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

Diff of /code/trunk/pcre_compile.c

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

revision 499 by ph10, Wed Mar 3 20:09:39 2010 UTC revision 773 by ph10, Wed Nov 30 18:10:27 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2010 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 88  so this number is very generous. Line 88  so this number is very generous.
88  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
89  remembering forward references to groups so that they can be filled in at the  remembering forward references to groups so that they can be filled in at the
90  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
91  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
92    filled up by repetitions of forward references, for example patterns like
93    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
94    that the workspace is expanded using malloc() in this situation. The value
95    below is therefore a minimum, and we put a maximum on it for safety. The
96    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
97    kicks in at the same number of forward references in all cases. */
98    
99  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
100    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
101    
102    /* The overrun tests check for a slightly smaller size so that they detect the
103    overrun before it actually does run off the end of the data block. */
104    
105    #define WORK_SIZE_SAFETY_MARGIN (100)
106    
107    
108  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 119  static const short int escapes[] = { Line 131  static const short int escapes[] = {
131       -ESC_H,                  0,       -ESC_H,                  0,
132       0,                       -ESC_K,       0,                       -ESC_K,
133       0,                       0,       0,                       0,
134       0,                       0,       -ESC_N,                  0,
135       -ESC_P,                  -ESC_Q,       -ESC_P,                  -ESC_Q,
136       -ESC_R,                  -ESC_S,       -ESC_R,                  -ESC_S,
137       0,                       0,       0,                       0,
# Line 166  static const short int escapes[] = { Line 178  static const short int escapes[] = {
178  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
179  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
180  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
181  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
182  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
183  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
184  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 183  string is built from string macros so th Line 195  string is built from string macros so th
195  platforms. */  platforms. */
196    
197  typedef struct verbitem {  typedef struct verbitem {
198    int   len;    int   len;                 /* Length of verb name */
199    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
200      int   op_arg;              /* Op when arg present, or -1 if not allowed */
201  } verbitem;  } verbitem;
202    
203  static const char verbnames[] =  static const char verbnames[] =
204      "\0"                       /* Empty name is a shorthand for MARK */
205      STRING_MARK0
206    STRING_ACCEPT0    STRING_ACCEPT0
207    STRING_COMMIT0    STRING_COMMIT0
208    STRING_F0    STRING_F0
# Line 197  static const char verbnames[] = Line 212  static const char verbnames[] =
212    STRING_THEN;    STRING_THEN;
213    
214  static const verbitem verbs[] = {  static const verbitem verbs[] = {
215    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
216    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
217    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
218    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
219    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
220    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
221    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
222      { 4, OP_SKIP,   OP_SKIP_ARG  },
223      { 4, OP_THEN,   OP_THEN_ARG  }
224  };  };
225    
226  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
# Line 251  static const int posix_class_maps[] = { Line 268  static const int posix_class_maps[] = {
268    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
269  };  };
270    
271    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
272    substitutes must be in the order of the names, defined above, and there are
273    both positive and negative cases. NULL means no substitute. */
274    
275    #ifdef SUPPORT_UCP
276    static const uschar *substitutes[] = {
277      (uschar *)"\\P{Nd}",    /* \D */
278      (uschar *)"\\p{Nd}",    /* \d */
279      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
280      (uschar *)"\\p{Xsp}",   /* \s */
281      (uschar *)"\\P{Xwd}",   /* \W */
282      (uschar *)"\\p{Xwd}"    /* \w */
283    };
284    
285    static const uschar *posix_substitutes[] = {
286      (uschar *)"\\p{L}",     /* alpha */
287      (uschar *)"\\p{Ll}",    /* lower */
288      (uschar *)"\\p{Lu}",    /* upper */
289      (uschar *)"\\p{Xan}",   /* alnum */
290      NULL,                   /* ascii */
291      (uschar *)"\\h",        /* blank */
292      NULL,                   /* cntrl */
293      (uschar *)"\\p{Nd}",    /* digit */
294      NULL,                   /* graph */
295      NULL,                   /* print */
296      NULL,                   /* punct */
297      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
298      (uschar *)"\\p{Xwd}",   /* word */
299      NULL,                   /* xdigit */
300      /* Negated cases */
301      (uschar *)"\\P{L}",     /* ^alpha */
302      (uschar *)"\\P{Ll}",    /* ^lower */
303      (uschar *)"\\P{Lu}",    /* ^upper */
304      (uschar *)"\\P{Xan}",   /* ^alnum */
305      NULL,                   /* ^ascii */
306      (uschar *)"\\H",        /* ^blank */
307      NULL,                   /* ^cntrl */
308      (uschar *)"\\P{Nd}",    /* ^digit */
309      NULL,                   /* ^graph */
310      NULL,                   /* ^print */
311      NULL,                   /* ^punct */
312      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
313      (uschar *)"\\P{Xwd}",   /* ^word */
314      NULL                    /* ^xdigit */
315    };
316    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
317    #endif
318    
319  #define STRING(a)  # a  #define STRING(a)  # a
320  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 263  the number of relocations needed when a Line 327  the number of relocations needed when a
327  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
328  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
329  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
330  because these strings are used only when there is a compilation error.  because these strings are used only when there is a compilation error.
331    
332  Each substring ends with \0 to insert a null character. This includes the final  Each substring ends with \0 to insert a null character. This includes the final
333  substring, so that the whole string ends with \0\0, which can be detected when  substring, so that the whole string ends with \0\0, which can be detected when
334  counting through. */  counting through. */
335    
336  static const char error_texts[] =  static const char error_texts[] =
# Line 314  static const char error_texts[] = Line 378  static const char error_texts[] =
378    /* 35 */    /* 35 */
379    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
380    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
381    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
382    "number after (?C is > 255\0"    "number after (?C is > 255\0"
383    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
384    /* 40 */    /* 40 */
# Line 336  static const char error_texts[] = Line 400  static const char error_texts[] =
400    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
401    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
402    /* 55 */    /* 55 */
403    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
404    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
405    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
406    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
407    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
408    /* 60 */    /* 60 */
409    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
410    "number is too big\0"    "number is too big\0"
# Line 348  static const char error_texts[] = Line 412  static const char error_texts[] =
412    "digit expected after (?+\0"    "digit expected after (?+\0"
413    "] is an invalid data character in JavaScript compatibility mode\0"    "] is an invalid data character in JavaScript compatibility mode\0"
414    /* 65 */    /* 65 */
415    "different names for subpatterns of the same number are not allowed\0";    "different names for subpatterns of the same number are not allowed\0"
416      "(*MARK) must have an argument\0"
417      "this version of PCRE is not compiled with PCRE_UCP support\0"
418      "\\c must be followed by an ASCII character\0"
419      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
420      /* 70 */
421      "internal error: unknown opcode in find_fixedlength()\0"
422      "\\N is not supported in a class\0"
423      "too many forward references\0"
424      ;
425    
426  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
427  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 484  static const unsigned char ebcdic_charta Line 557  static const unsigned char ebcdic_charta
557  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
558    
559  static BOOL  static BOOL
560    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
561      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
562    
563    
# Line 506  static const char * Line 579  static const char *
579  find_error_text(int n)  find_error_text(int n)
580  {  {
581  const char *s = error_texts;  const char *s = error_texts;
582  for (; n > 0; n--)  for (; n > 0; n--)
583    {    {
584    while (*s++ != 0) {};    while (*s++ != 0) {};
585    if (*s == 0) return "Error text not found (please report)";    if (*s == 0) return "Error text not found (please report)";
586    }    }
587  return s;  return s;
588  }  }
589    
590    
591  /*************************************************  /*************************************************
592    *           Expand the workspace                 *
593    *************************************************/
594    
595    /* This function is called during the second compiling phase, if the number of
596    forward references fills the existing workspace, which is originally a block on
597    the stack. A larger block is obtained from malloc() unless the ultimate limit
598    has been reached or the increase will be rather small.
599    
600    Argument: pointer to the compile data block
601    Returns:  0 if all went well, else an error number
602    */
603    
604    static int
605    expand_workspace(compile_data *cd)
606    {
607    uschar *newspace;
608    int newsize = cd->workspace_size * 2;
609    
610    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
611    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
612        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
613     return ERR72;
614    
615    newspace = (pcre_malloc)(newsize);
616    if (newspace == NULL) return ERR21;
617    
618    memcpy(newspace, cd->start_workspace, cd->workspace_size);
619    cd->hwm = (uschar *)newspace + (cd->hwm - cd->start_workspace);
620    if (cd->workspace_size > COMPILE_WORK_SIZE)
621      (pcre_free)((void *)cd->start_workspace);
622    cd->start_workspace = newspace;
623    cd->workspace_size = newsize;
624    return 0;
625    }
626    
627    
628    
629    /*************************************************
630    *            Check for counted repeat            *
631    *************************************************/
632    
633    /* This function is called when a '{' is encountered in a place where it might
634    start a quantifier. It looks ahead to see if it really is a quantifier or not.
635    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
636    where the ddds are digits.
637    
638    Arguments:
639      p         pointer to the first char after '{'
640    
641    Returns:    TRUE or FALSE
642    */
643    
644    static BOOL
645    is_counted_repeat(const uschar *p)
646    {
647    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
648    while ((digitab[*p] & ctype_digit) != 0) p++;
649    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
650    
651    if (*p++ != CHAR_COMMA) return FALSE;
652    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
653    
654    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
655    while ((digitab[*p] & ctype_digit) != 0) p++;
656    
657    return (*p == CHAR_RIGHT_CURLY_BRACKET);
658    }
659    
660    
661    
662    /*************************************************
663  *            Handle escapes                      *  *            Handle escapes                      *
664  *************************************************/  *************************************************/
665    
# Line 581  else Line 725  else
725    
726      case CHAR_l:      case CHAR_l:
727      case CHAR_L:      case CHAR_L:
728      case CHAR_N:      *errorcodeptr = ERR37;
729        break;
730    
731      case CHAR_u:      case CHAR_u:
732        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
733          {
734          /* In JavaScript, \u must be followed by four hexadecimal numbers.
735          Otherwise it is a lowercase u letter. */
736          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
737               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
738            {
739            c = 0;
740            for (i = 0; i < 4; ++i)
741              {
742              register int cc = *(++ptr);
743    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
744              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
745              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
746    #else           /* EBCDIC coding */
747              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
748              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
749    #endif
750              }
751            }
752          }
753        else
754          *errorcodeptr = ERR37;
755        break;
756    
757      case CHAR_U:      case CHAR_U:
758      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
759        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
760      break;      break;
761    
762      /* \g must be followed by one of a number of specific things:      /* In a character class, \g is just a literal "g". Outside a character
763        class, \g must be followed by one of a number of specific things:
764    
765      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
766      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 604  else Line 777  else
777      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
778    
779      case CHAR_g:      case CHAR_g:
780        if (isclass) break;
781      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
782        {        {
783        c = -ESC_g;        c = -ESC_g;
# Line 732  else Line 906  else
906      treated as a data character. */      treated as a data character. */
907    
908      case CHAR_x:      case CHAR_x:
909        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
910          {
911          /* In JavaScript, \x must be followed by two hexadecimal numbers.
912          Otherwise it is a lowercase x letter. */
913          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
914            {
915            c = 0;
916            for (i = 0; i < 2; ++i)
917              {
918              register int cc = *(++ptr);
919    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
920              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
921              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
922    #else           /* EBCDIC coding */
923              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
924              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
925    #endif
926              }
927            }
928          break;
929          }
930    
931      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
932        {        {
933        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
# Line 782  else Line 978  else
978      break;      break;
979    
980      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
981      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
982        coding is ASCII-specific, but then the whole concept of \cx is
983      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
984    
985      case CHAR_c:      case CHAR_c:
# Line 792  else Line 989  else
989        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
990        break;        break;
991        }        }
992    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
993  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
994          {
995          *errorcodeptr = ERR68;
996          break;
997          }
998      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
999      c ^= 0x40;      c ^= 0x40;
1000  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1001      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1002      c ^= 0xC0;      c ^= 0xC0;
1003  #endif  #endif
# Line 819  else Line 1020  else
1020      }      }
1021    }    }
1022    
1023    /* Perl supports \N{name} for character names, as well as plain \N for "not
1024    newline". PCRE does not support \N{name}. However, it does support
1025    quantification such as \N{2,3}. */
1026    
1027    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1028         !is_counted_repeat(ptr+2))
1029      *errorcodeptr = ERR37;
1030    
1031    /* If PCRE_UCP is set, we change the values for \d etc. */
1032    
1033    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
1034      c -= (ESC_DU - ESC_D);
1035    
1036    /* Set the pointer to the final character before returning. */
1037    
1038  *ptrptr = ptr;  *ptrptr = ptr;
1039  return c;  return c;
1040  }  }
# Line 919  return -1; Line 1135  return -1;
1135    
1136    
1137  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if (*p++ != CHAR_COMMA) return FALSE;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == CHAR_RIGHT_CURLY_BRACKET);  
 }  
   
