/[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 487 by ph10, Wed Jan 6 10:26:55 2010 UTC revision 746 by ph10, Tue Nov 15 15:07:02 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 92  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
# Line 119  static const short int escapes[] = { Line 124  static const short int escapes[] = {
124       -ESC_H,                  0,       -ESC_H,                  0,
125       0,                       -ESC_K,       0,                       -ESC_K,
126       0,                       0,       0,                       0,
127       0,                       0,       -ESC_N,                  0,
128       -ESC_P,                  -ESC_Q,       -ESC_P,                  -ESC_Q,
129       -ESC_R,                  -ESC_S,       -ESC_R,                  -ESC_S,
130       0,                       0,       0,                       0,
# Line 166  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  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,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  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 188  string is built from string macros so th
188  platforms. */  platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    int   len;    int   len;                 /* Length of verb name */
192    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static const char verbnames[] =  static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199    STRING_ACCEPT0    STRING_ACCEPT0
200    STRING_COMMIT0    STRING_COMMIT0
201    STRING_F0    STRING_F0
# Line 197  static const char verbnames[] = Line 205  static const char verbnames[] =
205    STRING_THEN;    STRING_THEN;
206    
207  static const verbitem verbs[] = {  static const verbitem verbs[] = {
208    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
209    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
210    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
212    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
213    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
214    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  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 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 263  the number of relocations needed when a Line 320  the number of relocations needed when a
320  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
321  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
322  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
323  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329  static const char error_texts[] =  static const char error_texts[] =
330    "no error\0"    "no error\0"
# Line 310  static const char error_texts[] = Line 371  static const char error_texts[] =
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255\0"    "number after (?C is > 255\0"
376    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
# Line 332  static const char error_texts[] = Line 393  static const char error_texts[] =
393    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\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"
399    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
400    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401    /* 60 */    /* 60 */
402    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
403    "number is too big\0"    "number is too big\0"
# Line 344  static const char error_texts[] = Line 405  static const char error_texts[] =
405    "digit expected after (?+\0"    "digit expected after (?+\0"
406    "] is an invalid data character in JavaScript compatibility mode\0"    "] is an invalid data character in JavaScript compatibility mode\0"
407    /* 65 */    /* 65 */
408    "different names for subpatterns of the same number are not allowed";    "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413      ;
414    
415  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
416  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 481  static const unsigned char ebcdic_charta Line 546  static const unsigned char ebcdic_charta
546  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
547    
548  static BOOL  static BOOL
549    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
550      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
551    
552    
# Line 503  static const char * Line 568  static const char *
568  find_error_text(int n)  find_error_text(int n)
569  {  {
570  const char *s = error_texts;  const char *s = error_texts;
571  for (; n > 0; n--) while (*s++ != 0) {};  for (; n > 0; n--)
572      {
573      while (*s++ != 0) {};
574      if (*s == 0) return "Error text not found (please report)";
575      }
576  return s;  return s;
577  }  }
578    
579    
580  /*************************************************  /*************************************************
581    *            Check for counted repeat            *
582    *************************************************/
583    
584    /* This function is called when a '{' is encountered in a place where it might
585    start a quantifier. It looks ahead to see if it really is a quantifier or not.
586    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
587    where the ddds are digits.
588    
589    Arguments:
590      p         pointer to the first char after '{'
591    
592    Returns:    TRUE or FALSE
593    */
594    
595    static BOOL
596    is_counted_repeat(const uschar *p)
597    {
598    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
599    while ((digitab[*p] & ctype_digit) != 0) p++;
600    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
601    
602    if (*p++ != CHAR_COMMA) return FALSE;
603    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
604    
605    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
606    while ((digitab[*p] & ctype_digit) != 0) p++;
607    
608    return (*p == CHAR_RIGHT_CURLY_BRACKET);
609    }
610    
611    
612    
613    /*************************************************
614  *            Handle escapes                      *  *            Handle escapes                      *
615  *************************************************/  *************************************************/
616    
# Line 574  else Line 676  else
676    
677      case CHAR_l:      case CHAR_l:
678      case CHAR_L:      case CHAR_L:
679      case CHAR_N:      *errorcodeptr = ERR37;
680        break;
681    
682      case CHAR_u:      case CHAR_u:
683        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
684          {
685          /* In JavaScript, \u must be followed by four hexadecimal numbers.
686          Otherwise it is a lowercase u letter. */
687          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
688               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
689            {
690            c = 0;
691            for (i = 0; i < 4; ++i)
692              {
693              register int cc = *(++ptr);
694    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
695              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
696              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
697    #else           /* EBCDIC coding */
698              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
699              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
700    #endif
701              }
702            }
703          }
704        else
705          *errorcodeptr = ERR37;
706        break;
707    
708      case CHAR_U:      case CHAR_U:
709      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
710        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
711      break;      break;
712    
713      /* \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
714        class, \g must be followed by one of a number of specific things:
715    
716      (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
717      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 597  else Line 728  else
728      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
729    
730      case CHAR_g:      case CHAR_g:
731        if (isclass) break;
732      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
733        {        {
734        c = -ESC_g;        c = -ESC_g;
# Line 725  else Line 857  else
857      treated as a data character. */      treated as a data character. */
858    
859      case CHAR_x:      case CHAR_x:
860        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
861          {
862          /* In JavaScript, \x must be followed by two hexadecimal numbers.
863          Otherwise it is a lowercase x letter. */
864          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
865            {
866            c = 0;
867            for (i = 0; i < 2; ++i)
868              {
869              register int cc = *(++ptr);
870    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
871              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
872              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
873    #else           /* EBCDIC coding */
874              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
875              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
876    #endif
877              }
878            }
879          break;
880          }
881    
882      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
883        {        {
884        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
# Line 775  else Line 929  else
929      break;      break;
930    
931      /* 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.
932      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
933        coding is ASCII-specific, but then the whole concept of \cx is
934      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
935    
936      case CHAR_c:      case CHAR_c:
# Line 785  else Line 940  else
940        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
941        break;        break;
942        }        }
943    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
944  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
945          {
946          *errorcodeptr = ERR68;
947          break;
948          }
949      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
950      c ^= 0x40;      c ^= 0x40;
951  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
952      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
953      c ^= 0xC0;      c ^= 0xC0;
954  #endif  #endif
# Line 812  else Line 971  else
971      }      }
972    }    }
973    
974    /* Perl supports \N{name} for character names, as well as plain \N for "not
975    newline". PCRE does not support \N{name}. However, it does support
976    quantification such as \N{2,3}. */
977    
978    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
979         !is_counted_repeat(ptr+2))
980      *errorcodeptr = ERR37;
981    
982    /* If PCRE_UCP is set, we change the values for \d etc. */
983    
984    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
985      c -= (ESC_DU - ESC_D);
986    
987    /* Set the pointer to the final character before returning. */
988    
989  *ptrptr = ptr;  *ptrptr = ptr;
990  return c;  return c;
991  }  }
# Line 912  return -1; Line 1086  return -1;
1086    
1087    
1088  /*************************************************  /*************************************************
 *            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);  
 }  
   
   
   
 /*************************************************  
1089  *         Read repeat counts                     *  *         Read repeat counts                     *
1090  *************************************************/  *************************************************/
1091    
# Line 1020  top-level call starts at the beginning o Line 1161  top-level call starts at the beginning o
1161  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
1162  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
1163  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
1164  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
1165  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1166  first pass. Recursion is used to keep track of subpatterns that reset the  
1167  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1168    that if (?< or (?' or (?P< is encountered, the name will be correctly
1169    terminated because that is checked in the first pass. There is now one call to
1170    this function in the first pass, to check for a recursive back reference by
1171    name (so that we can make the whole group atomic). In this case, we need check
1172    only up to the current position in the pattern, and that is still OK because
1173    and previous occurrences will have been checked. To make this work, the test
1174    for "end of pattern" is a check against cd->end_pattern in the main loop,
1175    instead of looking for a binary zero. This means that the special first-pass
1176    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1177    processing items within the loop are OK, because afterwards the main loop will
1178    terminate.)
1179    
1180  Arguments:  Arguments:
1181    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1031  Arguments: Line 1183  Arguments:
1183    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1184    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1185    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1186      utf8         TRUE if we are in UTF-8 mode
1187    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1188    
1189  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 1038  Returns:       the number of the named s Line 1191  Returns:       the number of the named s
1191    
1192  static int  static int
1193  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,
1194    BOOL xmode, int *count)    BOOL xmode, BOOL utf8, int *count)
1195  {  {
1196  uschar *ptr = *ptrptr;  uschar *ptr = *ptrptr;
1197  int start_count = *count;  int start_count = *count;
# Line 1050  dealing with. The very first call may no Line 1203  dealing with. The very first call may no
1203    
1204  if (ptr[0] == CHAR_LEFT_PARENTHESIS)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1205    {    {
1206    if (ptr[1] == CHAR_QUESTION_MARK &&    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1207        ptr[2] == CHAR_VERTICAL_LINE)  
1208      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1209    
1210      /* Handle a normal, unnamed capturing parenthesis. */
1211    
1212      else if (ptr[1] != CHAR_QUESTION_MARK)
1213        {
1214        *count += 1;
1215        if (name == NULL && *count == lorn) return *count;
1216        ptr++;
1217        }
1218    
1219      /* All cases now have (? at the start. Remember when we are in a group
1220      where the parenthesis numbers are duplicated. */
1221    
1222      else if (ptr[2] == CHAR_VERTICAL_LINE)
1223      {      {
1224      ptr += 3;      ptr += 3;
1225      dup_parens = TRUE;      dup_parens = TRUE;
1226      }      }
1227    
1228    /* Handle a normal, unnamed capturing parenthesis */    /* Handle comments; all characters are allowed until a ket is reached. */
1229    
1230    else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)    else if (ptr[2] == CHAR_NUMBER_SIGN)
1231      {      {
1232      *count += 1;      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1233      if (name == NULL && *count == lorn) return *count;      goto FAIL_EXIT;
     ptr++;  
1234      }      }
1235    
1236    /* 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
1237    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
1238    condition (there can't be any nested parens. */    condition (there can't be any nested parens). */
1239    
1240    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1241      {      {
# Line 1080  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1247  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1247        }        }
1248      }      }
1249    
1250    /* We have either (? or (* and not a condition */    /* Start with (? but not a condition. */
1251    
1252    else    else
1253      {      {
# Line 1109  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1276  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1276    }    }
1277    
1278  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1279  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1280    first-pass call when this value is temporarily adjusted to stop at the current
1281    position. So DO NOT change this to a test for binary zero. */
1282    
1283  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1284    {    {
1285    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1286    
# Line 1185  for (; *ptr != 0; ptr++) Line 1354  for (; *ptr != 0; ptr++)
1354    
1355    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1356      {      {
1357      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1358        while (*ptr != 0)
1359          {
1360          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1361          ptr++;
1362    #ifdef SUPPORT_UTF8
1363          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1364    #endif
1365          }
1366      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1367      continue;      continue;
1368      }      }
# Line 1194  for (; *ptr != 0; ptr++) Line 1371  for (; *ptr != 0; ptr++)
1371    
1372    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1373      {      {
1374      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1375      if (rc > 0) return rc;      if (rc > 0) return rc;
1376      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1377      }      }
# Line 1202  for (; *ptr != 0; ptr++) Line 1379  for (; *ptr != 0; ptr++)
1379    else if (*ptr == CHAR_RIGHT_PARENTHESIS)    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1380      {      {
1381      if (dup_parens && *count < hwm_count) *count = hwm_count;      if (dup_parens && *count < hwm_count) *count = hwm_count;
1382      *ptrptr = ptr;      goto FAIL_EXIT;
     return -1;  
1383      }      }
1384    
1385    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
# Line 1241  Arguments: Line 1417  Arguments:
1417    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1418    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1419    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1420      utf8         TRUE if we are in UTF-8 mode
1421    
1422  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1423  */  */
1424    
1425  static int  static int
1426  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1427      BOOL utf8)
1428  {  {
1429  uschar *ptr = (uschar *)cd->start_pattern;  uschar *ptr = (uschar *)cd->start_pattern;
1430  int count = 0;  int count = 0;
# Line 1259  matching closing parens. That is why we Line 1437  matching closing parens. That is why we
1437    
1438  for (;;)  for (;;)
1439    {    {
1440    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1441    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1442    }    }
1443    
# Line 1275  return rc; Line 1453  return rc;
1453    
1454  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1455  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
1456  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
1457  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
1458  assertions, and also the \b assertion; for others it does not.  does not.
