/[pcre]/code/trunk/pcre.c
ViewVC logotype

Diff of /code/trunk/pcre.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 11 by nigel, Sat Feb 24 21:38:17 2007 UTC revision 31 by nigel, Sat Feb 24 21:38:57 2007 UTC
# Line 9  the file Tech.Notes for some information Line 9  the file Tech.Notes for some information
9    
10  Written by: Philip Hazel <ph10@cam.ac.uk>  Written by: Philip Hazel <ph10@cam.ac.uk>
11    
12             Copyright (c) 1997 University of Cambridge             Copyright (c) 1997-1999 University of Cambridge
13    
14  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
15  Permission is granted to anyone to use this software for any purpose on any  Permission is granted to anyone to use this software for any purpose on any
# Line 25  restrictions: Line 25  restrictions:
25    
26  3. Altered versions must be plainly marked as such, and must not be  3. Altered versions must be plainly marked as such, and must not be
27     misrepresented as being the original software.     misrepresented as being the original software.
28    
29    4. If PCRE is embedded in any software that is released under the GNU
30       General Purpose Licence (GPL), then the terms of that licence shall
31       supersede any condition above with which it is incompatible.
32  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
33  */  */
34    
# Line 33  restrictions: Line 37  restrictions:
37    
38  /* #define DEBUG */  /* #define DEBUG */
39    
40  /* Use a macro for debugging printing, 'cause that eliminates the the use  /* Use a macro for debugging printing, 'cause that eliminates the use of #ifdef
41  of #ifdef inline, and there are *still* stupid compilers about that don't like  inline, and there are *still* stupid compilers about that don't like indented
42  indented pre-processor statements. I suppose it's only been 10 years... */  pre-processor statements. I suppose it's only been 10 years... */
43    
44  #ifdef DEBUG  #ifdef DEBUG
45  #define DPRINTF(p) printf p  #define DPRINTF(p) printf p
# Line 49  the external pcre header. */ Line 53  the external pcre header. */
53  #include "internal.h"  #include "internal.h"
54    
55    
56    /* Allow compilation as C++ source code, should anybody want to do that. */
57    
58    #ifdef __cplusplus
59    #define class pcre_class
60    #endif
61    
62    
63    /* Number of items on the nested bracket stacks at compile time. This should
64    not be set greater than 200. */
65    
66    #define BRASTACK_SIZE 200
67    
68    
69  /* Min and max values for the common repeats; for the maxima, 0 => infinity */  /* Min and max values for the common repeats; for the maxima, 0 => infinity */
70    
71  static char rep_min[] = { 0, 0, 1, 1, 0, 0 };  static const char rep_min[] = { 0, 0, 1, 1, 0, 0 };
72  static char rep_max[] = { 0, 0, 0, 0, 1, 1 };  static const char rep_max[] = { 0, 0, 0, 0, 1, 1 };
73    
74  /* Text forms of OP_ values and things, for debugging */  /* Text forms of OP_ values and things, for debugging (not all used) */
75    
76  #ifdef DEBUG  #ifdef DEBUG
77  static const char *OP_names[] = {  static const char *OP_names[] = {
78    "End", "\\A", "\\B", "\\b", "\\D", "\\d",    "End", "\\A", "\\B", "\\b", "\\D", "\\d",
79    "\\S", "\\s", "\\W", "\\w", "Cut", "\\Z", "^", "$", "Any", "chars",    "\\S", "\\s", "\\W", "\\w", "\\Z", "\\z",
80    "not",    "Opt", "^", "$", "Any", "chars", "not",
81    "*", "*?", "+", "+?", "?", "??", "{", "{", "{",    "*", "*?", "+", "+?", "?", "??", "{", "{", "{",
82    "*", "*?", "+", "+?", "?", "??", "{", "{", "{",    "*", "*?", "+", "+?", "?", "??", "{", "{", "{",
83    "*", "*?", "+", "+?", "?", "??", "{", "{", "{",    "*", "*?", "+", "+?", "?", "??", "{", "{", "{",
84    "*", "*?", "+", "+?", "?", "??", "{", "{",    "*", "*?", "+", "+?", "?", "??", "{", "{",
85    "class", "Ref",    "class", "Ref",
86    "Alt", "Ket", "KetRmax", "KetRmin", "Assert", "Assert not", "Once",    "Alt", "Ket", "KetRmax", "KetRmin", "Assert", "Assert not",
87      "AssertB", "AssertB not", "Reverse", "Once", "Cond", "Cref",
88    "Brazero", "Braminzero", "Bra"    "Brazero", "Braminzero", "Bra"
89  };  };
90  #endif  #endif
# Line 76  are simple data values; negative values Line 94  are simple data values; negative values
94  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
95  is invalid. */  is invalid. */
96    
97  static short int escapes[] = {  static const short int escapes[] = {
98      0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */      0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */
99      0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */      0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */
100    '@', -ESC_A, -ESC_B,      0, -ESC_D,      0,      0,      0,   /* @ - G */    '@', -ESC_A, -ESC_B,      0, -ESC_D,      0,      0,      0,   /* @ - G */
# Line 86  static short int escapes[] = { Line 104  static short int escapes[] = {
104    '`',      7, -ESC_b,      0, -ESC_d,     27,   '\f',      0,   /* ` - g */    '`',      7, -ESC_b,      0, -ESC_d,     27,   '\f',      0,   /* ` - g */
105      0,      0,      0,      0,      0,      0,   '\n',      0,   /* h - o */      0,      0,      0,      0,      0,      0,   '\n',      0,   /* h - o */
106      0,      0,   '\r', -ESC_s,   '\t',      0,      0, -ESC_w,   /* p - w */      0,      0,   '\r', -ESC_s,   '\t',      0,      0, -ESC_w,   /* p - w */
107      0,      0,      0                                            /* x - z */      0,      0, -ESC_z                                            /* x - z */
108  };  };
109    
110  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
111    
112  static BOOL compile_regex(int, int *, uschar **, const uschar **, const char **);  static BOOL
113      compile_regex(int, int, int *, uschar **, const uschar **, const char **,
114  /* Structure for passing "static" information around between the functions      BOOL, int, compile_data *);
 doing the matching, so that they are thread-safe. */  
   
 typedef struct match_data {  
   int    errorcode;             /* As it says */  
   int   *offset_vector;         /* Offset vector */  
   int    offset_end;            /* One past the end */  
   BOOL   offset_overflow;       /* Set if too many extractions */  
   BOOL   caseless;              /* Case-independent flag */  
   BOOL   runtime_caseless;      /* Caseless forced at run time */  
   BOOL   multiline;             /* Multiline flag */  
   BOOL   notbol;                /* NOTBOL flag */  
   BOOL   noteol;                /* NOTEOL flag */  
   BOOL   dotall;                /* Dot matches any char */  
   BOOL   endonly;               /* Dollar not before final \n */  
   const uschar *start_subject;  /* Start of the subject string */  
   const uschar *end_subject;    /* End of the subject string */  
   jmp_buf fail_env;             /* Environment for longjump() break out */  
   const uschar *end_match_ptr;  /* Subject position at end match */  
   int     end_offset_top;       /* Highwater mark at end of match */  
 } match_data;  
115    
116    
117    
# Line 133  void  (*pcre_free)(void *) = free; Line 131  void  (*pcre_free)(void *) = free;
131    
132    
133  /*************************************************  /*************************************************
134    *             Default character tables           *
135    *************************************************/
136    
137    /* A default set of character tables is included in the PCRE binary. Its source
138    is built by the maketables auxiliary program, which uses the default C ctypes
139    functions, and put in the file chartables.c. These tables are used by PCRE
140    whenever the caller of pcre_compile() does not provide an alternate set of
141    tables. */
142    
143    #include "chartables.c"
144    
145    
146    
147    /*************************************************
148  *          Return version string                 *  *          Return version string                 *
149  *************************************************/  *************************************************/
150    
# Line 210  while (length-- > 0) Line 222  while (length-- > 0)
222    
223    
224  /*************************************************  /*************************************************
 *         Check subpattern for empty operand     *  
 *************************************************/  
   
 /* This function checks a bracketed subpattern to see if any of the paths  
 through it could match an empty string. This is used to diagnose an error if  
 such a subpattern is followed by a quantifier with an unlimited upper bound.  
   
 Argument:  
   code      points to the opening bracket  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 could_be_empty(uschar *code)  
 {  
 do {  
   uschar *cc = code + 3;  
   
   /* Scan along the opcodes for this branch; as soon as we find something  
   that matches a non-empty string, break out and advance to test the next  
   branch. If we get to the end of the branch, return TRUE for the whole  
   sub-expression. */  
   
   for (;;)  
     {  
     /* Test an embedded subpattern; if it could not be empty, break the  
     loop. Otherwise carry on in the branch. */  
   
     if ((int)(*cc) >= OP_BRA || (int)(*cc) == OP_ONCE)  
       {  
       if (!could_be_empty(cc)) break;  
       do cc += (cc[1] << 8) + cc[2]; while (*cc == OP_ALT);  
       cc += 3;  
       }  
   
     else switch (*cc)  
       {  
       /* Reached end of a branch: the subpattern may match the empty string */  
   
       case OP_ALT:  
       case OP_KET:  
       case OP_KETRMAX:  
       case OP_KETRMIN:  
       return TRUE;  
   
       /* Skip over assertive subpatterns */  
   
       case OP_ASSERT:  
       case OP_ASSERT_NOT:  
       do cc += (cc[1] << 8) + cc[2]; while (*cc == OP_ALT);  
       cc += 3;  
       break;  
   
       /* Skip over things that don't match chars */  
   
       case OP_SOD:  
       case OP_EOD:  
       case OP_CIRC:  
       case OP_DOLL:  
       case OP_BRAZERO:  
       case OP_BRAMINZERO:  
       case OP_NOT_WORD_BOUNDARY:  
       case OP_WORD_BOUNDARY:  
       cc++;  
       break;  
   
       /* Skip over simple repeats with zero lower bound */  
   
       case OP_STAR:  
       case OP_MINSTAR:  
       case OP_QUERY:  
       case OP_MINQUERY:  
       case OP_NOTSTAR:  
       case OP_NOTMINSTAR:  
       case OP_NOTQUERY:  
       case OP_NOTMINQUERY:  
       case OP_TYPESTAR:  
       case OP_TYPEMINSTAR:  
       case OP_TYPEQUERY:  
       case OP_TYPEMINQUERY:  
       cc += 2;  
       break;  
   
       /* Skip over UPTOs (lower bound is zero) */  
   
       case OP_UPTO:  
       case OP_MINUPTO:  
       case OP_TYPEUPTO:  
       case OP_TYPEMINUPTO:  
       cc += 4;  
       break;  
   
       /* Check a class or a back reference for a zero minimum */  
   
       case OP_CLASS:  
       case OP_REF:  
       cc += (*cc == OP_REF)? 2 : 33;  
   
       switch (*cc)  
         {  
         case OP_CRSTAR:  
         case OP_CRMINSTAR:  
         case OP_CRQUERY:  
         case OP_CRMINQUERY:  
         cc++;  
         break;  
   
         case OP_CRRANGE:  
         case OP_CRMINRANGE:  
         if ((cc[1] << 8) + cc[2] != 0) goto NEXT_BRANCH;  
         cc += 3;  
         break;  
   
         default:  
         goto NEXT_BRANCH;  
         }  
       break;  
   
       /* Anything else matches at least one character */  
   
       default:  
       goto NEXT_BRANCH;  
       }  
     }  
   
   NEXT_BRANCH:  
   code += (code[1] << 8) + code[2];  
   }  
 while (*code == OP_ALT);  
   
 /* No branches match the empty string */  
   
 return FALSE;  
 }  
   
   
   
 /*************************************************  
225  *            Handle escapes                      *  *            Handle escapes                      *
226  *************************************************/  *************************************************/
227    
# Line 364  Arguments: Line 237  Arguments:
237    bracount   number of previous extracting brackets    bracount   number of previous extracting brackets
238    options    the options bits    options    the options bits
239    isclass    TRUE if inside a character class    isclass    TRUE if inside a character class
240      cd         pointer to char tables block
241    
242  Returns:     zero or positive => a data character  Returns:     zero or positive => a data character
243               negative => a special escape sequence               negative => a special escape sequence
# Line 372  Returns:     zero or positive => a data Line 246  Returns:     zero or positive => a data
246    
247  static int  static int
248  check_escape(const uschar **ptrptr, const char **errorptr, int bracount,  check_escape(const uschar **ptrptr, const char **errorptr, int bracount,
249    int options, BOOL isclass)    int options, BOOL isclass, compile_data *cd)
250  {  {
251  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
252  int c = *(++ptr) & 255;   /* Ensure > 0 on signed-char systems */  int c = *(++ptr) & 255;   /* Ensure > 0 on signed-char systems */
# Line 415  else Line 289  else
289        {        {
290        oldptr = ptr;        oldptr = ptr;
291        c -= '0';        c -= '0';
292        while ((pcre_ctypes[ptr[1]] & ctype_digit) != 0)        while ((cd->ctypes[ptr[1]] & ctype_digit) != 0)
293          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
294        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
295          {          {
# Line 441  else Line 315  else
315    
316      case '0':      case '0':
317      c -= '0';      c -= '0';
318      while(i++ < 2 && (pcre_ctypes[ptr[1]] & ctype_digit) != 0 &&      while(i++ < 2 && (cd->ctypes[ptr[1]] & ctype_digit) != 0 &&
319        ptr[1] != '8' && ptr[1] != '9')        ptr[1] != '8' && ptr[1] != '9')
320          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - '0';
321      break;      break;
# Line 450  else Line 324  else
324    
325      case 'x':      case 'x':
326      c = 0;      c = 0;
327      while (i++ < 2 && (pcre_ctypes[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (cd->ctypes[ptr[1]] & ctype_xdigit) != 0)
328        {        {
329        ptr++;        ptr++;
330        c = c * 16 + pcre_lcc[*ptr] -        c = c * 16 + cd->lcc[*ptr] -
331          (((pcre_ctypes[*ptr] & ctype_digit) != 0)? '0' : 'W');          (((cd->ctypes[*ptr] & ctype_digit) != 0)? '0' : 'W');
332        }        }
333      break;      break;
334    
# Line 468  else Line 342  else
342    
343      /* A letter is upper-cased; then the 0x40 bit is flipped */      /* A letter is upper-cased; then the 0x40 bit is flipped */
344    
345      if (c >= 'a' && c <= 'z') c = pcre_fcc[c];      if (c >= 'a' && c <= 'z') c = cd->fcc[c];
346      c ^= 0x40;      c ^= 0x40;
347      break;      break;
348    
349      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
350      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,
351      for Perl compatibility, it is a literal. */      for Perl compatibility, it is a literal. This code looks a bit odd, but
352        there used to be some cases other than the default, and there may be again
353        in future, so I haven't "optimized" it. */
354    
355      default:      default:
356      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
357        {        {
       case 'X':  
       c = -ESC_X;      /* This could be a lookup if it ever got into Perl */  
       break;  
   
