);      GETCHARINCTEST(c, eptr);
2565        {        {
2566        int chartype, script;        const pcre_uint32 *cp;
2567        int category = _pcre_ucp_findprop(c, &chartype, &script);        const ucd_record *prop = GET_UCD(c);
2568    
2569        switch(ecode[1])        switch(ecode[1])
2570          {          {
# Line 1614  for (;;) Line 2573  for (;;)
2573          break;          break;
2574    
2575          case PT_LAMP:          case PT_LAMP:
2576          if ((chartype == ucp_Lu ||          if ((prop->chartype == ucp_Lu ||
2577               chartype == ucp_Ll ||               prop->chartype == ucp_Ll ||
2578               chartype == ucp_Lt) == (op == OP_NOTPROP))               prop->chartype == ucp_Lt) == (op == OP_NOTPROP))
2579            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2580           break;          break;
2581    
2582          case PT_GC:          case PT_GC:
2583          if ((ecode[2] != category) == (op == OP_PROP))          if ((ecode[2] != PRIV(ucp_gentype)[prop->chartype]) == (op == OP_PROP))
2584            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2585          break;          break;
2586    
2587          case PT_PC:          case PT_PC:
2588          if ((ecode[2] != chartype) == (op == OP_PROP))          if ((ecode[2] != prop->chartype) == (op == OP_PROP))
2589            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2590          break;          break;
2591    
2592          case PT_SC:          case PT_SC:
2593          if ((ecode[2] != script) == (op == OP_PROP))          if ((ecode[2] != prop->script) == (op == OP_PROP))
2594              RRETURN(MATCH_NOMATCH);
2595            break;
2596    
2597            /* These are specials */
2598    
2599            case PT_ALNUM:
2600            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2601                 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == (op == OP_NOTPROP))
2602              RRETURN(MATCH_NOMATCH);
2603            break;
2604    
2605            case PT_SPACE:    /* Perl space */
2606            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2607                 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2608                   == (op == OP_NOTPROP))
2609              RRETURN(MATCH_NOMATCH);
2610            break;
2611    
2612            case PT_PXSPACE:  /* POSIX space */
2613            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2614                 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2615                 c == CHAR_FF || c == CHAR_CR)
2616                   == (op == OP_NOTPROP))
2617              RRETURN(MATCH_NOMATCH);
2618            break;
2619    
2620            case PT_WORD:
2621            if ((PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2622                 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2623                 c == CHAR_UNDERSCORE) == (op == OP_NOTPROP))
2624              RRETURN(MATCH_NOMATCH);
2625            break;
2626    
2627            case PT_CLIST:
2628            cp = PRIV(ucd_caseless_sets) + ecode[2];
2629            for (;;)
2630              {
2631              if (c < *cp)
2632                { if (op == OP_PROP) { RRETURN(MATCH_NOMATCH); } else break; }
2633              if (c == *cp++)
2634                { if (op == OP_PROP) break; else { RRETURN(MATCH_NOMATCH); } }
2635              }
2636            break;
2637    
2638            case PT_UCNC:
2639            if ((c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT ||
2640                 c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) ||
2641                 c >= 0xe000) == (op == OP_NOTPROP))
2642            RRETURN(MATCH_NOMATCH);            RRETURN(MATCH_NOMATCH);
2643          break;          break;
2644    
2645            /* This should never occur */
2646    
2647          default:          default:
2648          RRETURN(PCRE_ERROR_INTERNAL);          RRETURN(PCRE_ERROR_INTERNAL);
2649          }          }
# Line 1647  for (;;) Line 2656  for (;;)
2656      is in the binary; otherwise a compile-time error occurs. */      is in the binary; otherwise a compile-time error occurs. */
2657    
2658      case OP_EXTUNI:      case OP_EXTUNI:
2659      if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);      if (eptr >= md->end_subject)
2660      GETCHARINCTEST(c, eptr);        {
2661          SCHECK_PARTIAL();
2662          RRETURN(MATCH_NOMATCH);
2663          }
2664        else
2665        {        {
2666        int chartype, script;        int lgb, rgb;
2667        int category = _pcre_ucp_findprop(c, &chartype, &script);        GETCHARINCTEST(c, eptr);
2668        if (category == ucp_M) RRETURN(MATCH_NOMATCH);        lgb = UCD_GRAPHBREAK(c);
2669        while (eptr < md->end_subject)        while (eptr < md->end_subject)
2670          {          {
2671          int len = 1;          int len = 1;
2672          if (!utf8) c = *eptr; else          if (!utf) c = *eptr; else { GETCHARLEN(c, eptr, len); }
2673            {          rgb = UCD_GRAPHBREAK(c);
2674            GETCHARLEN(c, eptr, len);          if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break;
2675            }          lgb = rgb;