   
   
 /*************************************************  
1138  *         Read repeat counts                     *  *         Read repeat counts                     *
1139  *************************************************/  *************************************************/
1140    
# Line 1027  top-level call starts at the beginning o Line 1210  top-level call starts at the beginning o
1210  start at a parenthesis. It scans along a pattern's text looking for capturing  start at a parenthesis. It scans along a pattern's text looking for capturing
1211  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1212  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1213  returns when it reaches a given numbered subpattern. We know that if (?P< is  returns when it reaches a given numbered subpattern. Recursion is used to keep
1214  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1215  first pass. Recursion is used to keep track of subpatterns that reset the  
1216  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1217    that if (?< or (?' or (?P< is encountered, the name will be correctly
1218    terminated because that is checked in the first pass. There is now one call to
1219    this function in the first pass, to check for a recursive back reference by
1220    name (so that we can make the whole group atomic). In this case, we need check
1221    only up to the current position in the pattern, and that is still OK because
1222    and previous occurrences will have been checked. To make this work, the test
1223    for "end of pattern" is a check against cd->end_pattern in the main loop,
1224    instead of looking for a binary zero. This means that the special first-pass
1225    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1226    processing items within the loop are OK, because afterwards the main loop will
1227    terminate.)
1228    
1229  Arguments:  Arguments:
1230    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1038  Arguments: Line 1232  Arguments:
1232    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1233    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1234    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1235      utf8         TRUE if we are in UTF-8 mode
1236    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1237    
1238  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
# Line 1045  Returns:       the number of the named s Line 1240  Returns:       the number of the named s
1240    
1241  static int  static int
1242  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1243    BOOL xmode, int *count)    BOOL xmode, BOOL utf8, int *count)
1244  {  {
1245  uschar *ptr = *ptrptr;  uschar *ptr = *ptrptr;
1246  int start_count = *count;  int start_count = *count;
# Line 1057  dealing with. The very first call may no Line 1252  dealing with. The very first call may no
1252    
1253  if (ptr[0] == CHAR_LEFT_PARENTHESIS)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1254    {    {
1255    if (ptr[1] == CHAR_QUESTION_MARK &&    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1256        ptr[2] == CHAR_VERTICAL_LINE)  
1257      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1258    
1259      /* Handle a normal, unnamed capturing parenthesis. */
1260    
1261      else if (ptr[1] != CHAR_QUESTION_MARK)
1262        {
1263        *count += 1;
1264        if (name == NULL && *count == lorn) return *count;
1265        ptr++;
1266        }
1267    
1268      /* All cases now have (? at the start. Remember when we are in a group
1269      where the parenthesis numbers are duplicated. */
1270    
1271      else if (ptr[2] == CHAR_VERTICAL_LINE)
1272      {      {
1273      ptr += 3;      ptr += 3;
1274      dup_parens = TRUE;      dup_parens = TRUE;
1275      }      }
1276    
1277    /* Handle a normal, unnamed capturing parenthesis */    /* Handle comments; all characters are allowed until a ket is reached. */
1278    
1279    else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)    else if (ptr[2] == CHAR_NUMBER_SIGN)
1280      {      {
1281      *count += 1;      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1282      if (name == NULL && *count == lorn) return *count;      goto FAIL_EXIT;
     ptr++;  
1283      }      }
1284    
1285    /* Handle a condition. If it is an assertion, just carry on so that it    /* Handle a condition. If it is an assertion, just carry on so that it
1286    is processed as normal. If not, skip to the closing parenthesis of the    is processed as normal. If not, skip to the closing parenthesis of the
1287    condition (there can't be any nested parens. */    condition (there can't be any nested parens). */
1288    
1289    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1290      {      {
# Line 1087  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1296  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1296        }        }
1297      }      }
1298    
1299    /* We have either (? or (* and not a condition */    /* Start with (? but not a condition. */
1300    
1301    else    else
1302      {      {
# Line 1116  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1325  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1325    }    }
1326    
1327  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1328  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1329    first-pass call when this value is temporarily adjusted to stop at the current
1330    position. So DO NOT change this to a test for binary zero. */
1331    
1332  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1333    {    {
1334    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1335    
# Line 1192  for (; *ptr != 0; ptr++) Line 1403  for (; *ptr != 0; ptr++)
1403    
1404    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1405      {      {
1406      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1407        while (*ptr != 0)
1408          {
1409          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1410          ptr++;
1411    #ifdef SUPPORT_UTF8
1412          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1413    #endif
1414          }
1415      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1416      continue;      continue;
1417      }      }
# Line 1201  for (; *ptr != 0; ptr++) Line 1420  for (; *ptr != 0; ptr++)
1420    
1421    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1422      {      {
1423      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1424      if (rc > 0) return rc;      if (rc > 0) return rc;
1425      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1426      }      }
# Line 1209  for (; *ptr != 0; ptr++) Line 1428  for (; *ptr != 0; ptr++)
1428    else if (*ptr == CHAR_RIGHT_PARENTHESIS)    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1429      {      {
1430      if (dup_parens && *count < hwm_count) *count = hwm_count;      if (dup_parens && *count < hwm_count) *count = hwm_count;
1431      *ptrptr = ptr;      goto FAIL_EXIT;
     return -1;  
1432      }      }
1433    
1434    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
# Line 1248  Arguments: Line 1466  Arguments:
1466    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1467    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1468    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1469      utf8         TRUE if we are in UTF-8 mode
1470    
1471  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1472  */  */
1473    
1474  static int  static int
1475  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1476      BOOL utf8)
1477  {  {
1478  uschar *ptr = (uschar *)cd->start_pattern;  uschar *ptr = (uschar *)cd->start_pattern;
1479  int count = 0;  int count = 0;
# Line 1266  matching closing parens. That is why we Line 1486  matching closing parens. That is why we
1486    
1487  for (;;)  for (;;)
1488    {    {
1489    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1490    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1491    }    }
1492    
# Line 1282  return rc; Line 1502  return rc;
1502    
1503  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1504  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1505  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1506  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1507  assertions, and also the \b assertion; for others it does not.  does not.
1508    
1509  Arguments:  Arguments:
1510    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1511    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1512    
1513  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1514  */  */
1515    
1516  static const uschar*  static const uschar*
1517  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1518  {  {
1519  for (;;)  for (;;)
1520    {    {
1521    switch ((int)*code)    switch ((int)*code)
1522      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1523      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1524      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1525      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1359  and doing the check at the end; a flag s Line 1569  and doing the check at the end; a flag s
1569    
1570  Arguments:  Arguments:
1571    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1572    options  the compiling options    utf8     TRUE in UTF-8 mode
1573    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1574    cd       the "compile data" structure    cd       the "compile data" structure
1575    
1576  Returns:   the fixed length,  Returns:   the fixed length,
1577               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1578               or -2 if \C was encountered               or -2 if \C was encountered (in UTF-8 mode only)
1579               or -3 if an OP_RECURSE item was encountered and atend is FALSE               or -3 if an OP_RECURSE item was encountered and atend is FALSE
1580                 or -4 if an unknown opcode was encountered (internal error)
1581  */  */
1582    
1583  static int  static int
1584  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1585  {  {
1586  int length = -1;  int length = -1;
1587    
# Line 1387  for (;;) Line 1598  for (;;)
1598    register int op = *cc;    register int op = *cc;
1599    switch (op)    switch (op)
1600      {      {
1601        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1602        OP_BRA (normal non-capturing bracket) because the other variants of these
1603        opcodes are all concerned with unlimited repeated groups, which of course
1604        are not of fixed length. */
1605    
1606      case OP_CBRA:      case OP_CBRA:
1607      case OP_BRA:      case OP_BRA:
1608      case OP_ONCE:      case OP_ONCE:
1609        case OP_ONCE_NC:
1610      case OP_COND:      case OP_COND:
1611      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1612      if (d < 0) return d;      if (d < 0) return d;
1613      branchlength += d;      branchlength += d;
1614      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1615      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1616      break;      break;
1617    
1618      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested call.
1619      call. If it's ALT it is an alternation in a nested call. If it is      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1620      END it's the end of the outer call. All can be handled by the same code. */      an ALT. If it is END it's the end of the outer call. All can be handled by
1621        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1622        because they all imply an unlimited repeat. */
1623    
1624      case OP_ALT:      case OP_ALT:
1625      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1626      case OP_END:      case OP_END:
1627        case OP_ACCEPT:
1628        case OP_ASSERT_ACCEPT:
1629      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1630        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1631      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1423  for (;;) Line 1642  for (;;)
1642      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1643      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1644      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1645      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1646      if (d < 0) return d;      if (d < 0) return d;
1647      branchlength += d;      branchlength += d;
1648      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1440  for (;;) Line 1659  for (;;)
1659    
1660      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1661    
1662      case OP_REVERSE:      case OP_MARK:
1663        case OP_PRUNE_ARG:
1664        case OP_SKIP_ARG:
1665        case OP_THEN_ARG:
1666        cc += cc[1] + _pcre_OP_lengths[*cc];
1667        break;
1668    
1669        case OP_CALLOUT:
1670        case OP_CIRC:
1671        case OP_CIRCM:
1672        case OP_CLOSE:
1673        case OP_COMMIT:
1674      case OP_CREF:      case OP_CREF:
     case OP_NCREF:  
     case OP_RREF:  
     case OP_NRREF:  
1675      case OP_DEF:      case OP_DEF:
1676      case OP_OPT:      case OP_DOLL:
1677      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1678      case OP_EOD:      case OP_EOD:
1679      case OP_EODN:      case OP_EODN:
1680      case OP_CIRC:      case OP_FAIL:
1681      case OP_DOLL:      case OP_NCREF:
1682        case OP_NRREF:
1683      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1684        case OP_PRUNE:
1685        case OP_REVERSE:
1686        case OP_RREF:
1687        case OP_SET_SOM:
1688        case OP_SKIP:
1689        case OP_SOD:
1690        case OP_SOM:
1691        case OP_THEN:
1692      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1693      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
1694      break;      break;
# Line 1462  for (;;) Line 1696  for (;;)
1696      /* Handle literal characters */      /* Handle literal characters */
1697    
1698      case OP_CHAR:      case OP_CHAR:
1699      case OP_CHARNC:      case OP_CHARI:
1700      case OP_NOT:      case OP_NOT:
1701        case OP_NOTI:
1702      branchlength++;      branchlength++;
1703      cc += 2;      cc += 2;
1704  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1705      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1706  #endif  #endif
1707      break;      break;
1708    
# Line 1476  for (;;) Line 1710  for (;;)
1710      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1711    
1712      case OP_EXACT:      case OP_EXACT:
1713        case OP_EXACTI:
1714        case OP_NOTEXACT:
1715        case OP_NOTEXACTI:
1716      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1717      cc += 4;      cc += 4;
1718  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1719      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1720  #endif  #endif
1721      break;      break;
1722    
# Line 1497  for (;;) Line 1733  for (;;)
1733      cc += 2;      cc += 2;
1734      /* Fall through */      /* Fall through */
1735    
1736        case OP_HSPACE:
1737        case OP_VSPACE:
1738        case OP_NOT_HSPACE:
1739        case OP_NOT_VSPACE:
1740      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1741      case OP_DIGIT:      case OP_DIGIT:
1742      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1509  for (;;) Line 1749  for (;;)
1749      cc++;      cc++;
1750      break;      break;
1751    
1752      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1753        otherwise \C is coded as OP_ALLANY. */
1754    
1755      case OP_ANYBYTE:      case OP_ANYBYTE:
1756      return -2;      return -2;
# Line 1528  for (;;) Line 1769  for (;;)
1769    
1770      switch (*cc)      switch (*cc)
1771        {        {
1772          case OP_CRPLUS:
1773          case OP_CRMINPLUS:
1774        case OP_CRSTAR:        case OP_CRSTAR:
1775        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1776        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1548  for (;;) Line 1791  for (;;)
1791    
1792      /* Anything else is variable length */      /* Anything else is variable length */
1793    
1794      default:      case OP_ANYNL:
1795        case OP_BRAMINZERO:
1796        case OP_BRAPOS:
1797        case OP_BRAPOSZERO:
1798        case OP_BRAZERO:
1799        case OP_CBRAPOS:
1800        case OP_EXTUNI:
1801        case OP_KETRMAX:
1802        case OP_KETRMIN:
1803        case OP_KETRPOS:
1804        case OP_MINPLUS:
1805        case OP_MINPLUSI:
1806        case OP_MINQUERY:
1807        case OP_MINQUERYI:
1808        case OP_MINSTAR:
1809        case OP_MINSTARI:
1810        case OP_MINUPTO:
1811        case OP_MINUPTOI:
1812        case OP_NOTMINPLUS:
1813        case OP_NOTMINPLUSI:
1814        case OP_NOTMINQUERY:
1815        case OP_NOTMINQUERYI:
1816        case OP_NOTMINSTAR:
1817        case OP_NOTMINSTARI:
1818        case OP_NOTMINUPTO:
1819        case OP_NOTMINUPTOI:
1820        case OP_NOTPLUS:
1821        case OP_NOTPLUSI:
1822        case OP_NOTPOSPLUS:
1823        case OP_NOTPOSPLUSI:
1824        case OP_NOTPOSQUERY:
1825        case OP_NOTPOSQUERYI:
1826        case OP_NOTPOSSTAR:
1827        case OP_NOTPOSSTARI:
1828        case OP_NOTPOSUPTO:
1829        case OP_NOTPOSUPTOI:
1830        case OP_NOTQUERY:
1831        case OP_NOTQUERYI:
1832        case OP_NOTSTAR:
1833        case OP_NOTSTARI:
1834        case OP_NOTUPTO:
1835        case OP_NOTUPTOI:
1836        case OP_PLUS:
1837        case OP_PLUSI:
1838        case OP_POSPLUS:
1839        case OP_POSPLUSI:
1840        case OP_POSQUERY:
1841        case OP_POSQUERYI:
1842        case OP_POSSTAR:
1843        case OP_POSSTARI:
1844        case OP_POSUPTO:
1845        case OP_POSUPTOI:
1846        case OP_QUERY:
1847        case OP_QUERYI:
1848        case OP_REF:
1849        case OP_REFI:
1850        case OP_SBRA:
1851        case OP_SBRAPOS:
1852        case OP_SCBRA:
1853        case OP_SCBRAPOS:
1854        case OP_SCOND:
1855        case OP_SKIPZERO:
1856        case OP_STAR:
1857        case OP_STARI:
1858        case OP_TYPEMINPLUS:
1859        case OP_TYPEMINQUERY:
1860        case OP_TYPEMINSTAR:
1861        case OP_TYPEMINUPTO:
1862        case OP_TYPEPLUS:
1863        case OP_TYPEPOSPLUS:
1864        case OP_TYPEPOSQUERY:
1865        case OP_TYPEPOSSTAR:
1866        case OP_TYPEPOSUPTO:
1867        case OP_TYPEQUERY:
1868        case OP_TYPESTAR:
1869        case OP_TYPEUPTO:
1870        case OP_UPTO:
1871        case OP_UPTOI:
1872      return -1;      return -1;
1873    
1874        /* Catch unrecognized opcodes so that when new ones are added they
1875        are not forgotten, as has happened in the past. */
1876    
1877        default:
1878        return -4;
1879      }      }
1880    }    }
1881  /* Control never gets here */  /* Control never gets here */
# Line 1582  _pcre_find_bracket(const uschar *code, B Line 1908  _pcre_find_bracket(const uschar *code, B
1908  for (;;)  for (;;)
1909    {    {
1910    register int c = *code;    register int c = *code;
1911    
1912    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1913    
1914    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1600  for (;;) Line 1927  for (;;)
1927    
1928    /* Handle capturing bracket */    /* Handle capturing bracket */
1929    
1930    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1931               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1932      {      {
1933      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1934      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1609  for (;;) Line 1937  for (;;)
1937    
1938    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
1939    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
1940    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1941      must add in its length. */
1942    
1943    else    else