358        default:        default:
359        *errorptr = ERR3;        *errorptr = ERR3;
360        break;        break;
# Line 508  where the ddds are digits. Line 380  where the ddds are digits.
380    
381  Arguments:  Arguments:
382    p         pointer to the first char after '{'    p         pointer to the first char after '{'
383      cd        pointer to char tables block
384    
385  Returns:    TRUE or FALSE  Returns:    TRUE or FALSE
386  */  */
387    
388  static BOOL  static BOOL
389  is_counted_repeat(const uschar *p)  is_counted_repeat(const uschar *p, compile_data *cd)
390  {  {
391  if ((pcre_ctypes[*p++] & ctype_digit) == 0) return FALSE;  if ((cd->ctypes[*p++] & ctype_digit) == 0) return FALSE;
392  while ((pcre_ctypes[*p] & ctype_digit) != 0) p++;  while ((cd->ctypes[*p] & ctype_digit) != 0) p++;
393  if (*p == '}') return TRUE;  if (*p == '}') return TRUE;
394    
395  if (*p++ != ',') return FALSE;  if (*p++ != ',') return FALSE;
396  if (*p == '}') return TRUE;  if (*p == '}') return TRUE;
397    
398  if ((pcre_ctypes[*p++] & ctype_digit) == 0) return FALSE;  if ((cd->ctypes[*p++] & ctype_digit) == 0) return FALSE;
399  while ((pcre_ctypes[*p] & ctype_digit) != 0) p++;  while ((cd->ctypes[*p] & ctype_digit) != 0) p++;
400  return (*p == '}');  return (*p == '}');
401  }  }
402    
# Line 543  Arguments: Line 416  Arguments:
416    maxp       pointer to int for max    maxp       pointer to int for max
417               returned as -1 if no max               returned as -1 if no max
418    errorptr   points to pointer to error message    errorptr   points to pointer to error message
419      cd         pointer to character tables clock
420    
421  Returns:     pointer to '}' on success;  Returns:     pointer to '}' on success;
422               current ptr on error, with errorptr set               current ptr on error, with errorptr set
423  */  */
424    
425  static const uschar *  static const uschar *
426  read_repeat_counts(const uschar *p, int *minp, int *maxp, const char **errorptr)  read_repeat_counts(const uschar *p, int *minp, int *maxp,
427      const char **errorptr, compile_data *cd)
428  {  {
429  int min = 0;  int min = 0;
430  int max = -1;  int max = -1;
431    
432  while ((pcre_ctypes[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((cd->ctypes[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';
433    
434  if (*p == '}') max = min; else  if (*p == '}') max = min; else
435    {    {
436    if (*(++p) != '}')    if (*(++p) != '}')
437      {      {
438      max = 0;      max = 0;
439      while((pcre_ctypes[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((cd->ctypes[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';
440      if (max < min)      if (max < min)
441        {        {
442        *errorptr = ERR4;        *errorptr = ERR4;
# Line 586  return p; Line 461  return p;
461    
462    
463  /*************************************************  /*************************************************
464    *        Find the fixed length of a pattern      *
465    *************************************************/
466    
467    /* Scan a pattern and compute the fixed length of subject that will match it,
468    if the length is fixed. This is needed for dealing with backward assertions.
469    
470    Arguments:
471      code     points to the start of the pattern (the bracket)
472    
473    Returns:   the fixed length, or -1 if there is no fixed length
474    */
475    
476    static int
477    find_fixedlength(uschar *code)
478    {
479    int length = -1;
480    
481    register int branchlength = 0;
482    register uschar *cc = code + 3;
483    
484    /* Scan along the opcodes for this branch. If we get to the end of the
485    branch, check the length against that of the other branches. */
486    
487    for (;;)
488      {
489      int d;
490      register int op = *cc;
491      if (op >= OP_BRA) op = OP_BRA;
492    
493      switch (op)
494        {
495        case OP_BRA:
496        case OP_ONCE:
497        case OP_COND:
498        d = find_fixedlength(cc);
499        if (d < 0) return -1;
500        branchlength += d;
501        do cc += (cc[1] << 8) + cc[2]; while (*cc == OP_ALT);
502        cc += 3;
503        break;
504    
505        /* Reached end of a branch; if it's a ket it is the end of a nested
506        call. If it's ALT it is an alternation in a nested call. If it is
507        END it's the end of the outer call. All can be handled by the same code. */
508    
509        case OP_ALT:
510        case OP_KET:
511        case OP_KETRMAX:
512        case OP_KETRMIN:
513        case OP_END:
514        if (length < 0) length = branchlength;
515          else if (length != branchlength) return -1;
516        if (*cc != OP_ALT) return length;
517        cc += 3;
518        branchlength = 0;
519        break;
520    
521        /* Skip over assertive subpatterns */
522    
523        case OP_ASSERT:
524        case OP_ASSERT_NOT:
525        case OP_ASSERTBACK:
526        case OP_ASSERTBACK_NOT:
527        do cc += (cc[1] << 8) + cc[2]; while (*cc == OP_ALT);
528        cc += 3;
529        break;
530    
531        /* Skip over things that don't match chars */
532    
533        case OP_REVERSE:
534        cc++;
535    
536        case OP_CREF:
537        case OP_OPT:
538        cc++;
539        /* Fall through */
540    
541        case OP_SOD:
542        case OP_EOD:
543        case OP_EODN:
544        case OP_CIRC:
545        case OP_DOLL:
546        case OP_NOT_WORD_BOUNDARY:
547        case OP_WORD_BOUNDARY:
548        cc++;
549        break;
550    
551        /* Handle char strings */
552    
553        case OP_CHARS:
554        branchlength += *(++cc);
555        cc += *cc + 1;
556        break;
557    
558        /* Handle exact repetitions */
559    
560        case OP_EXACT:
561        case OP_TYPEEXACT:
562        branchlength += (cc[1] << 8) + cc[2];
563        cc += 4;
564        break;
565    
566        /* Handle single-char matchers */
567    
568        case OP_NOT_DIGIT:
569        case OP_DIGIT:
570        case OP_NOT_WHITESPACE:
571        case OP_WHITESPACE:
572        case OP_NOT_WORDCHAR:
573        case OP_WORDCHAR:
574        case OP_ANY:
575        branchlength++;
576        cc++;
577        break;
578    
579    
580        /* Check a class for variable quantification */
581    
582        case OP_CLASS:
583        cc += (*cc == OP_REF)? 2 : 33;
584    
585        switch (*cc)
586          {
587          case OP_CRSTAR:
588          case OP_CRMINSTAR:
589          case OP_CRQUERY:
590          case OP_CRMINQUERY:
591          return -1;
592    
593          case OP_CRRANGE:
594          case OP_CRMINRANGE:
595          if ((cc[1] << 8) + cc[2] != (cc[3] << 8) + cc[4]) return -1;
596          branchlength += (cc[1] << 8) + cc[2];
597          cc += 5;
598          break;
599    
600          default:
601          branchlength++;
602          }
603        break;
604    
605        /* Anything else is variable length */
606    
607        default:
608        return -1;
609        }
610      }
611    /* Control never gets here */
612    }
613    
614    
615    
616    
617    /*************************************************
618  *           Compile one branch                   *  *           Compile one branch                   *
619  *************************************************/  *************************************************/
620    
621  /* Scan the pattern, compiling it into the code vector.  /* Scan the pattern, compiling it into the code vector.
622    
623  Arguments:  Arguments:
624    options    the option bits    options      the option bits
625    bracket    points to number of brackets used    brackets     points to number of brackets used
626    code       points to the pointer to the current code point    code         points to the pointer to the current code point
627    ptrptr     points to the current pattern pointer    ptrptr       points to the current pattern pointer
628    errorptr   points to pointer to error message    errorptr     points to pointer to error message
629      optchanged   set to the value of the last OP_OPT item compiled
630      cd           contains pointers to tables
631    
632  Returns:     TRUE on success  Returns:       TRUE on success
633               FALSE, with *errorptr set on error                 FALSE, with *errorptr set on error
634  */  */
635    
636  static BOOL  static BOOL
637  compile_branch(int options, int *brackets, uschar **codeptr,  compile_branch(int options, int *brackets, uschar **codeptr,
638    const uschar **ptrptr, const char **errorptr)    const uschar **ptrptr, const char **errorptr, int *optchanged,
639      compile_data *cd)
640  {  {
641  int repeat_type, op_type;  int repeat_type, op_type;
642  int repeat_min, repeat_max;  int repeat_min, repeat_max;
643  int bravalue, length;  int bravalue, length;
644    int greedy_default, greedy_non_default;
645  register int c;  register int c;
646  register uschar *code = *codeptr;  register uschar *code = *codeptr;
647    uschar *tempcode;
648  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
649  const uschar *oldptr;  const uschar *tempptr;
650  uschar *previous = NULL;  uschar *previous = NULL;
651  uschar class[32];  uschar class[32];
652    
653    /* Set up the default and non-default settings for greediness */
654    
655    greedy_default = ((options & PCRE_UNGREEDY) != 0);
656    greedy_non_default = greedy_default ^ 1;
657    
658  /* Switch on next character until the end of the branch */  /* Switch on next character until the end of the branch */
659    
660  for (;; ptr++)  for (;; ptr++)
661    {    {
662    BOOL negate_class;    BOOL negate_class;
663    int  class_charcount;    int class_charcount;
664    int  class_lastchar;    int class_lastchar;
665      int newoptions;
666      int condref;
667    
668    c = *ptr;    c = *ptr;
669    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
670      {      {
671      if ((pcre_ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
672      if (c == '#')      if (c == '#')
673        {        {
674        while ((c = *(++ptr)) != 0 && c != '\n');        while ((c = *(++ptr)) != 0 && c != '\n');
# Line 672  for (;; ptr++) Line 713  for (;; ptr++)
713      previous = code;      previous = code;
714      *code++ = OP_CLASS;      *code++ = OP_CLASS;
715    
716      /* If the first character is '^', set the negation flag */      /* If the first character is '^', set the negation flag and skip it. */
717    
718      if ((c = *(++ptr)) == '^')      if ((c = *(++ptr)) == '^')
719        {        {
# Line 715  for (;; ptr++) Line 756  for (;; ptr++)
756    
757        if (c == '\\')        if (c == '\\')
758          {          {
759          c = check_escape(&ptr, errorptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorptr, *brackets, options, TRUE, cd);
760          if (-c == ESC_b) c = '\b';          if (-c == ESC_b) c = '\b';
761          else if (c < 0)          else if (c < 0)
762            {            {
763              register const uschar *cbits = cd->cbits;
764            class_charcount = 10;            class_charcount = 10;
765            switch (-c)            switch (-c)
766              {              {
767              case ESC_d:              case ESC_d:
768              for (c = 0; c < 32; c++) class[c] |= pcre_cbits[c+cbit_digit];              for (c = 0; c < 32; c++) class[c] |= cbits[c+cbit_digit];
769              continue;              continue;
770    
771              case ESC_D:              case ESC_D:
772              for (c = 0; c < 32; c++) class[c] |= ~pcre_cbits[c+cbit_digit];              for (c = 0; c < 32; c++) class[c] |= ~cbits[c+cbit_digit];
773              continue;              continue;
774    
775              case ESC_w:              case ESC_w:
776              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
777                class[c] |= (pcre_cbits[c] | pcre_cbits[c+cbit_word]);                class[c] |= (cbits[c+cbit_digit] | cbits[c+cbit_word]);
778              continue;              continue;
779    
780              case ESC_W:              case ESC_W:
781              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
782                class[c] |= ~(pcre_cbits[c] | pcre_cbits[c+cbit_word]);                class[c] |= ~(cbits[c+cbit_digit] | cbits[c+cbit_word]);
783              continue;              continue;
784    
785              case ESC_s:              case ESC_s:
786              for (c = 0; c < 32; c++) class[c] |= pcre_cbits[c+cbit_space];              for (c = 0; c < 32; c++) class[c] |= cbits[c+cbit_space];
787              continue;              continue;
788    
789              case ESC_S:              case ESC_S:
790              for (c = 0; c < 32; c++) class[c] |= ~pcre_cbits[c+cbit_space];              for (c = 0; c < 32; c++) class[c] |= ~cbits[c+cbit_space];
791              continue;              continue;
792    
793              default:              default:
# Line 777  for (;; ptr++) Line 819  for (;; ptr++)
819    
820          if (d == '\\')          if (d == '\\')
821            {            {
822            d = check_escape(&ptr, errorptr, *brackets, options, TRUE);            d = check_escape(&ptr, errorptr, *brackets, options, TRUE, cd);
823            if (d < 0)            if (d < 0)
824              {              {
825              if (d == -ESC_b) d = '\b'; else              if (d == -ESC_b) d = '\b'; else
# Line 799  for (;; ptr++) Line 841  for (;; ptr++)
841            class[c/8] |= (1 << (c&7));            class[c/8] |= (1 << (c&7));
842            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
843              {              {
844              int uc = pcre_fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
845              class[uc/8] |= (1 << (uc&7));              class[uc/8] |= (1 << (uc&7));
846              }              }
847            class_charcount++;                /* in case a one-char range */            class_charcount++;                /* in case a one-char range */
# Line 814  for (;; ptr++) Line 856  for (;; ptr++)
856        class [c/8] |= (1 << (c&7));        class [c/8] |= (1 << (c&7));
857        if ((options & PCRE_CASELESS) != 0)        if ((options & PCRE_CASELESS) != 0)
858          {          {
859          c = pcre_fcc[c];   /* flip case */          c = cd->fcc[c];   /* flip case */
860          class[c/8] |= (1 << (c&7));          class[c/8] |= (1 << (c&7));
861          }          }
862        class_charcount++;        class_charcount++;
# Line 861  for (;; ptr++) Line 903  for (;; ptr++)
903      /* Various kinds of repeat */      /* Various kinds of repeat */
904    
905      case '{':      case '{':
906      if (!is_counted_repeat(ptr+1)) goto NORMAL_CHAR;      if (!is_counted_repeat(ptr+1, cd)) goto NORMAL_CHAR;
907      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorptr, cd);
908      if (*errorptr != NULL) goto FAILED;      if (*errorptr != NULL) goto FAILED;
909      goto REPEAT;      goto REPEAT;
910    
# Line 887  for (;; ptr++) Line 929  for (;; ptr++)
929        goto FAILED;        goto FAILED;
930        }        }
931    
932      /* If the next character is '?' this is a minimizing repeat. Advance to the      /* If the next character is '?' this is a minimizing repeat, by default,
933        but if PCRE_UNGREEDY is set, it works the other way round. Advance to the
934      next character. */      next character. */
935    
936      if (ptr[1] == '?') { repeat_type = 1; ptr++; } else repeat_type = 0;      if (ptr[1] == '?')
937          { repeat_type = greedy_non_default; ptr++; }
938        else repeat_type = greedy_default;
939    
940      /* If the maximum is zero then the minimum must also be zero; Perl allows      /* If the maximum is zero then the minimum must also be zero; Perl allows
941      this case, so we do too - by simply omitting the item altogether. */      this case, so we do too - by simply omitting the item altogether. */
# Line 935  for (;; ptr++) Line 980  for (;; ptr++)
980      create a suitable repeat item. The code is shared with single-character      create a suitable repeat item. The code is shared with single-character
981      repeats by adding a suitable offset into repeat_type. */      repeats by adding a suitable offset into repeat_type. */
982    
983      else if ((int)*previous < OP_EOD || *previous == OP_ANY)      else if ((int)*previous < OP_EODN || *previous == OP_ANY)
984        {        {
985        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
986        c = *previous;        c = *previous;
# Line 979  for (;; ptr++) Line 1024  for (;; ptr++)
1024          /* If the mininum is 1 and the previous item was a character string,          /* If the mininum is 1 and the previous item was a character string,
1025          we either have to put back the item that got cancelled if the string          we either have to put back the item that got cancelled if the string
1026          length was 1, or add the character back onto the end of a longer          length was 1, or add the character back onto the end of a longer
1027          string. For a character type nothing need be done; it will just get put          string. For a character type nothing need be done; it will just get
1028          back naturally. */          put back naturally. Note that the final character is always going to
1029            get added below. */
1030    
1031          else if (*previous == OP_CHARS)          else if (*previous == OP_CHARS)
1032            {            {
1033            if (code == previous) code += 2; else previous[1]++;            if (code == previous) code += 2; else previous[1]++;
1034            }            }
1035    
1036            /*  For a single negated character we also have to put back the
1037            item that got cancelled. */
1038    
1039            else if (*previous == OP_NOT) code++;
1040    
1041          /* If the maximum is unlimited, insert an OP_STAR. */          /* If the maximum is unlimited, insert an OP_STAR. */
1042    
1043          if (repeat_max < 0)          if (repeat_max < 0)
# Line 1035  for (;; ptr++) Line 1086  for (;; ptr++)
1086        }        }
1087    
1088      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
1089      cases. If the maximum repeat count is unlimited, check that the bracket      cases. */
     group cannot match the empty string, and diagnose an error if it can. */  
1090    
1091      else if ((int)*previous >= OP_BRA)      else if ((int)*previous >= OP_BRA || (int)*previous == OP_ONCE ||
1092                 (int)*previous == OP_COND)
1093        {        {
1094        int i;        register int i;
1095          int ketoffset = 0;
1096        int len = code - previous;        int len = code - previous;
1097          uschar *bralink = NULL;
1098    
1099          /* If the maximum repeat count is unlimited, find the end of the bracket
1100          by scanning through from the start, and compute the offset back to it
1101          from the current code pointer. There may be an OP_OPT setting following
1102          the final KET, so we can't find the end just by going back from the code
1103          pointer. */
1104    
1105          if (repeat_max == -1)
1106            {
1107            register uschar *ket = previous;
1108            do ket += (ket[1] << 8) + ket[2]; while (*ket != OP_KET);
1109            ketoffset = code - ket;
1110            }
1111    
1112          /* The case of a zero minimum is special because of the need to stick
1113          OP_BRAZERO in front of it, and because the group appears once in the
1114          data, whereas in other cases it appears the minimum number of times. For
1115          this reason, it is simplest to treat this case separately, as otherwise
1116          the code gets far too mess. There are several special subcases when the
1117          minimum is zero. */
1118    
1119        if (repeat_max == -1 && could_be_empty(previous))        if (repeat_min == 0)
1120          {          {
1121          *errorptr = ERR10;          /* If the maximum is also zero, we just omit the group from the output
1122          goto FAILED;          altogether. */
1123    
1124            if (repeat_max == 0)
1125              {
1126              code = previous;
1127              previous = NULL;
1128              break;
1129              }
1130    
1131            /* If the maximum is 1 or unlimited, we just have to stick in the
1132            BRAZERO and do no more at this point. */
1133    
1134            if (repeat_max <= 1)
1135              {
1136              memmove(previous+1, previous, len);
1137              code++;
1138              *previous++ = OP_BRAZERO + repeat_type;
1139              }
1140    
1141            /* If the maximum is greater than 1 and limited, we have to replicate
1142            in a nested fashion, sticking OP_BRAZERO before each set of brackets.
1143            The first one has to be handled carefully because it's the original
1144            copy, which has to be moved up. The remainder can be handled by code
1145            that is common with the non-zero minimum case below. We just have to
1146            adjust the value or repeat_max, since one less copy is required. */
1147    
1148            else
1149              {
1150              int offset;
1151              memmove(previous+4, previous, len);
1152              code += 4;
1153              *previous++ = OP_BRAZERO + repeat_type;
1154              *previous++ = OP_BRA;
1155    
1156              /* We chain together the bracket offset fields that have to be
1157              filled in later when the ends of the brackets are reached. */
1158    
1159              offset = (bralink == NULL)? 0 : previous - bralink;
1160              bralink = previous;
1161              *previous++ = offset >> 8;
1162              *previous++ = offset & 255;
1163              }
1164    
1165            repeat_max--;
1166            }
1167    
1168          /* If the minimum is greater than zero, replicate the group as many
1169          times as necessary, and adjust the maximum to the number of subsequent
1170          copies that we need. */
1171    
1172          else
1173            {
1174            for (i = 1; i < repeat_min; i++)
1175              {
1176              memcpy(code, previous, len);
1177              code += len;
1178              }
1179            if (repeat_max > 0) repeat_max -= repeat_min;
1180            }
1181    
1182          /* This code is common to both the zero and non-zero minimum cases. If
1183          the maximum is limited, it replicates the group in a nested fashion,
1184          remembering the bracket starts on a stack. In the case of a zero minimum,
1185          the first one was set up above. In all cases the repeat_max now specifies
1186          the number of additional copies needed. */
1187    
1188          if (repeat_max >= 0)
1189            {
1190            for (i = repeat_max - 1; i >= 0; i--)
1191              {
1192              *code++ = OP_BRAZERO + repeat_type;
1193    
1194              /* All but the final copy start a new nesting, maintaining the
1195              chain of brackets outstanding. */
1196    
1197              if (i != 0)
1198                {
1199                int offset;
1200                *code++ = OP_BRA;
1201                offset = (bralink == NULL)? 0 : code - bralink;
1202                bralink = code;
1203                *code++ = offset >> 8;
1204                *code++ = offset & 255;
1205                }
1206    
1207              memcpy(code, previous, len);
1208              code += len;
1209              }
1210    
1211            /* Now chain through the pending brackets, and fill in their length
1212            fields (which are holding the chain links pro tem). */
1213    
1214            while (bralink != NULL)
1215              {
1216              int oldlinkoffset;
1217              int offset = code - bralink + 1;
1218              uschar *bra = code - offset;
1219              oldlinkoffset = (bra[1] << 8) + bra[2];
1220              bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
1221              *code++ = OP_KET;
1222              *code++ = bra[1] = offset >> 8;
1223              *code++ = bra[2] = (offset & 255);
1224              }
1225          }          }
1226    
1227          /* If the maximum is unlimited, set a repeater in the final copy. We
1228          can't just offset backwards from the current code point, because we
1229          don't know if there's been an options resetting after the ket. The
1230          correct offset was computed above. */
1231    
1232          else code[-ketoffset] = OP_KETRMAX + repeat_type;
1233    
1234    
1235    #ifdef NEVER
1236        /* If the minimum is greater than zero, and the maximum is unlimited or        /* If the minimum is greater than zero, and the maximum is unlimited or
1237        equal to the minimum, the first copy remains where it is, and is        equal to the minimum, the first copy remains where it is, and is
1238        replicated up to the minimum number of times. This case includes the +        replicated up to the minimum number of times. This case includes the +
# Line 1090  for (;; ptr++) Line 1274  for (;; ptr++)
1274            }            }
1275          }          }
1276    
1277        /* If the maximum is unlimited, set a repeater in the final copy. */        /* If the maximum is unlimited, set a repeater in the final copy. We
1278          can't just offset backwards from the current code point, because we
1279          don't know if there's been an options resetting after the ket. The
1280          correct offset was computed above. */
1281    
1282          if (repeat_max == -1) code[-ketoffset] = OP_KETRMAX + repeat_type;
1283    #endif
1284    
1285    
       if (repeat_max == -1) code[-3] = OP_KETRMAX + repeat_type;  
1286        }        }
1287    
1288      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
# Line 1109  for (;; ptr++) Line 1299  for (;; ptr++)
1299      break;      break;
1300    
1301    
1302      /* Start of nested bracket sub-expression, or comment or lookahead.      /* Start of nested bracket sub-expression, or comment or lookahead or
1303      First deal with special things that can come after a bracket; all are      lookbehind or option setting or condition. First deal with special things
1304      introduced by ?, and the appearance of any of them means that this is not a      that can come after a bracket; all are introduced by ?, and the appearance
1305      referencing group. They were checked for validity in the first pass over      of any of them means that this is not a referencing group. They were
1306      the string, so we don't have to check for syntax errors here.  */      checked for validity in the first pass over the string, so we don't have to
1307        check for syntax errors here.  */
1308    
1309      case '(':      case '(':
1310      previous = code;              /* Only real brackets can be repeated */      newoptions = options;
1311        condref = -1;
1312    
1313      if (*(++ptr) == '?')      if (*(++ptr) == '?')
1314        {        {
1315        bravalue = OP_BRA;        int set, unset;
1316          int *optset;
1317    
1318        switch (*(++ptr))        switch (*(++ptr))
1319          {          {
1320          case '#':          case '#':                 /* Comment; skip to ket */
         case 'i':  
         case 'm':  
         case 's':  
         case 'x':  
1321          ptr++;          ptr++;
1322          while (*ptr != ')') ptr++;          while (*ptr != ')') ptr++;
         previous = NULL;  
1323          continue;          continue;
1324    
1325          case ':':                 /* Non-extracting bracket */          case ':':                 /* Non-extracting bracket */
1326            bravalue = OP_BRA;
1327          ptr++;          ptr++;
1328          break;          break;
1329    
1330          case '=':                 /* Assertions can't be repeated */          case '(':
1331            bravalue = OP_COND;       /* Conditional group */
1332            if ((cd->ctypes[*(++ptr)] & ctype_digit) != 0)
1333              {
1334              condref = *ptr - '0';