1459    
1460  Arguments:  Arguments:
1461    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  
1462    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1463    
1464  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1465  */  */
1466    
1467  static const uschar*  static const uschar*
1468  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1469  {  {
1470  for (;;)  for (;;)
1471    {    {
1472    switch ((int)*code)    switch ((int)*code)
1473      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1474      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1475      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1476      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1352  and doing the check at the end; a flag s Line 1520  and doing the check at the end; a flag s
1520    
1521  Arguments:  Arguments:
1522    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1523    options  the compiling options    utf8     TRUE in UTF-8 mode
1524    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1525    cd       the "compile data" structure    cd       the "compile data" structure
1526    
# Line 1363  Returns:   the fixed length, Line 1531  Returns:   the fixed length,
1531  */  */
1532    
1533  static int  static int
1534  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1535  {  {
1536  int length = -1;  int length = -1;
1537    
# Line 1380  for (;;) Line 1548  for (;;)
1548    register int op = *cc;    register int op = *cc;
1549    switch (op)    switch (op)
1550      {      {
1551        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1552        OP_BRA (normal non-capturing bracket) because the other variants of these
1553        opcodes are all concerned with unlimited repeated groups, which of course
1554        are not of fixed length. They will cause a -1 response from the default
1555        case of this switch. */
1556    
1557      case OP_CBRA:      case OP_CBRA:
1558      case OP_BRA:      case OP_BRA:
1559      case OP_ONCE:      case OP_ONCE:
1560        case OP_ONCE_NC:
1561      case OP_COND:      case OP_COND:
1562      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1563      if (d < 0) return d;      if (d < 0) return d;
1564      branchlength += d;      branchlength += d;
1565      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1393  for (;;) Line 1568  for (;;)
1568    
1569      /* 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
1570      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is
1571      END it's the end of the outer call. All can be handled by the same code. */      END it's the end of the outer call. All can be handled by the same code.
1572        Note that we must not include the OP_KETRxxx opcodes here, because they
1573        all imply an unlimited repeat. */
1574    
1575      case OP_ALT:      case OP_ALT:
1576      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1577      case OP_END:      case OP_END:
1578      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1579        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1416  for (;;) Line 1591  for (;;)
1591      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1592      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1593      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1594      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1595      if (d < 0) return d;      if (d < 0) return d;
1596      branchlength += d;      branchlength += d;
1597      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1439  for (;;) Line 1614  for (;;)
1614      case OP_RREF:      case OP_RREF:
1615      case OP_NRREF:      case OP_NRREF:
1616      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1617      case OP_CALLOUT:      case OP_CALLOUT:
1618      case OP_SOD:      case OP_SOD:
1619      case OP_SOM:      case OP_SOM:
1620        case OP_SET_SOM:
1621      case OP_EOD:      case OP_EOD:
1622      case OP_EODN:      case OP_EODN:
1623      case OP_CIRC:      case OP_CIRC:
1624        case OP_CIRCM:
1625      case OP_DOLL:      case OP_DOLL:
1626        case OP_DOLLM:
1627      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1628      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1629      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1455  for (;;) Line 1632  for (;;)
1632      /* Handle literal characters */      /* Handle literal characters */
1633    
1634      case OP_CHAR:      case OP_CHAR:
1635      case OP_CHARNC:      case OP_CHARI:
1636      case OP_NOT:      case OP_NOT:
1637        case OP_NOTI:
1638      branchlength++;      branchlength++;
1639      cc += 2;      cc += 2;
1640  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1641      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];  
1642  #endif  #endif
1643      break;      break;
1644    
# Line 1469  for (;;) Line 1646  for (;;)
1646      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1647    
1648      case OP_EXACT:      case OP_EXACT:
1649        case OP_EXACTI:
1650      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1651      cc += 4;      cc += 4;
1652  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1653      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];  
1654  #endif  #endif
1655      break;      break;
1656    
# Line 1575  _pcre_find_bracket(const uschar *code, B Line 1752  _pcre_find_bracket(const uschar *code, B
1752  for (;;)  for (;;)
1753    {    {
1754    register int c = *code;    register int c = *code;
1755    
1756    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1757    
1758    /* 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 1593  for (;;) Line 1771  for (;;)
1771    
1772    /* Handle capturing bracket */    /* Handle capturing bracket */
1773    
1774    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1775               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1776      {      {
1777      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1778      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1602  for (;;) Line 1781  for (;;)
1781    
1782    /* 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
1783    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
1784    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1785      must add in its length. */
1786    
1787    else    else
1788      {      {
# Line 1626  for (;;) Line 1806  for (;;)
1806        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1807        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1808        break;        break;
1809    
1810          case OP_MARK:
1811          case OP_PRUNE_ARG:
1812          case OP_SKIP_ARG:
1813          code += code[1];
1814          break;
1815    
1816          case OP_THEN_ARG:
1817          code += code[1];
1818          break;
1819        }        }
1820    
1821      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1640  for (;;) Line 1830  for (;;)
1830      if (utf8) switch(c)      if (utf8) switch(c)
1831        {        {
1832        case OP_CHAR:        case OP_CHAR:
1833        case OP_CHARNC:        case OP_CHARI:
1834        case OP_EXACT:        case OP_EXACT:
1835          case OP_EXACTI:
1836        case OP_UPTO:        case OP_UPTO:
1837          case OP_UPTOI:
1838        case OP_MINUPTO:        case OP_MINUPTO:
1839          case OP_MINUPTOI:
1840        case OP_POSUPTO:        case OP_POSUPTO:
1841          case OP_POSUPTOI:
1842        case OP_STAR:        case OP_STAR:
1843          case OP_STARI:
1844        case OP_MINSTAR:        case OP_MINSTAR:
1845          case OP_MINSTARI:
1846        case OP_POSSTAR:        case OP_POSSTAR:
1847          case OP_POSSTARI:
1848        case OP_PLUS:        case OP_PLUS:
1849          case OP_PLUSI:
1850        case OP_MINPLUS:        case OP_MINPLUS:
1851          case OP_MINPLUSI:
1852        case OP_POSPLUS:        case OP_POSPLUS:
1853          case OP_POSPLUSI:
1854        case OP_QUERY:        case OP_QUERY:
1855          case OP_QUERYI:
1856        case OP_MINQUERY:        case OP_MINQUERY:
1857          case OP_MINQUERYI:
1858        case OP_POSQUERY:        case OP_POSQUERY:
1859          case OP_POSQUERYI:
1860        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1861        break;        break;
1862        }        }
# Line 1697  for (;;) Line 1900  for (;;)
1900    
1901    /* 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
1902    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
1903    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1904      must add in its length. */
1905    
1906    else    else
1907      {      {
# Line 1721  for (;;) Line 1925  for (;;)
1925        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1926        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1927        break;        break;
1928    
1929          case OP_MARK:
1930          case OP_PRUNE_ARG:
1931          case OP_SKIP_ARG:
1932          code += code[1];
1933          break;
1934    
1935          case OP_THEN_ARG:
1936          code += code[1];
1937          break;
1938        }        }
1939    
1940      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1735  for (;;) Line 1949  for (;;)
1949      if (utf8) switch(c)      if (utf8) switch(c)
1950        {        {
1951        case OP_CHAR:        case OP_CHAR:
1952        case OP_CHARNC:        case OP_CHARI:
1953        case OP_EXACT:        case OP_EXACT:
1954          case OP_EXACTI:
1955        case OP_UPTO:        case OP_UPTO:
1956          case OP_UPTOI:
1957        case OP_MINUPTO:        case OP_MINUPTO:
1958          case OP_MINUPTOI:
1959        case OP_POSUPTO:        case OP_POSUPTO:
1960          case OP_POSUPTOI:
1961        case OP_STAR:        case OP_STAR:
1962          case OP_STARI:
1963        case OP_MINSTAR:        case OP_MINSTAR:
1964          case OP_MINSTARI:
1965        case OP_POSSTAR:        case OP_POSSTAR:
1966          case OP_POSSTARI:
1967        case OP_PLUS:        case OP_PLUS:
1968          case OP_PLUSI:
1969        case OP_MINPLUS:        case OP_MINPLUS:
1970          case OP_MINPLUSI:
1971        case OP_POSPLUS:        case OP_POSPLUS:
1972          case OP_POSPLUSI:
1973        case OP_QUERY:        case OP_QUERY:
1974          case OP_QUERYI:
1975        case OP_MINQUERY:        case OP_MINQUERY:
1976          case OP_MINQUERYI:
1977        case OP_POSQUERY:        case OP_POSQUERY:
1978          case OP_POSQUERYI:
1979        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1980        break;        break;
1981        }        }
# Line 1777  Arguments: Line 2004  Arguments:
2004    code        points to start of search    code        points to start of search
2005    endcode     points to where to stop    endcode     points to where to stop
2006    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
2007      cd          contains pointers to tables etc.