         category = _pcre_ucp_findprop(c, &chartype, &script);  
         if (category != ucp_M) break;  
2676          eptr += len;          eptr += len;
2677          }          }
2678        }        }
2679        CHECK_PARTIAL();
2680      ecode++;      ecode++;
2681      break;      break;
2682  #endif  #endif  /* SUPPORT_UCP */
2683    
2684    
2685      /* 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 1679  for (;;) Line 2691  for (;;)
2691      loops). */      loops). */
2692    
2693      case OP_REF:      case OP_REF:
2694        {      case OP_REFI:
2695        offset = GET2(ecode, 1) << 1;               /* Doubled ref number */      caseless = op == OP_REFI;
2696        ecode += 3;                                 /* Advance past item */      offset = GET2(ecode, 1) << 1;               /* Doubled ref number */
2697        ecode += 1 + IMM2_SIZE;
2698    
2699        /* 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];  
2700    
2701        /* Set up for repetition, or handle the non-repeated case */      (a) In the default, Perl-compatible state, set the length negative;
2702        this ensures that every attempt at a match fails. We can't just fail
2703        here, because of the possibility of quantifiers with zero minima.
2704    
2705        switch (*ecode)      (b) If the JavaScript compatibility flag is set, set the length to zero
2706          {      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;  
2707    
2708          case OP_CRRANGE:      Otherwise, set the length to the length of what was matched by the
2709          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;  
2710    
2711          default:               /* No repeat follows */      if (offset >= offset_top || md->offset_vector[offset] < 0)
2712          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);        length = (md->jscript_compat)? 0 : -1;
2713          eptr += length;      else
2714          continue;              /* With the main loop */        length = md->offset_vector[offset+1] - md->offset_vector[offset];
2715    
2716        /* Set up for repetition, or handle the non-repeated case */
2717    
2718        switch (*ecode)
2719          {
2720          case OP_CRSTAR:
2721          case OP_CRMINSTAR:
2722          case OP_CRPLUS:
2723          case OP_CRMINPLUS:
2724          case OP_CRQUERY:
2725          case OP_CRMINQUERY:
2726          c = *ecode++ - OP_CRSTAR;
2727          minimize = (c & 1) != 0;
2728          min = rep_min[c];                 /* Pick up values from tables; */
2729          max = rep_max[c];                 /* zero for max => infinity */
2730          if (max == 0) max = INT_MAX;
2731          break;
2732    
2733          case OP_CRRANGE:
2734          case OP_CRMINRANGE:
2735          minimize = (*ecode == OP_CRMINRANGE);
2736          min = GET2(ecode, 1);
2737          max = GET2(ecode, 1 + IMM2_SIZE);
2738          if (max == 0) max = INT_MAX;
2739          ecode += 1 + 2 * IMM2_SIZE;
2740          break;
2741    
2742          default:               /* No repeat follows */
2743          if ((length = match_ref(offset, eptr, length, md, caseless)) < 0)
2744            {
2745            if (length == -2) eptr = md->end_subject;   /* Partial match */
2746            CHECK_PARTIAL();
2747            RRETURN(MATCH_NOMATCH);
2748          }          }
2749          eptr += length;
2750          continue;              /* With the main loop */
2751          }
2752    
2753        /* If the length of the reference is zero, just continue with the      /* Handle repeated back references. If the length of the reference is
2754        main loop. */      zero, just continue with the main loop. If the length is negative, it
2755        means the reference is unset in non-Java-compatible mode. If the minimum is
2756        zero, we can continue at the same level without recursion. For any other
2757        minimum, carrying on will result in NOMATCH. */
2758    
2759        if (length == 0) continue;      if (length == 0) continue;
2760        if (length < 0 && min == 0) continue;
2761    
2762        /* First, ensure the minimum number of matches are present. We get back      /* First, ensure the minimum number of matches are present. We get back
2763        the length of the reference string explicitly rather than passing the      the length of the reference string explicitly rather than passing the
2764        address of eptr, so that eptr can be a register variable. */      address of eptr, so that eptr can be a register variable. */
2765    
2766        for (i = 1; i <= min; i++)      for (i = 1; i <= min; i++)
2767          {
2768          int slength;
2769          if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2770          {          {
2771          if (!match_ref(offset, eptr, length, md, ims)) RRETURN(MATCH_NOMATCH);          if (slength == -2) eptr = md->end_subject;   /* Partial match */
2772          eptr += length;          CHECK_PARTIAL();
2773            RRETURN(MATCH_NOMATCH);
2774          }          }
2775          eptr += slength;
2776          }
2777    
2778        /* If min = max, continue at the same level without recursion.      /* If min = max, continue at the same level without recursion.