1335              while (*(++ptr) != ')') condref = condref*10 + *ptr - '0';
1336              ptr++;
1337              }
1338            else ptr--;
1339            break;
1340    
1341            case '=':                 /* Positive lookahead */
1342          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
1343          ptr++;          ptr++;
         previous = NULL;  
1344          break;          break;
1345    
1346          case '!':          case '!':                 /* Negative lookahead */
1347          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
1348          ptr++;          ptr++;
         previous = NULL;  
1349          break;          break;
1350    
1351          case '>':                         /* "Match once" brackets */          case '<':                 /* Lookbehinds */
1352          if ((options & PCRE_EXTRA) != 0)  /* Not yet standard */          switch (*(++ptr))
1353            {            {
1354            bravalue = OP_ONCE;            case '=':               /* Positive lookbehind */
1355              bravalue = OP_ASSERTBACK;
1356              ptr++;
1357              break;
1358    
1359              case '!':               /* Negative lookbehind */
1360              bravalue = OP_ASSERTBACK_NOT;
1361            ptr++;            ptr++;
           previous = NULL;  
1362            break;            break;
1363    
1364              default:                /* Syntax error */
1365              *errorptr = ERR24;
1366              goto FAILED;
1367            }            }
1368          /* Else fall through */          break;
1369    
1370          default:          case '>':                 /* One-time brackets */
1371          *errorptr = ERR12;          bravalue = OP_ONCE;
1372          goto FAILED;          ptr++;
1373            break;
1374    
1375            default:                  /* Option setting */
1376            set = unset = 0;
1377            optset = &set;
1378    
1379            while (*ptr != ')' && *ptr != ':')
1380              {
1381              switch (*ptr++)
1382                {
1383                case '-': optset = &unset; break;
1384    
1385                case 'i': *optset |= PCRE_CASELESS; break;
1386                case 'm': *optset |= PCRE_MULTILINE; break;
1387                case 's': *optset |= PCRE_DOTALL; break;
1388                case 'x': *optset |= PCRE_EXTENDED; break;
1389                case 'U': *optset |= PCRE_UNGREEDY; break;
1390                case 'X': *optset |= PCRE_EXTRA; break;
1391    
1392                default:
1393                *errorptr = ERR12;
1394                goto FAILED;
1395                }
1396              }
1397    
1398            /* Set up the changed option bits, but don't change anything yet. */
1399    
1400            newoptions = (options | set) & (~unset);
1401    
1402            /* If the options ended with ')' this is not the start of a nested
1403            group with option changes, so the options change at this level. At top
1404            level there is nothing else to be done (the options will in fact have
1405            been set from the start of compiling as a result of the first pass) but
1406            at an inner level we must compile code to change the ims options if
1407            necessary, and pass the new setting back so that it can be put at the
1408            start of any following branches, and when this group ends, a resetting
1409            item can be compiled. */
1410    
1411            if (*ptr == ')')
1412              {
1413              if ((options & PCRE_INGROUP) != 0 &&
1414                  (options & PCRE_IMS) != (newoptions & PCRE_IMS))
1415                {
1416                *code++ = OP_OPT;
1417                *code++ = *optchanged = newoptions & PCRE_IMS;
1418                }
1419              options = newoptions;  /* Change options at this level */
1420              previous = NULL;       /* This item can't be repeated */
1421              continue;              /* It is complete */
1422              }
1423    
1424            /* If the options ended with ':' we are heading into a nested group
1425            with possible change of options. Such groups are non-capturing and are
1426            not assertions of any kind. All we need to do is skip over the ':';
1427            the newoptions value is handled below. */
1428    
1429            bravalue = OP_BRA;
1430            ptr++;
1431          }          }
1432        }        }
1433    
1434      /* Else we have a referencing group */      /* Else we have a referencing group; adjust the opcode. */
1435    
1436      else      else
1437        {        {
# Line 1177  for (;; ptr++) Line 1443  for (;; ptr++)
1443        bravalue = OP_BRA + *brackets;        bravalue = OP_BRA + *brackets;
1444        }        }
1445    
1446      /* Process nested bracketed re; at end pointer is on the bracket. We copy      /* Process nested bracketed re. Assertions may not be repeated, but other
1447      code into a non-register variable in order to be able to pass its address      kinds can be. We copy code into a non-register variable in order to be able
1448      because some compilers complain otherwise. */      to pass its address because some compilers complain otherwise. Pass in a
1449        new setting for the ims options if they have changed. */
1450    
1451        previous = (bravalue >= OP_ONCE)? code : NULL;
1452      *code = bravalue;      *code = bravalue;
1453        tempcode = code;
1454    
1455        if (!compile_regex(
1456             options | PCRE_INGROUP,       /* Set for all nested groups */
1457             ((options & PCRE_IMS) != (newoptions & PCRE_IMS))?
1458               newoptions & PCRE_IMS : -1, /* Pass ims options if changed */
1459             brackets,                     /* Bracket level */
1460             &tempcode,                    /* Where to put code (updated) */
1461             &ptr,                         /* Input pointer (updated) */
1462             errorptr,                     /* Where to put an error message */
1463             (bravalue == OP_ASSERTBACK ||
1464              bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
1465             condref,                      /* Condition reference number */
1466             cd))                          /* Tables block */
1467          goto FAILED;
1468    
1469        /* At the end of compiling, code is still pointing to the start of the
1470        group, while tempcode has been updated to point past the end of the group
1471        and any option resetting that may follow it. The pattern pointer (ptr)
1472        is on the bracket. */
1473    
1474        /* If this is a conditional bracket, check that there are no more than
1475        two branches in the group. */
1476    
1477        if (bravalue == OP_COND)
1478        {        {
1479        uschar *mcode = code;        int branchcount = 0;
1480        if (!compile_regex(options, brackets, &mcode, &ptr, errorptr))        uschar *tc = code;
1481    
1482          do {
1483             branchcount++;
1484             tc += (tc[1] << 8) | tc[2];
1485             }
1486          while (*tc != OP_KET);
1487    
1488          if (branchcount > 2)
1489            {
1490            *errorptr = ERR27;
1491          goto FAILED;          goto FAILED;
1492        code = mcode;          }
1493        }        }
1494    
1495        /* Now update the main code pointer to the end of the group. */
1496    
1497        code = tempcode;
1498    
1499        /* Error if hit end of pattern */
1500    
1501      if (*ptr != ')')      if (*ptr != ')')
1502        {        {
1503        *errorptr = ERR14;        *errorptr = ERR14;
# Line 1201  for (;; ptr++) Line 1510  for (;; ptr++)
1510      for validity in the pre-compiling pass. */      for validity in the pre-compiling pass. */
1511    
1512      case '\\':      case '\\':
1513      oldptr = ptr;      tempptr = ptr;
1514      c = check_escape(&ptr, errorptr, *brackets, options, FALSE);      c = check_escape(&ptr, errorptr, *brackets, options, FALSE, cd);
1515    
1516      /* Handle metacharacters introduced by \. For ones like \d, the ESC_ values      /* Handle metacharacters introduced by \. For ones like \d, the ESC_ values
1517      are arranged to be the negation of the corresponding OP_values. For the      are arranged to be the negation of the corresponding OP_values. For the
# Line 1215  for (;; ptr++) Line 1524  for (;; ptr++)
1524        {        {
1525        if (-c >= ESC_REF)        if (-c >= ESC_REF)
1526          {          {
         int refnum = -c - ESC_REF;  
         if (*brackets < refnum)  
           {  
           *errorptr = ERR15;  
           goto FAILED;  
           }  
1527          previous = code;          previous = code;
1528          *code++ = OP_REF;          *code++ = OP_REF;
1529          *code++ = refnum;          *code++ = -c - ESC_REF;
1530          }          }
1531        else        else
1532          {          {
1533          previous = (-c > ESC_b && -c < ESC_X)? code : NULL;          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
1534          *code++ = -c;          *code++ = -c;
1535          }          }
1536        continue;        continue;
# Line 1235  for (;; ptr++) Line 1538  for (;; ptr++)
1538    
1539      /* Data character: reset and fall through */      /* Data character: reset and fall through */
1540    
1541      ptr = oldptr;      ptr = tempptr;
1542      c = '\\';      c = '\\';
1543    
1544      /* Handle a run of data characters until a metacharacter is encountered.      /* Handle a run of data characters until a metacharacter is encountered.
# Line 1253  for (;; ptr++) Line 1556  for (;; ptr++)
1556        {        {
1557        if ((options & PCRE_EXTENDED) != 0)        if ((options & PCRE_EXTENDED) != 0)
1558          {          {
1559          if ((pcre_ctypes[c] & ctype_space) != 0) continue;          if ((cd->ctypes[c] & ctype_space) != 0) continue;
1560          if (c == '#')          if (c == '#')
1561            {            {
1562            while ((c = *(++ptr)) != 0 && c != '\n');            while ((c = *(++ptr)) != 0 && c != '\n');
# Line 1268  for (;; ptr++) Line 1571  for (;; ptr++)
1571    
1572        if (c == '\\')        if (c == '\\')
1573          {          {
1574          oldptr = ptr;          tempptr = ptr;
1575          c = check_escape(&ptr, errorptr, *brackets, options, FALSE);          c = check_escape(&ptr, errorptr, *brackets, options, FALSE, cd);
1576          if (c < 0) { ptr = oldptr; break; }          if (c < 0) { ptr = tempptr; break; }
1577          }          }
1578    
1579        /* Ordinary character or single-char escape */        /* Ordinary character or single-char escape */
# Line 1281  for (;; ptr++) Line 1584  for (;; ptr++)
1584    
1585      /* This "while" is the end of the "do" above. */      /* This "while" is the end of the "do" above. */
1586    
1587      while (length < 255 && (pcre_ctypes[c = *(++ptr)] & ctype_meta) == 0);      while (length < 255 && (cd->ctypes[c = *(++ptr)] & ctype_meta) == 0);
1588    
1589      /* Compute the length and set it in the data vector, and advance to      /* Compute the length and set it in the data vector, and advance to
1590      the next state. */      the next state. */
1591    
1592      previous[1] = length;      previous[1] = length;
1593      ptr--;      if (length < 255) ptr--;
1594      break;      break;
1595      }      }
1596    }                   /* end of big loop */    }                   /* end of big loop */
# Line 1311  return FALSE; Line 1614  return FALSE;
1614  /* On entry, ptr is pointing past the bracket character, but on return  /* On entry, ptr is pointing past the bracket character, but on return
1615  it points to the closing bracket, or vertical bar, or end of string.  it points to the closing bracket, or vertical bar, or end of string.
1616  The code variable is pointing at the byte into which the BRA operator has been  The code variable is pointing at the byte into which the BRA operator has been
1617  stored.  stored. If the ims options are changed at the start (for a (?ims: group) or
1618    during any branch, we need to insert an OP_OPT item at the start of every
1619    following branch to ensure they get set correctly at run time, and also pass
1620    the new options into every subsequent branch compile.
1621    
1622  Argument:  Argument:
1623    options   the option bits    options     the option bits
1624    brackets  -> int containing the number of extracting brackets used    optchanged  new ims options to set as if (?ims) were at the start, or -1
1625    codeptr   -> the address of the current code pointer                 for no change
1626    ptrptr    -> the address of the current pattern pointer    brackets    -> int containing the number of extracting brackets used
1627    errorptr  -> pointer to error message    codeptr     -> the address of the current code pointer
1628      ptrptr      -> the address of the current pattern pointer
1629      errorptr    -> pointer to error message
1630      lookbehind  TRUE if this is a lookbehind assertion
1631      condref     > 0 for OPT_CREF setting at start of conditional group
1632      cd          points to the data block with tables pointers
1633    
1634  Returns:    TRUE on success  Returns:      TRUE on success
1635  */  */
1636    
1637  static BOOL  static BOOL
1638  compile_regex(int options, int *brackets, uschar **codeptr,  compile_regex(int options, int optchanged, int *brackets, uschar **codeptr,
1639    const uschar **ptrptr, const char **errorptr)    const uschar **ptrptr, const char **errorptr, BOOL lookbehind, int condref,
1640      compile_data *cd)
1641  {  {
1642  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
1643  uschar *code = *codeptr;  uschar *code = *codeptr;
1644    uschar *last_branch = code;
1645  uschar *start_bracket = code;  uschar *start_bracket = code;
1646    uschar *reverse_count = NULL;
1647    int oldoptions = options & PCRE_IMS;
1648    
1649    code += 3;
1650    
1651    /* At the start of a reference-based conditional group, insert the reference
1652    number as an OP_CREF item. */
1653    
1654    if (condref > 0)
1655      {
1656      *code++ = OP_CREF;
1657      *code++ = condref;
1658      }
1659    
1660    /* Loop for each alternative branch */
1661    
1662  for (;;)  for (;;)
1663    {    {
1664    int length;    int length;
   uschar *last_branch = code;  
1665    
1666    code += 3;    /* Handle change of options */
1667    if (!compile_branch(options, brackets, &code, &ptr, errorptr))  
1668      if (optchanged >= 0)
1669        {
1670        *code++ = OP_OPT;
1671        *code++ = optchanged;
1672        options = (options & ~PCRE_IMS) | optchanged;
1673        }
1674    
1675      /* Set up dummy OP_REVERSE if lookbehind assertion */
1676    
1677      if (lookbehind)
1678        {
1679        *code++ = OP_REVERSE;
1680        reverse_count = code;
1681        *code++ = 0;
1682        *code++ = 0;
1683        }
1684    
1685      /* Now compile the branch */
1686    
1687      if (!compile_branch(options,brackets,&code,&ptr,errorptr,&optchanged,cd))
1688      {      {
1689      *ptrptr = ptr;      *ptrptr = ptr;
1690      return FALSE;      return FALSE;
# Line 1349  for (;;) Line 1696  for (;;)
1696    last_branch[1] = length >> 8;    last_branch[1] = length >> 8;
1697    last_branch[2] = length & 255;    last_branch[2] = length & 255;
1698    
1699      /* If lookbehind, check that this branch matches a fixed-length string,
1700      and put the length into the OP_REVERSE item. Temporarily mark the end of
1701      the branch with OP_END. */
1702    
1703      if (lookbehind)
1704        {
1705        *code = OP_END;
1706        length = find_fixedlength(last_branch);
1707        DPRINTF(("fixed length = %d\n", length));
1708        if (length < 0)
1709          {
1710          *errorptr = ERR25;
1711          *ptrptr = ptr;
1712          return FALSE;
1713          }
1714        reverse_count[0] = (length >> 8);
1715        reverse_count[1] = length & 255;
1716        }
1717    
1718    /* Reached end of expression, either ')' or end of pattern. Insert a    /* Reached end of expression, either ')' or end of pattern. Insert a
1719    terminating ket and the length of the whole bracketed item, and return,    terminating ket and the length of the whole bracketed item, and return,
1720    leaving the pointer at the terminating char. */    leaving the pointer at the terminating char. If any of the ims options
1721      were changed inside the group, compile a resetting op-code following. */
1722    
1723    if (*ptr != '|')    if (*ptr != '|')
1724      {      {
# Line 1359  for (;;) Line 1726  for (;;)
1726      *code++ = OP_KET;      *code++ = OP_KET;
1727      *code++ = length >> 8;      *code++ = length >> 8;
1728      *code++ = length & 255;      *code++ = length & 255;
1729        if (optchanged >= 0)
1730          {
1731          *code++ = OP_OPT;
1732          *code++ = oldoptions;
1733          }
1734      *codeptr = code;      *codeptr = code;
1735      *ptrptr = ptr;      *ptrptr = ptr;
1736      return TRUE;      return TRUE;
# Line 1367  for (;;) Line 1739  for (;;)
1739    /* Another branch follows; insert an "or" node and advance the pointer. */    /* Another branch follows; insert an "or" node and advance the pointer. */
1740    
1741    *code = OP_ALT;    *code = OP_ALT;
1742      last_branch = code;
1743      code += 3;
1744    ptr++;    ptr++;
1745    }    }
1746  /* Control never reaches here */  /* Control never reaches here */
# Line 1374  for (;;) Line 1748  for (;;)
1748    
1749    
1750    
1751    
1752    /*************************************************
1753    *      Find first significant op code            *
1754    *************************************************/
1755    
1756    /* This is called by several functions that scan a compiled expression looking
1757    for a fixed first character, or an anchoring op code etc. It skips over things
1758    that do not influence this. For one application, a change of caseless option is
1759    important.
1760    
1761    Arguments:
1762      code       pointer to the start of the group
1763      options    pointer to external options
1764      optbit     the option bit whose changing is significant, or
1765                 zero if none are
1766      optstop    TRUE to return on option change, otherwise change the options
1767                   value and continue
1768    
1769    Returns:     pointer to the first significant opcode
1770    */
1771    
1772    static const uschar*
1773    first_significant_code(const uschar *code, int *options, int optbit,
1774      BOOL optstop)
1775    {
1776    for (;;)
1777      {
1778      switch ((int)*code)
1779        {
1780        case OP_OPT:
1781        if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))
1782          {
1783          if (optstop) return code;
1784          *options = (int)code[1];
1785          }
1786        code += 2;
1787        break;
1788    
1789        case OP_CREF:
1790        code += 2;
1791        break;
1792    
1793        case OP_ASSERT_NOT:
1794        case OP_ASSERTBACK:
1795        case OP_ASSERTBACK_NOT:
1796        do code += (code[1] << 8) + code[2]; while (*code == OP_ALT);
1797        code += 3;
1798        break;
1799    
1800        default:
1801        return code;
1802        }
1803      }
1804    /* Control never reaches here */
1805    }
1806    
1807    
1808    
1809    
1810  /*************************************************  /*************************************************
1811  *          Check for anchored expression         *  *          Check for anchored expression         *
1812  *************************************************/  *************************************************/
# Line 1388  A branch is also implicitly anchored if Line 1821  A branch is also implicitly anchored if
1821  the rest of the pattern at all possible matching points, so there is no point  the rest of the pattern at all possible matching points, so there is no point
1822  trying them again.  trying them again.
1823    
1824  Argument:  points to start of expression (the bracket)  Arguments:
1825  Returns:   TRUE or FALSE    code       points to start of expression (the bracket)
1826      options    points to the options setting
1827    
1828    Returns:     TRUE or FALSE
1829  */  */
1830    
1831  static BOOL  static BOOL
1832  is_anchored(register const uschar *code, BOOL multiline)  is_anchored(register const uschar *code, int *options)
1833  {  {
1834  do {  do {
1835     int op = (int)code[3];     const uschar *scode = first_significant_code(code + 3, options,
1836     if (op >= OP_BRA || op == OP_ASSERT || op == OP_ONCE)       PCRE_MULTILINE, FALSE);
1837       { if (!is_anchored(code+3, multiline)) return FALSE; }     register int op = *scode;
1838       if (op >= OP_BRA || op == OP_ASSERT || op == OP_ONCE || op == OP_COND)
1839         { if (!is_anchored(scode, options)) return FALSE; }
1840     else if (op == OP_TYPESTAR || op == OP_TYPEMINSTAR)     else if (op == OP_TYPESTAR || op == OP_TYPEMINSTAR)
1841       { if (code[4] != OP_ANY) return FALSE; }       { if (scode[1] != OP_ANY) return FALSE; }
1842     else if (op != OP_SOD && (multiline || op != OP_CIRC)) return FALSE;     else if (op != OP_SOD &&
1843               ((*options & PCRE_MULTILINE) != 0 || op != OP_CIRC))
1844         return FALSE;
1845     code += (code[1] << 8) + code[2];     code += (code[1] << 8) + code[2];
1846     }     }
1847  while (*code == OP_ALT);  while (*code == OP_ALT);
# Line 1425  static BOOL Line 1865  static BOOL
1865  is_startline(const uschar *code)  is_startline(const uschar *code)
1866  {  {
1867  do {  do {
1868     if ((int)code[3] >= OP_BRA || code[3] == OP_ASSERT)     const uschar *scode = first_significant_code(code + 3, NULL, 0, FALSE);
1869       { if (!is_startline(code+3)) return FALSE; }     register int op = *scode;
1870     else if (code[3] != OP_CIRC) return FALSE;     if (op >= OP_BRA || op == OP_ASSERT || op == OP_ONCE || op == OP_COND)
1871         { if (!is_startline(scode)) return FALSE; }
1872       else if (op != OP_CIRC) return FALSE;
1873     code += (code[1] << 8) + code[2];     code += (code[1] << 8) + code[2];
1874     }     }
1875  while (*code == OP_ALT);  while (*code == OP_ALT);
# Line 1446  Consider each alternative branch. If the Line 1888  Consider each alternative branch. If the
1888  a bracket all of whose alternatives start with the same char (recurse ad lib),  a bracket all of whose alternatives start with the same char (recurse ad lib),
1889  then we return that char, otherwise -1.  then we return that char, otherwise -1.
1890    
1891  Argument:  points to start of expression (the bracket)  Arguments:
1892  Returns:   -1 or the fixed first char    code       points to start of expression (the bracket)
1893      options    pointer to the options (used to check casing changes)
1894    
1895    Returns:     -1 or the fixed first char
1896  */  */
1897    
1898  static int  static int
1899  find_firstchar(uschar *code)  find_firstchar(const uschar *code, int *options)
1900  {  {
1901  register int c = -1;  register int c = -1;
1902  do  do {
1903    {     int d;
1904    register int charoffset = 4;     const uschar *scode = first_significant_code(code + 3, options,
1905         PCRE_CASELESS, TRUE);
1906    if ((int)code[3] >= OP_BRA || code[3] == OP_ASSERT)     register int op = *scode;
1907      {  
1908      register int d;     if (op >= OP_BRA) op = OP_BRA;
1909      if ((d = find_firstchar(code+3)) < 0) return -1;  
1910      if (c < 0) c = d; else if (c != d) return -1;     switch(op)
1911      }       {
1912         default:
1913    else switch(code[3])       return -1;
1914      {  
1915      default:       case OP_BRA:
1916      return -1;       case OP_ASSERT:
1917         case OP_ONCE:
1918      case OP_EXACT:       /* Fall through */       case OP_COND:
1919      charoffset++;       if ((d = find_firstchar(scode, options)) < 0) return -1;
1920         if (c < 0) c = d; else if (c != d) return -1;
1921      case OP_CHARS:       /* Fall through */       break;
1922      charoffset++;  
1923         case OP_EXACT:       /* Fall through */
1924         scode++;
1925    
1926         case OP_CHARS:       /* Fall through */
1927         scode++;
1928    
1929         case OP_PLUS:
1930         case OP_MINPLUS:
1931         if (c < 0) c = scode[1]; else if (c != scode[1]) return -1;
1932         break;
1933         }
1934    
1935      case OP_PLUS:     code += (code[1] << 8) + code[2];
1936      case OP_MINPLUS:     }
     if (c < 0) c = code[charoffset]; else if (c != code[charoffset]) return -1;  
     break;  
     }  
   code += (code[1] << 8) + code[2];  
   }  
1937  while (*code == OP_ALT);  while (*code == OP_ALT);
1938  return c;  return c;
1939  }  }
1940    
1941    
1942    
1943    
1944    
1945  /*************************************************  /*************************************************
1946  *        Compile a Regular Expression            *  *        Compile a Regular Expression            *
1947  *************************************************/  *************************************************/
# Line 1501  Arguments: Line 1954  Arguments:
1954    options      various option bits    options      various option bits
1955    errorptr     pointer to pointer to error text    errorptr     pointer to pointer to error text
1956    erroroffset  ptr offset in pattern where error was detected    erroroffset  ptr offset in pattern where error was detected
1957      tables       pointer to character tables or NULL
1958    
1959  Returns:       pointer to compiled data block, or NULL on error,  Returns:       pointer to compiled data block, or NULL on error,
1960                 with errorptr and erroroffset set                 with errorptr and erroroffset set
# Line 1508  Returns:       pointer to compiled data Line 1962  Returns:       pointer to compiled data
1962    
1963  pcre *  pcre *
1964  pcre_compile(const char *pattern, int options, const char **errorptr,  pcre_compile(const char *pattern, int options, const char **errorptr,
1965    int *erroroffset)    int *erroroffset, const unsigned char *tables)
1966  {  {
1967  real_pcre *re;  real_pcre *re;
 int spaces = 0;  
1968  int length = 3;      /* For initial BRA plus length */  int length = 3;      /* For initial BRA plus length */
1969  int runlength;  int runlength;
1970  int c, size;  int c, size;
1971  int bracount = 0;  int bracount = 0;
 int brastack[200];  
1972  int top_backref = 0;  int top_backref = 0;
1973    int branch_extra = 0;
1974    int branch_newextra;
1975  unsigned int brastackptr = 0;  unsigned int brastackptr = 0;
1976  uschar *code;  uschar *code;
1977  const uschar *ptr;  const uschar *ptr;
1978    compile_data compile_block;
1979    int brastack[BRASTACK_SIZE];
1980    uschar bralenstack[BRASTACK_SIZE];
1981    
1982  #ifdef DEBUG  #ifdef DEBUG
1983  uschar *code_base, *code_end;  uschar *code_base, *code_end;
# Line 1547  if ((options & ~PUBLIC_OPTIONS) != 0) Line 2004  if ((options & ~PUBLIC_OPTIONS) != 0)
2004    return NULL;    return NULL;
2005    }    }
2006    
2007    /* Set up pointers to the individual character tables */
2008    
2009    if (tables == NULL) tables = pcre_default_tables;
2010    compile_block.lcc = tables + lcc_offset;
2011    compile_block.fcc = tables + fcc_offset;
2012    compile_block.cbits = tables + cbits_offset;
2013    compile_block.ctypes = tables + ctypes_offset;
2014    
2015    /* Reflect pattern for debugging output */
2016    
2017  DPRINTF(("------------------------------------------------------------------\n"));  DPRINTF(("------------------------------------------------------------------\n"));
2018  DPRINTF(("%s\n", pattern));  DPRINTF(("%s\n", pattern));
2019    
# Line 1563  while ((c = *(++ptr)) != 0) Line 2030  while ((c = *(++ptr)) != 0)
2030    int min, max;    int min, max;
2031    int class_charcount;    int class_charcount;
2032    
2033    if ((pcre_ctypes[c] & ctype_space) != 0)    if ((options & PCRE_EXTENDED) != 0)
     {  
     if ((options & PCRE_EXTENDED) != 0) continue;  
     spaces++;  
     }  
   