2008    
2009  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2010  */  */
2011    
2012  static BOOL  static BOOL
2013  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
2014      compile_data *cd)
2015  {  {
2016  register int c;  register int c;
2017  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2018       code < endcode;       code < endcode;
2019       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2020    {    {
2021    const uschar *ccode;    const uschar *ccode;
2022    
# Line 1803  for (code = first_significant_code(code Line 2032  for (code = first_significant_code(code
2032      continue;      continue;
2033      }      }
2034    
2035      /* For a recursion/subroutine call, if its end has been reached, which
2036      implies a backward reference subroutine call, we can scan it. If it's a
2037      forward reference subroutine call, we can't. To detect forward reference
2038      we have to scan up the list that is kept in the workspace. This function is
2039      called only when doing the real compile, not during the pre-compile that
2040      measures the size of the compiled pattern. */
2041    
2042      if (c == OP_RECURSE)
2043        {
2044        const uschar *scode;
2045        BOOL empty_branch;
2046    
2047        /* Test for forward reference */
2048    
2049        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2050          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2051    
2052        /* Not a forward reference, test for completed backward reference */
2053    
2054        empty_branch = FALSE;
2055        scode = cd->start_code + GET(code, 1);
2056        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2057    
2058        /* Completed backwards reference */
2059    
2060        do
2061          {
2062          if (could_be_empty_branch(scode, endcode, utf8, cd))
2063            {
2064            empty_branch = TRUE;
2065            break;
2066            }
2067          scode += GET(scode, 1);
2068          }
2069        while (*scode == OP_ALT);
2070    
2071        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2072        continue;
2073        }
2074    
2075    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2076    
2077    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2078          c == OP_BRAPOSZERO)
2079      {      {
2080      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
2081      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1813  for (code = first_significant_code(code Line 2083  for (code = first_significant_code(code
2083      continue;      continue;
2084      }      }
2085    
2086      /* A nested group that is already marked as "could be empty" can just be
2087      skipped. */
2088    
2089      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2090          c == OP_SCBRA || c == OP_SCBRAPOS)
2091        {
2092        do code += GET(code, 1); while (*code == OP_ALT);
2093        c = *code;
2094        continue;
2095        }
2096    
2097    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2098    
2099    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2100          c == OP_CBRA || c == OP_CBRAPOS ||
2101          c == OP_ONCE || c == OP_ONCE_NC ||
2102          c == OP_COND)
2103      {      {
2104      BOOL empty_branch;      BOOL empty_branch;
2105      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1831  for (code = first_significant_code(code Line 2115  for (code = first_significant_code(code
2115        empty_branch = FALSE;        empty_branch = FALSE;
2116        do        do
2117          {          {
2118          if (!empty_branch && could_be_empty_branch(code, endcode, utf8))          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2119            empty_branch = TRUE;            empty_branch = TRUE;
2120          code += GET(code, 1);          code += GET(code, 1);
2121          }          }
# Line 1902  for (code = first_significant_code(code Line 2186  for (code = first_significant_code(code
2186      case OP_ALLANY:      case OP_ALLANY:
2187      case OP_ANYBYTE:      case OP_ANYBYTE:
2188      case OP_CHAR:      case OP_CHAR:
2189      case OP_CHARNC:      case OP_CHARI:
2190      case OP_NOT:      case OP_NOT:
2191        case OP_NOTI:
2192      case OP_PLUS:      case OP_PLUS:
2193      case OP_MINPLUS:      case OP_MINPLUS:
2194      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1943  for (code = first_significant_code(code Line 2228  for (code = first_significant_code(code
2228      case OP_KET:      case OP_KET:
2229      case OP_KETRMAX:      case OP_KETRMAX:
2230      case OP_KETRMIN:      case OP_KETRMIN:
2231        case OP_KETRPOS:
2232      case OP_ALT:      case OP_ALT:
2233      return TRUE;      return TRUE;
2234    
# Line 1951  for (code = first_significant_code(code Line 2237  for (code = first_significant_code(code
2237    
2238  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2239      case OP_STAR:      case OP_STAR:
2240        case OP_STARI:
2241      case OP_MINSTAR:      case OP_MINSTAR:
2242        case OP_MINSTARI:
2243      case OP_POSSTAR:      case OP_POSSTAR:
2244        case OP_POSSTARI:
2245      case OP_QUERY:      case OP_QUERY:
2246        case OP_QUERYI:
2247      case OP_MINQUERY:      case OP_MINQUERY:
2248        case OP_MINQUERYI:
2249      case OP_POSQUERY:      case OP_POSQUERY:
2250        case OP_POSQUERYI:
2251      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2252      break;      break;
2253    
2254      case OP_UPTO:      case OP_UPTO:
2255        case OP_UPTOI:
2256      case OP_MINUPTO:      case OP_MINUPTO:
2257        case OP_MINUPTOI:
2258      case OP_POSUPTO:      case OP_POSUPTO:
2259        case OP_POSUPTOI:
2260      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2261      break;      break;
2262  #endif  #endif
2263    
2264        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2265        string. */
2266    
2267        case OP_MARK:
2268        case OP_PRUNE_ARG:
2269        case OP_SKIP_ARG:
2270        code += code[1];
2271        break;
2272    
2273        case OP_THEN_ARG:
2274        code += code[1];
2275        break;
2276    
2277        /* None of the remaining opcodes are required to match a character. */
2278    
2279        default:
2280        break;
2281      }      }
2282    }    }
2283    
# Line 1981  return TRUE; Line 2294  return TRUE;
2294  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
2295  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,
2296  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.
2297    This function is called only during the real compile, not during the
2298    pre-compile.
2299    
2300  Arguments:  Arguments:
2301    code        points to start of the recursion    code        points to start of the recursion
2302    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2303    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2304    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2305      cd          pointers to tables etc
2306    
2307  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2308  */  */
2309    
2310  static BOOL  static BOOL
2311  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2312    BOOL utf8)    BOOL utf8, compile_data *cd)
2313  {  {
2314  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2315    {    {
2316    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2317      return FALSE;      return FALSE;
2318    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2319    }    }
# Line 2030  where Perl recognizes it as the POSIX cl Line 2346  where Perl recognizes it as the POSIX cl
2346  "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,
2347  I think.  I think.
2348    
2349    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2350    It seems that the appearance of a nested POSIX class supersedes an apparent
2351    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2352    a digit.
2353    
2354    In Perl, unescaped square brackets may also appear as part of class names. For
2355    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2356    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2357    seem right at all. PCRE does not allow closing square brackets in POSIX class
2358    names.
2359    
2360  Arguments:  Arguments:
2361    ptr      pointer to the initial [    ptr      pointer to the initial [
2362    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2044  int terminator;          /* Don't combin Line 2371  int terminator;          /* Don't combin
2371  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2372  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2373    {    {
2374    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2375        ptr++;
2376      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2377      else
2378      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2379      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2380        {        {
2381        *endptr = ptr;        *endptr = ptr;
2382        return TRUE;        return TRUE;
2383        }        }
2384        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2385             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2386              ptr[1] == CHAR_EQUALS_SIGN) &&
2387            check_posix_syntax(ptr, endptr))
2388          return FALSE;
2389      }      }
2390    }    }
2391  return FALSE;  return FALSE;
# Line 2180  auto_callout(uschar *code, const uschar Line 2514  auto_callout(uschar *code, const uschar
2514  {  {
2515  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2516  *code++ = 255;  *code++ = 255;
2517  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2518  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2519  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2520  }  }
2521    
# Line 2206  Returns:             nothing Line 2540  Returns:             nothing
2540  static void  static void
2541  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2542  {  {
2543  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2544  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2545  }  }
2546    
# Line 2256  for (++c; c <= d; c++) Line 2590  for (++c; c <= d; c++)
2590    
2591  return TRUE;  return TRUE;
2592  }  }
2593    
2594    
2595    
2596    /*************************************************
2597    *        Check a character and a property        *
2598    *************************************************/
2599    
2600    /* This function is called by check_auto_possessive() when a property item
2601    is adjacent to a fixed character.
2602    
2603    Arguments:
2604      c            the character
2605      ptype        the property type
2606      pdata        the data for the type
2607      negated      TRUE if it's a negated property (\P or \p{^)
2608    
2609    Returns:       TRUE if auto-possessifying is OK
2610    */
2611    
2612    static BOOL
2613    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2614    {
2615    const ucd_record *prop = GET_UCD(c);
2616    switch(ptype)
2617      {
2618      case PT_LAMP:
2619      return (prop->chartype == ucp_Lu ||
2620              prop->chartype == ucp_Ll ||
2621              prop->chartype == ucp_Lt) == negated;
2622    
2623      case PT_GC:
2624      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2625    
2626      case PT_PC:
2627      return (pdata == prop->chartype) == negated;
2628    
2629      case PT_SC:
2630      return (pdata == prop->script) == negated;
2631    
2632      /* These are specials */
2633    
2634      case PT_ALNUM:
2635      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2636              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2637    
2638      case PT_SPACE:    /* Perl space */
2639      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2640              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2641              == negated;
2642    
2643      case PT_PXSPACE:  /* POSIX space */
2644      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2645              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2646              c == CHAR_FF || c == CHAR_CR)
2647              == negated;
2648    
2649      case PT_WORD:
2650      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2651              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2652              c == CHAR_UNDERSCORE) == negated;
2653      }
2654    return FALSE;
2655    }
2656  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2657    
2658    
# Line 2269  whether the next thing could possibly ma Line 2666  whether the next thing could possibly ma
2666  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2667    
2668  Arguments:  Arguments:
2669    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2670    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2671    ptr           next character in pattern    ptr           next character in pattern
2672    options       options bits    options       options bits
2673    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2281  Returns:        TRUE if possessifying is Line 2676  Returns:        TRUE if possessifying is
2676  */  */
2677    
2678  static BOOL  static BOOL
2679  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2680    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2681  {  {
2682  int next;  int c, next;
2683    int op_code = *previous++;
2684    
2685  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2686    
# Line 2295  if ((options & PCRE_EXTENDED) != 0) Line 2691  if ((options & PCRE_EXTENDED) != 0)
2691      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2692      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2693        {        {
2694        while (*(++ptr) != 0)        ptr++;
2695          while (*ptr != 0)
2696            {
2697          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2698            ptr++;
2699    #ifdef SUPPORT_UTF8
2700            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2701    #endif
2702            }
2703        }        }
2704      else break;      else break;
2705      }      }
# Line 2332  if ((options & PCRE_EXTENDED) != 0) Line 2735  if ((options & PCRE_EXTENDED) != 0)
2735      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2736      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2737        {        {
2738        while (*(++ptr) != 0)        ptr++;
2739          while (*ptr != 0)
2740            {
2741          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2742            ptr++;
2743    #ifdef SUPPORT_UTF8
2744            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2745    #endif
2746            }
2747        }        }
2748      else break;      else break;
2749      }      }
# Line 2345  if (*ptr == CHAR_ASTERISK || *ptr == CHA Line 2755  if (*ptr == CHAR_ASTERISK || *ptr == CHA
2755    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)
2756      return FALSE;      return FALSE;
2757    
2758  /* 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
2759  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. */  
2760    
2761  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2762    {    {
2763    case OP_CHAR:    case OP_CHAR:
2764  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2765    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2766  #else  #else
2767    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2768  #endif  #endif
2769    return item != next;    return c != next;
2770    
2771    /* 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
2772    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
2773    high-valued characters. */    high-valued characters. */
2774    
2775    case OP_CHARNC:    case OP_CHARI:
2776  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2777    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2778    #else
2779      c = *previous;
2780  #endif  #endif
2781    if (item == next) return FALSE;    if (c == next) return FALSE;
2782  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2783    if (utf8)    if (utf8)
2784      {      {
# Line 2382  if (next >= 0) switch(op_code) Line 2789  if (next >= 0) switch(op_code)
2789  #else  #else
2790      othercase = NOTACHAR;      othercase = NOTACHAR;
2791  #endif  #endif