1944      {      {
# Line 1633  for (;;) Line 1962  for (;;)
1962        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1963        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1964        break;        break;
1965    
1966          case OP_MARK:
1967          case OP_PRUNE_ARG:
1968          case OP_SKIP_ARG:
1969          code += code[1];
1970          break;
1971    
1972          case OP_THEN_ARG:
1973          code += code[1];
1974          break;
1975        }        }
1976    
1977      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1647  for (;;) Line 1986  for (;;)
1986      if (utf8) switch(c)      if (utf8) switch(c)
1987        {        {
1988        case OP_CHAR:        case OP_CHAR:
1989        case OP_CHARNC:        case OP_CHARI:
1990        case OP_EXACT:        case OP_EXACT:
1991          case OP_EXACTI:
1992        case OP_UPTO:        case OP_UPTO:
1993          case OP_UPTOI:
1994        case OP_MINUPTO:        case OP_MINUPTO:
1995          case OP_MINUPTOI:
1996        case OP_POSUPTO:        case OP_POSUPTO:
1997          case OP_POSUPTOI:
1998        case OP_STAR:        case OP_STAR:
1999          case OP_STARI:
2000        case OP_MINSTAR:        case OP_MINSTAR:
2001          case OP_MINSTARI:
2002        case OP_POSSTAR:        case OP_POSSTAR:
2003          case OP_POSSTARI:
2004        case OP_PLUS:        case OP_PLUS:
2005          case OP_PLUSI:
2006        case OP_MINPLUS:        case OP_MINPLUS:
2007          case OP_MINPLUSI:
2008        case OP_POSPLUS:        case OP_POSPLUS:
2009          case OP_POSPLUSI:
2010        case OP_QUERY:        case OP_QUERY:
2011          case OP_QUERYI:
2012        case OP_MINQUERY:        case OP_MINQUERY:
2013          case OP_MINQUERYI:
2014        case OP_POSQUERY:        case OP_POSQUERY:
2015          case OP_POSQUERYI:
2016        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
2017        break;        break;
2018        }        }
# Line 1704  for (;;) Line 2056  for (;;)
2056    
2057    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
2058    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
2059    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2060      must add in its length. */
2061    
2062    else    else
2063      {      {
# Line 1728  for (;;) Line 2081  for (;;)
2081        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2082        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2083        break;        break;
2084    
2085          case OP_MARK:
2086          case OP_PRUNE_ARG:
2087          case OP_SKIP_ARG:
2088          code += code[1];
2089          break;
2090    
2091          case OP_THEN_ARG:
2092          code += code[1];
2093          break;
2094        }        }
2095    
2096      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1742  for (;;) Line 2105  for (;;)
2105      if (utf8) switch(c)      if (utf8) switch(c)
2106        {        {
2107        case OP_CHAR:        case OP_CHAR:
2108        case OP_CHARNC:        case OP_CHARI:
2109        case OP_EXACT:        case OP_EXACT:
2110          case OP_EXACTI:
2111        case OP_UPTO:        case OP_UPTO:
2112          case OP_UPTOI:
2113        case OP_MINUPTO:        case OP_MINUPTO:
2114          case OP_MINUPTOI:
2115        case OP_POSUPTO:        case OP_POSUPTO:
2116          case OP_POSUPTOI:
2117        case OP_STAR:        case OP_STAR:
2118          case OP_STARI:
2119        case OP_MINSTAR:        case OP_MINSTAR:
2120          case OP_MINSTARI:
2121        case OP_POSSTAR:        case OP_POSSTAR:
2122          case OP_POSSTARI:
2123        case OP_PLUS:        case OP_PLUS:
2124          case OP_PLUSI:
2125        case OP_MINPLUS:        case OP_MINPLUS:
2126          case OP_MINPLUSI:
2127        case OP_POSPLUS:        case OP_POSPLUS:
2128          case OP_POSPLUSI:
2129        case OP_QUERY:        case OP_QUERY:
2130          case OP_QUERYI:
2131        case OP_MINQUERY:        case OP_MINQUERY:
2132          case OP_MINQUERYI:
2133        case OP_POSQUERY:        case OP_POSQUERY:
2134          case OP_POSQUERYI:
2135        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
2136        break;        break;
2137        }        }
# Line 1784  Arguments: Line 2160  Arguments:
2160    code        points to start of search    code        points to start of search
2161    endcode     points to where to stop    endcode     points to where to stop
2162    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
2163      cd          contains pointers to tables etc.
2164    
2165  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2166  */  */
2167    
2168  static BOOL  static BOOL
2169  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
2170      compile_data *cd)
2171  {  {
2172  register int c;  register int c;
2173  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2174       code < endcode;       code < endcode;
2175       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2176    {    {
2177    const uschar *ccode;    const uschar *ccode;
2178    
# Line 1810  for (code = first_significant_code(code Line 2188  for (code = first_significant_code(code
2188      continue;      continue;
2189      }      }
2190    
2191      /* For a recursion/subroutine call, if its end has been reached, which
2192      implies a backward reference subroutine call, we can scan it. If it's a
2193      forward reference subroutine call, we can't. To detect forward reference
2194      we have to scan up the list that is kept in the workspace. This function is
2195      called only when doing the real compile, not during the pre-compile that
2196      measures the size of the compiled pattern. */
2197    
2198      if (c == OP_RECURSE)
2199        {
2200        const uschar *scode;
2201        BOOL empty_branch;
2202    
2203        /* Test for forward reference */
2204    
2205        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2206          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2207    
2208        /* Not a forward reference, test for completed backward reference */
2209    
2210        empty_branch = FALSE;
2211        scode = cd->start_code + GET(code, 1);
2212        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2213    
2214        /* Completed backwards reference */
2215    
2216        do
2217          {
2218          if (could_be_empty_branch(scode, endcode, utf8, cd))
2219            {
2220            empty_branch = TRUE;
2221            break;
2222            }
2223          scode += GET(scode, 1);
2224          }
2225        while (*scode == OP_ALT);
2226    
2227        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2228        continue;
2229        }
2230    
2231    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2232    
2233    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2234          c == OP_BRAPOSZERO)
2235      {      {
2236      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
2237      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1820  for (code = first_significant_code(code Line 2239  for (code = first_significant_code(code
2239      continue;      continue;
2240      }      }
2241    
2242      /* A nested group that is already marked as "could be empty" can just be
2243      skipped. */
2244    
2245      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2246          c == OP_SCBRA || c == OP_SCBRAPOS)
2247        {
2248        do code += GET(code, 1); while (*code == OP_ALT);
2249        c = *code;
2250        continue;
2251        }
2252    
2253    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2254    
2255    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2256          c == OP_CBRA || c == OP_CBRAPOS ||
2257          c == OP_ONCE || c == OP_ONCE_NC ||
2258          c == OP_COND)
2259      {      {
2260      BOOL empty_branch;      BOOL empty_branch;
2261      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1838  for (code = first_significant_code(code Line 2271  for (code = first_significant_code(code
2271        empty_branch = FALSE;        empty_branch = FALSE;
2272        do        do
2273          {          {
2274          if (!empty_branch && could_be_empty_branch(code, endcode, utf8))          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2275            empty_branch = TRUE;            empty_branch = TRUE;
2276          code += GET(code, 1);          code += GET(code, 1);
2277          }          }
# Line 1909  for (code = first_significant_code(code Line 2342  for (code = first_significant_code(code
2342      case OP_ALLANY:      case OP_ALLANY:
2343      case OP_ANYBYTE:      case OP_ANYBYTE:
2344      case OP_CHAR:      case OP_CHAR:
2345      case OP_CHARNC:      case OP_CHARI:
2346      case OP_NOT:      case OP_NOT:
2347        case OP_NOTI:
2348      case OP_PLUS:      case OP_PLUS:
2349      case OP_MINPLUS:      case OP_MINPLUS:
2350      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1950  for (code = first_significant_code(code Line 2384  for (code = first_significant_code(code
2384      case OP_KET:      case OP_KET:
2385      case OP_KETRMAX:      case OP_KETRMAX:
2386      case OP_KETRMIN:      case OP_KETRMIN:
2387        case OP_KETRPOS:
2388      case OP_ALT:      case OP_ALT:
2389      return TRUE;      return TRUE;
2390    
# Line 1958  for (code = first_significant_code(code Line 2393  for (code = first_significant_code(code
2393    
2394  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2395      case OP_STAR:      case OP_STAR:
2396        case OP_STARI:
2397      case OP_MINSTAR:      case OP_MINSTAR:
2398        case OP_MINSTARI:
2399      case OP_POSSTAR:      case OP_POSSTAR:
2400        case OP_POSSTARI:
2401      case OP_QUERY:      case OP_QUERY:
2402        case OP_QUERYI:
2403      case OP_MINQUERY:      case OP_MINQUERY:
2404        case OP_MINQUERYI:
2405      case OP_POSQUERY:      case OP_POSQUERY:
2406        case OP_POSQUERYI:
2407      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2408      break;      break;
2409    
2410      case OP_UPTO:      case OP_UPTO:
2411        case OP_UPTOI:
2412      case OP_MINUPTO:      case OP_MINUPTO:
2413        case OP_MINUPTOI:
2414      case OP_POSUPTO:      case OP_POSUPTO:
2415        case OP_POSUPTOI:
2416      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2417      break;      break;
2418  #endif  #endif
2419    
2420        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2421        string. */
2422    
2423        case OP_MARK:
2424        case OP_PRUNE_ARG:
2425        case OP_SKIP_ARG:
2426        code += code[1];
2427        break;
2428    
2429        case OP_THEN_ARG:
2430        code += code[1];
2431        break;
2432    
2433        /* None of the remaining opcodes are required to match a character. */
2434    
2435        default:
2436        break;
2437      }      }
2438    }    }
2439    
# Line 1988  return TRUE; Line 2450  return TRUE;
2450  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2451  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2452  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2453    This function is called only during the real compile, not during the
2454    pre-compile.
2455    
2456  Arguments:  Arguments:
2457    code        points to start of the recursion    code        points to start of the recursion
2458    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2459    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2460    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2461      cd          pointers to tables etc
2462    
2463  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2464  */  */
2465    
2466  static BOOL  static BOOL
2467  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2468    BOOL utf8)    BOOL utf8, compile_data *cd)
2469  {  {
2470  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2471    {    {
2472    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2473      return FALSE;      return FALSE;
2474    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2475    }    }
# Line 2037  where Perl recognizes it as the POSIX cl Line 2502  where Perl recognizes it as the POSIX cl
2502  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2503  I think.  I think.
2504    
2505    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2506    It seems that the appearance of a nested POSIX class supersedes an apparent
2507    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2508    a digit.
2509    
2510    In Perl, unescaped square brackets may also appear as part of class names. For
2511    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2512    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2513    seem right at all. PCRE does not allow closing square brackets in POSIX class
2514    names.
2515    
2516  Arguments:  Arguments:
2517    ptr      pointer to the initial [    ptr      pointer to the initial [
2518    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2051  int terminator;          /* Don't combin Line 2527  int terminator;          /* Don't combin
2527  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2528  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2529    {    {
2530    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2531        ptr++;
2532      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2533      else
2534      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2535      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2536        {        {
2537        *endptr = ptr;        *endptr = ptr;
2538        return TRUE;        return TRUE;
2539        }        }
2540        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2541             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2542              ptr[1] == CHAR_EQUALS_SIGN) &&
2543            check_posix_syntax(ptr, endptr))
2544          return FALSE;
2545      }      }
2546    }    }
2547  return FALSE;  return FALSE;
# Line 2187  auto_callout(uschar *code, const uschar Line 2670  auto_callout(uschar *code, const uschar
2670  {  {
2671  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2672  *code++ = 255;  *code++ = 255;
2673  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2674  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2675  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2676  }  }
2677    
# Line 2213  Returns:             nothing Line 2696  Returns:             nothing
2696  static void  static void
2697  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2698  {  {
2699  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2700  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2701  }  }
2702    
# Line 2263  for (++c; c <= d; c++) Line 2746  for (++c; c <= d; c++)
2746    
2747  return TRUE;  return TRUE;
2748  }  }
2749    
2750    
2751    
2752    /*************************************************
2753    *        Check a character and a property        *
2754    *************************************************/
2755    
2756    /* This function is called by check_auto_possessive() when a property item
2757    is adjacent to a fixed character.
2758    
2759    Arguments:
2760      c            the character
2761      ptype        the property type
2762      pdata        the data for the type
2763      negated      TRUE if it's a negated property (\P or \p{^)
2764    
2765    Returns:       TRUE if auto-possessifying is OK
2766    */
2767    
2768    static BOOL
2769    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2770    {
2771    const ucd_record *prop = GET_UCD(c);
2772    switch(ptype)
2773      {
2774      case PT_LAMP:
2775      return (prop->chartype == ucp_Lu ||
2776              prop->chartype == ucp_Ll ||
2777              prop->chartype == ucp_Lt) == negated;
2778    
2779      case PT_GC:
2780      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2781    
2782      case PT_PC:
2783      return (pdata == prop->chartype) == negated;
2784    
2785      case PT_SC:
2786      return (pdata == prop->script) == negated;
2787    
2788      /* These are specials */
2789    
2790      case PT_ALNUM:
2791      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2792              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2793    
2794      case PT_SPACE:    /* Perl space */
2795      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2796              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2797              == negated;
2798    
2799      case PT_PXSPACE:  /* POSIX space */
2800      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2801              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2802              c == CHAR_FF || c == CHAR_CR)
2803              == negated;
2804    
2805      case PT_WORD:
2806      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2807              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2808              c == CHAR_UNDERSCORE) == negated;
2809      }
2810    return FALSE;
2811    }
2812  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2813    
2814    
# Line 2276  whether the next thing could possibly ma Line 2822  whether the next thing could possibly ma
2822  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2823    
2824  Arguments:  Arguments:
2825    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2826    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2827    ptr           next character in pattern    ptr           next character in pattern
2828    options       options bits    options       options bits
2829    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2288  Returns:        TRUE if possessifying is Line 2832  Returns:        TRUE if possessifying is
2832  */  */
2833    
2834  static BOOL  static BOOL
2835  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2836    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2837  {  {
2838  int next;  int c, next;
2839    int op_code = *previous++;
2840    
2841  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2842    
# Line 2302  if ((options & PCRE_EXTENDED) != 0) Line 2847  if ((options & PCRE_EXTENDED) != 0)
2847      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2848      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2849        {        {
2850        while (*(++ptr) != 0)        ptr++;
2851          while (*ptr != 0)
2852            {
2853          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2854            ptr++;
2855    #ifdef SUPPORT_UTF8
2856            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2857    #endif
2858            }
2859        }        }
2860      else break;      else break;
2861      }      }
# Line 2339  if ((options & PCRE_EXTENDED) != 0) Line 2891  if ((options & PCRE_EXTENDED) != 0)
2891      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2892      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2893        {        {
2894        while (*(++ptr) != 0)        ptr++;
2895          while (*ptr != 0)
2896            {
2897          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2898            ptr++;
2899    #ifdef SUPPORT_UTF8
2900            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2901    #endif
2902            }
2903        }        }
2904      else break;      else break;
2905      }      }
# Line 2352  if (*ptr == CHAR_ASTERISK || *ptr == CHA Line 2911  if (*ptr == CHAR_ASTERISK || *ptr == CHA
2911    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2912      return FALSE;      return FALSE;
2913    
2914  /* Now compare the next item with the previous opcode. If the previous is a  /* Now compare the next item with the previous opcode. First, handle cases when
2915  positive single character match, "item" either contains the character or, if  the next item is a character. */
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
   