   if (c == '#' && (options & PCRE_EXTENDED) != 0)  
2034      {      {
2035      while ((c = *(++ptr)) != 0 && c != '\n');      if ((compile_block.ctypes[c] & ctype_space) != 0) continue;
2036      continue;      if (c == '#')
2037          {
2038          while ((c = *(++ptr)) != 0 && c != '\n');
2039          continue;
2040          }
2041      }      }
2042    
2043    switch(c)    switch(c)
# Line 1585  while ((c = *(++ptr)) != 0) Line 2050  while ((c = *(++ptr)) != 0)
2050      case '\\':      case '\\':
2051        {        {
2052        const uschar *save_ptr = ptr;        const uschar *save_ptr = ptr;
2053        c = check_escape(&ptr, errorptr, bracount, options, FALSE);        c = check_escape(&ptr, errorptr, bracount, options, FALSE, &compile_block);
2054        if (*errorptr != NULL) goto PCRE_ERROR_RETURN;        if (*errorptr != NULL) goto PCRE_ERROR_RETURN;
2055        if (c >= 0)        if (c >= 0)
2056          {          {
# Line 1605  while ((c = *(++ptr)) != 0) Line 2070  while ((c = *(++ptr)) != 0)
2070        int refnum = -c - ESC_REF;        int refnum = -c - ESC_REF;
2071        if (refnum > top_backref) top_backref = refnum;        if (refnum > top_backref) top_backref = refnum;
2072        length++;   /* For single back reference */        length++;   /* For single back reference */
2073        if (ptr[1] == '{' && is_counted_repeat(ptr+2))        if (ptr[1] == '{' && is_counted_repeat(ptr+2, &compile_block))
2074          {          {
2075          ptr = read_repeat_counts(ptr+2, &min, &max, errorptr);          ptr = read_repeat_counts(ptr+2, &min, &max, errorptr, &compile_block);
2076          if (*errorptr != NULL) goto PCRE_ERROR_RETURN;          if (*errorptr != NULL) goto PCRE_ERROR_RETURN;
2077          if ((min == 0 && (max == 1 || max == -1)) ||          if ((min == 0 && (max == 1 || max == -1)) ||
2078            (min == 1 && max == -1))            (min == 1 && max == -1))
# Line 1631  while ((c = *(++ptr)) != 0) Line 2096  while ((c = *(++ptr)) != 0)
2096      or back reference. */      or back reference. */
2097    
2098      case '{':      case '{':
2099      if (!is_counted_repeat(ptr+1)) goto NORMAL_CHAR;      if (!is_counted_repeat(ptr+1, &compile_block)) goto NORMAL_CHAR;
2100      ptr = read_repeat_counts(ptr+1, &min, &max, errorptr);      ptr = read_repeat_counts(ptr+1, &min, &max, errorptr, &compile_block);
2101      if (*errorptr != NULL) goto PCRE_ERROR_RETURN;      if (*errorptr != NULL) goto PCRE_ERROR_RETURN;
2102      if ((min == 0 && (max == 1 || max == -1)) ||      if ((min == 0 && (max == 1 || max == -1)) ||
2103        (min == 1 && max == -1))        (min == 1 && max == -1))
# Line 1646  while ((c = *(++ptr)) != 0) Line 2111  while ((c = *(++ptr)) != 0)
2111      if (ptr[1] == '?') ptr++;      if (ptr[1] == '?') ptr++;
2112      continue;      continue;
2113    
2114      /* An alternation contains an offset to the next branch or ket. */      /* An alternation contains an offset to the next branch or ket. If any ims
2115        options changed in the previous branch(es), and/or if we are in a
2116        lookbehind assertion, extra space will be needed at the start of the
2117        branch. This is handled by branch_extra. */
2118    
2119      case '|':      case '|':
2120      length += 3;      length += 3 + branch_extra;
2121      continue;      continue;
2122    
2123      /* A character class uses 33 characters. Don't worry about character types      /* A character class uses 33 characters. Don't worry about character types
# Line 1663  while ((c = *(++ptr)) != 0) Line 2132  while ((c = *(++ptr)) != 0)
2132        {        {
2133        if (*ptr == '\\')        if (*ptr == '\\')
2134          {          {
2135          int ch = check_escape(&ptr, errorptr, bracount, options, TRUE);          int ch = check_escape(&ptr, errorptr, bracount, options, TRUE,
2136              &compile_block);
2137          if (*errorptr != NULL) goto PCRE_ERROR_RETURN;          if (*errorptr != NULL) goto PCRE_ERROR_RETURN;
2138          if (-ch == ESC_b) class_charcount++; else class_charcount = 10;          if (-ch == ESC_b) class_charcount++; else class_charcount = 10;
2139          }          }
# Line 1680  while ((c = *(++ptr)) != 0) Line 2150  while ((c = *(++ptr)) != 0)
2150    
2151        /* A repeat needs either 1 or 5 bytes. */        /* A repeat needs either 1 or 5 bytes. */
2152    
2153        if (*ptr != 0 && ptr[1] == '{' && is_counted_repeat(ptr+2))        if (*ptr != 0 && ptr[1] == '{' && is_counted_repeat(ptr+2, &compile_block))
2154          {          {
2155          ptr = read_repeat_counts(ptr+2, &min, &max, errorptr);          ptr = read_repeat_counts(ptr+2, &min, &max, errorptr, &compile_block);
2156          if (*errorptr != NULL) goto PCRE_ERROR_RETURN;          if (*errorptr != NULL) goto PCRE_ERROR_RETURN;
2157          if ((min == 0 && (max == 1 || max == -1)) ||          if ((min == 0 && (max == 1 || max == -1)) ||
2158            (min == 1 && max == -1))            (min == 1 && max == -1))
# Line 1695  while ((c = *(++ptr)) != 0) Line 2165  while ((c = *(++ptr)) != 0)
2165    
2166      /* Brackets may be genuine groups or special things */      /* Brackets may be genuine groups or special things */
2167    
2168      case '(':      case '(':
2169        branch_newextra = 0;
2170    
2171        /* Handle special forms of bracket, which all start (? */
2172    
2173        if (ptr[1] == '?')
2174          {
2175          int set, unset;
2176          int *optset;
2177    
2178          switch (c = ptr[2])
2179            {
2180            /* Skip over comments entirely */
2181            case '#':
2182            ptr += 3;
2183            while (*ptr != 0 && *ptr != ')') ptr++;
2184            if (*ptr == 0)
2185              {
2186              *errorptr = ERR18;
2187              goto PCRE_ERROR_RETURN;
2188              }
2189            continue;
2190    
2191            /* Non-referencing groups and lookaheads just move the pointer on, and
2192            then behave like a non-special bracket, except that they don't increment
2193            the count of extracting brackets. Ditto for the "once only" bracket,
2194            which is in Perl from version 5.005. */
2195    
2196            case ':':
2197            case '=':
2198            case '!':
2199            case '>':
2200            ptr += 2;
2201            break;
2202    
2203            /* Lookbehinds are in Perl from version 5.005 */
2204    
2205            case '<':
2206            if (ptr[3] == '=' || ptr[3] == '!')
2207              {
2208              ptr += 3;
2209              branch_newextra = 3;
2210              length += 3;         /* For the first branch */
2211              break;
2212              }
2213            *errorptr = ERR24;
2214            goto PCRE_ERROR_RETURN;
2215    
2216            /* Conditionals are in Perl from version 5.005. The bracket must either
2217            be followed by a number (for bracket reference) or by an assertion
2218            group. */
2219    
2220            case '(':
2221            if ((compile_block.ctypes[ptr[3]] & ctype_digit) != 0)
2222              {
2223              ptr += 4;
2224              length += 2;
2225              while ((compile_block.ctypes[*ptr] & ctype_digit) != 0) ptr++;
2226              if (*ptr != ')')
2227                {
2228                *errorptr = ERR26;
2229                goto PCRE_ERROR_RETURN;
2230                }
2231              }
2232            else   /* An assertion must follow */
2233              {
2234              ptr++;   /* Can treat like ':' as far as spacing is concerned */
2235    
2236              if (ptr[2] != '?' || strchr("=!<", ptr[3]) == NULL)
2237                {
2238                ptr += 2;    /* To get right offset in message */
2239                *errorptr = ERR28;
2240                goto PCRE_ERROR_RETURN;
2241                }
2242              }
2243            break;
2244    
2245            /* Else loop checking valid options until ) is met. Anything else is an
2246            error. If we are without any brackets, i.e. at top level, the settings
2247            act as if specified in the options, so massage the options immediately.
2248            This is for backward compatibility with Perl 5.004. */
2249    
2250            default:
2251            set = unset = 0;
2252            optset = &set;
2253            ptr += 2;
2254    
2255            for (;; ptr++)
2256              {
2257              c = *ptr;
2258              switch (c)
2259                {
2260                case 'i':
2261                *optset |= PCRE_CASELESS;
2262                continue;
2263    
2264                case 'm':
2265                *optset |= PCRE_MULTILINE;
2266                continue;
2267    
2268                case 's':
2269                *optset |= PCRE_DOTALL;
2270                continue;
2271    
2272                case 'x':
2273                *optset |= PCRE_EXTENDED;
2274                continue;
2275    
2276                case 'X':
2277                *optset |= PCRE_EXTRA;
2278                continue;
2279    
2280                case 'U':
2281                *optset |= PCRE_UNGREEDY;
2282                continue;
2283    
2284      /* Handle special forms of bracket, which all start (? */              case '-':
2285                optset = &unset;
2286                continue;
2287    
2288      if (ptr[1] == '?') switch (c = ptr[2])              /* A termination by ')' indicates an options-setting-only item;
2289        {              this is global at top level; otherwise nothing is done here and
2290        /* Skip over comments entirely */              it is handled during the compiling process on a per-bracket-group
2291        case '#':              basis. */
       ptr += 3;  
       while (*ptr != 0 && *ptr != ')') ptr++;  
       if (*ptr == 0)  
         {  
         *errorptr = ERR18;  
         goto PCRE_ERROR_RETURN;  
         }  
       continue;  
2292    
2293        /* Non-referencing groups and lookaheads just move the pointer on, and              case ')':
2294        then behave like a non-special bracket, except that they don't increment              if (brastackptr == 0)
2295        the count of extracting brackets. */                {
2296                  options = (options | set) & (~unset);
2297        case ':':                set = unset = 0;     /* To save length */
2298        case '=':                }
2299        case '!':              /* Fall through */
       ptr += 2;  
       break;  
2300    
2301        /* Ditto for the "once only" bracket, allowed only if the extra bit              /* A termination by ':' indicates the start of a nested group with
2302        is set. */              the given options set. This is again handled at compile time, but
2303                we must allow for compiled space if any of the ims options are
2304                set. We also have to allow for resetting space at the end of
2305                the group, which is why 4 is added to the length and not just 2.
2306                If there are several changes of options within the same group, this
2307                will lead to an over-estimate on the length, but this shouldn't
2308                matter very much. We also have to allow for resetting options at
2309                the start of any alternations, which we do by setting
2310                branch_newextra to 2. */
2311    
2312        case '>':              case ':':
2313        if ((options & PCRE_EXTRA) != 0)              if (((set|unset) & PCRE_IMS) != 0)
2314          {                {
2315          ptr += 2;                length += 4;
2316          break;                branch_newextra = 2;
2317          }                }
2318        /* Else fall thourh */              goto END_OPTIONS;
2319    
2320        /* Else loop setting valid options until ) is met. Anything else is an              /* Unrecognized option character */
       error. */  
2321    
2322        default:              default:
2323        ptr += 2;              *errorptr = ERR12;
2324        for (;; ptr++)              goto PCRE_ERROR_RETURN;
2325          {              }
         if ((c = *ptr) == 'i')  
           {  
           options |= PCRE_CASELESS;  
           continue;  
           }  
         else if ((c = *ptr) == 'm')  
           {  
           options |= PCRE_MULTILINE;  
           continue;  
           }  
         else if (c == 's')  
           {  
           options |= PCRE_DOTALL;  
           continue;  
2326            }            }
2327          else if (c == 'x')  
2328            /* If we hit a closing bracket, that's it - this is a freestanding
2329            option-setting. We need to ensure that branch_extra is updated if
2330            necessary. The only values branch_newextra can have here are 0 or 2.
2331            If the value is 2, then branch_extra must either be 2 or 5, depending
2332            on whether this is a lookbehind group or not. */
2333    
2334            END_OPTIONS:
2335            if (c == ')')
2336            {            {
2337            options |= PCRE_EXTENDED;            if (branch_newextra == 2 && (branch_extra == 0 || branch_extra == 3))
2338            length -= spaces;          /* Already counted spaces */              branch_extra += branch_newextra;
2339            continue;            continue;
2340            }            }
         else if (c == ')') break;  
2341    
2342          *errorptr = ERR12;          /* If options were terminated by ':' control comes here. Fall through
2343          goto PCRE_ERROR_RETURN;          to handle the group below. */
2344          }          }
       continue;                      /* End of this bracket handling */  
2345        }        }
2346    
2347      /* Extracting brackets must be counted so we can process escapes in a      /* Extracting brackets must be counted so we can process escapes in a
# Line 1775  while ((c = *(++ptr)) != 0) Line 2350  while ((c = *(++ptr)) != 0)
2350      else bracount++;      else bracount++;
2351    
2352      /* Non-special forms of bracket. Save length for computing whole length      /* Non-special forms of bracket. Save length for computing whole length
2353      at end if there's a repeat that requires duplication of the group. */      at end if there's a repeat that requires duplication of the group. Also
2354        save the current value of branch_extra, and start the new group with
2355        the new value. If non-zero, this will either be 2 for a (?imsx: group, or 3
2356        for a lookbehind assertion. */
2357    
2358      if (brastackptr >= sizeof(brastack)/sizeof(int))      if (brastackptr >= sizeof(brastack)/sizeof(int))
2359        {        {
# Line 1783  while ((c = *(++ptr)) != 0) Line 2361  while ((c = *(++ptr)) != 0)
2361        goto PCRE_ERROR_RETURN;        goto PCRE_ERROR_RETURN;
2362        }        }
2363    
2364        bralenstack[brastackptr] = branch_extra;
2365        branch_extra = branch_newextra;
2366    
2367      brastack[brastackptr++] = length;      brastack[brastackptr++] = length;
2368      length += 3;      length += 3;
2369      continue;      continue;
# Line 1790  while ((c = *(++ptr)) != 0) Line 2371  while ((c = *(++ptr)) != 0)
2371      /* Handle ket. Look for subsequent max/min; for certain sets of values we      /* Handle ket. Look for subsequent max/min; for certain sets of values we
2372      have to replicate this bracket up to that many times. If brastackptr is      have to replicate this bracket up to that many times. If brastackptr is
2373      0 this is an unmatched bracket which will generate an error, but take care      0 this is an unmatched bracket which will generate an error, but take care
2374      not to try to access brastack[-1]. */      not to try to access brastack[-1] when computing the length and restoring
2375        the branch_extra value. */
2376    
2377      case ')':      case ')':
2378      length += 3;      length += 3;
2379        {        {
2380        int minval = 1;        int minval = 1;
2381        int maxval = 1;        int maxval = 1;
2382        int duplength = (brastackptr > 0)? length - brastack[--brastackptr] : 0;        int duplength;
2383    
2384          if (brastackptr > 0)
2385            {
2386            duplength = length - brastack[--brastackptr];
2387            branch_extra = bralenstack[brastackptr];
2388            }
2389          else duplength = 0;
2390    
2391        /* Leave ptr at the final char; for read_repeat_counts this happens        /* Leave ptr at the final char; for read_repeat_counts this happens
2392        automatically; for the others we need an increment. */        automatically; for the others we need an increment. */
2393    
2394        if ((c = ptr[1]) == '{' && is_counted_repeat(ptr+2))        if ((c = ptr[1]) == '{' && is_counted_repeat(ptr+2, &compile_block))
2395          {          {
2396          ptr = read_repeat_counts(ptr+2, &minval, &maxval, errorptr);          ptr = read_repeat_counts(ptr+2, &minval, &maxval, errorptr,
2397              &compile_block);
2398          if (*errorptr != NULL) goto PCRE_ERROR_RETURN;          if (*errorptr != NULL) goto PCRE_ERROR_RETURN;
2399          }          }
2400        else if (c == '*') { minval = 0; maxval = -1; ptr++; }        else if (c == '*') { minval = 0; maxval = -1; ptr++; }
2401        else if (c == '+') { maxval = -1; ptr++; }        else if (c == '+') { maxval = -1; ptr++; }
2402        else if (c == '?') { minval = 0; ptr++; }        else if (c == '?') { minval = 0; ptr++; }
2403    
2404        /* If there is a minimum > 1 we have to replicate up to minval-1 times;        /* If the minimum is zero, we have to allow for an OP_BRAZERO before the
2405        if there is a limited maximum we have to replicate up to maxval-1 times        group, and if the maximum is greater than zero, we have to replicate
2406        and allow for a BRAZERO item before each optional copy, as we also have        maxval-1 times; each replication acquires an OP_BRAZERO plus a nesting
2407        to do before the first copy if the minimum is zero. */        bracket set - hence the 7. */
2408    
2409        if (minval == 0) length++;        if (minval == 0)
2410          else if (minval > 1) length += (minval - 1) * duplength;          {
2411        if (maxval > minval) length += (maxval - minval) * (duplength + 1);          length++;
2412            if (maxval > 0) length += (maxval - 1) * (duplength + 7);
2413            }
2414    
2415          /* When the minimum is greater than zero, 1 we have to replicate up to
2416          minval-1 times, with no additions required in the copies. Then, if
2417          there is a limited maximum we have to replicate up to maxval-1 times
2418          allowing for a BRAZERO item before each optional copy and nesting
2419          brackets for all but one of the optional copies. */
2420    
2421          else
2422            {
2423            length += (minval - 1) * duplength;
2424            if (maxval > minval)   /* Need this test as maxval=-1 means no limit */
2425              length += (maxval - minval) * (duplength + 7) - 6;
2426            }
2427        }        }
2428      continue;      continue;
2429    
# Line 1833  while ((c = *(++ptr)) != 0) Line 2438  while ((c = *(++ptr)) != 0)
2438      runlength = 0;      runlength = 0;
2439      do      do
2440        {        {
2441        if ((pcre_ctypes[c] & ctype_space) != 0)        if ((options & PCRE_EXTENDED) != 0)
         {  
         if ((options & PCRE_EXTENDED) != 0) continue;  
         spaces++;  
         }  
   