2792      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2793      }      }
2794    else    else
2795  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2796    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2797    
2798    /* 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
2799      opcodes are not used for multi-byte characters, because they are coded using
2800      an XCLASS instead. */
2801    
2802    case OP_NOT:    case OP_NOT:
2803    if (item == next) return TRUE;    return (c = *previous) == next;
2804    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2805      case OP_NOTI:
2806      if ((c = *previous) == next) return TRUE;
2807  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2808    if (utf8)    if (utf8)
2809      {      {
# Line 2403  if (next >= 0) switch(op_code) Line 2814  if (next >= 0) switch(op_code)
2814  #else  #else
2815      othercase = NOTACHAR;      othercase = NOTACHAR;
2816  #endif  #endif
2817      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2818      }      }
2819    else    else
2820  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2821    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2822    
2823      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2824      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2825    
2826    case OP_DIGIT:    case OP_DIGIT:
2827    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2450  if (next >= 0) switch(op_code) Line 2864  if (next >= 0) switch(op_code)
2864      case 0x202f:      case 0x202f:
2865      case 0x205f:      case 0x205f:
2866      case 0x3000:      case 0x3000:
2867      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2868      default:      default:
2869      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2870      }      }
2871    
2872      case OP_ANYNL:
2873    case OP_VSPACE:    case OP_VSPACE:
2874    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2875    switch(next)    switch(next)
# Line 2466  if (next >= 0) switch(op_code) Line 2881  if (next >= 0) switch(op_code)
2881      case 0x85:      case 0x85:
2882      case 0x2028:      case 0x2028:
2883      case 0x2029:      case 0x2029:
2884      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2885      default:      default:
2886      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2887      }      }
2888    
2889    #ifdef SUPPORT_UCP
2890      case OP_PROP:
2891      return check_char_prop(next, previous[0], previous[1], FALSE);
2892    
2893      case OP_NOTPROP:
2894      return check_char_prop(next, previous[0], previous[1], TRUE);
2895    #endif
2896    
2897    default:    default:
2898    return FALSE;    return FALSE;
2899    }    }
2900    
2901    
2902  /* 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
2903    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2904    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2905    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2906    replaced by OP_PROP codes when PCRE_UCP is set. */
2907    
2908  switch(op_code)  switch(op_code)
2909    {    {
2910    case OP_CHAR:    case OP_CHAR:
2911    case OP_CHARNC:    case OP_CHARI:
2912  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2913    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2914    #else
2915      c = *previous;
2916  #endif  #endif
2917    switch(-next)    switch(-next)
2918      {      {
2919      case ESC_d:      case ESC_d:
2920      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2921    
2922      case ESC_D:      case ESC_D:
2923      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2924    
2925      case ESC_s:      case ESC_s:
2926      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2927    
2928      case ESC_S:      case ESC_S:
2929      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2930    
2931      case ESC_w:      case ESC_w:
2932      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2933    
2934      case ESC_W:      case ESC_W:
2935      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2936    
2937      case ESC_h:      case ESC_h:
2938      case ESC_H:      case ESC_H:
2939      switch(item)      switch(c)
2940        {        {
2941        case 0x09:        case 0x09:
2942        case 0x20:        case 0x20:
# Line 2535  switch(op_code) Line 2964  switch(op_code)
2964    
2965      case ESC_v:      case ESC_v:
2966      case ESC_V:      case ESC_V:
2967      switch(item)      switch(c)
2968        {        {
2969        case 0x0a:        case 0x0a:
2970        case 0x0b:        case 0x0b:
# Line 2549  switch(op_code) Line 2978  switch(op_code)
2978        return -next == ESC_v;        return -next == ESC_v;
2979        }        }
2980    
2981        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2982        their substitutions and process them. The result will always be either
2983        -ESC_p or -ESC_P. Then fall through to process those values. */
2984    
2985    #ifdef SUPPORT_UCP
2986        case ESC_du:
2987        case ESC_DU:
2988        case ESC_wu:
2989        case ESC_WU:
2990        case ESC_su:
2991        case ESC_SU:
2992          {
2993          int temperrorcode = 0;
2994          ptr = substitutes[-next - ESC_DU];
2995          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2996          if (temperrorcode != 0) return FALSE;
2997          ptr++;    /* For compatibility */
2998          }
2999        /* Fall through */
3000    
3001        case ESC_p:
3002        case ESC_P:
3003          {
3004          int ptype, pdata, errorcodeptr;
3005          BOOL negated;
3006    
3007          ptr--;      /* Make ptr point at the p or P */
3008          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3009          if (ptype < 0) return FALSE;
3010          ptr++;      /* Point past the final curly ket */
3011    
3012          /* If the property item is optional, we have to give up. (When generated
3013          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3014          to the original \d etc. At this point, ptr will point to a zero byte. */
3015    
3016          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3017            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3018              return FALSE;
3019    
3020          /* Do the property check. */
3021    
3022          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3023          }
3024    #endif
3025    
3026      default:      default:
3027      return FALSE;      return FALSE;
3028      }      }
3029    
3030      /* In principle, support for Unicode properties should be integrated here as
3031      well. It means re-organizing the above code so as to get hold of the property
3032      values before switching on the op-code. However, I wonder how many patterns
3033      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3034      these op-codes are never generated.) */
3035    
3036    case OP_DIGIT:    case OP_DIGIT:
3037    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3038           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3039    
3040    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3041    return next == -ESC_d;    return next == -ESC_d;
3042    
3043    case OP_WHITESPACE:    case OP_WHITESPACE:
3044    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3045    
3046    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3047    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3048    
3049    case OP_HSPACE:    case OP_HSPACE:
3050    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3051             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3052    
3053    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3054    return next == -ESC_h;    return next == -ESC_h;
3055    
3056    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3057      case OP_ANYNL:
3058    case OP_VSPACE:    case OP_VSPACE:
3059    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3060    
3061    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3062    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3063    
3064    case OP_WORDCHAR:    case OP_WORDCHAR:
3065    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3066             next == -ESC_v || next == -ESC_R;
3067    
3068    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3069    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2612  Arguments: Line 3095  Arguments:
3095    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3096    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3097    bcptr          points to current branch chain    bcptr          points to current branch chain
3098      cond_depth     conditional nesting depth
3099    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3100    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3101                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2623  Returns:         TRUE on success Line 3107  Returns:         TRUE on success
3107  static BOOL  static BOOL
3108  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3109    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3110    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3111  {  {
3112  int repeat_type, op_type;  int repeat_type, op_type;
3113  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2632  int greedy_default, greedy_non_default; Line 3116  int greedy_default, greedy_non_default;
3116  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3117  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3118  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3119  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3120  int after_manual_callout = 0;  int after_manual_callout = 0;
3121  int length_prevgroup = 0;  int length_prevgroup = 0;
3122  register int c;  register int c;
# Line 2644  BOOL inescq = FALSE; Line 3128  BOOL inescq = FALSE;
3128  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3129  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3130  const uschar *tempptr;  const uschar *tempptr;
3131    const uschar *nestptr = NULL;
3132  uschar *previous = NULL;  uschar *previous = NULL;
3133  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3134  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3135  uschar classbits[32];  uschar classbits[32];
3136    
3137    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3138    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3139    dynamically as we process the pattern. */
3140    
3141  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3142  BOOL class_utf8;  BOOL class_utf8;
3143  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2657  uschar *class_utf8data_base; Line 3146  uschar *class_utf8data_base;
3146  uschar utf8_char[6];  uschar utf8_char[6];
3147  #else  #else
3148  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3149  #endif  #endif
3150    
3151  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 2708  for (;; ptr++) Line 3196  for (;; ptr++)
3196    int subfirstbyte;    int subfirstbyte;
3197    int terminator;    int terminator;
3198    int mclength;    int mclength;
3199      int tempbracount;
3200    uschar mcbuffer[8];    uschar mcbuffer[8];
3201    
3202    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3203    
3204    c = *ptr;    c = *ptr;
3205    
3206      /* If we are at the end of a nested substitution, revert to the outer level
3207      string. Nesting only happens one level deep. */
3208    
3209      if (c == 0 && nestptr != NULL)
3210        {
3211        ptr = nestptr;
3212        nestptr = NULL;
3213        c = *ptr;
3214        }
3215    
3216    /* 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
3217    previous cycle of this loop. */    previous cycle of this loop. */
3218    
# Line 2722  for (;; ptr++) Line 3221  for (;; ptr++)
3221  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
3222      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3223  #endif  #endif
3224      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3225        {        {
3226        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3227        goto FAILED;        goto FAILED;
# Line 2744  for (;; ptr++) Line 3243  for (;; ptr++)
3243        goto FAILED;        goto FAILED;
3244        }        }
3245    
3246      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3247      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3248    
3249      /* 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
# Line 2771  for (;; ptr++) Line 3270  for (;; ptr++)
3270    /* 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
3271    reference list. */    reference list. */
3272    
3273    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3274      {      {
3275      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3276      goto FAILED;      goto FAILED;
# Line 2819  for (;; ptr++) Line 3318  for (;; ptr++)
3318      previous_callout = NULL;      previous_callout = NULL;
3319      }      }
3320    
3321    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3322    
3323    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3324      {      {
3325      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3326      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3327        {        {
3328        while (*(++ptr) != 0)        ptr++;
3329          while (*ptr != 0)
3330          {          {
3331          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3332            ptr++;
3333    #ifdef SUPPORT_UTF8
3334            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3335    #endif
3336          }          }
3337        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3338    
# Line 2862  for (;; ptr++) Line 3366  for (;; ptr++)
3366          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3367          goto FAILED;          goto FAILED;
3368          }          }
3369        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3370        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3371        }        }
3372      return TRUE;      return TRUE;
# Line 2873  for (;; ptr++) Line 3377  for (;; ptr++)
3377      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3378    
3379      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3380        previous = NULL;
3381      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3382        {        {
3383        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3384          *code++ = OP_CIRCM;
3385        }        }
3386      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3387      break;      break;
3388    
3389      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3390      previous = NULL;      previous = NULL;
3391      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3392      break;      break;
3393    
3394      /* 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 3067  for (;; ptr++) Line 3572  for (;; ptr++)
3572            ptr++;            ptr++;
3573            }            }
3574    
3575          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3576          if (posix_class < 0)          if (posix_class < 0)
3577            {            {
3578            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 3081  for (;; ptr++) Line 3586  for (;; ptr++)
3586          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3587            posix_class = 0;            posix_class = 0;
3588    
3589          /* 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
3590          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3591          subtract bits that may be in the main map already. At the end we or the  
3592          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3593            if ((options & PCRE_UCP) != 0)
3594              {
3595              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3596              if (posix_substitutes[pc] != NULL)
3597                {
3598                nestptr = tempptr + 1;
3599                ptr = posix_substitutes[pc] - 1;
3600                continue;
3601                }
3602              }
3603    #endif
3604            /* In the non-UCP case, we build the bit map for the POSIX class in a
3605            chunk of local store because we may be adding and subtracting from it,
3606            and we don't want to subtract bits that may be in the main map already.