2779        They are not both allowed to be zero. */      They are not both allowed to be zero. */
2780    
2781        if (min == max) continue;      if (min == max) continue;
2782    
2783        /* If minimizing, keep trying and advancing the pointer */      /* If minimizing, keep trying and advancing the pointer */
2784    
2785        if (minimize)      if (minimize)
2786          {
2787          for (fi = min;; fi++)
2788          {          {
2789          for (fi = min;; fi++)          int slength;
2790            RMATCH(eptr, ecode, offset_top, md, eptrb, RM14);
2791            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2792            if (fi >= max) RRETURN(MATCH_NOMATCH);
2793            if ((slength = match_ref(offset, eptr, length, md, caseless)) < 0)
2794            {            {
2795            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM14);            if (slength == -2) eptr = md->end_subject;   /* Partial match */
2796            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            CHECK_PARTIAL();
2797            if (fi >= max || !match_ref(offset, eptr, length, md, ims))            RRETURN(MATCH_NOMATCH);
             RRETURN(MATCH_NOMATCH);  
           eptr += length;  
2798            }            }
2799          /* Control never gets here */          eptr += slength;
2800          }          }
2801          /* Control never gets here */
2802          }
2803    
2804        /* If maximizing, find the longest string and work backwards */      /* If maximizing, find the longest string and work backwards */
2805    
2806        else      else
2807          {
2808          pp = eptr;
2809          for (i = min; i < max; i++)
2810          {          {
2811          pp = eptr;          int slength;
2812          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)  
2813            {            {
2814            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM15);            /* Can't use CHECK_PARTIAL because we don't want to update eptr in
2815            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            the soft partial matching case. */
2816            eptr -= length;  
2817              if (slength == -2 && md->partial != 0 &&
2818                  md->end_subject > md->start_used_ptr)
2819                {
2820                md->hitend = TRUE;
2821                if (md->partial > 1) RRETURN(PCRE_ERROR_PARTIAL);
2822                }
2823              break;
2824            }            }
2825          RRETURN(MATCH_NOMATCH);          eptr += slength;
2826            }
2827    
2828          while (eptr >= pp)
2829            {
2830            RMATCH(eptr, ecode, offset_top, md, eptrb, RM15);
2831            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2832            eptr -= length;
2833          }          }
2834          RRETURN(MATCH_NOMATCH);
2835        }        }
2836      /* Control never gets here */      /* Control never gets here */
2837    
   
   
2838      /* Match a bit-mapped character class, possibly repeatedly. This op code is      /* Match a bit-mapped character class, possibly repeatedly. This op code is
2839      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,
2840      and either the matching is caseful, or the characters are in the range      and either the matching is caseful, or the characters are in the range
# Line 1796  for (;;) Line 2849  for (;;)
2849      case OP_NCLASS:      case OP_NCLASS:
2850      case OP_CLASS:      case OP_CLASS:
2851        {        {
2852          /* The data variable is saved across frames, so the byte map needs to
2853          be stored there. */
2854    #define BYTE_MAP ((pcre_uint8 *)data)
2855        data = ecode + 1;                /* Save for matching */        data = ecode + 1;                /* Save for matching */
2856        ecode += 33;                     /* Advance past the item */        ecode += 1 + (32 / sizeof(pcre_uchar)); /* Advance past the item */
2857    
2858        switch (*ecode)        switch (*ecode)
2859          {          {
# Line 1818  for (;;) Line 2874  for (;;)
2874          case OP_CRMINRANGE:          case OP_CRMINRANGE:
2875          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
2876          min = GET2(ecode, 1);          min = GET2(ecode, 1);
2877          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
2878          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
2879          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