       if (c == '#' && (options & PCRE_EXTENDED) != 0)  
2442          {          {
2443          while ((c = *(++ptr)) != 0 && c != '\n');          if ((compile_block.ctypes[c] & ctype_space) != 0) continue;
2444          continue;          if (c == '#')
2445              {
2446              while ((c = *(++ptr)) != 0 && c != '\n');
2447              continue;
2448              }
2449          }          }
2450    
2451        /* Backslash may introduce a data char or a metacharacter; stop the        /* Backslash may introduce a data char or a metacharacter; stop the
# Line 1851  while ((c = *(++ptr)) != 0) Line 2454  while ((c = *(++ptr)) != 0)
2454        if (c == '\\')        if (c == '\\')
2455          {          {
2456          const uschar *saveptr = ptr;          const uschar *saveptr = ptr;
2457          c = check_escape(&ptr, errorptr, bracount, options, FALSE);          c = check_escape(&ptr, errorptr, bracount, options, FALSE,
2458              &compile_block);
2459          if (*errorptr != NULL) goto PCRE_ERROR_RETURN;          if (*errorptr != NULL) goto PCRE_ERROR_RETURN;
2460          if (c < 0) { ptr = saveptr; break; }          if (c < 0) { ptr = saveptr; break; }
2461          }          }
# Line 1863  while ((c = *(++ptr)) != 0) Line 2467  while ((c = *(++ptr)) != 0)
2467    
2468      /* This "while" is the end of the "do" above. */      /* This "while" is the end of the "do" above. */
2469    
2470      while (runlength < 255 && (pcre_ctypes[c = *(++ptr)] & ctype_meta) == 0);      while (runlength < 255 &&
2471          (compile_block.ctypes[c = *(++ptr)] & ctype_meta) == 0);
2472    
2473      ptr--;      ptr--;
2474      length += runlength;      length += runlength;
# Line 1898  if (re == NULL) Line 2503  if (re == NULL)
2503    
2504  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
2505  re->options = options;  re->options = options;
2506    re->tables = tables;
2507    
2508  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
2509  error, *errorptr will be set non-NULL, so we don't need to look at the result  error, *errorptr will be set non-NULL, so we don't need to look at the result
# Line 1907  ptr = (const uschar *)pattern; Line 2513  ptr = (const uschar *)pattern;
2513  code = re->code;  code = re->code;
2514  *code = OP_BRA;  *code = OP_BRA;
2515  bracount = 0;  bracount = 0;
2516  (void)compile_regex(options, &bracount, &code, &ptr, errorptr);  (void)compile_regex(options, -1, &bracount, &code, &ptr, errorptr, FALSE, -1,
2517      &compile_block);
2518  re->top_bracket = bracount;  re->top_bracket = bracount;
2519  re->top_backref = top_backref;  re->top_backref = top_backref;
2520    
# Line 1924  if debugging, leave the test till after Line 2531  if debugging, leave the test till after
2531  if (code - re->code > length) *errorptr = ERR23;  if (code - re->code > length) *errorptr = ERR23;
2532  #endif  #endif
2533    
2534    /* Give an error if there's back reference to a non-existent capturing
2535    subpattern. */
2536    
2537    if (top_backref > re->top_bracket) *errorptr = ERR15;
2538    
2539  /* Failed to compile */  /* Failed to compile */
2540    
2541  if (*errorptr != NULL)  if (*errorptr != NULL)
# Line 1942  to set the PCRE_STARTLINE flag if all br Line 2554  to set the PCRE_STARTLINE flag if all br
2554    
2555  if ((options & PCRE_ANCHORED) == 0)  if ((options & PCRE_ANCHORED) == 0)
2556    {    {
2557    if (is_anchored(re->code, (options & PCRE_MULTILINE) != 0))    int temp_options = options;
2558      if (is_anchored(re->code, &temp_options))
2559      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
2560    else    else
2561      {      {
2562      int ch = find_firstchar(re->code);      int ch = find_firstchar(re->code, &temp_options);
2563      if (ch >= 0)      if (ch >= 0)
2564        {        {
2565        re->first_char = ch;        re->first_char = ch;
# Line 1961  if ((options & PCRE_ANCHORED) == 0) Line 2574  if ((options & PCRE_ANCHORED) == 0)
2574    
2575  #ifdef DEBUG  #ifdef DEBUG
2576    
2577  printf("Length = %d top_bracket = %d top_backref=%d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
2578    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
2579    
2580  if (re->options != 0)  if (re->options != 0)
2581    {    {
2582    printf("%s%s%s%s%s%s%s\n",    printf("%s%s%s%s%s%s%s%s\n",
2583      ((re->options & PCRE_ANCHORED) != 0)? "anchored " : "",      ((re->options & PCRE_ANCHORED) != 0)? "anchored " : "",
2584      ((re->options & PCRE_CASELESS) != 0)? "caseless " : "",      ((re->options & PCRE_CASELESS) != 0)? "caseless " : "",
2585      ((re->options & PCRE_EXTENDED) != 0)? "extended " : "",      ((re->options & PCRE_EXTENDED) != 0)? "extended " : "",
2586      ((re->options & PCRE_MULTILINE) != 0)? "multiline " : "",      ((re->options & PCRE_MULTILINE) != 0)? "multiline " : "",
2587      ((re->options & PCRE_DOTALL) != 0)? "dotall " : "",      ((re->options & PCRE_DOTALL) != 0)? "dotall " : "",
2588      ((re->options & PCRE_DOLLAR_ENDONLY) != 0)? "endonly " : "",      ((re->options & PCRE_DOLLAR_ENDONLY) != 0)? "endonly " : "",
2589      ((re->options & PCRE_EXTRA) != 0)? "extra " : "");      ((re->options & PCRE_EXTRA) != 0)? "extra " : "",
2590        ((re->options & PCRE_UNGREEDY) != 0)? "ungreedy " : "");
2591    }    }
2592    
2593  if ((re->options & PCRE_FIRSTSET) != 0)  if ((re->options & PCRE_FIRSTSET) != 0)
# Line 1999  while (code < code_end) Line 2613  while (code < code_end)
2613    
2614    else switch(*code)    else switch(*code)
2615      {      {
2616        case OP_OPT:
2617        printf(" %.2x %s", code[1], OP_names[*code]);
2618        code++;
2619        break;
2620    
2621        case OP_COND:
2622        printf("%3d Cond", (code[1] << 8) + code[2]);
2623        code += 2;
2624        break;
2625    
2626        case OP_CREF:
2627        printf(" %.2d %s", code[1], OP_names[*code]);
2628        code++;
2629        break;
2630    
2631      case OP_CHARS:      case OP_CHARS:
2632      charlength = *(++code);      charlength = *(++code);
2633      printf("%3d ", charlength);      printf("%3d ", charlength);
# Line 2012  while (code < code_end) Line 2641  while (code < code_end)
2641      case OP_KET:      case OP_KET:
2642      case OP_ASSERT:      case OP_ASSERT:
2643      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
2644        case OP_ASSERTBACK:
2645        case OP_ASSERTBACK_NOT:
2646      case OP_ONCE:      case OP_ONCE:
2647      printf("%3d %s", (code[1] << 8) + code[2], OP_names[*code]);      printf("%3d %s", (code[1] << 8) + code[2], OP_names[*code]);
2648      code += 2;      code += 2;
2649      break;      break;
2650    
2651        case OP_REVERSE:
2652        printf("%3d %s", (code[1] << 8) + code[2], OP_names[*code]);
2653        code += 2;
2654        break;
2655    
2656      case OP_STAR:      case OP_STAR:
2657      case OP_MINSTAR:      case OP_MINSTAR:
2658      case OP_PLUS:      case OP_PLUS:
# Line 2092  while (code < code_end) Line 2728  while (code < code_end)
2728      case OP_CLASS:      case OP_CLASS:
2729        {        {
2730        int i, min, max;        int i, min, max;
   