3607            At the end we or the result into the bit map that is being built. */
3608    
3609          posix_class *= 3;          posix_class *= 3;
3610    
# Line 3128  for (;; ptr++) Line 3648  for (;; ptr++)
3648    
3649        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3650        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
3651        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
3652        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
3653        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
3654        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3655          PCRE_EXTRA is set. */
3656    
3657        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3658          {          {
3659          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3660          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3661    
3662          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 */
         else if (-c == ESC_X) c = CHAR_X;   /* \X is literal X in a class */  
         else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */  
3663          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3664            {            {
3665            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3157  for (;; ptr++) Line 3676  for (;; ptr++)
3676            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3677            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3678    
3679            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3680              {              {
3681    #ifdef SUPPORT_UCP
3682                case ESC_du:     /* These are the values given for \d etc */
3683                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3684                case ESC_wu:     /* escape sequence with an appropriate \p */
3685                case ESC_WU:     /* or \P to test Unicode properties instead */
3686                case ESC_su:     /* of the default ASCII testing. */
3687                case ESC_SU:
3688                nestptr = ptr;
3689                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3690                class_charcount -= 2;                /* Undo! */
3691                continue;
3692    #endif
3693              case ESC_d:              case ESC_d:
3694              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3695              continue;              continue;
# Line 3179  for (;; ptr++) Line 3708  for (;; ptr++)
3708              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3709              continue;              continue;
3710    
3711                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3712                if it was previously set by something earlier in the character
3713                class. */
3714    
3715              case ESC_s:              case ESC_s:
3716              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3717              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3718                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3719              continue;              continue;
3720    
3721              case ESC_S:              case ESC_S:
# Line 3190  for (;; ptr++) Line 3724  for (;; ptr++)
3724              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3725              continue;              continue;
3726    
3727              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)  
             {  
3728              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3729              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3730              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 3227  for (;; ptr++) Line 3748  for (;; ptr++)
3748                }                }
3749  #endif  #endif
3750              continue;              continue;
             }  
3751    
3752            if (-c == ESC_H)              case ESC_H:
             {  
3753              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3754                {                {
3755                int x = 0xff;                int x = 0xff;
# Line 3272  for (;; ptr++) Line 3791  for (;; ptr++)
3791                }                }
3792  #endif  #endif
3793              continue;              continue;
             }  
3794    
3795            if (-c == ESC_v)              case ESC_v:
             {  
3796              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3797              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3798              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3291  for (;; ptr++) Line 3808  for (;; ptr++)
3808                }                }
3809  #endif  #endif
3810              continue;              continue;
             }  
3811    
3812            if (-c == ESC_V)              case ESC_V:
             {  
3813              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3814                {                {
3815                int x = 0xff;                int x = 0xff;
# Line 3324  for (;; ptr++) Line 3839  for (;; ptr++)
3839                }                }
3840  #endif  #endif
3841              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3842    
3843  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3844            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3845              {              case ESC_P:
3846              BOOL negated;                {
3847              int pdata;                BOOL negated;
3848              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3849              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3850              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3851              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3852                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3853              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3854              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3855              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3856              continue;                class_charcount -= 2;   /* Not a < 256 character */
3857              }                continue;
3858                  }
3859  #endif  #endif
3860            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3861            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3862            treated as literals. */              treated as literals. */
3863    
3864            if ((options & PCRE_EXTRA) != 0)              default:
3865              {              if ((options & PCRE_EXTRA) != 0)
3866              *errorcodeptr = ERR7;                {
3867              goto FAILED;                *errorcodeptr = ERR7;
3868                  goto FAILED;
3869                  }
3870                class_charcount -= 2;  /* Undo the default count from above */
3871                c = *ptr;              /* Get the final character and fall through */
3872                break;
3873              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3874            }            }
3875    
3876          /* 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 3425  for (;; ptr++) Line 3940  for (;; ptr++)
3940            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3941            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3942    
3943            /* \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 */  
3944    
3945            if (d < 0)            if (d < 0)
3946              {              {
3947              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  
3948                {                {
3949                ptr = oldptr;                ptr = oldptr;
3950                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3598  for (;; ptr++) Line 4110  for (;; ptr++)
4110          }          }
4111        }        }
4112    
4113      /* 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.
4114        If we are at the end of an internal nested string, revert to the outer
4115        string. */
4116    
4117        while (((c = *(++ptr)) != 0 ||
4118               (nestptr != NULL &&
4119                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4120               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4121    
4122      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
4123    
4124      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4125        {        {
4126        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4127        goto FAILED;        goto FAILED;
4128        }        }
4129    
   
 /* 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  
   
   
4130      /* 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
4131      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
4132      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 3634  we set the flag only if there is a liter Line 4134  we set the flag only if there is a liter
4134    
4135      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
4136      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4137      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4138      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4139    
4140      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
4141      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.
4142      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
4143      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
4144      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
4145      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4146    
4147  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4148      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3653  we set the flag only if there is a liter Line 4153  we set the flag only if there is a liter
4153        {        {
4154        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4155    
4156        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4157    
4158        if (negate_class)        if (negate_class)
4159          {          {
4160          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4161          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4162          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4163          *code++ = class_lastchar;          *code++ = class_lastchar;
4164          break;          break;
4165          }          }
# Line 3690  we set the flag only if there is a liter Line 4190  we set the flag only if there is a liter
4190    
4191      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4192      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4193      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
4194      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
4195      (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
4196      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
4197        actual compiled code. */
4198    
4199  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4200      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4201        {        {
4202        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4203        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3722  we set the flag only if there is a liter Line 4223  we set the flag only if there is a liter
4223        }        }
4224  #endif  #endif
4225    
4226      /* 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
4227      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
4228      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
4229      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4230        negating it if necessary. */
4231    
4232      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4233      if (negate_class)      if (negate_class)
# Line 3785  we set the flag only if there is a liter Line 4287  we set the flag only if there is a liter
4287      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4288      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4289    
4290      /* 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
4291      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4292    
4293      tempcode = previous;      tempcode = previous;
4294    
# Line 3809  we set the flag only if there is a liter Line 4311  we set the flag only if there is a liter
4311        }        }
4312      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4313    
4314        /* If previous was a recursion call, wrap it in atomic brackets so that
4315        previous becomes the atomic group. All recursions were so wrapped in the
4316        past, but it no longer happens for non-repeated recursions. In fact, the
4317        repeated ones could be re-implemented independently so as not to need this,
4318        but for the moment we rely on the code for repeating groups. */
4319    
4320        if (*previous == OP_RECURSE)
4321          {
4322          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4323          *previous = OP_ONCE;
4324          PUT(previous, 1, 2 + 2*LINK_SIZE);
4325          previous[2 + 2*LINK_SIZE] = OP_KET;
4326          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4327          code += 2 + 2 * LINK_SIZE;
4328          length_prevgroup = 3 + 3*LINK_SIZE;
4329    
4330          /* When actually compiling, we need to check whether this was a forward
4331          reference, and if so, adjust the offset. */
4332    
4333          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4334            {
4335            int offset = GET(cd->hwm, -LINK_SIZE);
4336            if (offset == previous + 1 - cd->start_code)
4337              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4338            }
4339          }
4340    
4341        /* Now handle repetition for the different types of item. */
4342    
4343      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4344      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
4345      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
4346      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
4347      instead.  */      instead.  */
4348    
4349      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4350        {        {
4351          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4352    
4353        /* 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
4354        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
4355        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 3849  we set the flag only if there is a liter Line 4382  we set the flag only if there is a liter
4382    
4383        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4384            repeat_max < 0 &&            repeat_max < 0 &&
4385            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4386          {          {
4387          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4388          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3862  we set the flag only if there is a liter Line 4394  we set the flag only if there is a liter
4394      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4395      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-
4396      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4397      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
4398      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4399    
4400      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4401        {        {
4402        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4403        c = previous[1];        c = previous[1];
4404        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4405            repeat_max < 0 &&            repeat_max < 0 &&
4406            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4407          {          {
4408          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4409          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3895  we set the flag only if there is a liter Line 4427  we set the flag only if there is a liter
4427    
4428        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4429            repeat_max < 0 &&            repeat_max < 0 &&
4430            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4431          {          {
4432          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4433          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4064  we set the flag only if there is a liter Line 4596  we set the flag only if there is a liter
4596  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4597               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4598  #endif  #endif
4599               *previous == OP_REF)               *previous == OP_REF ||
4600                 *previous == OP_REFI)
4601        {        {
4602        if (repeat_max == 0)        if (repeat_max == 0)
4603          {          {
# Line 4098  we set the flag only if there is a liter Line 4631  we set the flag only if there is a liter
4631        }        }
4632    
4633      /* 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
4634      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4635        opcodes such as BRA and CBRA, as this is the place where they get converted
4636        into the more special varieties such as BRAPOS and SBRA. A test for >=
4637        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4638        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4639        repetition of assertions, but now it does, for Perl compatibility. */
4640    
4641      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4642        {        {
4643        register int i;        register int i;
4644        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4645        uschar *bralink = NULL;        uschar *bralink = NULL;
4646          uschar *brazeroptr = NULL;
4647    
4648        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4649          we just ignore the repeat. */
4650    
4651        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4652          {          goto END_REPEAT;
         *errorcodeptr = ERR55;  
         goto FAILED;  
         }  
4653    
4654        /* If the maximum repeat count is unlimited, find the end of the bracket        /* There is no sense in actually repeating assertions. The only potential
4655        by scanning through from the start, and compute the offset back to it        use of repetition is in cases when the assertion is optional. Therefore,
4656        from the current code pointer. There may be an OP_OPT setting following        if the minimum is greater than zero, just ignore the repeat. If the
4657        the final KET, so we can't find the end just by going back from the code        maximum is not not zero or one, set it to 1. */
4658        pointer. */  
4659          if (*previous < OP_ONCE)    /* Assertion */
4660        if (repeat_max == -1)          {
4661          {          if (repeat_min > 0) goto END_REPEAT;
4662          register uschar *ket = previous;          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = code - ket;  
4663          }          }
4664    
4665        /* 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 4147  we set the flag only if there is a liter Line 4680  we set the flag only if there is a liter
4680          **   goto END_REPEAT;          **   goto END_REPEAT;
4681          **   }          **   }
4682    
4683          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
4684          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
4685          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
4686          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4687            selectively.
4688    
4689          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
4690          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 4170  we set the flag only if there is a liter Line 4704  we set the flag only if there is a liter
4704              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4705              goto END_REPEAT;              goto END_REPEAT;
4706              }              }
4707              brazeroptr = previous;    /* Save for possessive optimizing */
4708            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4709            }            }
4710    
# Line 4194  we set the flag only if there is a liter Line 4729  we set the flag only if there is a liter
4729            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4730            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4731    
4732            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4733            bralink = previous;            bralink = previous;
4734            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4735            }            }
# Line 4303  we set the flag only if there is a liter Line 4838  we set the flag only if there is a liter
4838              {              {
4839              int offset;              int offset;
4840              *code++ = OP_BRA;              *code++ = OP_BRA;
4841              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4842              bralink = code;              bralink = code;
4843              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4844              }              }
# Line 4324  we set the flag only if there is a liter Line 4859  we set the flag only if there is a liter
4859          while (bralink != NULL)          while (bralink != NULL)
4860            {            {
4861            int oldlinkoffset;            int oldlinkoffset;
4862            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4863            uschar *bra = code - offset;            uschar *bra = code - offset;
4864            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4865            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4334  we set the flag only if there is a liter Line 4869  we set the flag only if there is a liter
4869            }            }
4870          }          }
4871    
4872        /* 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
4873        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4874        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
4875        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4876          deal with possessive ONCEs specially.
4877    
4878          Otherwise, if the quantifier was possessive, we convert the BRA code to
4879          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4880          at runtime to detect this kind of subpattern at both the start and at the
4881          end.) The use of special opcodes makes it possible to reduce greatly the
4882          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4883          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4884          the default action below, of wrapping everything inside atomic brackets,
4885          does not happen.