2880          break;          break;
2881    
2882          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 1830  for (;;) Line 2886  for (;;)
2886    
2887        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
2888    
2889  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2890        /* UTF-8 mode */        if (utf)
       if (utf8)  
2891          {          {
2892          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2893            {            {
2894            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2895                {
2896                SCHECK_PARTIAL();
2897                RRETURN(MATCH_NOMATCH);
2898                }
2899            GETCHARINC(c, eptr);            GETCHARINC(c, eptr);
2900            if (c > 255)            if (c > 255)
2901              {              {
2902              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);              if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2903              }              }
2904            else            else
2905              {              if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
             if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
             }  
2906            }            }
2907          }          }
2908        else        else
2909  #endif  #endif
2910        /* Not UTF-8 mode */        /* Not UTF mode */
2911          {          {
2912          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
2913            {            {
2914            if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (eptr >= md->end_subject)
2915                {
2916                SCHECK_PARTIAL();
2917                RRETURN(MATCH_NOMATCH);
2918                }
2919            c = *eptr++;            c = *eptr++;
2920            if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2921              if (c > 255)
2922                {
2923                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2924                }
2925              else
2926    #endif
2927                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2928            }            }
2929          }          }
2930    
# Line 1870  for (;;) Line 2938  for (;;)
2938    
2939        if (minimize)        if (minimize)
2940          {          {
2941  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2942          /* UTF-8 mode */          if (utf)
         if (utf8)  
2943            {            {
2944            for (fi = min;; fi++)            for (fi = min;; fi++)
2945              {              {
2946              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM16);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM16);
2947              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2948              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2949                if (eptr >= md->end_subject)
2950                  {
2951                  SCHECK_PARTIAL();
2952                  RRETURN(MATCH_NOMATCH);
2953                  }
2954              GETCHARINC(c, eptr);              GETCHARINC(c, eptr);
2955              if (c > 255)              if (c > 255)
2956                {                {
2957                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);                if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2958                }                }
2959              else              else
2960                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
               if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  
               }  
2961              }              }
2962            }            }
2963          else          else
2964  #endif  #endif
2965          /* Not UTF-8 mode */          /* Not UTF mode */
2966            {            {
2967            for (fi = min;; fi++)            for (fi = min;; fi++)
2968              {              {
2969              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM17);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM17);
2970              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
2971              if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);              if (fi >= max) RRETURN(MATCH_NOMATCH);
2972                if (eptr >= md->end_subject)
2973                  {
2974                  SCHECK_PARTIAL();
2975                  RRETURN(MATCH_NOMATCH);
2976                  }
2977              c = *eptr++;              c = *eptr++;