2731        code++;        code++;
2732        printf("    [");        printf("    [");
2733    
# Line 2176  return (pcre *)re; Line 2811  return (pcre *)re;
2811    
2812    
2813  /*************************************************  /*************************************************
 *        Match a character type                  *  
 *************************************************/  
   
 /* Not used in all the places it might be as it's sometimes faster  
 to put the code inline.  
   
 Arguments:  
   type        the character type  
   c           the character  
   dotall      the dotall flag  
   
 Returns:      TRUE if character is of the type  
 */  
   
 static BOOL  
 match_type(int type, int c, BOOL dotall)  
 {  
   
 #ifdef DEBUG  
 if (isprint(c)) printf("matching subject %c against ", c);  
   else printf("matching subject \\x%02x against ", c);  
 printf("%s\n", OP_names[type]);  
 #endif  
   
 switch(type)  
   {  
   case OP_ANY:            return dotall || c != '\n';  
   case OP_NOT_DIGIT:      return (pcre_ctypes[c] & ctype_digit) == 0;  
   case OP_DIGIT:          return (pcre_ctypes[c] & ctype_digit) != 0;  
   case OP_NOT_WHITESPACE: return (pcre_ctypes[c] & ctype_space) == 0;  
   case OP_WHITESPACE:     return (pcre_ctypes[c] & ctype_space) != 0;  
   case OP_NOT_WORDCHAR:   return (pcre_ctypes[c] & ctype_word) == 0;  
   case OP_WORDCHAR:       return (pcre_ctypes[c] & ctype_word) != 0;  
   }  
 return FALSE;  
 }  
   
   
   