4886    
4887        Then, when we are doing the actual compile phase, check to see whether        Then, when we are doing the actual compile phase, check to see whether
4888        this group is a non-atomic one that could match an empty string. If so,        this group is one that could match an empty string. If so, convert the
4889        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so that runtime
4890        that runtime checking can be done. [This check is also applied to        checking can be done. [This check is also applied to ONCE groups at
4891        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4892    
4893        else        else
4894          {          {
4895          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4896          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4897          *ketcode = OP_KETRMAX + repeat_type;  
4898          if (lengthptr == NULL && *bracode != OP_ONCE)          if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
4899                possessive_quantifier) *bracode = OP_BRA;
4900    
4901            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
4902              *ketcode = OP_KETRMAX + repeat_type;
4903            else
4904            {            {
4905            uschar *scode = bracode;            if (possessive_quantifier)
           do  
4906              {              {
4907              if (could_be_empty_branch(scode, ketcode, utf8))              *bracode += 1;                   /* Switch to xxxPOS opcodes */
4908                *ketcode = OP_KETRPOS;
4909                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4910                possessive_quantifier = FALSE;
4911                }
4912              else *ketcode = OP_KETRMAX + repeat_type;
4913    
4914              if (lengthptr == NULL)
4915                {
4916                uschar *scode = bracode;
4917                do
4918                {                {
4919                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4920                break;                  {
4921                    *bracode += OP_SBRA - OP_BRA;
4922                    break;
4923                    }
4924                  scode += GET(scode, 1);
4925                }                }
4926              scode += GET(scode, 1);              while (*scode == OP_ALT);
4927              }              }
           while (*scode == OP_ALT);  
4928            }            }
4929          }          }
4930        }        }
# Line 4383  we set the flag only if there is a liter Line 4945  we set the flag only if there is a liter
4945        }        }
4946    
4947      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4948      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4949      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4950      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4951      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
4952      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4953      tempcode, not at previous, which might be the first part of a string whose  
4954      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4955        just above, so possessive_quantifier is always FALSE for them at this
4956        stage.
4957    
4958        Note that the repeated item starts at tempcode, not at previous, which
4959        might be the first part of a string whose (former) last char we repeated.
4960    
4961      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
4962      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 4412  we set the flag only if there is a liter Line 4979  we set the flag only if there is a liter
4979  #endif  #endif
4980          }          }
4981    
4982        len = code - tempcode;        len = (int)(code - tempcode);
4983        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4984          {          {
4985          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4420  we set the flag only if there is a liter Line 4987  we set the flag only if there is a liter
4987          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4988          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4989    
4990          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4991          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4992          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4993          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4994    
4995          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4996          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4997          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4998          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4999    
5000            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
5001            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
5002            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
5003            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
5004    
5005            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
5006            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
5007            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
5008            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
5009    
5010            /* Because we are moving code along, we must ensure that any
5011            pending recursive references are updated. */
5012    
5013          default:          default:
5014            *code = OP_END;
5015            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5016          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
5017          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
5018          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4466  we set the flag only if there is a liter Line 5048  we set the flag only if there is a liter
5048    
5049      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
5050    
5051      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
5052             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
5053        {        {
5054        int i, namelen;        int i, namelen;
5055          int arglen = 0;
5056        const char *vn = verbnames;        const char *vn = verbnames;
5057        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
5058          const uschar *arg = NULL;
5059        previous = NULL;        previous = NULL;
5060        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5061          namelen = (int)(ptr - name);
5062    
5063          /* It appears that Perl allows any characters whatsoever, other than
5064          a closing parenthesis, to appear in arguments, so we no longer insist on
5065          letters, digits, and underscores. */
5066    
5067        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5068          {          {
5069          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
5070          goto FAILED;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5071            arglen = (int)(ptr - arg);
5072          }          }
5073    
5074        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
5075          {          {
5076          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
5077          goto FAILED;          goto FAILED;
5078          }          }
5079        namelen = ptr - name;  
5080          /* Scan the table of verb names */
5081    
5082        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
5083          {          {
5084          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5085              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5086            {            {
5087            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5088              ASSERT_ACCEPT if in an assertion. */
5089    
5090            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5091              {              {
5092              open_capitem *oc;              open_capitem *oc;
5093                if (arglen != 0)
5094                  {
5095                  *errorcodeptr = ERR59;
5096                  goto FAILED;
5097                  }
5098              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5099              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5100                {                {
5101                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5102                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5103                }                }
5104                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5105    
5106                /* Do not set firstbyte after *ACCEPT */
5107                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5108              }              }
5109            *code++ = verbs[i].op;  
5110            break;            /* Handle other cases with/without an argument */
5111    
5112              else if (arglen == 0)
5113                {
5114                if (verbs[i].op < 0)   /* Argument is mandatory */
5115                  {
5116                  *errorcodeptr = ERR66;
5117                  goto FAILED;
5118                  }
5119                *code = verbs[i].op;
5120                if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
5121                }
5122    
5123              else
5124                {
5125                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
5126                  {
5127                  *errorcodeptr = ERR59;
5128                  goto FAILED;
5129                  }
5130                *code = verbs[i].op_arg;
5131                if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
5132                *code++ = arglen;
5133                memcpy(code, arg, arglen);
5134                code += arglen;
5135                *code++ = 0;
5136                }
5137    
5138              break;  /* Found verb, exit loop */
5139            }            }
5140    
5141          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
5142          }          }
5143        if (i < verbcount) continue;  
5144        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
5145          *errorcodeptr = ERR60;          /* Verb not recognized */
5146        goto FAILED;        goto FAILED;
5147        }        }
5148    
# Line 4626  we set the flag only if there is a liter Line 5261  we set the flag only if there is a liter
5261                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5262            ptr++;            ptr++;
5263            }            }
5264          namelen = ptr - name;          namelen = (int)(ptr - name);
5265    
5266          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
5267              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
# Line 4687  we set the flag only if there is a liter Line 5322  we set the flag only if there is a liter
5322          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5323    
5324          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5325                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5326            {            {
5327            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5328            code[1+LINK_SIZE]++;            code[1+LINK_SIZE]++;
# Line 4755  we set the flag only if there is a liter Line 5390  we set the flag only if there is a liter
5390          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5391          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5392          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5393            cd->assert_depth += 1;
5394          ptr++;          ptr++;
5395          break;          break;
5396    
# Line 4769  we set the flag only if there is a liter Line 5405  we set the flag only if there is a liter
5405            continue;            continue;
5406            }            }
5407          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5408            cd->assert_depth += 1;
5409          break;          break;
5410    
5411    
# Line 4778  we set the flag only if there is a liter Line 5415  we set the flag only if there is a liter
5415            {            {
5416            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5417            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5418              cd->assert_depth += 1;
5419            ptr += 2;            ptr += 2;
5420            break;            break;
5421    
5422            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5423            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5424              cd->assert_depth += 1;
5425            ptr += 2;            ptr += 2;
5426            break;            break;
5427    
# Line 4822  we set the flag only if there is a liter Line 5461  we set the flag only if there is a liter
5461              goto FAILED;              goto FAILED;
5462              }              }
5463            *code++ = n;            *code++ = n;
5464            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5465            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5466            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5467            }            }
5468          previous = NULL;          previous = NULL;
# Line 4856  we set the flag only if there is a liter Line 5495  we set the flag only if there is a liter
5495            name = ++ptr;            name = ++ptr;
5496    
5497            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5498            namelen = ptr - name;            namelen = (int)(ptr - name);
5499    
5500            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5501    
# Line 4986  we set the flag only if there is a liter Line 5625  we set the flag only if there is a liter
5625          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5626          name = ++ptr;          name = ++ptr;
5627          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5628          namelen = ptr - name;          namelen = (int)(ptr - name);
5629    
5630          /* 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
5631          reference number. */          a dummy reference number, because it was not used in the first pass.
5632            However, with the change of recursive back references to be atomic,
5633            we have to look for the number so that this state can be identified, as
5634            otherwise the incorrect length is computed. If it's not a backwards
5635            reference, the dummy number will do. */
5636    
5637          if (lengthptr != NULL)          if (lengthptr != NULL)
5638            {            {
5639              const uschar *temp;
5640    
5641            if (namelen == 0)            if (namelen == 0)
5642              {              {
5643              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5008  we set the flag only if there is a liter Line 5653  we set the flag only if there is a liter
5653              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5654              goto FAILED;              goto FAILED;
5655              }              }
5656            recno = 0;  
5657              /* The name table does not exist in the first pass, so we cannot
5658              do a simple search as in the code below. Instead, we have to scan the
5659              pattern to find the number. It is important that we scan it only as
5660              far as we have got because the syntax of named subpatterns has not
5661              been checked for the rest of the pattern, and find_parens() assumes
5662              correct syntax. In any case, it's a waste of resources to scan
5663              further. We stop the scan at the current point by temporarily
5664              adjusting the value of cd->endpattern. */
5665    
5666              temp = cd->end_pattern;
5667              cd->end_pattern = ptr;
5668              recno = find_parens(cd, name, namelen,
5669                (options & PCRE_EXTENDED) != 0, utf8);
5670              cd->end_pattern = temp;
5671              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5672            }            }
5673    
5674          /* 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 5033  we set the flag only if there is a liter Line 5693  we set the flag only if there is a liter
5693              }              }
5694            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5695                      find_parens(cd, name, namelen,                      find_parens(cd, name, namelen,
5696                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5697              {              {
5698              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5699              goto FAILED;              goto FAILED;
# Line 5144  we set the flag only if there is a liter Line 5804  we set the flag only if there is a liter
5804              if (called == NULL)              if (called == NULL)
5805                {                {
5806                if (find_parens(cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5807                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5808                  {                  {
5809                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5810                  goto FAILED;                  goto FAILED;
5811                  }                  }
5812    
5813                  /* Fudge the value of "called" so that when it is inserted as an
5814                  offset below, what it actually inserted is the reference number
5815                  of the group. Then remember the forward reference. */
5816    
5817                called = cd->start_code + recno;                called = cd->start_code + recno;
5818                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5819                }                }
5820    
5821              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
5822              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
5823              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
5824                must not, however, do this check if we are in a conditional
5825                subpattern because the condition might be testing for recursion in
5826                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
5827                Forever loops are also detected at runtime, so those that occur in
5828                conditional subpatterns will be picked up then. */
5829    
5830              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
5831                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5832                {                {
5833                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5834                goto FAILED;                goto FAILED;
5835                }                }
5836              }              }
5837    
5838            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
           "once" brackets. Set up a "previous group" length so that a  
           subsequent quantifier will work. */  
   
           *code = OP_ONCE;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
5839    
5840            *code = OP_RECURSE;            *code = OP_RECURSE;
5841            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);  
5842            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
   
           length_prevgroup = 3 + 3*LINK_SIZE;  
5843            }            }
5844    
5845          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5242  we set the flag only if there is a liter Line 5900  we set the flag only if there is a liter
5900          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
5901          both phases.          both phases.