2978              if ((data[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);  #ifndef COMPILE_PCRE8
2979                if (c > 255)
2980                  {
2981                  if (op == OP_CLASS) RRETURN(MATCH_NOMATCH);
2982                  }
2983                else
2984    #endif
2985                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) RRETURN(MATCH_NOMATCH);
2986              }              }
2987            }            }
2988          /* Control never gets here */          /* Control never gets here */
# Line 1912  for (;;) Line 2994  for (;;)
2994          {          {
2995          pp = eptr;          pp = eptr;
2996    
2997  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2998          /* UTF-8 mode */          if (utf)
         if (utf8)  
2999            {            {
3000            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3001              {              {
3002              int len = 1;              int len = 1;
3003              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3004                  {
3005                  SCHECK_PARTIAL();
3006                  break;
3007                  }
3008              GETCHARLEN(c, eptr, len);              GETCHARLEN(c, eptr, len);
3009              if (c > 255)              if (c > 255)
3010                {                {
3011                if (op == OP_CLASS) break;                if (op == OP_CLASS) break;
3012                }                }
3013              else              else
3014                {                if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
               if ((data[c/8] & (1 << (c&7))) == 0) break;  
               }  
3015              eptr += len;              eptr += len;
3016              }              }
3017            for (;;)            for (;;)
3018              {              {
3019              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM18);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM18);
3020              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3021              if (eptr-- == pp) break;        /* Stop if tried at original pos */              if (eptr-- == pp) break;        /* Stop if tried at original pos */
3022              BACKCHAR(eptr);              BACKCHAR(eptr);
# Line 1941  for (;;) Line 3024  for (;;)
3024            }            }
3025          else          else
3026  #endif  #endif
3027            /* Not UTF-8 mode */            /* Not UTF mode */
3028            {            {
3029            for (i = min; i < max; i++)            for (i = min; i < max; i++)
3030              {              {
3031              if (eptr >= md->end_subject) break;              if (eptr >= md->end_subject)
3032                  {
3033                  SCHECK_PARTIAL();
3034                  break;
3035                  }
3036              c = *eptr;              c = *eptr;
3037              if ((data[c/8] & (1 << (c&7))) == 0) break;  #ifndef COMPILE_PCRE8
3038                if (c > 255)
3039                  {
3040                  if (op == OP_CLASS) break;
3041                  }
3042                else
3043    #endif
3044                  if ((BYTE_MAP[c/8] & (1 << (c&7))) == 0) break;
3045              eptr++;              eptr++;
3046              }              }
3047            while (eptr >= pp)            while (eptr >= pp)
3048              {              {
3049              RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM19);              RMATCH(eptr, ecode, offset_top, md, eptrb, RM19);
3050              if (rrc != MATCH_NOMATCH) RRETURN(rrc);              if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3051              eptr--;              eptr--;
3052              }              }
# Line 1960  for (;;) Line 3054  for (;;)
3054    
3055          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3056          }          }
3057    #undef BYTE_MAP
3058        }        }
3059      /* Control never gets here */      /* Control never gets here */
3060    
3061    
3062      /* Match an extended character class. This opcode is encountered only      /* Match an extended character class. This opcode is encountered only
3063      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
3064        mode, because Unicode properties are supported in non-UTF-8 mode. */
3065    
3066  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3067      case OP_XCLASS:      case OP_XCLASS:
3068        {        {
3069        data = ecode + 1 + LINK_SIZE;                /* Save for matching */        data = ecode + 1 + LINK_SIZE;                /* Save for matching */