 /*************************************************  
2814  *          Match a back-reference                *  *          Match a back-reference                *
2815  *************************************************/  *************************************************/
2816    
2817  /* If a back reference hasn't been set, the match fails.  /* If a back reference hasn't been set, the length that is passed is greater
2818    than the number of characters left in the string, so the match fails.
2819    
2820  Arguments:  Arguments:
2821    number      reference number    offset      index into the offset vector
2822    eptr        points into the subject    eptr        points into the subject
2823    length      length to be matched    length      length to be matched
2824    md          points to match data block    md          points to match data block
2825      ims         the ims flags
2826    
2827  Returns:      TRUE if matched  Returns:      TRUE if matched
2828  */  */
2829    
2830  static BOOL  static BOOL
2831  match_ref(int number, register const uschar *eptr, int length, match_data *md)  match_ref(int offset, register const uschar *eptr, int length, match_data *md,
2832      int ims)
2833  {  {
2834  const uschar *p = md->start_subject + md->offset_vector[number];  const uschar *p = md->start_subject + md->offset_vector[offset];
2835    
2836  #ifdef DEBUG  #ifdef DEBUG
2837  if (eptr >= md->end_subject)  if (eptr >= md->end_subject)
# Line 2250  printf("\n"); Line 2848  printf("\n");
2848    
2849  /* Always fail if not enough characters left */  /* Always fail if not enough characters left */
2850    
2851  if (length > md->end_subject - p) return FALSE;  if (length > md->end_subject - eptr) return FALSE;
2852    
2853  /* Separate the caselesss case for speed */  /* Separate the caselesss case for speed */
2854    
2855  if (md->caseless)  if ((ims & PCRE_CASELESS) != 0)
2856    { while (length-- > 0) if (pcre_lcc[*p++] != pcre_lcc[*eptr++]) return FALSE; }    {
2857      while (length-- > 0)
2858        if (md->lcc[*p++] != md->lcc[*eptr++]) return FALSE;
2859      }
2860  else  else
2861    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }    { while (length-- > 0) if (*p++ != *eptr++) return FALSE; }
2862    
# Line 2268  return TRUE; Line 2869  return TRUE;
2869  *         Match from current position            *  *         Match from current position            *
2870  *************************************************/  *************************************************/
2871    
2872  /* On entry ecode points to the first opcode, and eptr to the first character.  /* On entry ecode points to the first opcode, and eptr to the first character
2873    in the subject string, while eptrb holds the value of eptr at the start of the
2874    last bracketed group - used for breaking infinite loops matching zero-length
2875    strings.
2876    
2877  Arguments:  Arguments:
2878     eptr        pointer in subject     eptr        pointer in subject
2879     ecode       position in code     ecode       position in code
2880     offset_top  current top pointer     offset_top  current top pointer
2881     md          pointer to "static" info for the match     md          pointer to "static" info for the match
2882       ims         current /i, /m, and /s options
2883       condassert  TRUE if called to check a condition assertion
2884       eptrb       eptr at start of last bracket
2885    
2886  Returns:       TRUE if matched  Returns:       TRUE if matched
2887  */  */
2888    
2889  static BOOL  static BOOL
2890  match(register const uschar *eptr, register const uschar *ecode, int offset_top,  match(register const uschar *eptr, register const uschar *ecode,
2891    match_data *md)    int offset_top, match_data *md, int ims, BOOL condassert, const uschar *eptrb)
2892  {  {
2893    int original_ims = ims;   /* Save for resetting on ')' */
2894    
2895  for (;;)  for (;;)
2896    {    {
2897      int op = (int)*ecode;
2898    int min, max, ctype;    int min, max, ctype;
2899    register int i;    register int i;
2900    register int c;    register int c;
2901    BOOL minimize = FALSE;    BOOL minimize = FALSE;
2902    
2903    /* Opening bracket. Check the alternative branches in turn, failing if none    /* Opening capturing bracket. If there is space in the offset vector, save
2904    match. We have to set the start offset if required and there is space    the current subject position in the working slot at the top of the vector. We
2905    in the offset vector so that it is available for subsequent back references    mustn't change the current values of the data slot, because they may be set
2906    if the bracket matches. However, if the bracket fails, we must put back the    from a previous iteration of this group, and be referred to by a reference
2907    previous value of both offsets in case they were set by a previous copy of    inside the group.
2908    the same bracket. Don't worry about setting the flag for the error case here;  
2909    that is handled in the code for KET. */    If the bracket fails to match, we need to restore this value and also the
2910      values of the final offsets, in case they were set by a previous iteration of
2911      the same bracket.
2912    
2913      If there isn't enough space in the offset vector, treat this as if it were a
2914      non-capturing bracket. Don't worry about setting the flag for the error case
2915      here; that is handled in the code for KET. */
2916    
2917    if ((int)*ecode >= OP_BRA)    if (op > OP_BRA)
2918      {      {
2919      int number = (*ecode - OP_BRA) << 1;      int number = op - OP_BRA;
2920      int save_offset1 = 0, save_offset2 = 0;      int offset = number << 1;
2921    
2922      DPRINTF(("start bracket %d\n", number/2));  #ifdef DEBUG
2923        printf("start bracket %d subject=", number);
2924        pchars(eptr, 16, TRUE, md);
2925        printf("\n");
2926    #endif
2927    
2928      if (number > 0 && number < md->offset_end)      if (offset < md->offset_max)
2929        {        {
2930        save_offset1 = md->offset_vector[number];        int save_offset1 = md->offset_vector[offset];
2931        save_offset2 = md->offset_vector[number+1];        int save_offset2 = md->offset_vector[offset+1];
2932        md->offset_vector[number] = eptr - md->start_subject;        int save_offset3 = md->offset_vector[md->offset_end - number];
2933    
2934          DPRINTF(("saving %d %d %d\n", save_offset1, save_offset2, save_offset3));
2935          md->offset_vector[md->offset_end - number] = eptr - md->start_subject;
2936    
2937          do
2938            {
2939            if (match(eptr, ecode+3, offset_top, md, ims, FALSE, eptr)) return TRUE;
2940            ecode += (ecode[1] << 8) + ecode[2];
2941            }
2942          while (*ecode == OP_ALT);
2943    
2944          DPRINTF(("bracket %d failed\n", number));
2945    
2946        DPRINTF(("saving %d %d\n", save_offset1, save_offset2));        md->offset_vector[offset] = save_offset1;
2947          md->offset_vector[offset+1] = save_offset2;
2948          md->offset_vector[md->offset_end - number] = save_offset3;
2949          return FALSE;
2950        }        }
2951    
2952      /* Recurse for all the alternatives. */      /* Insufficient room for saving captured contents */
2953    
2954        else op = OP_BRA;
2955        }
2956    
2957      /* Other types of node can be handled by a switch */
2958    
2959      switch(op)
2960        {
2961        case OP_BRA:     /* Non-capturing bracket: optimized */
2962        DPRINTF(("start bracket 0\n"));
2963      do      do
2964        {        {
2965        if (match(eptr, ecode+3, offset_top, md)) return TRUE;        if (match(eptr, ecode+3, offset_top, md, ims, FALSE, eptr)) return TRUE;
2966        ecode += (ecode[1] << 8) + ecode[2];        ecode += (ecode[1] << 8) + ecode[2];
2967        }        }
2968      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
2969        DPRINTF(("bracket 0 failed\n"));
2970        return FALSE;
2971    
2972        /* Conditional group: compilation checked that there are no more than
2973        two branches. If the condition is false, skipping the first branch takes us
2974        past the end if there is only one branch, but that's OK because that is
2975        exactly what going to the ket would do. */
2976    
2977        case OP_COND:
2978        if (ecode[3] == OP_CREF)         /* Condition is extraction test */
2979          {
2980          int offset = ecode[4] << 1;    /* Doubled reference number */
2981          return match(eptr,
2982            ecode + ((offset < offset_top && md->offset_vector[offset] >= 0)?
2983              5 : 3 + (ecode[1] << 8) + ecode[2]),
2984            offset_top, md, ims, FALSE, eptr);
2985          }
2986    
2987      DPRINTF(("bracket %d failed\n", number/2));      /* The condition is an assertion. Call match() to evaluate it - setting
2988        the final argument TRUE causes it to stop at the end of an assertion. */
2989    
2990      if (number > 0 && number < md->offset_end)      else
2991        {        {
2992        md->offset_vector[number] = save_offset1;        if (match(eptr, ecode+3, offset_top, md, ims, TRUE, NULL))
2993        md->offset_vector[number+1] = save_offset2;          {
2994            ecode += 3 + (ecode[4] << 8) + ecode[5];
2995            while (*ecode == OP_ALT) ecode += (ecode[1] << 8) + ecode[2];
2996            }
2997          else ecode += (ecode[1] << 8) + ecode[2];
2998          return match(eptr, ecode+3, offset_top, md, ims, FALSE, eptr);
2999        }        }
3000        /* Control never reaches here */
3001    
3002      return FALSE;      /* Skip over conditional reference data if encountered (should not be) */
     }  
3003    
3004    /* Other types of node can be handled by a switch */      case OP_CREF:
3005        ecode += 2;
3006        break;
3007    
3008        /* End of the pattern */
3009    
   switch(*ecode)  
     {  
3010      case OP_END:      case OP_END:
3011      md->end_match_ptr = eptr;          /* Record where we ended */      md->end_match_ptr = eptr;          /* Record where we ended */
3012      md->end_offset_top = offset_top;   /* and how many extracts were taken */      md->end_offset_top = offset_top;   /* and how many extracts were taken */
3013      return TRUE;      return TRUE;
3014    
3015      /* The equivalent of Prolog's "cut" - if the rest doesn't match, the      /* Change option settings */
     whole thing doesn't match, so we have to get out via a longjmp(). */  
3016    
3017      case OP_CUT:      case OP_OPT:
3018      if (match(eptr, ecode+1, offset_top, md)) return TRUE;      ims = ecode[1];
3019      longjmp(md->fail_env, 1);      ecode += 2;
3020        DPRINTF(("ims set to %02x\n", ims));
3021        break;
3022    
3023      /* Assertion brackets. Check the alternative branches in turn - the      /* Assertion brackets. Check the alternative branches in turn - the
3024      matching won't pass the KET for an assertion. If any one branch matches,      matching won't pass the KET for an assertion. If any one branch matches,
3025      the assertion is true. */      the assertion is true. Lookbehind assertions have an OP_REVERSE item at the
3026        start of each branch to move the current point backwards, so the code at
3027        this level is identical to the lookahead case. */
3028    
3029      case OP_ASSERT:      case OP_ASSERT:
3030        case OP_ASSERTBACK:
3031      do      do
3032        {        {
3033        if (match(eptr, ecode+3, offset_top, md)) break;        if (match(eptr, ecode+3, offset_top, md, ims, FALSE, NULL)) break;
3034        ecode += (ecode[1] << 8) + ecode[2];        ecode += (ecode[1] << 8) + ecode[2];
3035        }        }
3036      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
3037      if (*ecode == OP_KET) return FALSE;      if (*ecode == OP_KET) return FALSE;
3038    
3039        /* If checking an assertion for a condition, return TRUE. */
3040    
3041        if (condassert) return TRUE;
3042    
3043      /* Continue from after the assertion, updating the offsets high water      /* Continue from after the assertion, updating the offsets high water
3044      mark, since extracts may have been taken during the assertion. */      mark, since extracts may have been taken during the assertion. */
3045    
# Line 2374  for (;;) Line 3051  for (;;)
3051      /* Negative assertion: all branches must fail to match */      /* Negative assertion: all branches must fail to match */
3052    
3053      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
3054        case OP_ASSERTBACK_NOT:
3055      do      do
3056        {        {
3057        if (match(eptr, ecode+3, offset_top, md)) return FALSE;        if (match(eptr, ecode+3, offset_top, md, ims, FALSE, NULL)) return FALSE;
3058        ecode += (ecode[1] << 8) + ecode[2];        ecode += (ecode[1] << 8) + ecode[2];
3059        }        }
3060      while (*ecode == OP_ALT);      while (*ecode == OP_ALT);
3061    
3062        if (condassert) return TRUE;
3063      ecode += 3;      ecode += 3;
3064      continue;      continue;
3065    
3066        /* Move the subject pointer back. This occurs only at the start of
3067        each branch of a lookbehind assertion. If we are too close to the start to
3068        move back, this match function fails. */
3069    
3070        case OP_REVERSE:
3071        eptr -= (ecode[1] << 8) + ecode[2];
3072        if (eptr < md->start_subject) return FALSE;
3073        ecode += 3;
3074        break;
3075    
3076    
3077      /* "Once" brackets are like assertion brackets except that after a match,      /* "Once" brackets are like assertion brackets except that after a match,
3078      the point in the subject string is not moved back. Thus there can never be      the point in the subject string is not moved back. Thus there can never be
3079      a move back into the brackets. Check the alternative branches in turn - the      a move back into the brackets. Check the alternative branches in turn - the
3080      matching won't pass the KET for this kind of subpattern. If any one branch      matching won't pass the KET for this kind of subpattern. If any one branch
3081      matches, we carry on, leaving the subject pointer. */      matches, we carry on as at the end of a normal bracket, leaving the subject
3082        pointer. */
3083    
3084      case OP_ONCE:      case OP_ONCE:
     do  
3085        {        {
3086        if (match(eptr, ecode+3, offset_top, md)) break;        const uschar *prev = ecode;
       ecode += (ecode[1] << 8) + ecode[2];  
       }  
     while (*ecode == OP_ALT);  
     if (*ecode == OP_KET) return FALSE;  
3087    
3088      /* Continue as from after the assertion, updating the offsets high water        do
3089      mark, since extracts may have been taken. */          {
3090            if (match(eptr, ecode+3, offset_top, md, ims, FALSE, eptr)) break;
3091            ecode += (ecode[1] << 8) + ecode[2];
3092            }
3093          while (*ecode == OP_ALT);
3094    
3095      do ecode += (ecode[1] << 8) + ecode[2]; while (*ecode == OP_ALT);        /* If hit the end of the group (which could be repeated), fail */
3096      ecode += 3;  
3097      offset_top = md->end_offset_top;        if (*ecode != OP_ONCE && *ecode != OP_ALT) return FALSE;
3098      eptr = md->end_match_ptr;  
3099      continue;        /* Continue as from after the assertion, updating the offsets high water
3100          mark, since extracts may have been taken. */
3101    
3102          do ecode += (ecode[1] << 8) + ecode[2]; while (*ecode == OP_ALT);
3103    
3104          offset_top = md->end_offset_top;
3105          eptr = md->end_match_ptr;
3106    
3107          /* For a non-repeating ket, just continue at this level. This also
3108          happens for a repeating ket if no characters were matched in the group.
3109          This is the forcible breaking of infinite loops as implemented in Perl
3110          5.005. If there is an options reset, it will get obeyed in the normal
3111          course of events. */
3112    
3113          if (*ecode == OP_KET || eptr == eptrb)
3114            {
3115            ecode += 3;
3116            break;
3117            }
3118    
3119          /* The repeating kets try the rest of the pattern or restart from the
3120          preceding bracket, in the appropriate order. We need to reset any options
3121          that changed within the bracket before re-running it, so check the next
3122          opcode. */
3123    
3124          if (ecode[3] == OP_OPT)
3125            {
3126            ims = (ims & ~PCRE_IMS) | ecode[4];
3127            DPRINTF(("ims set to %02x at group repeat\n", ims));
3128            }
3129    
3130          if (*ecode == OP_KETRMIN)
3131            {
3132            if (match(eptr, ecode+3, offset_top, md, ims, FALSE, eptr) ||
3133                match(eptr, prev, offset_top, md, ims, FALSE, eptr)) return TRUE;
3134            }
3135          else  /* OP_KETRMAX */
3136            {
3137            if (match(eptr, prev, offset_top, md, ims, FALSE, eptr) ||
3138                match(eptr, ecode+3, offset_top, md, ims, FALSE, eptr)) return TRUE;
3139            }
3140          }
3141        return FALSE;
3142    
3143      /* 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
3144      bracketed group and go to there. */      bracketed group and go to there. */
# Line 2423  for (;;) Line 3156  for (;;)
3156      case OP_BRAZERO:      case OP_BRAZERO:
3157        {        {
3158        const uschar *next = ecode+1;        const uschar *next = ecode+1;
3159        if (match(eptr, next, offset_top, md)) return TRUE;        if (match(eptr, next, offset_top, md, ims, FALSE, eptr)) return TRUE;
3160        do next += (next[1] << 8) + next[2]; while (*next == OP_ALT);        do next += (next[1] << 8) + next[2]; while (*next == OP_ALT);
3161        ecode = next + 3;        ecode = next + 3;
3162        }        }
# Line 2433  for (;;) Line 3166  for (;;)
3166        {        {
3167        const uschar *next = ecode+1;        const uschar *next = ecode+1;
3168        do next += (next[1] << 8) + next[2]; while (*next == OP_ALT);        do next += (next[1] << 8) + next[2]; while (*next == OP_ALT);
3169        if (match(eptr, next+3, offset_top, md)) return TRUE;        if (match(eptr, next+3, offset_top, md, ims, FALSE, eptr)) return TRUE;
3170        ecode++;        ecode++;
3171        }        }
3172      break;;      break;
3173    
3174      /* End of a group, repeated or non-repeating. If we are at the end of      /* End of a group, repeated or non-repeating. If we are at the end of
3175      an assertion "group", stop matching and return TRUE, but record the      an assertion "group", stop matching and return TRUE, but record the
3176      current high water mark for use by positive assertions. */      current high water mark for use by positive assertions. Do this also
3177        for the "once" (not-backup up) groups. */
3178    
3179      case OP_KET:      case OP_KET:
3180      case OP_KETRMIN:      case OP_KETRMIN:
3181      case OP_KETRMAX:      case OP_KETRMAX:
3182        {        {
       int number;  
3183        const uschar *prev = ecode - (ecode[1] << 8) - ecode[2];        const uschar *prev = ecode - (ecode[1] << 8) - ecode[2];
3184    
3185        if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT || *prev == OP_ONCE)        if (*prev == OP_ASSERT || *prev == OP_ASSERT_NOT ||
3186              *prev == OP_ASSERTBACK || *prev == OP_ASSERTBACK_NOT ||
3187              *prev == OP_ONCE)
3188          {          {
3189          md->end_match_ptr = eptr;      /* For ONCE */          md->end_match_ptr = eptr;      /* For ONCE */
3190          md->end_offset_top = offset_top;          md->end_offset_top = offset_top;
3191          return TRUE;          return TRUE;
3192          }          }
3193    
3194        /* In all other cases we have to check the group number back at the        /* In all other cases except a conditional group we have to check the
3195        start and if necessary complete handling an extraction by setting the        group number back at the start and if necessary complete handling an
3196        final offset and bumping the high water mark. */        extraction by setting the offsets and bumping the high water mark. */
3197    
3198        number = (*prev - OP_BRA) << 1;        if (*prev != OP_COND)
3199            {
3200            int number = *prev - OP_BRA;
3201            int offset = number << 1;
3202    
3203        DPRINTF(("end bracket %d\n", number/2));          DPRINTF(("end bracket %d\n", number));
3204    
3205        if (number > 0)          if (number > 0)
         {  
         if (number >= md->offset_end) md->offset_overflow = TRUE; else  
3206            {            {
3207            md->offset_vector[number+1] = eptr - md->start_subject;            if (offset >= md->offset_max) md->offset_overflow = TRUE; else
3208            if (offset_top <= number) offset_top = number + 2;              {
3209                md->offset_vector[offset] =
3210                  md->offset_vector[md->offset_end - number];
3211                md->offset_vector[offset+1] = eptr - md->start_subject;
3212                if (offset_top <= offset) offset_top = offset + 2;
3213                }
3214            }            }
3215          }          }
3216    
3217        /* For a non-repeating ket, just advance to the next node and continue at        /* Reset the value of the ims flags, in case they got changed during
3218        this level. */        the group. */
3219    
3220        if (*ecode == OP_KET)        ims = original_ims;
3221          DPRINTF(("ims reset to %02x\n", ims));
3222    
3223          /* For a non-repeating ket, just continue at this level. This also
3224          happens for a repeating ket if no characters were matched in the group.
3225          This is the forcible breaking of infinite loops as implemented in Perl
3226          5.005. If there is an options reset, it will get obeyed in the normal
3227          course of events. */
3228    
3229          if (*ecode == OP_KET || eptr == eptrb)
3230          {          {
3231          ecode += 3;          ecode += 3;
3232          break;          break;
# Line 2487  for (;;) Line 3237  for (;;)
3237    
3238        if (*ecode == OP_KETRMIN)        if (*ecode == OP_KETRMIN)
3239          {          {
3240          if (match(eptr, ecode+3, offset_top, md) ||          if (match(eptr, ecode+3, offset_top, md, ims, FALSE, eptr) ||
3241              match(eptr, prev, offset_top, md)) return TRUE;              match(eptr, prev, offset_top, md, ims, FALSE, eptr)) return TRUE;
3242          }          }
3243        else  /* OP_KETRMAX */        else  /* OP_KETRMAX */
3244          {          {
3245          if (match(eptr, prev, offset_top, md) ||          if (match(eptr, prev, offset_top, md, ims, FALSE, eptr) ||
3246              match(eptr, ecode+3, offset_top, md)) return TRUE;              match(eptr, ecode+3, offset_top, md, ims, FALSE, eptr)) return TRUE;
3247          }          }
3248        }        }
3249      return FALSE;      return FALSE;
# Line 2502  for (;;) Line 3252  for (;;)
3252    
3253      case OP_CIRC:      case OP_CIRC:
3254      if (md->notbol && eptr == md->start_subject) return FALSE;      if (md->notbol && eptr == md->start_subject) return FALSE;
3255      if (md->multiline)      if ((ims & PCRE_MULTILINE) != 0)
3256        {        {
3257        if (eptr != md->start_subject && eptr[-1] != '\n') return FALSE;        if (eptr != md->start_subject && eptr[-1] != '\n') return FALSE;
3258        ecode++;        ecode++;
# Line 2517  for (;;) Line 3267  for (;;)
3267      ecode++;      ecode++;
3268      break;      break;
3269    
3270      /* Assert before internal newline if multiline, or before      /* Assert before internal newline if multiline, or before a terminating
3271      a terminating newline unless endonly is set, else end of subject unless      newline unless endonly is set, else end of subject unless noteol is set. */
     noteol is set. */  
3272    
3273      case OP_DOLL:      case OP_DOLL:
3274      if (md->noteol && eptr >= md->end_subject) return FALSE;      if ((ims & PCRE_MULTILINE) != 0)
     if (md->multiline)  
3275        {        {
3276        if (eptr < md->end_subject && *eptr != '\n') return FALSE;        if (eptr < md->end_subject) { if (*eptr != '\n') return FALSE; }
3277            else { if (md->noteol) return FALSE; }
3278        ecode++;        ecode++;
3279        break;        break;
3280        }        }
3281      else if (!md->endonly)      else
3282        {        {
3283        if (eptr < md->end_subject - 1 ||        if (md->noteol) return FALSE;
3284           (eptr == md->end_subject - 1 && *eptr != '\n')) return FALSE;        if (!md->endonly)
3285        ecode++;          {
3286        break;          if (eptr < md->end_subject - 1 ||
3287               (eptr == md->end_subject - 1 && *eptr != '\n')) return FALSE;
3288    
3289            ecode++;
3290            break;
3291            }
3292        }        }
3293      /* ... else fall through */      /* ... else fall through */
3294    
3295      /* End of subject assertion */      /* End of subject assertion (\z) */
3296    
3297      case OP_EOD:      case OP_EOD:
3298      if (eptr < md->end_subject) return FALSE;      if (eptr < md->end_subject) return FALSE;
3299      ecode++;      ecode++;
3300      break;      break;
3301    
3302        /* End of subject or ending \n assertion (\Z) */
3303    
3304        case OP_EODN:
3305        if (eptr < md->end_subject - 1 ||
3306           (eptr == md->end_subject - 1 && *eptr != '\n')) return FALSE;
3307        ecode++;
3308        break;
3309    
3310      /* Word boundary assertions */      /* Word boundary assertions */
3311    
3312      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
3313      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
3314        {        {
3315        BOOL prev_is_word = (eptr != md->start_subject) &&        BOOL prev_is_word = (eptr != md->start_subject) &&
3316          ((pcre_ctypes[eptr[-1]] & ctype_word) != 0);          ((md->ctypes[eptr[-1]] & ctype_word) != 0);
3317        BOOL cur_is_word = (eptr < md->end_subject) &&        BOOL cur_is_word = (eptr < md->end_subject) &&
3318          ((pcre_ctypes[*eptr] & ctype_word) != 0);          ((md->ctypes[*eptr] & ctype_word) != 0);
3319        if ((*ecode++ == OP_WORD_BOUNDARY)?        if ((*ecode++ == OP_WORD_BOUNDARY)?
3320             cur_is_word == prev_is_word : cur_is_word != prev_is_word)             cur_is_word == prev_is_word : cur_is_word != prev_is_word)
3321          return FALSE;          return FALSE;
# Line 2563  for (;;) Line 3325  for (;;)
3325      /* Match a single character type; inline for speed */      /* Match a single character type; inline for speed */
3326    
3327      case OP_ANY:      case OP_ANY:
3328      if (!md->dotall && eptr < md->end_subject && *eptr == '\n') return FALSE;      if ((ims & PCRE_DOTALL) == 0 && eptr < md->end_subject && *eptr == '\n')
3329          return FALSE;
3330      if (eptr++ >= md->end_subject) return FALSE;      if (eptr++ >= md->end_subject) return FALSE;
3331      ecode++;      ecode++;
3332      break;      break;
3333    
3334      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
3335      if (eptr >= md->end_subject || (pcre_ctypes[*eptr++] & ctype_digit) != 0)      if (eptr >= md->end_subject ||
3336           (md->ctypes[*eptr++] & ctype_digit) != 0)
3337        return FALSE;        return FALSE;
3338      ecode++;      ecode++;
3339      break;      break;
3340    
3341      case OP_DIGIT:      case OP_DIGIT:
3342      if (eptr >= md->end_subject || (pcre_ctypes[*eptr++] & ctype_digit) == 0)      if (eptr >= md->end_subject ||
3343           (md->ctypes[*eptr++] & ctype_digit) == 0)
3344        return FALSE;        return FALSE;
3345      ecode++;      ecode++;
3346      break;      break;
3347    
3348      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
3349      if (eptr >= md->end_subject || (pcre_ctypes[*eptr++] & ctype_space) != 0)      if (eptr >= md->end_subject ||
3350           (md->ctypes[*eptr++] & ctype_space) != 0)
3351        return FALSE;        return FALSE;
3352      ecode++;      ecode++;
3353      break;      break;
3354    
3355      case OP_WHITESPACE:      case OP_WHITESPACE:
3356      if (eptr >= md->end_subject || (pcre_ctypes[*eptr++] & ctype_space) == 0)      if (eptr >= md->end_subject ||
3357           (md->ctypes[*eptr++] & ctype_space) == 0)
3358        return FALSE;        return FALSE;
3359      ecode++;      ecode++;
3360      break;      break;
3361    
3362      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
3363      if (eptr >= md->end_subject || (pcre_ctypes[*eptr++] & ctype_word) != 0)      if (eptr >= md->end_subject ||
3364           (md->ctypes[*eptr++] & ctype_word) != 0)
3365        return FALSE;        return FALSE;
3366      ecode++;      ecode++;
3367      break;      break;
3368    
3369      case OP_WORDCHAR:      case OP_WORDCHAR:
3370      if (eptr >= md->end_subject || (pcre_ctypes[*eptr++] & ctype_word) == 0)      if (eptr >= md->end_subject ||
3371           (md->ctypes[*eptr++] & ctype_word) == 0)
3372        return FALSE;        return FALSE;
3373      ecode++;      ecode++;
3374      break;      break;
# Line 2615  for (;;) Line 3384  for (;;)
3384      case OP_REF:      case OP_REF:
3385        {        {
3386        int length;        int length;
3387        int number = ecode[1] << 1;                /* Doubled reference number */        int offset = ecode[1] << 1;                /* Doubled reference number */
3388        ecode += 2;                                /* Advance past the item */        ecode += 2;                                /* Advance past the item */
3389    
3390        if (number >= offset_top || md->offset_vector[number] < 0)        /* If the reference is unset, set the length to be longer than the amount
3391          {        of subject left; this ensures that every attempt at a match fails. We
3392          md->errorcode = PCRE_ERROR_BADREF;        can't just fail here, because of the possibility of quantifiers with zero
3393          return FALSE;        minima. */
3394          }  
3395          length = (offset >= offset_top || md->offset_vector[offset] < 0)?
3396            md->end_subject - eptr + 1 :
3397            md->offset_vector[offset+1] - md->offset_vector[offset];
3398    
3399        length = md->offset_vector[number+1] - md->offset_vector[number];        /* Set up for repetition, or handle the non-repeated case */
3400    
3401        switch (*ecode)        switch (*ecode)
3402          {          {
# Line 2651  for (;;) Line 3423  for (;;)
3423          break;          break;
3424    
3425          default:               /* No repeat follows */          default:               /* No repeat follows */
3426          if (!match_ref(number, eptr, length, md)) return FALSE;          if (!match_ref(offset, eptr, length, md, ims)) return FALSE;
3427          eptr += length;          eptr += length;
3428          continue;              /* With the main loop */          continue;              /* With the main loop */
3429          }          }
# Line 2667  for (;;) Line 3439  for (;;)
3439    
3440        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3441          {          {
3442          if (!match_ref(number, eptr, length, md)) return FALSE;          if (!match_ref(offset, eptr, length, md, ims)) return FALSE;
3443          eptr += length;          eptr += length;
3444          }          }
3445    
# Line 2682  for (;;) Line 3454  for (;;)
3454          {          {
3455          for (i = min;; i++)          for (i = min;; i++)
3456            {            {
3457            if (match(eptr, ecode, offset_top, md)) return TRUE;            if (match(eptr, ecode, offset_top, md, ims, FALSE, eptrb))
3458            if (i >= max || !match_ref(number, eptr, length, md))              return TRUE;
3459              if (i >= max || !match_ref(offset, eptr, length, md, ims))
3460              return FALSE;              return FALSE;
3461            eptr += length;            eptr += length;
3462            }            }
# Line 2697  for (;;) Line 3470  for (;;)
3470          const uschar *pp = eptr;          const uschar *pp = eptr;
3471          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3472            {            {
3473            if (!match_ref(number, eptr, length, md)) break;            if (!match_ref(offset, eptr, length, md, ims)) break;
3474            eptr += length;            eptr += length;
3475            }            }
3476          while (eptr >= pp)          while (eptr >= pp)
3477            {            {
3478            if (match(eptr, ecode, offset_top, md)) return TRUE;            if (match(eptr, ecode, offset_top, md, ims, FALSE, eptrb))
3479                return TRUE;
3480            eptr -= length;            eptr -= length;
3481            }            }
3482          return FALSE;          return FALSE;
# Line 2710  for (;;) Line 3484  for (;;)
3484        }        }
3485      /* Control never gets here */      /* Control never gets here */
3486    
3487    
3488    
3489      /* Match a character class, possibly repeatedly. Look past the end of the      /* Match a character class, possibly repeatedly. Look past the end of the
3490      item to see if there is repeat information following. Then obey similar      item to see if there is repeat information following. Then obey similar
3491      code to character type repeats - written out again for speed. If caseless      code to character type repeats - written out again for speed. */
     matching was set at runtime but not at compile time, we have to check both  
     versions of a character. */  
3492    
3493      case OP_CLASS:      case OP_CLASS:
3494        {        {
# Line 2746  for (;;) Line 3520  for (;;)
3520          break;          break;
3521    
3522          default:               /* No repeat follows */          default:               /* No repeat follows */
3523          if (eptr >= md->end_subject) return FALSE;          min = max = 1;
3524          c = *eptr++;          break;
         if ((data[c/8] & (1 << (c&7))) != 0) continue;    /* With main loop */  
         if (md->runtime_caseless)  
           {  
           c = pcre_fcc[c];  
           if ((data[c/8] & (1 << (c&7))) != 0) continue;  /* With main loop */  
           }  
         return FALSE;  
3525          }          }
3526    
3527        /* First, ensure the minimum number of matches are present. */        /* First, ensure the minimum number of matches are present. */
# Line 2764  for (;;) Line 3531  for (;;)
3531          if (eptr >= md->end_subject) return FALSE;          if (eptr >= md->end_subject) return FALSE;
3532          c = *eptr++;          c = *eptr++;
3533          if ((data[c/8] & (1 << (c&7))) != 0) continue;          if ((data[c/8] & (1 << (c&7))) != 0) continue;
         if (md->runtime_caseless)  
           {  
           c = pcre_fcc[c];  
           if ((data[c/8] & (1 << (c&7))) != 0) continue;  
           }  
3534          return FALSE;          return FALSE;
3535          }          }
3536    
# Line 2784  for (;;) Line 3546  for (;;)
3546          {          {
3547          for (i = min;; i++)          for (i = min;; i++)
3548            {            {
3549            if (match(eptr, ecode, offset_top, md)) return TRUE;            if (match(eptr, ecode, offset_top, md, ims, FALSE, eptrb))
3550                return TRUE;
3551            if (i >= max || eptr >= md->end_subject) return FALSE;            if (i >= max || eptr >= md->end_subject) return FALSE;
3552            c = *eptr++;            c = *eptr++;
3553            if ((data[c/8] & (1 << (c&7))) != 0) continue;            if ((data[c/8] & (1 << (c&7))) != 0) continue;
           if (md->runtime_caseless)  
             {  
             c = pcre_fcc[c];  
             if ((data[c/8] & (1 << (c&7))) != 0) continue;  
             }  
3554            return FALSE;            return FALSE;
3555            }            }
3556          /* Control never gets here */          /* Control never gets here */
# Line 2808  for (;;) Line 3566  for (;;)
3566            if (eptr >= md->end_subject) break;            if (eptr >= md->end_subject) break;
3567            c = *eptr;            c = *eptr;
3568            if ((data[c/8] & (1 << (c&7))) != 0) continue;            if ((data[c/8] & (1 << (c&7))) != 0) continue;
           if (md->runtime_caseless)  
             {  
             c = pcre_fcc[c];  
             if ((data[c/8] & (1 << (c&7))) != 0) continue;  
             }  
3569            break;            break;
3570            }            }
3571    
3572          while (eptr >= pp)          while (eptr >= pp)
3573            if (match(eptr--, ecode, offset_top, md)) return TRUE;            if (match(eptr--, ecode, offset_top, md, ims, FALSE, eptrb))
3574                return TRUE;
3575          return FALSE;          return FALSE;
3576          }          }
3577        }        }
# Line 2844  for (;;) Line 3598  for (;;)
3598  #endif  #endif
3599    
3600        if (length > md->end_subject - eptr) return FALSE;        if (length > md->end_subject - eptr) return FALSE;
3601        if (md->caseless)        if ((ims & PCRE_CASELESS) != 0)
3602          {          {
3603          while (length-- > 0) if (pcre_lcc[*ecode++] != pcre_lcc[*eptr++]) return FALSE;          while (length-- > 0)
3604              if (md->lcc[*ecode++] != md->lcc[*eptr++])
3605                return FALSE;
3606          }          }
3607        else        else
3608          {          {
# Line 2901  for (;;) Line 3657  for (;;)
3657      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", c, min, max,      DPRINTF(("matching %c{%d,%d} against subject %.*s\n", c, min, max,
3658        max, eptr));        max, eptr));
3659    
3660      if (md->caseless)      if ((ims & PCRE_CASELESS) != 0)
3661        {        {
3662        c = pcre_lcc[c];        c = md->lcc[c];
3663        for (i = 1; i <= min; i++) if (c != pcre_lcc[*eptr++]) return FALSE;        for (i = 1; i <= min; i++)
3664            if (c != md->lcc[*eptr++]) return FALSE;
3665        if (min == max) continue;        if (min == max) continue;
3666        if (minimize)        if (minimize)
3667          {          {
3668          for (i = min;; i++)          for (i = min;; i++)
3669            {            {
3670            if (match(eptr, ecode, offset_top, md)) return TRUE;            if (match(eptr, ecode, offset_top, md, ims, FALSE, eptrb))
3671            if (i >= max || eptr >= md->end_subject || c != pcre_lcc[*eptr++])              return TRUE;
3672              if (i >= max || eptr >= md->end_subject ||
3673                  c != md->lcc[*eptr++])
3674              return FALSE;              return FALSE;
3675            }            }
3676          /* Control never gets here */          /* Control never gets here */
# Line 2921  for (;;) Line 3680  for (;;)
3680          const uschar *pp = eptr;          const uschar *pp = eptr;
3681          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3682            {            {
3683            if (eptr >= md->end_subject || c != pcre_lcc[*eptr]) break;            if (eptr >= md->end_subject || c != md->lcc[*eptr]) break;
3684            eptr++;            eptr++;
3685            }            }
3686          while (eptr >= pp)          while (eptr >= pp)
3687            if (match(eptr--, ecode, offset_top, md)) return TRUE;            if (match(eptr--, ecode, offset_top, md, ims, FALSE, eptrb))
3688                return TRUE;
3689          return FALSE;          return FALSE;
3690          }          }
3691        /* Control never gets here */        /* Control never gets here */
# Line 2941  for (;;) Line 3701  for (;;)
3701          {          {
3702          for (i = min;; i++)          for (i = min;; i++)
3703            {            {
3704            if (match(eptr, ecode, offset_top, md)) return TRUE;            if (match(eptr, ecode, offset_top, md, ims, FALSE, eptrb))
3705                return TRUE;
3706            if (i >= max || eptr >= md->end_subject || c != *eptr++) return FALSE;            if (i >= max || eptr >= md->end_subject || c != *eptr++) return FALSE;
3707            }            }
3708          /* Control never gets here */          /* Control never gets here */
# Line 2955  for (;;) Line 3716  for (;;)
3716            eptr++;            eptr++;
3717            }            }
3718          while (eptr >= pp)          while (eptr >= pp)
3719           if (match(eptr--, ecode, offset_top, md)) return TRUE;           if (match(eptr--, ecode, offset_top, md, ims, FALSE, eptrb))
3720               return TRUE;
3721          return FALSE;          return FALSE;
3722          }          }
3723        }        }
# Line 2966  for (;;) Line 3728  for (;;)
3728      case OP_NOT:      case OP_NOT:
3729      if (eptr >= md->end_subject) return FALSE;      if (eptr >= md->end_subject) return FALSE;
3730      ecode++;      ecode++;
3731      if (md->caseless)      if ((ims & PCRE_CASELESS) != 0)
3732        {        {
3733        if (pcre_lcc[*ecode++] == pcre_lcc[*eptr++]) return FALSE;        if (md->lcc[*ecode++] == md->lcc[*eptr++]) return FALSE;
3734        }        }
3735      else      else
3736        {        {
# Line 3026  for (;;) Line 3788  for (;;)
3788      DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", c, min, max,      DPRINTF(("negative matching %c{%d,%d} against subject %.*s\n", c, min, max,
3789        max, eptr));        max, eptr));
3790    
3791      if (md->caseless)      if ((ims & PCRE_CASELESS) != 0)
3792        {        {
3793        c = pcre_lcc[c];        c = md->lcc[c];
3794        for (i = 1; i <= min; i++) if (c == pcre_lcc[*eptr++]) return FALSE;        for (i = 1; i <= min; i++)
3795            if (c == md->lcc[*eptr++]) return FALSE;
3796        if (min == max) continue;        if (min == max) continue;
3797        if (minimize)        if (minimize)
3798          {          {
3799          for (i = min;; i++)          for (i = min;; i++)
3800            {            {
3801            if (match(eptr, ecode, offset_top, md)) return TRUE;            if (match(eptr, ecode, offset_top, md, ims, FALSE, eptrb))
3802            if (i >= max || eptr >= md->end_subject || c == pcre_lcc[*eptr++])              return TRUE;
3803              if (i >= max || eptr >= md->end_subject ||
3804                  c == md->lcc[*eptr++])
3805              return FALSE;              return FALSE;
3806            }            }
3807          /* Control never gets here */          /* Control never gets here */
# Line 3046  for (;;) Line 3811  for (;;)
3811          const uschar *pp = eptr;          const uschar *pp = eptr;
3812          for (i = min; i < max; i++)          for (i = min; i < max; i++)
3813            {            {
3814            if (eptr >= md->end_subject || c == pcre_lcc[*eptr]) break;            if (eptr >= md->end_subject || c == md->lcc[*eptr]) break;
3815            eptr++;            eptr++;
3816            }            }
3817          while (eptr >= pp)          while (eptr >= pp)
3818            if (match(eptr--, ecode, offset_top, md)) return TRUE;            if (match(eptr--, ecode, offset_top, md, ims, FALSE, eptrb))
3819                return TRUE;
3820          return FALSE;          return FALSE;
3821          }          }
3822        /* Control never gets here */        /* Control never gets here */
# Line 3066  for (;;) Line 3832  for (;;)
3832          {          {
3833          for (i = min;; i++)          for (i = min;; i++)
3834            {            {
3835            if (match(eptr, ecode, offset_top, md)) return TRUE;            if (match(eptr, ecode, offset_top, md, ims, FALSE, eptrb))
3836                return TRUE;
3837            if (i >= max || eptr >= md->end_subject || c == *eptr++) return FALSE;            if (i >= max || eptr >= md->end_subject || c == *eptr++) return FALSE;
3838            }            }
3839          /* Control never gets here */          /* Control never gets here */
# Line 3080  for (;;) Line 3847  for (;;)
3847            eptr++;            eptr++;
3848            }            }
3849          while (eptr >= pp)          while (eptr >= pp)
3850           if (match(eptr--, ecode, offset_top, md)) return TRUE;           if (match(eptr--, ecode, offset_top, md, ims, FALSE, eptrb))
3851               return TRUE;
3852          return FALSE;          return FALSE;
3853          }          }
3854        }        }
# Line 3130  for (;;) Line 3898  for (;;)
3898      if (min > 0) switch(ctype)      if (min > 0) switch(ctype)
3899        {        {
3900        case OP_ANY:        case OP_ANY:
3901        if (!md->dotall)        if ((ims & PCRE_DOTALL) == 0)
3902          { for (i = 1; i <= min; i++) if (*eptr++ == '\n') return FALSE; }          { for (i = 1; i <= min; i++) if (*eptr++ == '\n') return FALSE; }
3903        else eptr += min;        else eptr += min;
3904        break;        break;
3905    
3906        case OP_NOT_DIGIT:        case OP_NOT_DIGIT:
3907        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3908          if ((pcre_ctypes[*eptr++] & ctype_digit) != 0) return FALSE;          if ((md->ctypes[*eptr++] & ctype_digit) != 0) return FALSE;
3909        break;        break;
3910    
3911        case OP_DIGIT:        case OP_DIGIT:
3912        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3913          if ((pcre_ctypes[*eptr++] & ctype_digit) == 0) return FALSE;          if ((md->ctypes[*eptr++] & ctype_digit) == 0) return FALSE;
3914        break;        break;
3915    
3916        case OP_NOT_WHITESPACE:        case OP_NOT_WHITESPACE:
3917        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3918          if ((pcre_ctypes[*eptr++] & ctype_space) != 0) return FALSE;          if ((md->ctypes[*eptr++] & ctype_space) != 0) return FALSE;
3919        break;        break;
3920    
3921        case OP_WHITESPACE:        case OP_WHITESPACE:
3922        for (i = 1; i <= min; i++)        for (i = 1; i <= min; i++)
3923          if ((pcre_ctypes[*eptr++] & ctype_space) == 0) return FALSE;          if ((md->ctypes[*eptr++] & ctype_space) == 0) return FALSE;
3924        break;        break;
3925    
3926        case OP_NOT_WORDCHAR:        case OP_NOT_WORDCHAR:
3927        for (i = 1; i <= min; i++) if ((pcre_ctypes[*eptr++] & ctype_word) != 0)        for (i = 1; i <= min; i++)
3928          return FALSE;          if ((md->ctypes[*eptr++] & ctype_word) != 0)
3929              return FALSE;
3930        break;        break;
3931    
3932        case OP_WORDCHAR:        case OP_WORDCHAR:
3933        for (i = 1; i <= min; i++) if ((pcre_ctypes[*eptr++] & ctype_word) == 0)        for (i = 1; i <= min; i++)
3934          return FALSE;          if ((md->ctypes[*eptr++] & ctype_word) == 0)
3935              return FALSE;
3936        break;        break;
3937        }        }
3938    
# Line 3171  for (;;) Line 3941  for (;;)
3941      if (min == max) continue;      if (min == max) continue;
3942    
3943      /* If minimizing, we have to test the rest of the pattern before each      /* If minimizing, we have to test the rest of the pattern before each
3944      subsequent match, so inlining isn't much help; just use the function. */      subsequent match. */
3945    
3946      if (minimize)      if (minimize)
3947        {        {
3948        for (i = min;; i++)        for (i = min;; i++)
3949          {          {
3950          if (match(eptr, ecode, offset_top, md)) return TRUE;          if (match(eptr, ecode, offset_top, md, ims, FALSE, eptrb)) return TRUE;
3951          if (i >= max || eptr >= md->end_subject ||          if (i >= max || eptr >= md->end_subject) return FALSE;
3952            !match_type(ctype, *eptr++, md->dotall))  
3953              return FALSE;          c = *eptr++;
3954            switch(ctype)
3955              {
3956              case OP_ANY:
3957              if ((ims & PCRE_DOTALL) == 0 && c == '\n') return FALSE;
3958              break;
3959    
3960              case OP_NOT_DIGIT:
3961              if ((md->ctypes[c] & ctype_digit) != 0) return FALSE;
3962              break;
3963    
3964              case OP_DIGIT:
3965              if ((md->ctypes[c] & ctype_digit) == 0) return FALSE;
3966              break;
3967    
3968              case OP_NOT_WHITESPACE:
3969              if ((md->ctypes[c] & ctype_space) != 0) return FALSE;
3970              break;
3971    
3972              case OP_WHITESPACE:
3973              if  ((md->ctypes[c] & ctype_space) == 0) return FALSE;
3974              break;
3975    
3976              case OP_NOT_WORDCHAR:
3977              if ((md->ctypes[c] & ctype_word) != 0) return FALSE;
3978              break;
3979    
3980              case OP_WORDCHAR:
3981              if ((md->ctypes[c] & ctype_word) == 0) return FALSE;
3982              break;
3983              }
3984          }          }
3985        /* Control never gets here */        /* Control never gets here */
3986        }        }
# Line 3194  for (;;) Line 3994  for (;;)
3994        switch(ctype)        switch(ctype)
3995          {          {
3996          case OP_ANY:          case OP_ANY:
3997          if (!md->dotall)          if ((ims & PCRE_DOTALL) == 0)
3998            {            {
3999            for (i = min; i < max; i++)            for (i = min; i < max; i++)
4000              {              {
# Line 3213  for (;;) Line 4013  for (;;)
4013          case OP_NOT_DIGIT:          case OP_NOT_DIGIT:
4014          for (i = min; i < max; i++)          for (i = min; i < max; i++)
4015            {            {
4016            if (eptr >= md->end_subject || (pcre_ctypes[*eptr] & ctype_digit) != 0)            if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_digit) != 0)
4017              break;              break;
4018            eptr++;            eptr++;
4019            }            }
# Line 3222  for (;;) Line 4022  for (;;)
4022          case OP_DIGIT:          case OP_DIGIT:
4023          for (i = min; i < max; i++)          for (i = min; i < max; i++)
4024            {            {
4025            if (eptr >= md->end_subject || (pcre_ctypes[*eptr] & ctype_digit) == 0)            if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_digit) == 0)
4026              break;              break;
4027            eptr++;            eptr++;
4028            }            }
# Line 3231  for (;;) Line 4031  for (;;)
4031          case OP_NOT_WHITESPACE:          case OP_NOT_WHITESPACE:
4032          for (i = min; i < max; i++)          for (i = min; i < max; i++)
4033            {            {
4034            if (eptr >= md->end_subject || (pcre_ctypes[*eptr] & ctype_space) != 0)            if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_space) != 0)
4035              break;              break;
4036            eptr++;            eptr++;
4037            }            }
# Line 3240  for (;;) Line 4040  for (;;)
4040          case OP_WHITESPACE:          case OP_WHITESPACE:
4041          for (i = min; i < max; i++)          for (i = min; i < max; i++)
4042            {            {
4043            if (eptr >= md->end_subject || (pcre_ctypes[*eptr] & ctype_space) == 0)            if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_space) == 0)
4044              break;              break;
4045            eptr++;            eptr++;
4046            }            }
# Line 3249  for (;;) Line 4049  for (;;)
4049          case OP_NOT_WORDCHAR:          case OP_NOT_WORDCHAR:
4050          for (i = min; i < max; i++)          for (i = min; i < max; i++)
4051            {            {
4052            if (eptr >= md->end_subject || (pcre_ctypes[*eptr] & ctype_word) != 0)            if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_word) != 0)
4053              break;              break;
4054            eptr++;            eptr++;
4055            }            }
# Line 3258  for (;;) Line 4058  for (;;)
4058          case OP_WORDCHAR:          case OP_WORDCHAR:
4059          for (i = min; i < max; i++)          for (i = min; i < max; i++)
4060            {            {
4061            if (eptr >= md->end_subject || (pcre_ctypes[*eptr] & ctype_word) == 0)            if (eptr >= md->end_subject || (md->ctypes[*eptr] & ctype_word) == 0)
4062              break;              break;
4063            eptr++;            eptr++;
4064            }            }
# Line 3266  for (;;) Line 4066  for (;;)
4066          }          }
4067    
4068        while (eptr >= pp)        while (eptr >= pp)
4069          if (match(eptr--, ecode, offset_top, md)) return TRUE;          if (match(eptr--, ecode, offset_top, md, ims, FALSE, eptrb))
4070              return TRUE;
4071        return FALSE;        return FALSE;
4072        }        }
4073      /* Control never gets here */      /* Control never gets here */
# Line 3289  for (;;) Line 4090  for (;;)
4090    
4091    
4092    
 /*************************************************  
 *         Segregate setjmp()                     *  
 *************************************************/  
   