 /* Handle cases when the next item is a character. */  
2916    
2917  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2918    {    {
2919    case OP_CHAR:    case OP_CHAR:
2920  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2921    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2922  #else  #else
2923    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2924  #endif  #endif
2925    return item != next;    return c != next;
2926    
2927    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
2928    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
2929    high-valued characters. */    high-valued characters. */
2930    
2931    case OP_CHARNC:    case OP_CHARI:
2932  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2933    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2934    #else
2935      c = *previous;
2936  #endif  #endif
2937    if (item == next) return FALSE;    if (c == next) return FALSE;
2938  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2939    if (utf8)    if (utf8)
2940      {      {
# Line 2389  if (next >= 0) switch(op_code) Line 2945  if (next >= 0) switch(op_code)
2945  #else  #else
2946      othercase = NOTACHAR;      othercase = NOTACHAR;
2947  #endif  #endif
2948      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2949      }      }
2950    else    else
2951  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2952    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2953    
2954    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2955      opcodes are not used for multi-byte characters, because they are coded using
2956      an XCLASS instead. */
2957    
2958    case OP_NOT:    case OP_NOT:
2959    if (item == next) return TRUE;    return (c = *previous) == next;
2960    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2961      case OP_NOTI:
2962      if ((c = *previous) == next) return TRUE;
2963  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2964    if (utf8)    if (utf8)
2965      {      {
# Line 2410  if (next >= 0) switch(op_code) Line 2970  if (next >= 0) switch(op_code)
2970  #else  #else
2971      othercase = NOTACHAR;      othercase = NOTACHAR;
2972  #endif  #endif
2973      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2974      }      }
2975    else    else
2976  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2977    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2978    
2979      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2980      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2981    
2982    case OP_DIGIT:    case OP_DIGIT:
2983    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2457  if (next >= 0) switch(op_code) Line 3020  if (next >= 0) switch(op_code)
3020      case 0x202f:      case 0x202f:
3021      case 0x205f:      case 0x205f:
3022      case 0x3000:      case 0x3000:
3023      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
3024      default:      default:
3025      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
3026      }      }
3027    
3028      case OP_ANYNL:
3029    case OP_VSPACE:    case OP_VSPACE:
3030    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3031    switch(next)    switch(next)
# Line 2473  if (next >= 0) switch(op_code) Line 3037  if (next >= 0) switch(op_code)
3037      case 0x85:      case 0x85:
3038      case 0x2028:      case 0x2028:
3039      case 0x2029:      case 0x2029:
3040      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
3041      default:      default:
3042      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
3043      }      }
3044    
3045    #ifdef SUPPORT_UCP
3046      case OP_PROP:
3047      return check_char_prop(next, previous[0], previous[1], FALSE);
3048    
3049      case OP_NOTPROP:
3050      return check_char_prop(next, previous[0], previous[1], TRUE);
3051    #endif
3052    
3053    default:    default:
3054    return FALSE;    return FALSE;
3055    }    }
3056    
3057    
3058  /* Handle the case when the next item is \d, \s, etc. */  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3059    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3060    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3061    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3062    replaced by OP_PROP codes when PCRE_UCP is set. */
3063    
3064  switch(op_code)  switch(op_code)
3065    {    {
3066    case OP_CHAR:    case OP_CHAR:
3067    case OP_CHARNC:    case OP_CHARI:
3068  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3069    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3070    #else
3071      c = *previous;
3072  #endif  #endif
3073    switch(-next)    switch(-next)
3074      {      {
3075      case ESC_d:      case ESC_d:
3076      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3077    
3078      case ESC_D:      case ESC_D:
3079      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3080    
3081      case ESC_s:      case ESC_s:
3082      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3083    
3084      case ESC_S:      case ESC_S:
3085      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3086    
3087      case ESC_w:      case ESC_w:
3088      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3089    
3090      case ESC_W:      case ESC_W:
3091      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3092    
3093      case ESC_h:      case ESC_h:
3094      case ESC_H:      case ESC_H:
3095      switch(item)      switch(c)
3096        {        {
3097        case 0x09:        case 0x09:
3098        case 0x20:        case 0x20:
# Line 2542  switch(op_code) Line 3120  switch(op_code)
3120    
3121      case ESC_v:      case ESC_v:
3122      case ESC_V:      case ESC_V:
3123      switch(item)      switch(c)
3124        {        {
3125        case 0x0a:        case 0x0a:
3126        case 0x0b:        case 0x0b:
# Line 2556  switch(op_code) Line 3134  switch(op_code)
3134        return -next == ESC_v;        return -next == ESC_v;
3135        }        }
3136    
3137        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3138        their substitutions and process them. The result will always be either
3139        -ESC_p or -ESC_P. Then fall through to process those values. */
3140    
3141    #ifdef SUPPORT_UCP
3142        case ESC_du:
3143        case ESC_DU:
3144        case ESC_wu:
3145        case ESC_WU:
3146        case ESC_su:
3147        case ESC_SU:
3148          {
3149          int temperrorcode = 0;
3150          ptr = substitutes[-next - ESC_DU];
3151          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3152          if (temperrorcode != 0) return FALSE;
3153          ptr++;    /* For compatibility */
3154          }
3155        /* Fall through */
3156    
3157        case ESC_p:
3158        case ESC_P:
3159          {
3160          int ptype, pdata, errorcodeptr;
3161          BOOL negated;
3162    
3163          ptr--;      /* Make ptr point at the p or P */
3164          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3165          if (ptype < 0) return FALSE;
3166          ptr++;      /* Point past the final curly ket */
3167    
3168          /* If the property item is optional, we have to give up. (When generated
3169          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3170          to the original \d etc. At this point, ptr will point to a zero byte. */
3171    
3172          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3173            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3174              return FALSE;
3175    
3176          /* Do the property check. */
3177    
3178          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3179          }
3180    #endif
3181    
3182      default:      default:
3183      return FALSE;      return FALSE;
3184      }      }
3185    
3186      /* In principle, support for Unicode properties should be integrated here as
3187      well. It means re-organizing the above code so as to get hold of the property
3188      values before switching on the op-code. However, I wonder how many patterns
3189      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3190      these op-codes are never generated.) */
3191    
3192    case OP_DIGIT:    case OP_DIGIT:
3193    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3194           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3195    
3196    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3197    return next == -ESC_d;    return next == -ESC_d;
3198    
3199    case OP_WHITESPACE:    case OP_WHITESPACE:
3200    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3201    
3202    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3203    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3204    
3205    case OP_HSPACE:    case OP_HSPACE:
3206    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3207             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3208    
3209    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3210    return next == -ESC_h;    return next == -ESC_h;
3211    
3212    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3213      case OP_ANYNL:
3214    case OP_VSPACE:    case OP_VSPACE:
3215    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3216    
3217    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3218    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3219    
3220    case OP_WORDCHAR:    case OP_WORDCHAR:
3221    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3222             next == -ESC_v || next == -ESC_R;
3223    
3224    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3225    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2619  Arguments: Line 3251  Arguments:
3251    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3252    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3253    bcptr          points to current branch chain    bcptr          points to current branch chain
3254      cond_depth     conditional nesting depth
3255    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3256    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3257                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2630  Returns:         TRUE on success Line 3263  Returns:         TRUE on success
3263  static BOOL  static BOOL
3264  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3265    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3266    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3267  {  {
3268  int repeat_type, op_type;  int repeat_type, op_type;
3269  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2639  int greedy_default, greedy_non_default; Line 3272  int greedy_default, greedy_non_default;
3272  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3273  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3274  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3275  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3276  int after_manual_callout = 0;  int after_manual_callout = 0;
3277  int length_prevgroup = 0;  int length_prevgroup = 0;
3278  register int c;  register int c;
# Line 2651  BOOL inescq = FALSE; Line 3284  BOOL inescq = FALSE;
3284  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3285  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3286  const uschar *tempptr;  const uschar *tempptr;
3287    const uschar *nestptr = NULL;
3288  uschar *previous = NULL;  uschar *previous = NULL;
3289  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3290  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3291  uschar classbits[32];  uschar classbits[32];
3292    
3293    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3294    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3295    dynamically as we process the pattern. */
3296    
3297  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3298  BOOL class_utf8;  BOOL class_utf8;
3299  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2664  uschar *class_utf8data_base; Line 3302  uschar *class_utf8data_base;
3302  uschar utf8_char[6];  uschar utf8_char[6];
3303  #else  #else
3304  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3305  #endif  #endif
3306    
3307  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 2715  for (;; ptr++) Line 3352  for (;; ptr++)
3352    int subfirstbyte;    int subfirstbyte;
3353    int terminator;    int terminator;
3354    int mclength;    int mclength;
3355      int tempbracount;
3356    uschar mcbuffer[8];    uschar mcbuffer[8];
3357    
3358    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3359    
3360    c = *ptr;    c = *ptr;
3361    
3362      /* If we are at the end of a nested substitution, revert to the outer level
3363      string. Nesting only happens one level deep. */
3364    
3365      if (c == 0 && nestptr != NULL)
3366        {
3367        ptr = nestptr;
3368        nestptr = NULL;
3369        c = *ptr;
3370        }
3371    
3372    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
3373    previous cycle of this loop. */    previous cycle of this loop. */
3374    
# Line 2729  for (;; ptr++) Line 3377  for (;; ptr++)
3377  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
3378      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3379  #endif  #endif
3380      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + cd->workspace_size -
3381            WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
3382        {        {
3383        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3384        goto FAILED;        goto FAILED;
# Line 2751  for (;; ptr++) Line 3400  for (;; ptr++)
3400        goto FAILED;        goto FAILED;
3401        }        }
3402    
3403      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3404      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, (int)(code - last_code),
3405          c));
3406    
3407      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
3408      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
# Line 2778  for (;; ptr++) Line 3428  for (;; ptr++)
3428    /* In the real compile phase, just check the workspace used by the forward    /* In the real compile phase, just check the workspace used by the forward
3429    reference list. */    reference list. */
3430    
3431    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + cd->workspace_size -
3432               WORK_SIZE_SAFETY_MARGIN)
3433      {      {
3434      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3435      goto FAILED;      goto FAILED;
# Line 2826  for (;; ptr++) Line 3477  for (;; ptr++)
3477      previous_callout = NULL;      previous_callout = NULL;
3478      }      }
3479    
3480    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3481    
3482    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3483      {      {
3484      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3485      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3486        {        {
3487        while (*(++ptr) != 0)        ptr++;
3488          while (*ptr != 0)
3489          {          {
3490          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3491            ptr++;
3492    #ifdef SUPPORT_UTF8
3493            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3494    #endif
3495          }          }
3496        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3497    
# Line 2869  for (;; ptr++) Line 3525  for (;; ptr++)
3525          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3526          goto FAILED;          goto FAILED;
3527          }          }
3528        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3529        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3530        }        }
3531      return TRUE;      return TRUE;
# Line 2880  for (;; ptr++) Line 3536  for (;; ptr++)
3536      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3537    
3538      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3539        previous = NULL;
3540      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3541        {        {
3542        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3543          *code++ = OP_CIRCM;
3544        }        }
3545      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3546      break;      break;
3547    
3548      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3549      previous = NULL;      previous = NULL;
3550      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3551      break;      break;
3552    
3553      /* There can never be a first char if '.' is first, whatever happens about      /* There can never be a first char if '.' is first, whatever happens about
# Line 3074  for (;; ptr++) Line 3731  for (;; ptr++)
3731            ptr++;            ptr++;
3732            }            }
3733    
3734          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3735          if (posix_class < 0)          if (posix_class < 0)
3736            {            {
3737            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 3088  for (;; ptr++) Line 3745  for (;; ptr++)
3745          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3746            posix_class = 0;            posix_class = 0;
3747    
3748          /* We build the bit map for the POSIX class in a chunk of local store          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3749          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3750          subtract bits that may be in the main map already. At the end we or the  
3751          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3752            if ((options & PCRE_UCP) != 0)
3753              {
3754              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3755              if (posix_substitutes[pc] != NULL)
3756                {
3757                nestptr = tempptr + 1;
3758                ptr = posix_substitutes[pc] - 1;
3759                continue;
3760                }
3761              }
3762    #endif
3763            /* In the non-UCP case, we build the bit map for the POSIX class in a
3764            chunk of local store because we may be adding and subtracting from it,
3765            and we don't want to subtract bits that may be in the main map already.
3766            At the end we or the result into the bit map that is being built. */
3767    
3768          posix_class *= 3;          posix_class *= 3;
3769    
# Line 3135  for (;; ptr++) Line 3807  for (;; ptr++)
3807    
3808        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3809        of the specials, which just set a flag. The sequence \b is a special        of the specials, which just set a flag. The sequence \b is a special
3810        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace. We
3811        Elsewhere it marks a word boundary. Other escapes have preset maps ready        assume that other escapes have more than one character in them, so set
3812        to 'or' into the one we are building. We assume they have more than one        class_charcount bigger than one. Unrecognized escapes fall through and
3813        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3814          PCRE_EXTRA is set. */
3815    
3816        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3817          {          {
3818          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3819          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3820    
3821          if (-c == ESC_b) c = CHAR_BS;       /* \b is backspace in a class */          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
3822          else if (-c == ESC_X) c = CHAR_X;   /* \X is literal X in a class */          else if (-c == ESC_N)            /* \N is not supported in a class */
3823          else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */            {
3824              *errorcodeptr = ERR71;
3825              goto FAILED;
3826              }
3827          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3828            {            {
3829            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3164  for (;; ptr++) Line 3840  for (;; ptr++)
3840            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3841            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3842    
3843            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3844              {              {
3845    #ifdef SUPPORT_UCP
3846                case ESC_du:     /* These are the values given for \d etc */
3847                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3848                case ESC_wu:     /* escape sequence with an appropriate \p */
3849                case ESC_WU:     /* or \P to test Unicode properties instead */
3850                case ESC_su:     /* of the default ASCII testing. */
3851                case ESC_SU:
3852                nestptr = ptr;
3853                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3854                class_charcount -= 2;                /* Undo! */
3855                continue;
3856    #endif
3857              case ESC_d:              case ESC_d:
3858              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3859              continue;              continue;
# Line 3186  for (;; ptr++) Line 3872  for (;; ptr++)
3872              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3873              continue;              continue;
3874    
3875                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3876                if it was previously set by something earlier in the character
3877                class. */
3878    
3879              case ESC_s:              case ESC_s:
3880              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3881              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3882                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3883              continue;              continue;
3884    
3885              case ESC_S:              case ESC_S:
# Line 3197  for (;; ptr++) Line 3888  for (;; ptr++)
3888              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3889              continue;              continue;
3890    
3891              default:    /* Not recognized; fall through */              case ESC_h:
             break;      /* Need "default" setting to stop compiler warning. */  
             }  
   