# Line 1992  for (;;) Line 3088  for (;;)
3088          case OP_CRMINRANGE:          case OP_CRMINRANGE:
3089          minimize = (*ecode == OP_CRMINRANGE);          minimize = (*ecode == OP_CRMINRANGE);
3090          min = GET2(ecode, 1);          min = GET2(ecode, 1);
3091          max = GET2(ecode, 3);          max = GET2(ecode, 1 + IMM2_SIZE);
3092          if (max == 0) max = INT_MAX;          if (max == 0) max = INT_MAX;
3093          ecode += 5;          ecode += 1 + 2 * IMM2_SIZE;
3094          break;          break;
3095    
3096          default:               /* No repeat follows */          default:               /* No repeat follows */
# Line 2006  for (;;) Line 3102  for (;;)
3102    
3103        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3104          {          {
3105          if (eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);          if (eptr >= md->end_subject)
3106          GETCHARINC(c, eptr);            {
3107          if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);            SCHECK_PARTIAL();
3108              RRETURN(MATCH_NOMATCH);
3109              }
3110            GETCHARINCTEST(c, eptr);
3111            if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3112          }          }
3113    
3114        /* If max == min we can continue with the main loop without the        /* If max == min we can continue with the main loop without the
# Line 2023  for (;;) Line 3123  for (;;)
3123          {          {
3124          for (fi = min;; fi++)          for (fi = min;; fi++)
3125            {            {
3126            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM20);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM20);
3127            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3128            if (fi >= max || eptr >= md->end_subject) RRETURN(MATCH_NOMATCH);            if (fi >= max) RRETURN(MATCH_NOMATCH);
3129            GETCHARINC(c, eptr);            if (eptr >= md->end_subject)
3130            if (!_pcre_xclass(c, data)) RRETURN(MATCH_NOMATCH);              {
3131                SCHECK_PARTIAL();
3132                RRETURN(MATCH_NOMATCH);
3133                }
3134              GETCHARINCTEST(c, eptr);
3135              if (!PRIV(xclass)(c, data, utf)) RRETURN(MATCH_NOMATCH);
3136            }            }
3137          /* Control never gets here */          /* Control never gets here */
3138          }          }
# Line 2040  for (;;) Line 3145  for (;;)
3145          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3146            {            {
3147            int len = 1;            int len = 1;
3148            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject)
3149            GETCHARLEN(c, eptr, len);              {
3150            if (!_pcre_xclass(c, data)) break;              SCHECK_PARTIAL();
3151                break;
3152                }
3153    #ifdef SUPPORT_UTF
3154              GETCHARLENTEST(c, eptr, len);
3155    #else
3156              c = *eptr;
3157    #endif
3158              if (!PRIV(xclass)(c, data, utf)) break;
3159            eptr += len;            eptr += len;
3160            }            }
3161          for(;;)          for(;;)
3162            {            {
3163            RMATCH(eptr, ecode, offset_top, md, ims, eptrb, 0, RM21);            RMATCH(eptr, ecode, offset_top, md, eptrb, RM21);
3164            if (rrc != MATCH_NOMATCH) RRETURN(rrc);            if (rrc != MATCH_NOMATCH) RRETURN(rrc);
3165            if (eptr-- == pp) break;        /* Stop if tried at original pos */            if (eptr-- == pp) break;        /* Stop if tried at original pos */
3166            BACKCHAR(eptr)  #ifdef SUPPORT_UTF
3167              if (utf) BACKCHAR(eptr);
3168    #endif
3169            }            }
3170          RRETURN(MATCH_NOMATCH);          RRETURN(MATCH_NOMATCH);
3171          }          }
# Line 2062  for (;;) Line 3177  for (;;)
3177      /* Match a single character, casefully */      /* Match a single character, casefully */
3178    
3179      case OP_CHAR:      case OP_CHAR:
3180  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3181      if (utf8)      if (utf)
3182        {        {
3183        length = 1;        length = 1;
3184        ecode++;        ecode++;
3185        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3186        if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);        if (length > md->end_subject - eptr)