5902    
5903          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
5904          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. */  
5905    
5906          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5907            {            {
# Line 5255  we set the flag only if there is a liter Line 5912  we set the flag only if there is a liter
5912              }              }
5913            else            else
5914              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
5915              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5916              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5917              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5918              }              }
5919    
5920            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
5921            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). */  
5922    
5923            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
5924            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5285  we set the flag only if there is a liter Line 5935  we set the flag only if there is a liter
5935          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
5936        }       /* End of (? handling */        }       /* End of (? handling */
5937    
5938      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
5939      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
5940      brackets. */      brackets. */
5941    
5942      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 5304  we set the flag only if there is a liter Line 5954  we set the flag only if there is a liter
5954        skipbytes = 2;        skipbytes = 2;
5955        }        }
5956    
5957      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
5958      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
5959      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
5960      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. */  
5961    
5962      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                      /* For handling repetition */
5963      *code = bravalue;      *code = bravalue;
5964      tempcode = code;      tempcode = code;
5965      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
5966      length_prevgroup = 0;              /* Initialize for pre-compile phase */      tempbracount = cd->bracount;          /* Save value before bracket */
5967        length_prevgroup = 0;                 /* Initialize for pre-compile phase */
5968    
5969      if (!compile_regex(      if (!compile_regex(
5970           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
5971           options & PCRE_IMS,           /* The previous ims option state */           &tempcode,                       /* Where to put code (updated) */
5972           &tempcode,                    /* Where to put code (updated) */           &ptr,                            /* Input pointer (updated) */
5973           &ptr,                         /* Input pointer (updated) */           errorcodeptr,                    /* Where to put an error message */
          errorcodeptr,                 /* Where to put an error message */  
5974           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
5975            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
5976           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
5977           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
5978           &subfirstbyte,                /* For possible first char */           cond_depth +
5979           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
5980           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
5981           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
5982           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
5983             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
5984             (lengthptr == NULL)? NULL :      /* Actual compile phase */
5985               &length_prevgroup              /* Pre-compile phase */
5986           ))           ))
5987        goto FAILED;        goto FAILED;
5988    
5989        /* If this was an atomic group and there are no capturing groups within it,
5990        generate OP_ONCE_NC instead of OP_ONCE. */
5991    
5992        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
5993          *code = OP_ONCE_NC;
5994    
5995        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5996          cd->assert_depth -= 1;
5997    
5998      /* 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
5999      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.
6000      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
6001    
6002      /* 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
6003      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
6004      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
6005      not be available. */      not be available. */
# Line 5405  we set the flag only if there is a liter Line 6064  we set the flag only if there is a liter
6064          goto FAILED;          goto FAILED;
6065          }          }
6066        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
6067        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
6068        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
6069        *code++ = OP_KET;        *code++ = OP_KET;
6070        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5478  we set the flag only if there is a liter Line 6137  we set the flag only if there is a liter
6137    
6138      /* ===================================================================*/      /* ===================================================================*/
6139      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
6140      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
6141      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
6142      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
6143      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
6144      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
6145        ever created. */
6146    
6147      case CHAR_BACKSLASH:      case CHAR_BACKSLASH:
6148      tempptr = ptr;      tempptr = ptr;
# Line 5572  we set the flag only if there is a liter Line 6232  we set the flag only if there is a liter
6232          }          }
6233    
6234        /* \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).
6235        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6236    
6237        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6238          {          {
6239            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6240              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6241              {
6242              *errorcodeptr = ERR69;
6243              break;
6244              }
6245          is_recurse = FALSE;          is_recurse = FALSE;
6246          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6247            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
# Line 5590  we set the flag only if there is a liter Line 6255  we set the flag only if there is a liter
6255    
6256        if (-c >= ESC_REF)        if (-c >= ESC_REF)
6257          {          {
6258            open_capitem *oc;
6259          recno = -c - ESC_REF;          recno = -c - ESC_REF;
6260    
6261          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6262          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6263          previous = code;          previous = code;
6264          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6265          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6266          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6267          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
6268    
6269            /* Check to see if this back reference is recursive, that it, it
6270            is inside the group that it references. A flag is set so that the
6271            group can be made atomic. */
6272    
6273            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6274              {
6275              if (oc->number == recno)
6276                {
6277                oc->flag = TRUE;
6278                break;
6279                }
6280              }
6281          }          }
6282    
6283        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5628  we set the flag only if there is a liter Line 6307  we set the flag only if there is a liter
6307  #endif  #endif
6308    
6309        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
6310        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
6311          situation when PCRE_UCP is not set. When it *is* set, we substitute
6312          Unicode property tests. */
6313    
6314        else        else
6315          {          {
6316          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
6317          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
6318              {
6319              nestptr = ptr + 1;                   /* Where to resume */
6320              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
6321              }
6322            else
6323    #endif
6324              {
6325              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6326              *code++ = -c;
6327              }
6328          }          }
6329        continue;        continue;
6330        }        }
# Line 5678  we set the flag only if there is a liter Line 6369  we set the flag only if there is a liter
6369    
6370      ONE_CHAR:      ONE_CHAR:
6371      previous = code;      previous = code;
6372      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6373      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6374    
6375      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 5707  we set the flag only if there is a liter Line 6398  we set the flag only if there is a liter
6398        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6399        }        }
6400    
6401      /* 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
6402      1 or the matching is caseful. */      1 or the matching is caseful. */
6403    
6404      else      else
# Line 5742  return FALSE; Line 6433  return FALSE;
6433  /* 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
6434  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
6435  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.  
   
6436  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
6437  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
6438  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6439    
6440  Arguments:  Arguments:
6441    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  
6442    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6443    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6444    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
6445    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6446    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6447    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6448      cond_depth     depth of nesting for conditional subpatterns
6449    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6450    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6451    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 5771  Returns:         TRUE on success Line 6457  Returns:         TRUE on success
6457  */  */
6458    
6459  static BOOL  static BOOL
6460  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6461    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6462    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6463    int *lengthptr)    compile_data *cd, int *lengthptr)
6464  {  {
6465  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6466  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 5788  int branchfirstbyte, branchreqbyte; Line 6474  int branchfirstbyte, branchreqbyte;
6474  int length;  int length;
6475  int orig_bracount;  int orig_bracount;
6476  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6477  branch_chain bc;  branch_chain bc;
6478    
6479  bc.outer = bcptr;  bc.outer = bcptr;
# Line 5811  them global. It tests the value of lengt Line 6496  them global. It tests the value of lengt
6496  pre-compile phase to find out whether anything has yet been compiled or not. */  pre-compile phase to find out whether anything has yet been compiled or not. */
6497    
6498  /* 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
6499  so that we can detect them if (*ACCEPT) is encountered. */  so that we can detect them if (*ACCEPT) is encountered. This is also used to
6500    detect groups that contain recursive back references to themselves. Note that
6501    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6502    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6503    
6504  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6505    {    {
6506    capnumber = GET2(code, 1 + LINK_SIZE);    capnumber = GET2(code, 1 + LINK_SIZE);
6507    capitem.number = capnumber;    capitem.number = capnumber;
6508    capitem.next = cd->open_caps;    capitem.next = cd->open_caps;
6509      capitem.flag = FALSE;
6510    cd->open_caps = &capitem;    cd->open_caps = &capitem;
6511    }    }
6512    
# Line 5836  for (;;) Line 6525  for (;;)
6525    
6526    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6527    
   /* 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;  
     }  
   
6528    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6529    
6530    if (lookbehind)    if (lookbehind)
# Line 5859  for (;;) Line 6539  for (;;)
6539    into the length. */    into the length. */
6540    
6541    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6542          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6543            (lengthptr == NULL)? NULL : &length))
6544      {      {
6545      *ptrptr = ptr;      *ptrptr = ptr;
6546      return FALSE;      return FALSE;
6547      }      }
6548    
   /* 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;  
   
6549    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6550    has fewer than the rest. */    has fewer than the rest. */
6551    
# Line 5934  for (;;) Line 6606  for (;;)
6606        {        {
6607        int fixed_length;        int fixed_length;
6608        *code = OP_END;        *code = OP_END;
6609        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6610            FALSE, cd);
6611        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6612        if (fixed_length == -3)        if (fixed_length == -3)
6613          {          {
# Line 5955  for (;;) Line 6628  for (;;)
6628    of offsets, with the field in the BRA item now becoming an offset to the    of offsets, with the field in the BRA item now becoming an offset to the
6629    first alternative. If there are no alternatives, it points to the end of the    first alternative. If there are no alternatives, it points to the end of the
6630    group. The length in the terminating ket is always the length of the whole    group. The length in the terminating ket is always the length of the whole
6631    bracketed item. If any of the ims options were changed inside the group,    bracketed item. Return leaving the pointer at the terminating char. */
   compile a resetting op-code following, except at the very end of the pattern.  
   Return leaving the pointer at the terminating char. */  
6632    
6633    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6634      {      {
6635      if (lengthptr == NULL)      if (lengthptr == NULL)
6636        {        {
6637        int branch_length = code - last_branch;        int branch_length = (int)(code - last_branch);
6638        do        do
6639          {          {
6640          int prev_length = GET(last_branch, 1);          int prev_length = GET(last_branch, 1);
# Line 5974  for (;;) Line 6645  for (;;)
6645        while (branch_length > 0);        while (branch_length > 0);
6646        }        }
6647    
     /* If it was a capturing subpattern, remove it from the chain. */  
   
     if (capnumber > 0) cd->open_caps = cd->open_caps->next;  
   
6648      /* Fill in the ket */      /* Fill in the ket */
6649    
6650      *code = OP_KET;      *code = OP_KET;
6651      PUT(code, 1, code - start_bracket);      PUT(code, 1, (int)(code - start_bracket));
6652      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6653    
6654      /* Reset options if needed. */      /* If it was a capturing subpattern, check to see if it contained any
6655        recursive back references. If so, we must wrap it in atomic brackets.
6656        In any event, remove the block from the chain. */
6657    
6658      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)      if (capnumber > 0)
6659        {        {
6660        *code++ = OP_OPT;        if (cd->open_caps->flag)
6661        *code++ = oldims;          {
6662        length += 2;          memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6663              code - start_bracket);
6664            *start_bracket = OP_ONCE;
6665            code += 1 + LINK_SIZE;
6666            PUT(start_bracket, 1, (int)(code - start_bracket));
6667            *code = OP_KET;
6668            PUT(code, 1, (int)(code - start_bracket));
6669            code += 1 + LINK_SIZE;
6670            length += 2 + 2*LINK_SIZE;
6671            }
6672          cd->open_caps = cd->open_caps->next;
6673        }        }
6674    
6675      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
# Line 6032  for (;;) Line 6711  for (;;)
6711    else    else
6712      {      {
6713      *code = OP_ALT;      *code = OP_ALT;
6714      PUT(code, 1, code - last_branch);      PUT(code, 1, (int)(code - last_branch));
6715      bc.current_branch = last_branch = code;      bc.current_branch = last_branch = code;
6716      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6717      }      }
# Line 6052  for (;;) Line 6731  for (;;)
6731  /* Try to find out if this is an anchored regular expression. Consider each  /* Try to find out if this is an anchored regular expression. Consider each
6732  alternative branch. If they all start with OP_SOD or OP_CIRC, or with a bracket  alternative branch. If they all start with OP_SOD or OP_CIRC, or with a bracket
6733  all of whose alternatives start with OP_SOD or OP_CIRC (recurse ad lib), then  all of whose alternatives start with OP_SOD or OP_CIRC (recurse ad lib), then
6734  it's anchored. However, if this is a multiline pattern, then only OP_SOD  it's anchored. However, if this is a multiline pattern, then only OP_SOD will
6735  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6736    
6737  We can also consider a regex to be anchored if OP_SOM starts all its branches.  We can also consider a regex to be anchored if OP_SOM starts all its branches.