           /* In the pre-compile phase, just do the recognition. */  
   
           else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||  
                    c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;  
   
           /* We need to deal with \H, \h, \V, and \v in both phases because  
           they use extra memory. */  
   
           if (-c == ESC_h)  
             {  
3892              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3893              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3894              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 3234  for (;; ptr++) Line 3912  for (;; ptr++)
3912                }                }
3913  #endif  #endif
3914              continue;              continue;
             }  
3915    
3916            if (-c == ESC_H)              case ESC_H:
             {  
3917              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3918                {                {
3919                int x = 0xff;                int x = 0xff;
# Line 3279  for (;; ptr++) Line 3955  for (;; ptr++)
3955                }                }
3956  #endif  #endif
3957              continue;              continue;
             }  
3958    
3959            if (-c == ESC_v)              case ESC_v:
             {  
3960              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3961              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3962              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3298  for (;; ptr++) Line 3972  for (;; ptr++)
3972                }                }
3973  #endif  #endif
3974              continue;              continue;
             }  
3975    
3976            if (-c == ESC_V)              case ESC_V:
             {  
3977              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3978                {                {
3979                int x = 0xff;                int x = 0xff;
# Line 3331  for (;; ptr++) Line 4003  for (;; ptr++)
4003                }                }
4004  #endif  #endif
4005              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
4006    
4007  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4008            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
4009              {              case ESC_P:
4010              BOOL negated;                {
4011              int pdata;                BOOL negated;
4012              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
4013              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
4014              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
4015              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
4016                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
4017              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
4018              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
4019              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
4020              continue;                class_charcount -= 2;   /* Not a < 256 character */
4021              }                continue;
4022                  }
4023  #endif  #endif
4024            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
4025            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
4026            treated as literals. */              treated as literals. */
4027    
4028            if ((options & PCRE_EXTRA) != 0)              default:
4029              {              if ((options & PCRE_EXTRA) != 0)
4030              *errorcodeptr = ERR7;                {
4031              goto FAILED;                *errorcodeptr = ERR7;
4032                  goto FAILED;
4033                  }
4034                class_charcount -= 2;  /* Undo the default count from above */
4035                c = *ptr;              /* Get the final character and fall through */
4036                break;
4037              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
4038            }            }
4039    
4040          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
# Line 3432  for (;; ptr++) Line 4104  for (;; ptr++)
4104            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
4105            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
4106    
4107            /* \b is backspace; \X is literal X; \R is literal R; any other            /* \b is backspace; any other special means the '-' was literal */
           special means the '-' was literal */  
4108    
4109            if (d < 0)            if (d < 0)
4110              {              {
4111              if (d == -ESC_b) d = CHAR_BS;              if (d == -ESC_b) d = CHAR_BS; else
             else if (d == -ESC_X) d = CHAR_X;  
             else if (d == -ESC_R) d = CHAR_R; else  
4112                {                {
4113                ptr = oldptr;                ptr = oldptr;
4114                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3605  for (;; ptr++) Line 4274  for (;; ptr++)
4274          }          }
4275        }        }
4276    
4277      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" far above.
4278        If we are at the end of an internal nested string, revert to the outer
4279        string. */
4280    
4281        while (((c = *(++ptr)) != 0 ||
4282               (nestptr != NULL &&
4283                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4284               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4285    
4286      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
4287    
4288      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4289        {        {
4290        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4291        goto FAILED;        goto FAILED;
4292        }        }
4293    
   
 /* This code has been disabled because it would mean that \s counts as  
 an explicit \r or \n reference, and that's not really what is wanted. Now  
 we set the flag only if there is a literal "\r" or "\n" in the class. */  
   