3187        while (length-- > 0) if (*ecode++ != *eptr++) RRETURN(MATCH_NOMATCH);          {
3188            CHECK_PARTIAL();             /* Not SCHECK_PARTIAL() */
3189            RRETURN(MATCH_NOMATCH);
3190            }
3191          while (length-- > 0) if (*ecode++ != RAWUCHARINC(eptr)) RRETURN(MATCH_NOMATCH);
3192        }        }
3193      else      else
3194  #endif  #endif
3195        /* Not UTF mode */
     /* Non-UTF-8 mode */  
3196        {        {
3197        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (md->end_subject - eptr < 1)
3198            {
3199            SCHECK_PARTIAL();            /* This one can use SCHECK_PARTIAL() */
3200            RRETURN(MATCH_NOMATCH);
3201            }
3202        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);        if (ecode[1] != *eptr++) RRETURN(MATCH_NOMATCH);
3203        ecode += 2;        ecode += 2;
3204        }        }
3205      break;      break;
3206    
3207      /* Match a single character, caselessly */      /* Match a single character, caselessly. If we are at the end of the
3208        subject, give up immediately. */
3209    
3210        case OP_CHARI:
3211        if (eptr >= md->end_subject)
3212          {
3213          SCHECK_PARTIAL();
3214          RRETURN(MATCH_NOMATCH);
3215          }
3216    
3217      case OP_CHARNC:  #ifdef SUPPORT_UTF
3218  #ifdef SUPPORT_UTF8      if (utf)
     if (utf8)  
3219        {        {
3220        length = 1;        length = 1;
3221        ecode++;        ecode++;
3222        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
3223    
       if (length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
   
3224        /* 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
3225        can use the fast lookup table. */        we know that its other case must also be one byte long, so we can use the
3226          fast lookup table. We know that there is at least one byte left in the
3227          subject. */
3228    
3229        if (fc < 128)        if (fc < 128)
3230          {          {
3231          if (md->lcc[*ecode++] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);          pcre_uint32 cc = RAWUCHAR(eptr);
3232            if (md->lcc[fc] != TABLE_GET(cc, md->lcc, cc)) RRETURN(MATCH_NOMATCH);
3233            ecode++;
3234            eptr++;
3235          }          }
3236    
3237        /* Otherwise we must pick up the subject character */        /* Otherwise we must pick up the subject character. Note that we cannot
3238          use the value of "length" to check for sufficient bytes left, because the
3239          other case of the character may have more or fewer bytes.  */
3240    
3241        else        else
3242          {          {
3243          unsigned int dc;          pcre_uint32 dc;
3244          GETCHARINC(dc, eptr);          GETCHARINC(dc, eptr);
3245          ecode += length;          ecode += length;
3246    
# Line 2116  for (;;) Line 3250  for (;;)
3250          if (fc != dc)          if (fc != dc)
3251            {            {
3252  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3253            if (dc != _pcre_ucp_othercase(fc))            if (dc != UCD_OTHERCASE(fc))
3254  #endif  #endif
3255              RRETURN(MATCH_NOMATCH);              RRETURN(MATCH_NOMATCH);
3256            }            }
3257          }          }
3258        }        }
3259      else      else
3260  #endif   /* SUPPORT_UTF8 */  #endif   /* SUPPORT_UTF */
3261    
3262      /* Non-UTF-8 mode */      /* Not UTF mode */
3263        {        {
3264        if (md->end_subject - eptr < 1) RRETURN(MATCH_NOMATCH);        if (TABLE_GET(ecode[1], md->lcc, ecode[1])
3265        if (md->lcc[ecode[1]] != md->lcc[*eptr++]) RRETURN(MATCH_NOMATCH);            != TABLE_GET(*eptr, md->lcc, *eptr)) RRETURN(MATCH_NOMATCH);
3266          eptr++;
3267        ecode += 2;        ecode += 2;
3268        }        }
3269      break;      break;
# Line 2136  for (;;) Line 3271  for (;;)
3271      /* Match a single character repeatedly. */      /* Match a single character repeatedly. */
3272    
3273      case OP_EXACT:      case OP_EXACT:
3274        case OP_EXACTI:
3275      min = max = GET2(ecode, 1);      min = max = GET2(ecode, 1);
3276      ecode += 3;      ecode += 1 + IMM2_SIZE;
3277      goto REPEATCHAR;      goto REPEATCHAR;
3278    
3279      case OP_POSUPTO:      case OP_POSUPTO:
3280        case OP_POSUPTOI:
3281      possessive = TRUE;      possessive = TRUE;
3282      /* Fall through */      /* Fall through */
3283    