 /* The -Wall option of gcc gives warnings for all local variables when setjmp()  
 is used, even if the coding conforms to the rules of ANSI C. To avoid this, we  
 hide it in a separate function. This is called only when PCRE_EXTRA is set,  
 since it's needed only for the extension \X option, and with any luck, a good  
 compiler will spot the tail recursion and compile it efficiently.  
   
 Arguments:  
    eptr        pointer in subject  
    ecode       position in code  
    offset_top  current top pointer  
    md          pointer to "static" info for the match  
   
 Returns:       TRUE if matched  
 */  
   
 static BOOL  
 match_with_setjmp(const uschar *eptr, const uschar *ecode, int offset_top,  
   match_data *match_block)  
 {  
 return setjmp(match_block->fail_env) == 0 &&  
       match(eptr, ecode, offset_top, match_block);  
 }  
   
   
4093    
4094  /*************************************************  /*************************************************
4095  *         Execute a Regular Expression           *  *         Execute a Regular Expression           *
# Line 3347  pcre_exec(const pcre *external_re, const Line 4120  pcre_exec(const pcre *external_re, const
4120  {  {
4121  int resetcount, ocount;  int resetcount, ocount;
4122  int first_char = -1;  int first_char = -1;
4123    int ims = 0;
4124  match_data match_block;  match_data match_block;
4125  const uschar *start_bits = NULL;  const uschar *start_bits = NULL;
4126  const uschar *start_match = (const uschar *)subject;  const uschar *start_match = (const uschar *)subject;
# Line 3367  match_block.start_subject = (const uscha Line 4141  match_block.start_subject = (const uscha
4141  match_block.end_subject = match_block.start_subject + length;  match_block.end_subject = match_block.start_subject + length;
4142  end_subject = match_block.end_subject;  end_subject = match_block.end_subject;
4143    
4144  match_block.caseless  = ((re->options | options) & PCRE_CASELESS) != 0;  match_block.endonly = (re->options & PCRE_DOLLAR_ENDONLY) != 0;
 match_block.runtime_caseless = match_block.caseless &&  
   (re->options & PCRE_CASELESS) == 0;  
   
 match_block.multiline = ((re->options | options) & PCRE_MULTILINE) != 0;  
 match_block.dotall    = ((re->options | options) & PCRE_DOTALL) != 0;  
 match_block.endonly   = ((re->options | options) & PCRE_DOLLAR_ENDONLY) != 0;  
4145    
4146  match_block.notbol = (options & PCRE_NOTBOL) != 0;  match_block.notbol = (options & PCRE_NOTBOL) != 0;
4147  match_block.noteol = (options & PCRE_NOTEOL) != 0;  match_block.noteol = (options & PCRE_NOTEOL) != 0;
4148    
4149  match_block.errorcode = PCRE_ERROR_NOMATCH;     /* Default error */  match_block.errorcode = PCRE_ERROR_NOMATCH;     /* Default error */
4150    
4151    match_block.lcc = re->tables + lcc_offset;
4152    match_block.ctypes = re->tables + ctypes_offset;
4153    
4154    /* The ims options can vary during the matching as a result of the presence
4155    of (?ims) items in the pattern. They are kept in a local variable so that
4156    restoring at the exit of a group is easy. */
4157    
4158    ims = re->options & (PCRE_CASELESS|PCRE_MULTILINE|PCRE_DOTALL);
4159    
4160  /* If the expression has got more back references than the offsets supplied can  /* If the expression has got more back references than the offsets supplied can
4161  hold, we get a temporary bit of working store to use during the matching.  hold, we get a temporary bit of working store to use during the matching.
4162  Otherwise, we can use the vector supplied, rounding down its size to a multiple  Otherwise, we can use the vector supplied, rounding down its size to a multiple
4163  of 2. */  of 3. */
4164    
4165  ocount = offsetcount & (-2);  ocount = offsetcount - (offsetcount % 3);
4166  if (re->top_backref > 0 && re->top_backref >= ocount/2)  
4167    if (re->top_backref > 0 && re->top_backref >= ocount/3)
4168    {    {
4169    ocount = re->top_backref * 2 + 2;    ocount = re->top_backref * 3 + 3;
4170    match_block.offset_vector = (pcre_malloc)(ocount * sizeof(int));    match_block.offset_vector = (int *)(pcre_malloc)(ocount * sizeof(int));
4171    if (match_block.offset_vector == NULL) return PCRE_ERROR_NOMEMORY;    if (match_block.offset_vector == NULL) return PCRE_ERROR_NOMEMORY;
4172    using_temporary_offsets = TRUE;    using_temporary_offsets = TRUE;
4173    DPRINTF(("Got memory to hold back references\n"));    DPRINTF(("Got memory to hold back references\n"));
# Line 3397  if (re->top_backref > 0 && re->top_backr Line 4175  if (re->top_backref > 0 && re->top_backr
4175  else match_block.offset_vector = offsets;  else match_block.offset_vector = offsets;
4176    
4177  match_block.offset_end = ocount;  match_block.offset_end = ocount;
4178    match_block.offset_max = (2*ocount)/3;
4179  match_block.offset_overflow = FALSE;  match_block.offset_overflow = FALSE;
4180    
4181  /* Compute the minimum number of offsets that we need to reset each time. Doing  /* Compute the minimum number of offsets that we need to reset each time. Doing
# Line 3406  in the pattern. */ Line 4185  in the pattern. */
4185  resetcount = 2 + re->top_bracket * 2;  resetcount = 2 + re->top_bracket * 2;
4186  if (resetcount > offsetcount) resetcount = ocount;  if (resetcount > offsetcount) resetcount = ocount;
4187    
4188  /* If MULTILINE is set at exec time but was not set at compile time, and the  /* Reset the working variable associated with each extraction. These should
4189  anchored flag is set, we must re-check because a setting provoked by ^ in the  never be used unless previously set, but they get saved and restored, and so we
4190  pattern is not right in multi-line mode. Calling is_anchored() again here does  initialize them to avoid reading uninitialized locations. */
 the right check, because multiline is now set. If it now yields FALSE, the  
 expression must have had ^ starting some of its branches. Check to see if  
 that is true for *all* branches, and if so, set the startline flag. */  
4191    
4192  if (match_block. multiline && anchored && (re->options & PCRE_MULTILINE) == 0 &&  if (match_block.offset_vector != NULL)
     !is_anchored(re->code, match_block.multiline))  
4193    {    {
4194    anchored = FALSE;    register int *iptr = match_block.offset_vector + ocount;
4195    if (is_startline(re->code)) startline = TRUE;    register int *iend = iptr - resetcount/2 + 1;
4196      while (--iptr >= iend) *iptr = -1;
4197    }    }
4198    
4199  /* Set up the first character to match, if available. The first_char value is  /* Set up the first character to match, if available. The first_char value is
4200  never set for an anchored regular expression, but the anchoring may be forced  never set for an anchored regular expression, but the anchoring may be forced
4201  at run time, so we have to test for anchoring. The first char may be unset for  at run time, so we have to test for anchoring. The first char may be unset for
4202  an unanchored pattern, of course. If there's no first char and the pattern was  an unanchored pattern, of course. If there's no first char and the pattern was
4203  studied, the may be a bitmap of possible first characters. However, we can  studied, there may be a bitmap of possible first characters. */
 use this only if the caseless state of the studying was correct. */  
4204    
4205  if (!anchored)  if (!anchored)
4206    {    {
4207    if ((re->options & PCRE_FIRSTSET) != 0)    if ((re->options & PCRE_FIRSTSET) != 0)
4208      {      {
4209      first_char = re->first_char;      first_char = re->first_char;
4210      if (match_block.caseless) first_char = pcre_lcc[first_char];      if ((ims & PCRE_CASELESS) != 0) first_char = match_block.lcc[first_char];
4211      }      }
4212    else    else
4213      if (!startline && extra != NULL &&      if (!startline && extra != NULL &&
4214        (extra->options & PCRE_STUDY_MAPPED) != 0 &&        (extra->options & PCRE_STUDY_MAPPED) != 0)
       ((extra->options & PCRE_STUDY_CASELESS) != 0) == match_block.caseless)  
4215          start_bits = extra->start_bits;          start_bits = extra->start_bits;
4216    }    }
4217    
# Line 3457  do Line 4231  do
4231    
4232    if (first_char >= 0)    if (first_char >= 0)
4233      {      {
4234      if (match_block.caseless)      if ((ims & PCRE_CASELESS) != 0)
4235        while (start_match < end_subject && pcre_lcc[*start_match] != first_char)        while (start_match < end_subject &&
4236                 match_block.lcc[*start_match] != first_char)
4237          start_match++;          start_match++;
4238      else      else
4239        while (start_match < end_subject && *start_match != first_char)        while (start_match < end_subject && *start_match != first_char)
# Line 3498  do Line 4273  do
4273    there were too many extractions, set the return code to zero. In the case    there were too many extractions, set the return code to zero. In the case
4274    where we had to get some local store to hold offsets for backreferences, copy    where we had to get some local store to hold offsets for backreferences, copy
4275    those back references that we can. In this case there need not be overflow    those back references that we can. In this case there need not be overflow
4276    if certain parts of the pattern were not used.    if certain parts of the pattern were not used. */
   
   Before starting the match, we have to set up a longjmp() target to enable  
   the "cut" operation to fail a match completely without backtracking. This  
   is done in a separate function to avoid compiler warnings. We need not do  
   it unless PCRE_EXTRA is set, since only in that case is the "cut" operation  
   enabled. */  
4277    
4278    if ((re->options & PCRE_EXTRA) != 0)    if (!match(start_match, re->code, 2, &match_block, ims, FALSE, start_match))
4279      {      continue;
     if (!match_with_setjmp(start_match, re->code, 2, &match_block))  
       continue;  
     }  
   else if (!match(start_match, re->code, 2, &match_block)) continue;  
4280    
4281    /* Copy the offset information from temporary store if necessary */    /* Copy the offset information from temporary store if necessary */
4282    
# Line 3541  do Line 4306  do
4306    DPRINTF((">>>> returning %d\n", rc));    DPRINTF((">>>> returning %d\n", rc));
4307    return rc;    return rc;
4308    }    }
4309    
4310    /* This "while" is the end of the "do" above */
4311    
4312  while (!anchored &&  while (!anchored &&
4313         match_block.errorcode == PCRE_ERROR_NOMATCH &&         match_block.errorcode == PCRE_ERROR_NOMATCH &&
4314         start_match++ < end_subject);         start_match++ < end_subject);

Legend:
Removed from v.11  
changed lines
  Added in v.31

  ViewVC Help
Powered by ViewVC 1.1.5