 #if 0  
     /* Remember whether \r or \n are in this class */  
   
     if (negate_class)  
       {  
       if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;  
       }  
     else  
       {  
       if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;  
       }  
 #endif  
   
   
4294      /* If class_charcount is 1, we saw precisely one character whose value is      /* If class_charcount is 1, we saw precisely one character whose value is
4295      less than 256. As long as there were no characters >= 128 and there was no      less than 256. As long as there were no characters >= 128 and there was no
4296      use of \p or \P, in other words, no use of any XCLASS features, we can      use of \p or \P, in other words, no use of any XCLASS features, we can
# Line 3641  we set the flag only if there is a liter Line 4298  we set the flag only if there is a liter
4298    
4299      In UTF-8 mode, we can optimize the negative case only if there were no      In UTF-8 mode, we can optimize the negative case only if there were no
4300      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4301      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4302      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4303    
4304      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
4305      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4306      that OP_NOT does not support multibyte characters. In the positive case, it      Note that OP_NOT[I] does not support multibyte characters. In the positive
4307      can cause firstbyte to be set. Otherwise, there can be no first char if      case, it can cause firstbyte to be set. Otherwise, there can be no first
4308      this item is first, whatever repeat count may follow. In the case of      char if this item is first, whatever repeat count may follow. In the case
4309      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4310    
4311  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4312      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3660  we set the flag only if there is a liter Line 4317  we set the flag only if there is a liter
4317        {        {
4318        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4319    
4320        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4321    
4322        if (negate_class)        if (negate_class)
4323          {          {
4324          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4325          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4326          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4327          *code++ = class_lastchar;          *code++ = class_lastchar;
4328          break;          break;
4329          }          }
# Line 3697  we set the flag only if there is a liter Line 4354  we set the flag only if there is a liter
4354    
4355      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4356      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4357      such as \S in the class, because in that case all characters > 255 are in      such as \S in the class, and PCRE_UCP is not set, because in that case all
4358      the class, so any that were explicitly given as well can be ignored. If      characters > 255 are in the class, so any that were explicitly given as
4359      (when there are explicit characters > 255 that must be listed) there are no      well can be ignored. If (when there are explicit characters > 255 that must
4360      characters < 256, we can omit the bitmap in the actual compiled code. */      be listed) there are no characters < 256, we can omit the bitmap in the
4361        actual compiled code. */
4362    
4363  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4364      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4365        {        {
4366        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4367        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3729  we set the flag only if there is a liter Line 4387  we set the flag only if there is a liter
4387        }        }
4388  #endif  #endif
4389    
4390      /* If there are no characters > 255, set the opcode to OP_CLASS or      /* If there are no characters > 255, or they are all to be included or
4391      OP_NCLASS, depending on whether the whole class was negated and whether      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
4392      there were negative specials such as \S in the class. Then copy the 32-byte      whole class was negated and whether there were negative specials such as \S
4393      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4394        negating it if necessary. */
4395    
4396      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4397      if (negate_class)      if (negate_class)
# Line 3792  we set the flag only if there is a liter Line 4451  we set the flag only if there is a liter
4451      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4452      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4453    
4454      /* Save start of previous item, in case we have to move it up to make space      /* Save start of previous item, in case we have to move it up in order to
4455      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4456    
4457      tempcode = previous;      tempcode = previous;
4458    
# Line 3816  we set the flag only if there is a liter Line 4475  we set the flag only if there is a liter
4475        }        }
4476      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4477    
4478        /* If previous was a recursion call, wrap it in atomic brackets so that
4479        previous becomes the atomic group. All recursions were so wrapped in the
4480        past, but it no longer happens for non-repeated recursions. In fact, the
4481        repeated ones could be re-implemented independently so as not to need this,
4482        but for the moment we rely on the code for repeating groups. */
4483    
4484        if (*previous == OP_RECURSE)
4485          {
4486          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4487          *previous = OP_ONCE;
4488          PUT(previous, 1, 2 + 2*LINK_SIZE);
4489          previous[2 + 2*LINK_SIZE] = OP_KET;
4490          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4491          code += 2 + 2 * LINK_SIZE;
4492          length_prevgroup = 3 + 3*LINK_SIZE;
4493    
4494          /* When actually compiling, we need to check whether this was a forward
4495          reference, and if so, adjust the offset. */
4496    
4497          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4498            {
4499            int offset = GET(cd->hwm, -LINK_SIZE);
4500            if (offset == previous + 1 - cd->start_code)
4501              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4502            }
4503          }
4504    
4505        /* Now handle repetition for the different types of item. */
4506    
4507      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4508      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
4509      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
4510      the first thing in a branch because the x will have gone into firstbyte      the first thing in a branch because the x will have gone into firstbyte
4511      instead.  */      instead.  */
4512    
4513      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4514        {        {
4515          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4516    
4517        /* Deal with UTF-8 characters that take up more than one byte. It's        /* Deal with UTF-8 characters that take up more than one byte. It's
4518        easier to write this out separately than try to macrify it. Use c to        easier to write this out separately than try to macrify it. Use c to
4519        hold the length of the character in bytes, plus 0x80 to flag that it's a        hold the length of the character in bytes, plus 0x80 to flag that it's a
# Line 3856  we set the flag only if there is a liter Line 4546  we set the flag only if there is a liter
4546    
4547        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4548            repeat_max < 0 &&            repeat_max < 0 &&
4549            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4550          {          {
4551          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4552          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3869  we set the flag only if there is a liter Line 4558  we set the flag only if there is a liter
4558      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4559      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4560      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4561      repeat_type. We can also test for auto-possessification. OP_NOT is      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4562      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4563    
4564      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4565        {        {
4566        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4567        c = previous[1];        c = previous[1];
4568        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4569            repeat_max < 0 &&            repeat_max < 0 &&
4570            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4571          {          {
4572          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4573          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3902  we set the flag only if there is a liter Line 4591  we set the flag only if there is a liter
4591    
4592        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4593            repeat_max < 0 &&            repeat_max < 0 &&
4594            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4595          {          {
4596          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4597          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4071  we set the flag only if there is a liter Line 4760  we set the flag only if there is a liter
4760  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4761               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4762  #endif  #endif
4763               *previous == OP_REF)               *previous == OP_REF ||
4764                 *previous == OP_REFI)
4765        {        {
4766        if (repeat_max == 0)        if (repeat_max == 0)
4767          {          {
# Line 4105  we set the flag only if there is a liter Line 4795  we set the flag only if there is a liter
4795        }        }
4796    
4797      /* 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
4798      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4799        opcodes such as BRA and CBRA, as this is the place where they get converted
4800        into the more special varieties such as BRAPOS and SBRA. A test for >=
4801        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4802        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4803        repetition of assertions, but now it does, for Perl compatibility. */
4804    
4805      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4806        {        {
4807        register int i;        register int i;
4808        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4809        uschar *bralink = NULL;        uschar *bralink = NULL;
4810          uschar *brazeroptr = NULL;
4811    
4812        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4813          we just ignore the repeat. */
4814    
4815        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4816          {          goto END_REPEAT;
4817          *errorcodeptr = ERR55;  
4818          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4819          }        use of repetition is in cases when the assertion is optional. Therefore,
4820          if the minimum is greater than zero, just ignore the repeat. If the
4821          maximum is not not zero or one, set it to 1. */
4822    
4823        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4824        by scanning through from the start, and compute the offset back to it          {
4825        from the current code pointer. There may be an OP_OPT setting following          if (repeat_min > 0) goto END_REPEAT;
4826        the final KET, so we can't find the end just by going back from the code          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
       pointer. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = code - ket;  
4827          }          }
4828    
4829        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 4154  we set the flag only if there is a liter Line 4844  we set the flag only if there is a liter
4844          **   goto END_REPEAT;          **   goto END_REPEAT;
4845          **   }          **   }
4846    
4847          However, that fails when a group is referenced as a subroutine from          However, that fails when a group or a subgroup within it is referenced
4848          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it          as a subroutine from elsewhere in the pattern, so now we stick in
4849          so that it is skipped on execution. As we don't have a list of which          OP_SKIPZERO in front of it so that it is skipped on execution. As we
4850          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4851            selectively.
4852    
4853          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4854          and do no more at this point. However, we do need to adjust any          and do no more at this point. However, we do need to adjust any
# Line 4177  we set the flag only if there is a liter Line 4868  we set the flag only if there is a liter
4868              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4869              goto END_REPEAT;              goto END_REPEAT;
4870              }              }
4871              brazeroptr = previous;    /* Save for possessive optimizing */
4872            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4873            }            }
4874    
# Line 4201  we set the flag only if there is a liter Line 4893  we set the flag only if there is a liter
4893            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4894            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4895    
4896            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4897            bralink = previous;            bralink = previous;
4898            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4899            }            }
# Line 4239  we set the flag only if there is a liter Line 4931  we set the flag only if there is a liter
4931              *lengthptr += delta;              *lengthptr += delta;
4932              }              }
4933    
4934            /* This is compiling for real */            /* This is compiling for real. If there is a set first byte for
4935              the group, and we have not yet set a "required byte", set it. Make
4936              sure there is enough workspace for copying forward references before
4937              doing the copy. */
4938    
4939            else            else
4940              {              {
4941              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
4942    
4943              for (i = 1; i < repeat_min; i++)              for (i = 1; i < repeat_min; i++)
4944                {                {
4945                uschar *hc;                uschar *hc;
4946                uschar *this_hwm = cd->hwm;                uschar *this_hwm = cd->hwm;
4947                memcpy(code, previous, len);                memcpy(code, previous, len);
4948    
4949                  while (cd->hwm > cd->start_workspace + cd->workspace_size -
4950                         WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
4951                    {
4952                    int save_offset = save_hwm - cd->start_workspace;
4953                    int this_offset = this_hwm - cd->start_workspace;
4954                    *errorcodeptr = expand_workspace(cd);
4955                    if (*errorcodeptr != 0) goto FAILED;
4956                    save_hwm = (uschar *)cd->start_workspace + save_offset;
4957                    this_hwm = (uschar *)cd->start_workspace + this_offset;
4958                    }
4959    
4960                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
4961                  {                  {
4962                  PUT(cd->hwm, 0, GET(hc, 0) + len);                  PUT(cd->hwm, 0, GET(hc, 0) + len);
# Line 4278  we set the flag only if there is a liter Line 4986  we set the flag only if there is a liter
4986          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4987          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4988          a 64-bit integer type when available, otherwise double. */          a 64-bit integer type when available, otherwise double. */
4989    
4990          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4991            {            {
4992            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
# Line 4310  we set the flag only if there is a liter Line 5018  we set the flag only if there is a liter
5018              {              {
5019              int offset;              int offset;
5020              *code++ = OP_BRA;              *code++ = OP_BRA;
5021              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
5022              bralink = code;              bralink = code;
5023              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
5024              }              }
5025    
5026            memcpy(code, previous, len);            memcpy(code, previous, len);
5027    
5028              /* Ensure there is enough workspace for forward references before
5029              copying them. */
5030    
5031              while (cd->hwm > cd->start_workspace + cd->workspace_size -
5032                     WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5033                {
5034                int save_offset = save_hwm - cd->start_workspace;
5035                int this_offset = this_hwm - cd->start_workspace;
5036                *errorcodeptr = expand_workspace(cd);
5037                if (*errorcodeptr != 0) goto FAILED;
5038                save_hwm = (uschar *)cd->start_workspace + save_offset;
5039                this_hwm = (uschar *)cd->start_workspace + this_offset;
5040                }
5041    
5042            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5043              {              {
5044              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
# Line 4331  we set the flag only if there is a liter Line 5054  we set the flag only if there is a liter
5054          while (bralink != NULL)          while (bralink != NULL)
5055            {            {
5056            int oldlinkoffset;            int oldlinkoffset;
5057            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
5058            uschar *bra = code - offset;            uschar *bra = code - offset;
5059            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
5060            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4341  we set the flag only if there is a liter Line 5064  we set the flag only if there is a liter
5064            }            }
5065          }          }
5066    
5067        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. For
5068        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
5069        don't know if there's been an options resetting after the ket. The        ONCE brackets can be converted into non-capturing brackets, as the
5070        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
5071          deal with possessive ONCEs specially.
5072        Then, when we are doing the actual compile phase, check to see whether  
5073        this group is a non-atomic one that could match an empty string. If so,        Otherwise, when we are doing the actual compile phase, check to see
5074          whether this group is one that could match an empty string. If so,
5075        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
5076        that runtime checking can be done. [This check is also applied to        that runtime checking can be done. [This check is also applied to ONCE
5077        atomic groups at runtime, but in a different way.] */        groups at runtime, but in a different way.]
5078    
5079          Then, if the quantifier was possessive and the bracket is not a
5080          conditional, we convert the BRA code to the POS form, and the KET code to
5081          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
5082          subpattern at both the start and at the end.) The use of special opcodes
5083          makes it possible to reduce greatly the stack usage in pcre_exec(). If
5084          the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
5085    
5086          Then, if the minimum number of matches is 1 or 0, cancel the possessive
5087          flag so that the default action below, of wrapping everything inside
5088          atomic brackets, does not happen. When the minimum is greater than 1,
5089          there will be earlier copies of the group, and so we still have to wrap
5090          the whole thing. */
5091    
5092        else        else
5093          {          {
5094          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
5095          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
5096          *ketcode = OP_KETRMAX + repeat_type;  
5097          if (lengthptr == NULL && *bracode != OP_ONCE)          /* Convert possessive ONCE brackets to non-capturing */
5098    