6738  This is the code for \G, which means "match at start of match position, taking  This is the code for \G, which means "match at start of match position, taking
# Line 6074  of the more common cases more precisely. Line 6753  of the more common cases more precisely.
6753    
6754  Arguments:  Arguments:
6755    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6756    bracket_map    a bitmap of which brackets we are inside while testing; this    bracket_map    a bitmap of which brackets we are inside while testing; this
6757                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6758                    the less precise approach                    the less precise approach
# Line 6084  Returns:     TRUE or FALSE Line 6762  Returns:     TRUE or FALSE
6762  */  */
6763    
6764  static BOOL  static BOOL
6765  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6766    unsigned int backref_map)    unsigned int backref_map)
6767  {  {
6768  do {  do {
6769     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6770       options, PCRE_MULTILINE, FALSE);       FALSE);
6771     register int op = *scode;     register int op = *scode;
6772    
6773     /* Non-capturing brackets */     /* Non-capturing brackets */
6774    
6775     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6776           op == OP_SBRA || op == OP_SBRAPOS)
6777       {       {
6778       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6779       }       }
6780    
6781     /* Capturing brackets */     /* Capturing brackets */
6782    
6783     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6784                op == OP_SCBRA || op == OP_SCBRAPOS)
6785       {       {
6786       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6787       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6788       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6789       }       }
6790    
6791     /* Other brackets */     /* Other brackets */
6792    
6793     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC ||
6794                op == OP_COND)
6795       {       {
6796       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6797       }       }
6798    
6799     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
# Line 6127  do { Line 6808  do {
6808    
6809     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6810    
6811     else if (op != OP_SOD && op != OP_SOM &&     else if (op != OP_SOD && op != OP_SOM && op != OP_CIRC) return FALSE;
            ((*options & PCRE_MULTILINE) != 0 || op != OP_CIRC))  
      return FALSE;  
6812     code += GET(code, 1);     code += GET(code, 1);
6813     }     }
6814  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6165  is_startline(const uschar *code, unsigne Line 6844  is_startline(const uschar *code, unsigne
6844  {  {
6845  do {  do {
6846     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6847       NULL, 0, FALSE);       FALSE);
6848     register int op = *scode;     register int op = *scode;
6849    
6850     /* If we are at the start of a conditional assertion group, *both* the     /* If we are at the start of a conditional assertion group, *both* the
# Line 6192  do { Line 6871  do {
6871         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6872         break;         break;
6873         }         }
6874       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
6875       op = *scode;       op = *scode;
6876       }       }
6877    
6878     /* Non-capturing brackets */     /* Non-capturing brackets */
6879    
6880     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6881           op == OP_SBRA || op == OP_SBRAPOS)
6882       {       {
6883       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6884       }       }
6885    
6886     /* Capturing brackets */     /* Capturing brackets */
6887    
6888     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6889                op == OP_SCBRA || op == OP_SCBRAPOS)
6890       {       {
6891       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6892       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6214  do { Line 6895  do {
6895    
6896     /* Other brackets */     /* Other brackets */
6897    
6898     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC)
6899       {       {
6900       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6901       }       }
# Line 6229  do { Line 6910  do {
6910    
6911     /* Check for explicit circumflex */     /* Check for explicit circumflex */
6912    
6913     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
6914    
6915     /* Move on to the next alternative */     /* Move on to the next alternative */
6916    
# Line 6255  we return that char, otherwise -1. Line 6936  we return that char, otherwise -1.
6936    
6937  Arguments:  Arguments:
6938    code       points to start of expression (the bracket)    code       points to start of expression (the bracket)
   options    pointer to the options (used to check casing changes)  
6939    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
6940    
6941  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
6942  */  */
6943    
6944  static int  static int
6945  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
6946  {  {
6947  register int c = -1;  register int c = -1;
6948  do {  do {
6949     int d;     int d;
6950     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
6951       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
6952       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
6953     register int op = *scode;     register int op = *scode;
6954    
6955     switch(op)     switch(op)
# Line 6277  do { Line 6958  do {
6958       return -1;       return -1;
6959    
6960       case OP_BRA:       case OP_BRA:
6961         case OP_BRAPOS:
6962       case OP_CBRA:       case OP_CBRA:
6963         case OP_SCBRA:
6964         case OP_CBRAPOS:
6965         case OP_SCBRAPOS:
6966       case OP_ASSERT:       case OP_ASSERT:
6967       case OP_ONCE:       case OP_ONCE:
6968         case OP_ONCE_NC:
6969       case OP_COND:       case OP_COND:
6970       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
6971         return -1;         return -1;
6972       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
6973       break;       break;
6974    
6975       case OP_EXACT:       /* Fall through */       case OP_EXACT:
6976       scode += 2;       scode += 2;
6977         /* Fall through */
6978    
6979       case OP_CHAR:       case OP_CHAR:
      case OP_CHARNC:  
6980       case OP_PLUS:       case OP_PLUS:
6981       case OP_MINPLUS:       case OP_MINPLUS:
6982       case OP_POSPLUS:       case OP_POSPLUS:
6983       if (!inassert) return -1;       if (!inassert) return -1;
6984       if (c < 0)       if (c < 0) c = scode[1];
6985         {         else if (c != scode[1]) return -1;
6986         c = scode[1];       break;
6987         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
6988         }       case OP_EXACTI:
6989       else if (c != scode[1]) return -1;       scode += 2;
6990         /* Fall through */
6991    
6992         case OP_CHARI:
6993         case OP_PLUSI:
6994         case OP_MINPLUSI:
6995         case OP_POSPLUSI:
6996         if (!inassert) return -1;
6997         if (c < 0) c = scode[1] | REQ_CASELESS;
6998           else if (c != scode[1]) return -1;
6999       break;       break;
7000       }       }
7001    
# Line 6351  int length = 1;  /* For final END opcode Line 7046  int length = 1;  /* For final END opcode
7046  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
7047  int errorcode = 0;  int errorcode = 0;
7048  int skipatstart = 0;  int skipatstart = 0;
7049  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8;
7050  size_t size;  size_t size;
7051  uschar *code;  uschar *code;
7052  const uschar *codestart;  const uschar *codestart;
# Line 6421  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7116  while (ptr[skipatstart] == CHAR_LEFT_PAR
7116    
7117    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
7118      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
7119      else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
7120        { skipatstart += 6; options |= PCRE_UCP; continue; }
7121      else if (strncmp((char *)(ptr+skipatstart+2), STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
7122        { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
7123    
7124    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
7125      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6445  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7144  while (ptr[skipatstart] == CHAR_LEFT_PAR
7144    else break;    else break;
7145    }    }
7146    
7147  /* Can't support UTF8 unless PCRE has been compiled to include the code. */  utf8 = (options & PCRE_UTF8) != 0;
7148    
7149    /* Can't support UTF8 unless PCRE has been compiled to include the code. The
7150    return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7151    release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7152    not used here. */
7153    
7154  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7155  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7156       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7157    {    {
7158    errorcode = ERR44;    errorcode = ERR44;
7159    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6462  if (utf8) Line 7166  if (utf8)
7166    }    }
7167  #endif  #endif
7168    
7169    /* Can't support UCP unless PCRE has been compiled to include the code. */
7170    
7171    #ifndef SUPPORT_UCP
7172    if ((options & PCRE_UCP) != 0)
7173      {
7174      errorcode = ERR67;
7175      goto PCRE_EARLY_ERROR_RETURN;
7176      }
7177    #endif
7178    
7179  /* Check validity of \R options. */  /* Check validity of \R options. */
7180    
7181  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
7182         (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
7183    {    {
7184    case 0:    errorcode = ERR56;
7185    case PCRE_BSR_ANYCRLF:    goto PCRE_EARLY_ERROR_RETURN;
   case PCRE_BSR_UNICODE:  
   break;  
   default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;  
7186    }    }
7187    
7188  /* Handle different types of newline. The three bits give seven cases. The  /* Handle different types of newline. The three bits give seven cases. The
# Line 6555  outside can help speed up starting point Line 7267  outside can help speed up starting point
7267  ptr += skipatstart;  ptr += skipatstart;
7268  code = cworkspace;  code = cworkspace;
7269  *code = OP_BRA;  *code = OP_BRA;
7270  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7271    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7272  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7273    
7274  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 6590  regex compiled on a system with 4-byte p Line 7301  regex compiled on a system with 4-byte p
7301  pointers. */  pointers. */
7302    
7303  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
7304  re->size = size;  re->size = (int)size;
7305  re->options = cd->external_options;  re->options = cd->external_options;
7306  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7307  re->dummy1 = 0;  re->dummy1 = 0;
# Line 6611  field; this time it's used for rememberi Line 7322  field; this time it's used for rememberi
7322  */  */
7323    
7324  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7325    cd->assert_depth = 0;
7326  cd->bracount = 0;  cd->bracount = 0;
7327  cd->names_found = 0;  cd->names_found = 0;
7328  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 6629  of the function here. */ Line 7341  of the function here. */
7341  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7342  code = (uschar *)codestart;  code = (uschar *)codestart;
7343  *code = OP_BRA;  *code = OP_BRA;
7344  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0, 0,
7345    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7346  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7347  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7348  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7349    
7350  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */  if (cd->had_accept) reqbyte = REQ_NONE;   /* Must disable after (*ACCEPT) */
7351    
7352  /* If not reached end of pattern on success, there's an excess bracket. */  /* If not reached end of pattern on success, there's an excess bracket. */
7353    
# Line 6661  while (errorcode == 0 && cd->hwm > cwork Line 7373  while (errorcode == 0 && cd->hwm > cwork
7373    recno = GET(codestart, offset);    recno = GET(codestart, offset);
7374    groupptr = _pcre_find_bracket(codestart, utf8, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
7375    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
7376      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, (int)(groupptr - codestart));
7377    }    }
7378    
7379  /* Give an error if there's back reference to a non-existent capturing  /* Give an error if there's back reference to a non-existent capturing
# Line 6696  if (cd->check_lookbehind) Line 7408  if (cd->check_lookbehind)
7408        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7409        int end_op = *be;        int end_op = *be;
7410        *be = OP_END;        *be = OP_END;
7411        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7412            cd);
7413        *be = end_op;        *be = end_op;
7414        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7415        if (fixed_length < 0)        if (fixed_length < 0)
# Line 6716  if (errorcode != 0) Line 7429  if (errorcode != 0)
7429    {    {
7430    (pcre_free)(re);    (pcre_free)(re);
7431    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
7432    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = (int)(ptr - (const uschar *)pattern);
7433    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
7434    *errorptr = find_error_text(errorcode);    *errorptr = find_error_text(errorcode);
7435    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
# Line 6735  start with ^. and also when all branches Line 7448  start with ^. and also when all branches
7448    
7449  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7450    {    {
7451    int temp_options = re->options;   /* May get changed during these scans */    if (is_anchored(codestart, 0, cd->backref_map))
   if (is_anchored(codestart, &temp_options, 0, cd->backref_map))  
7452      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7453    else    else
7454      {      {
7455      if (firstbyte < 0)      if (firstbyte < 0)
7456        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7457      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7458        {        {
7459        int ch = firstbyte & 255;        int ch = firstbyte & 255;
# Line 6771  if (reqbyte >= 0 && Line 7483  if (reqbyte >= 0 &&
7483  case when building a production library. */  case when building a production library. */
7484    
7485  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
   
7486  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
7487    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
7488    

Legend:
Removed from v.487  
changed lines
  Added in v.746

  ViewVC Help
Powered by ViewVC 1.1.5