3284      case OP_UPTO:      case OP_UPTO:
3285        case OP_UPTOI:
3286      case OP_MINUPTO:      case OP_MINUPTO:
3287        case OP_MINUPTOI:
3288      min = 0;      min = 0;
3289      max = GET2(ecode, 1);      max = GET2(ecode, 1);
3290      minimize = *ecode == OP_MINUPTO;      minimize = *ecode == OP_MINUPTO || *ecode == OP_MINUPTOI;
3291      ecode += 3;      ecode += 1 + IMM2_SIZE;
3292      goto REPEATCHAR;      goto REPEATCHAR;
3293    
3294      case OP_POSSTAR:      case OP_POSSTAR:
3295        case OP_POSSTARI:
3296      possessive = TRUE;      possessive = TRUE;
3297      min = 0;      min = 0;
3298      max = INT_MAX;      max = INT_MAX;
# Line 2160  for (;;) Line 3300  for (;;)
3300      goto REPEATCHAR;      goto REPEATCHAR;
3301    
3302      case OP_POSPLUS:      case OP_POSPLUS:
3303        case OP_POSPLUSI:
3304      possessive = TRUE;      possessive = TRUE;
3305      min = 1;      min = 1;
3306      max = INT_MAX;      max = INT_MAX;
# Line 2167  for (;;) Line 3308  for (;;)
3308      goto REPEATCHAR;      goto REPEATCHAR;
3309    
3310      case OP_POSQUERY:      case OP_POSQUERY:
3311        case OP_POSQUERYI:
3312      possessive = TRUE;      possessive = TRUE;
3313      min = 0;      min = 0;
3314      max = 1;      max = 1;
# Line 2174  for (;;) Line 3316  for (;;)
3316      goto REPEATCHAR;      goto REPEATCHAR;
3317    
3318      case OP_STAR:      case OP_STAR:
3319        case OP_STARI:
3320      case OP_MINSTAR:      case OP_MINSTAR:
3321        case OP_MINSTARI:
3322      case OP_PLUS:      case OP_PLUS:
3323        case OP_PLUSI:
3324      case OP_MINPLUS:      case OP_MINPLUS:
3325        case OP_MINPLUSI:
3326      case OP_QUERY:      case OP_QUERY:
3327        case OP_QUERYI:
3328      case OP_MINQUERY:      case OP_MINQUERY:
3329      c = *ecode++ - OP_STAR;      case OP_MINQUERYI:
3330        c = *ecode++ - ((op < OP_STARI)? OP_STAR : OP_STARI);
3331      minimize = (c & 1) != 0;      minimize = (c & 1) != 0;
3332      min = rep_min[c];                 /* Pick up values from tables; */      min = rep_min[c];                 /* Pick up values from tables; */
3333      max = rep_max[c];                 /* zero for max => infinity */      max = rep_max[c];                 /* zero for max => infinity */
3334      if (max == 0) max = INT_MAX;      if (max == 0) max = INT_MAX;
3335    
3336      /* 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. */  
3337    
3338      REPEATCHAR:      REPEATCHAR:
3339  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3340      if (utf8)      if (utf)
3341        {        {
3342        length = 1;        length = 1;
3343        charptr = ecode;        charptr = ecode;
3344        GETCHARLEN(fc, ecode, length);        GETCHARLEN(fc, ecode, length);
       if (min * length > md->end_subject - eptr) RRETURN(MATCH_NOMATCH);  
3345        ecode += length;        ecode += length;
3346    
3347        /* Handle multibyte character matching specially here. There is        /* Handle multibyte character matching specially here. There is
# Line 2205  for (;;) Line 3350  for (;;)
3350        if (length > 1)        if (length > 1)
3351          {          {
3352  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3353          unsigned int othercase;          pcre_uint32 othercase;
3354          if ((ims & PCRE_CASELESS) != 0 &&          if (op >= OP_STARI &&     /* Caseless */
3355              (othercase = _pcre_ucp_othercase(fc)) != NOTACHAR)              (othercase = UCD_OTHERCASE(fc)) != fc)
3356            oclength = _pcre_ord2utf8(othercase, occhars);            oclength = PRIV(ord2utf)(othercase, occhars);
3357          else oclength = 0;          else oclength = 0;
3358  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3359    
3360          for (i = 1; i <= min; i++)          for (i = 1; i <= min; i++)
3361            {            {
3362            if (memcmp(eptr, charptr, length) == 0) eptr += length;            if (eptr <= md->end_subject - length &&
3363                memcmp(eptr, charptr, IN_UCHARS(length)) == 0) eptr += length;
3364  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3365            /* Need braces because of following else */            else if (oclength > 0 &&
3366            else if (oclength == 0) { RRETURN(MATCH_NOMATCH); }                     eptr <= md->end_subject - oclength &&
3367                       memcmp(eptr, occhars, IN_UCHARS(oclength)) == 0) eptr += oclength;
3368    #endif  /* SUPPORT_UCP */
3369            else            else
3370