5099            if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5100                possessive_quantifier) *bracode = OP_BRA;
5101    
5102            /* For non-possessive ONCE brackets, all we need to do is to
5103            set the KET. */
5104    
5105            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5106              *ketcode = OP_KETRMAX + repeat_type;
5107    
5108            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5109            converted to non-capturing above). */
5110    
5111            else
5112            {            {
5113            uschar *scode = bracode;            /* In the compile phase, check for empty string matching. */
5114            do  
5115              if (lengthptr == NULL)
5116              {              {
5117              if (could_be_empty_branch(scode, ketcode, utf8))              uschar *scode = bracode;
5118                do
5119                {                {
5120                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
5121                break;                  {
5122                    *bracode += OP_SBRA - OP_BRA;
5123                    break;
5124                    }
5125                  scode += GET(scode, 1);
5126                }                }
5127              scode += GET(scode, 1);              while (*scode == OP_ALT);
5128              }              }
5129            while (*scode == OP_ALT);  
5130              /* Handle possessive quantifiers. */
5131    
5132              if (possessive_quantifier)
5133                {
5134                /* For COND brackets, we wrap the whole thing in a possessively
5135                repeated non-capturing bracket, because we have not invented POS
5136                versions of the COND opcodes. Because we are moving code along, we
5137                must ensure that any pending recursive references are updated. */
5138    
5139                if (*bracode == OP_COND || *bracode == OP_SCOND)
5140                  {
5141                  int nlen = (int)(code - bracode);
5142                  *code = OP_END;
5143                  adjust_recurse(bracode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5144                  memmove(bracode + 1+LINK_SIZE, bracode, nlen);
5145                  code += 1 + LINK_SIZE;
5146                  nlen += 1 + LINK_SIZE;
5147                  *bracode = OP_BRAPOS;
5148                  *code++ = OP_KETRPOS;
5149                  PUTINC(code, 0, nlen);
5150                  PUT(bracode, 1, nlen);
5151                  }
5152    
5153                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5154    
5155                else
5156                  {
5157                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5158                  *ketcode = OP_KETRPOS;
5159                  }
5160    
5161                /* If the minimum is zero, mark it as possessive, then unset the
5162                possessive flag when the minimum is 0 or 1. */
5163    
5164                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5165                if (repeat_min < 2) possessive_quantifier = FALSE;
5166                }
5167    
5168              /* Non-possessive quantifier */
5169    
5170              else *ketcode = OP_KETRMAX + repeat_type;
5171            }            }
5172          }          }
5173        }        }
# Line 4390  we set the flag only if there is a liter Line 5188  we set the flag only if there is a liter
5188        }        }
5189    
5190      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
5191      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
5192      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
5193      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
5194      The '+' notation is just syntactic sugar, taken from Sun's Java package,      notation is just syntactic sugar, taken from Sun's Java package, but the
5195      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
5196      tempcode, not at previous, which might be the first part of a string whose  
5197      (former) last char we repeated.      Some (but not all) possessively repeated subpatterns have already been
5198        completely handled in the code just above. For them, possessive_quantifier
5199        is always FALSE at this stage.
5200    
5201        Note that the repeated item starts at tempcode, not at previous, which
5202        might be the first part of a string whose (former) last char we repeated.
5203    
5204      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
5205      an 'upto' may follow. We skip over an 'exact' item, and then test the      an 'upto' may follow. We skip over an 'exact' item, and then test the
# Line 4419  we set the flag only if there is a liter Line 5222  we set the flag only if there is a liter
5222  #endif  #endif
5223          }          }
5224    
5225        len = code - tempcode;        len = (int)(code - tempcode);
5226        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
5227          {          {
5228          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4427  we set the flag only if there is a liter Line 5230  we set the flag only if there is a liter
5230          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
5231          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
5232    
5233          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
5234          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
5235          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
5236          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
5237    
5238          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
5239          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
5240          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
5241          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
5242    
5243            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
5244            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
5245            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
5246            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
5247    
5248            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
5249            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
5250            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
5251            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
5252    
5253          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
5254          pending recursive references are updated. */          pending recursive references are updated. */
5255    
5256          default:          default:
5257          *code = OP_END;          *code = OP_END;
5258          adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);          adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
# Line 4478  we set the flag only if there is a liter Line 5291  we set the flag only if there is a liter
5291    
5292      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
5293    
5294      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
5295             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
5296        {        {
5297        int i, namelen;        int i, namelen;
5298          int arglen = 0;
5299        const char *vn = verbnames;        const char *vn = verbnames;
5300        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
5301          const uschar *arg = NULL;
5302        previous = NULL;        previous = NULL;
5303        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5304          namelen = (int)(ptr - name);
5305    
5306          /* It appears that Perl allows any characters whatsoever, other than
5307          a closing parenthesis, to appear in arguments, so we no longer insist on
5308          letters, digits, and underscores. */
5309    
5310        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5311          {          {
5312          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
5313          goto FAILED;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5314            arglen = (int)(ptr - arg);
5315          }          }
5316    
5317        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
5318          {          {
5319          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
5320          goto FAILED;          goto FAILED;
5321          }          }
5322        namelen = ptr - name;  
5323          /* Scan the table of verb names */
5324    
5325        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
5326          {          {
5327          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5328              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5329            {            {
5330            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5331              ASSERT_ACCEPT if in an assertion. */
5332    
5333            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5334              {              {
5335              open_capitem *oc;              open_capitem *oc;
5336                if (arglen != 0)
5337                  {
5338                  *errorcodeptr = ERR59;
5339                  goto FAILED;
5340                  }
5341              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5342              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5343                {                {
5344                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5345                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5346                }                }
5347                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5348    
5349                /* Do not set firstbyte after *ACCEPT */
5350                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5351              }              }
5352            *code++ = verbs[i].op;  
5353            break;            /* Handle other cases with/without an argument */
5354    
5355              else if (arglen == 0)
5356                {
5357                if (verbs[i].op < 0)   /* Argument is mandatory */
5358                  {
5359                  *errorcodeptr = ERR66;
5360                  goto FAILED;
5361                  }
5362                *code = verbs[i].op;
5363                if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
5364                }
5365    
5366              else
5367                {
5368                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
5369                  {
5370                  *errorcodeptr = ERR59;
5371                  goto FAILED;
5372                  }
5373                *code = verbs[i].op_arg;
5374                if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
5375                *code++ = arglen;
5376                memcpy(code, arg, arglen);
5377                code += arglen;
5378                *code++ = 0;
5379                }
5380    
5381              break;  /* Found verb, exit loop */
5382            }            }
5383    
5384          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
5385          }          }
5386        if (i < verbcount) continue;  
5387        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
5388          *errorcodeptr = ERR60;          /* Verb not recognized */
5389        goto FAILED;        goto FAILED;
5390        }        }
5391    
# Line 4638  we set the flag only if there is a liter Line 5504  we set the flag only if there is a liter
5504                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5505            ptr++;            ptr++;
5506            }            }
5507          namelen = ptr - name;          namelen = (int)(ptr - name);
5508    
5509          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
5510              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
# Line 4699  we set the flag only if there is a liter Line 5565  we set the flag only if there is a liter
5565          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5566    
5567          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5568                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5569            {            {
5570            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5571            code[1+LINK_SIZE]++;            code[1+LINK_SIZE]++;
# Line 4767  we set the flag only if there is a liter Line 5633  we set the flag only if there is a liter
5633          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5634          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5635          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5636            cd->assert_depth += 1;
5637          ptr++;          ptr++;
5638          break;          break;
5639    
# Line 4781  we set the flag only if there is a liter Line 5648  we set the flag only if there is a liter
5648            continue;            continue;
5649            }            }
5650          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5651            cd->assert_depth += 1;
5652          break;          break;
5653    
5654    
# Line 4790  we set the flag only if there is a liter Line 5658  we set the flag only if there is a liter
5658            {            {
5659            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5660            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5661              cd->assert_depth += 1;
5662            ptr += 2;            ptr += 2;
5663            break;            break;
5664    
5665            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5666            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5667              cd->assert_depth += 1;
5668            ptr += 2;            ptr += 2;
5669            break;            break;
5670    
# Line 4816  we set the flag only if there is a liter Line 5686  we set the flag only if there is a liter
5686    
5687          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5688          case CHAR_C:                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
5689          previous_callout = code;  /* Save for later completion */          previous_callout = code;     /* Save for later completion */
5690          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1;    /* Skip one item before completing */
5691          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
5692            {            {
5693            int n = 0;            int n = 0;
# Line 4834  we set the flag only if there is a liter Line 5704  we set the flag only if there is a liter
5704              goto FAILED;              goto FAILED;
5705              }              }
5706            *code++ = n;            *code++ = n;
5707            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5708            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5709            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5710            }            }
5711          previous = NULL;          previous = NULL;
# Line 4868  we set the flag only if there is a liter Line 5738  we set the flag only if there is a liter
5738            name = ++ptr;            name = ++ptr;
5739    
5740            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5741            namelen = ptr - name;            namelen = (int)(ptr - name);
5742    
5743            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5744    
# Line 4998  we set the flag only if there is a liter Line 5868  we set the flag only if there is a liter
5868          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5869          name = ++ptr;          name = ++ptr;
5870          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5871          namelen = ptr - name;          namelen = (int)(ptr - name);
5872    
5873          /* In the pre-compile phase, do a syntax check and set a dummy          /* In the pre-compile phase, do a syntax check. We used to just set
5874          reference number. */          a dummy reference number, because it was not used in the first pass.
5875            However, with the change of recursive back references to be atomic,
5876            we have to look for the number so that this state can be identified, as
5877            otherwise the incorrect length is computed. If it's not a backwards
5878            reference, the dummy number will do. */
5879    
5880          if (lengthptr != NULL)          if (lengthptr != NULL)
5881            {            {
5882              const uschar *temp;
5883    
5884            if (namelen == 0)            if (namelen == 0)
5885              {              {
5886              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5020  we set the flag only if there is a liter Line 5896  we set the flag only if there is a liter
5896              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5897              goto FAILED;              goto FAILED;
5898              }              }
5899            recno = 0;  
5900              /* The name table does not exist in the first pass, so we cannot
5901              do a simple search as in the code below. Instead, we have to scan the
5902              pattern to find the number. It is important that we scan it only as
5903              far as we have got because the syntax of named subpatterns has not
5904              been checked for the rest of the pattern, and find_parens() assumes
5905              correct syntax. In any case, it's a waste of resources to scan
5906              further. We stop the scan at the current point by temporarily
5907              adjusting the value of cd->endpattern. */
5908    
5909              temp = cd->end_pattern;
5910              cd->end_pattern = ptr;
5911              recno = find_parens(cd, name, namelen,
5912                (options & PCRE_EXTENDED) != 0, utf8);
5913              cd->end_pattern = temp;
5914              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5915            }            }
5916    
5917          /* In the real compile, seek the name in the table. We check the name          /* In the real compile, seek the name in the table. We check the name
# Line 5045  we set the flag only if there is a liter Line 5936  we set the flag only if there is a liter
5936              }              }
5937            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5938                      find_parens(cd, name, namelen,                      find_parens(cd, name, namelen,
5939                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5940              {              {
5941              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5942              goto FAILED;              goto FAILED;
# Line 5156  we set the flag only if there is a liter Line 6047  we set the flag only if there is a liter
6047              if (called == NULL)              if (called == NULL)
6048                {                {
6049                if (find_parens(cd, NULL, recno,                if (find_parens(cd, NULL, recno,
6050                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
6051                  {                  {
6052                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
6053                  goto FAILED;                  goto FAILED;
6054                  }                  }
6055    
6056                /* Fudge the value of "called" so that when it is inserted as an                /* Fudge the value of "called" so that when it is inserted as an
6057                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
6058                of the group. */                of the group. Then remember the forward reference. */
6059    
6060                called = cd->start_code + recno;                called = cd->start_code + recno;
6061                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                if (cd->hwm >= cd->start_workspace + cd->workspace_size -
6062                      WORK_SIZE_SAFETY_MARGIN)
6063                    {
6064                    *errorcodeptr = expand_workspace(cd);
6065                    if (*errorcodeptr != 0) goto FAILED;
6066                    }
6067                  PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
6068                }                }
6069    
6070              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
6071              this is a recursive call. We check to see if this is a left              this is a recursive call. We check to see if this is a left
6072              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
6073                must not, however, do this check if we are in a conditional
6074                subpattern because the condition might be testing for recursion in
6075                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
6076                Forever loops are also detected at runtime, so those that occur in
6077                conditional subpatterns will be picked up then. */
6078    
6079              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
6080                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
6081                {                {
6082                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
6083                goto FAILED;                goto FAILED;
6084                }                }
6085              }              }
6086    
6087            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. It does not have a set first
6088            "once" brackets. Set up a "previous group" length so that a            byte (relevant if it is repeated, because it will then be wrapped
6089            subsequent quantifier will work. */            with ONCE brackets). */
   
           *code = OP_ONCE;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
6090    
6091            *code = OP_RECURSE;            *code = OP_RECURSE;
6092            PUT(code, 1, called - cd->start_code);            PUT(code, 1, (int)(called - cd->start_code));
           code += 1 + LINK_SIZE;  
   
           *code = OP_KET;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
6093            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
6094              groupsetfirstbyte = FALSE;
           length_prevgroup = 3 + 3*LINK_SIZE;  
6095            }            }
6096    
6097          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5259  we set the flag only if there is a liter Line 6152  we set the flag only if there is a liter
6152          is necessary to ensure we correctly detect the start of the pattern in          is necessary to ensure we correctly detect the start of the pattern in
6153          both phases.          both phases.
6154    
6155          If we are not at the pattern start, compile code to change the ims          If we are not at the pattern start, reset the greedy defaults and the
6156          options if this setting actually changes any of them, and reset the          case value for firstbyte and reqbyte. */
         greedy defaults and the case value for firstbyte and reqbyte. */  
6157    
6158          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
6159            {            {
# Line 5272  we set the flag only if there is a liter Line 6164  we set the flag only if there is a liter
6164              }              }
6165            else            else
6166              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
6167              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
6168              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
6169              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
6170              }              }
6171    
6172            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
6173            in subsequent branches. When not at the start of the pattern, this            in subsequent branches. */
           information is also necessary so that a resetting item can be  
           compiled at the end of a group (if we are in a group). */  
6174    
6175            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
6176            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5302  we set the flag only if there is a liter Line 6187  we set the flag only if there is a liter
6187          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
6188        }       /* End of (? handling */        }       /* End of (? handling */
6189    
6190      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
6191      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
6192      brackets. */      brackets. */
6193    
6194      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 5321  we set the flag only if there is a liter Line 6206  we set the flag only if there is a liter
6206        skipbytes = 2;        skipbytes = 2;
6207        }        }
6208    
6209      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
6210      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a      but this was changed for Perl compatibility, so all kinds can now be
6211      non-register variable in order to be able to pass its address because some      repeated. We copy code into a non-register variable (tempcode) in order to
6212      compilers complain otherwise. Pass in a new setting for the ims options if      be able to pass its address because some compilers complain otherwise. */
     they have changed. */  
6213    
6214      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                      /* For handling repetition */
6215      *code = bravalue;      *code = bravalue;
6216      tempcode = code;      tempcode = code;
6217      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
6218      length_prevgroup = 0;              /* Initialize for pre-compile phase */      tempbracount = cd->bracount;          /* Save value before bracket */
6219        length_prevgroup = 0;                 /* Initialize for pre-compile phase */
6220    
6221      if (!compile_regex(      if (!compile_regex(
6222           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
6223           options & PCRE_IMS,           /* The previous ims option state */           &tempcode,                       /* Where to put code (updated) */
6224           &tempcode,                    /* Where to put code (updated) */           &ptr,                            /* Input pointer (updated) */
6225           &ptr,                         /* Input pointer (updated) */           errorcodeptr,                    /* Where to put an error message */
          errorcodeptr,                 /* Where to put an error message */  
6226           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
6227            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
6228           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
6229           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
6230           &subfirstbyte,                /* For possible first char */           cond_depth +
6231           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
6232           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
6233           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
6234           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
6235             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
6236             (lengthptr == NULL)? NULL :      /* Actual compile phase */
6237               &length_prevgroup              /* Pre-compile phase */
6238           ))           ))
6239        goto FAILED;        goto FAILED;
6240    
6241        /* If this was an atomic group and there are no capturing groups within it,
6242        generate OP_ONCE_NC instead of OP_ONCE. */
6243    
6244        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
6245          *code = OP_ONCE_NC;
6246    
6247        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
6248          cd->assert_depth -= 1;
6249    
6250      /* At the end of compiling, code is still pointing to the start of the      /* At the end of compiling, code is still pointing to the start of the
6251      group, while tempcode has been updated to point past the end of the group      group, while tempcode has been updated to point past the end of the group.
6252      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
6253    
6254      /* If this is a conditional bracket, check that there are no more than      If this is a conditional bracket, check that there are no more than
6255      two branches in the group, or just one if it's a DEFINE group. We do this      two branches in the group, or just one if it's a DEFINE group. We do this
6256      in the real compile phase, not in the pre-pass, where the whole group may      in the real compile phase, not in the pre-pass, where the whole group may
6257      not be available. */      not be available. */
# Line 5422  we set the flag only if there is a liter Line 6316  we set the flag only if there is a liter
6316          goto FAILED;          goto FAILED;
6317          }          }
6318        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
6319        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
6320        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
6321        *code++ = OP_KET;        *code++ = OP_KET;
6322        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5495  we set the flag only if there is a liter Line 6389  we set the flag only if there is a liter
6389    
6390      /* ===================================================================*/      /* ===================================================================*/
6391      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
6392      are arranged to be the negation of the corresponding OP_values. For the      are arranged to be the negation of the corresponding OP_values in the
6393      back references, the values are ESC_REF plus the reference number. Only      default case when PCRE_UCP is not set. For the back references, the values
6394      back references and those types that consume a character may be repeated.      are ESC_REF plus the reference number. Only back references and those types
6395      We can test for values between ESC_b and ESC_Z for the latter; this may      that consume a character may be repeated. We can test for values between
6396      have to change if any new ones are ever created. */      ESC_b and ESC_Z for the latter; this may have to change if any new ones are
6397        ever created. */
6398    
6399      case CHAR_BACKSLASH:      case CHAR_BACKSLASH:
6400      tempptr = ptr;      tempptr = ptr;
# Line 5589  we set the flag only if there is a liter Line 6484  we set the flag only if there is a liter
6484          }          }
6485    
6486        /* \k<name> or \k'name' is a back reference by name (Perl syntax).        /* \k<name> or \k'name' is a back reference by name (Perl syntax).
6487        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6488    
6489        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6490          {          {
6491            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6492              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6493              {
6494              *errorcodeptr = ERR69;
6495              break;
6496              }
6497          is_recurse = FALSE;          is_recurse = FALSE;
6498          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6499            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
# Line 5613  we set the flag only if there is a liter Line 6513  we set the flag only if there is a liter
6513          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6514          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6515          previous = code;          previous = code;
6516          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6517          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6518          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6519          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
# Line 5659  we set the flag only if there is a liter Line 6559  we set the flag only if there is a liter
6559  #endif  #endif
6560    
6561        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
6562        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
6563          situation when PCRE_UCP is not set. When it *is* set, we substitute
6564          Unicode property tests. */
6565    
6566        else        else
6567          {          {
6568          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
6569          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
6570              {
6571              nestptr = ptr + 1;                   /* Where to resume */
6572              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
6573              }
6574            else
6575    #endif
6576            /* In non-UTF-8 mode, we turn \C into OP_ALLANY instead of OP_ANYBYTE
6577            so that it works in DFA mode and in lookbehinds. */
6578    
6579              {
6580              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6581              *code++ = (!utf8 && c == -ESC_C)? OP_ALLANY : -c;
6582              }
6583          }          }
6584        continue;        continue;
6585        }        }
# Line 5709  we set the flag only if there is a liter Line 6624  we set the flag only if there is a liter
6624    
6625      ONE_CHAR:      ONE_CHAR:
6626      previous = code;      previous = code;
6627      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6628      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6629    
6630      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 5738  we set the flag only if there is a liter Line 6653  we set the flag only if there is a liter
6653        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6654        }        }
6655    
6656      /* firstbyte was previously set; we can set reqbyte only the length is      /* firstbyte was previously set; we can set reqbyte only if the length is
6657      1 or the matching is caseful. */      1 or the matching is caseful. */
6658    
6659      else      else
# Line 5773  return FALSE; Line 6688  return FALSE;
6688  /* On entry, ptr is pointing past the bracket character, but on return it  /* On entry, ptr is pointing past the bracket character, but on return it
6689  points to the closing bracket, or vertical bar, or end of string. The code  points to the closing bracket, or vertical bar, or end of string. The code
6690  variable is pointing at the byte into which the BRA operator has been stored.  variable is pointing at the byte into which the BRA operator has been stored.
 If the ims options are changed at the start (for a (?ims: group) or during any  
 branch, we need to insert an OP_OPT item at the start of every following branch  
 to ensure they get set correctly at run time, and also pass the new options  
 into every subsequent branch compile.  
   
6691  This function is used during the pre-compile phase when we are trying to find  This function is used during the pre-compile phase when we are trying to find
6692  out the amount of memory needed, as well as during the real compile phase. The  out the amount of memory needed, as well as during the real compile phase. The
6693  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6694    
6695  Arguments:  Arguments:
6696    options        option bits, including any changes for this subpattern    options        option bits, including any changes for this subpattern
   oldims         previous settings of ims option bits  
6697    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6698    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6699    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
6700    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6701    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6702    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6703      cond_depth     depth of nesting for conditional subpatterns
6704    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6705    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6706    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 5802  Returns:         TRUE on success Line 6712  Returns:         TRUE on success
6712  */  */
6713    
6714  static BOOL  static BOOL
6715  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6716    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6717    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6718    int *lengthptr)    compile_data *cd, int *lengthptr)
6719  {  {
6720  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6721  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 5819  int branchfirstbyte, branchreqbyte; Line 6729  int branchfirstbyte, branchreqbyte;
6729  int length;  int length;
6730  int orig_bracount;  int orig_bracount;
6731  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6732  branch_chain bc;  branch_chain bc;
6733    
6734  bc.outer = bcptr;  bc.outer = bcptr;
# Line 5843  pre-compile phase to find out whether an Line 6752  pre-compile phase to find out whether an
6752    
6753  /* If this is a capturing subpattern, add to the chain of open capturing items  /* If this is a capturing subpattern, add to the chain of open capturing items
6754  so that we can detect them if (*ACCEPT) is encountered. This is also used to  so that we can detect them if (*ACCEPT) is encountered. This is also used to
6755  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6756    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6757    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6758    
6759  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6760    {    {
# Line 5869  for (;;) Line 6780  for (;;)
6780    
6781    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6782    
   /* Handle a change of ims options at the start of the branch */  
   
   if ((options & PCRE_IMS) != oldims)  
     {  
     *code++ = OP_OPT;  
     *code++ = options & PCRE_IMS;  
     length += 2;  
     }  
   
6783    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6784    
6785    if (lookbehind)    if (lookbehind)
# Line 5892  for (;;) Line 6794  for (;;)
6794    into the length. */    into the length. */
6795    
6796    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6797          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6798            (lengthptr == NULL)? NULL : &length))
6799      {      {
6800      *ptrptr = ptr;      *ptrptr = ptr;
6801      return FALSE;      return FALSE;
6802      }      }
6803    
   /* If the external options have changed during this branch, it means that we  
   are at the top level, and a leading option setting has been encountered. We  
   need to re-set the original option values to take account of this so that,  
   during the pre-compile phase, we know to allow for a re-set at the start of  
   subsequent branches. */  
   
   if (old_external_options != cd->external_options)  
     oldims = cd->external_options & PCRE_IMS;  
   
6804    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6805    has fewer than the rest. */    has fewer than the rest. */
6806    
# Line 5967  for (;;) Line 6861  for (;;)
6861        {        {
6862        int fixed_length;        int fixed_length;
6863        *code = OP_END;        *code = OP_END;
6864        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6865            FALSE, cd);
6866        DPRINTF(("fixed length = %d\n", fixed_length));       &