/[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 503 by ph10, Sun Mar 7 17:35:52 2010 UTC revision 835 by ph10, Wed Dec 28 16:10:09 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  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  substring, so that the whole string ends with \0\0, which can be detected when
327  counting through. */  counting through. */
328    
329  static const char error_texts[] =  static const char error_texts[] =
# Line 314  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 336  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 348  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\0";    "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      /* 70 */
414      "internal error: unknown opcode in find_fixedlength()\0"
415      ;
416    
417  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
418  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 484  static const unsigned char ebcdic_charta Line 548  static const unsigned char ebcdic_charta
548  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
549    
550  static BOOL  static BOOL
551    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
552      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
553    
554    
# Line 506  static const char * Line 570  static const char *
570  find_error_text(int n)  find_error_text(int n)
571  {  {
572  const char *s = error_texts;  const char *s = error_texts;
573  for (; n > 0; n--)  for (; n > 0; n--)
574    {    {
575    while (*s++ != 0) {};    while (*s++ != 0) {};
576    if (*s == 0) return "Error text not found (please report)";    if (*s == 0) return "Error text not found (please report)";
577    }    }
578  return s;  return s;
579  }  }
580    
581    
582  /*************************************************  /*************************************************
583    *            Check for counted repeat            *
584    *************************************************/
585    
586    /* This function is called when a '{' is encountered in a place where it might
587    start a quantifier. It looks ahead to see if it really is a quantifier or not.
588    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
589    where the ddds are digits.
590    
591    Arguments:
592      p         pointer to the first char after '{'
593    
594    Returns:    TRUE or FALSE
595    */
596    
597    static BOOL
598    is_counted_repeat(const uschar *p)
599    {
600    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
601    while ((digitab[*p] & ctype_digit) != 0) p++;
602    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
603    
604    if (*p++ != CHAR_COMMA) return FALSE;
605    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
606    
607    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
608    while ((digitab[*p] & ctype_digit) != 0) p++;
609    
610    return (*p == CHAR_RIGHT_CURLY_BRACKET);
611    }
612    
613    
614    
615    /*************************************************
616  *            Handle escapes                      *  *            Handle escapes                      *
617  *************************************************/  *************************************************/
618    
# Line 581  else Line 678  else
678    
679      case CHAR_l:      case CHAR_l:
680      case CHAR_L:      case CHAR_L:
681      case CHAR_N:      *errorcodeptr = ERR37;
682        break;
683    
684      case CHAR_u:      case CHAR_u:
685        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
686          {
687          /* In JavaScript, \u must be followed by four hexadecimal numbers.
688          Otherwise it is a lowercase u letter. */
689          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
690               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
691            {
692            c = 0;
693            for (i = 0; i < 4; ++i)
694              {
695              register int cc = *(++ptr);
696    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
697              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
698              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
699    #else           /* EBCDIC coding */
700              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
701              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
702    #endif
703              }
704            }
705          }
706        else
707          *errorcodeptr = ERR37;
708        break;
709    
710      case CHAR_U:      case CHAR_U:
711      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
712        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
713      break;      break;
714    
715      /* \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
716        class, \g must be followed by one of a number of specific things:
717    
718      (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
719      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 604  else Line 730  else
730      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
731    
732      case CHAR_g:      case CHAR_g:
733        if (isclass) break;
734      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
735        {        {
736        c = -ESC_g;        c = -ESC_g;
# Line 732  else Line 859  else
859      treated as a data character. */      treated as a data character. */
860    
861      case CHAR_x:      case CHAR_x:
862        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
863          {
864          /* In JavaScript, \x must be followed by two hexadecimal numbers.
865          Otherwise it is a lowercase x letter. */
866          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
867            {
868            c = 0;
869            for (i = 0; i < 2; ++i)
870              {
871              register int cc = *(++ptr);
872    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
873              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
874              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
875    #else           /* EBCDIC coding */
876              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
877              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
878    #endif
879              }
880            }
881          break;
882          }
883    
884      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
885        {        {
886        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
# Line 782  else Line 931  else
931      break;      break;
932    
933      /* 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.
934      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
935        coding is ASCII-specific, but then the whole concept of \cx is
936      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
937    
938      case CHAR_c:      case CHAR_c:
# Line 792  else Line 942  else
942        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
943        break;        break;
944        }        }
945    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
946  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
947          {
948          *errorcodeptr = ERR68;
949          break;
950          }
951      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
952      c ^= 0x40;      c ^= 0x40;
953  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
954      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
955      c ^= 0xC0;      c ^= 0xC0;
956  #endif  #endif
# Line 819  else Line 973  else
973      }      }
974    }    }
975    
976    /* Perl supports \N{name} for character names, as well as plain \N for "not
977    newline". PCRE does not support \N{name}. However, it does support
978    quantification such as \N{2,3}. */
979    
980    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
981         !is_counted_repeat(ptr+2))
982      *errorcodeptr = ERR37;
983    
984    /* If PCRE_UCP is set, we change the values for \d etc. */
985    
986    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
987      c -= (ESC_DU - ESC_D);
988    
989    /* Set the pointer to the final character before returning. */
990    
991  *ptrptr = ptr;  *ptrptr = ptr;
992  return c;  return c;
993  }  }
# Line 919  return -1; Line 1088  return -1;
1088    
1089    
1090  /*************************************************  /*************************************************
 *            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);  
 }  
   
   
   
 /*************************************************  
1091  *         Read repeat counts                     *  *         Read repeat counts                     *
1092  *************************************************/  *************************************************/
1093    
# Line 1027  top-level call starts at the beginning o Line 1163  top-level call starts at the beginning o
1163  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
1164  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
1165  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
1166  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
1167  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1168  first pass. Recursion is used to keep track of subpatterns that reset the  
1169  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1170    that if (?< or (?' or (?P< is encountered, the name will be correctly
1171    terminated because that is checked in the first pass. There is now one call to
1172    this function in the first pass, to check for a recursive back reference by
1173    name (so that we can make the whole group atomic). In this case, we need check
1174    only up to the current position in the pattern, and that is still OK because
1175    and previous occurrences will have been checked. To make this work, the test
1176    for "end of pattern" is a check against cd->end_pattern in the main loop,
1177    instead of looking for a binary zero. This means that the special first-pass
1178    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1179    processing items within the loop are OK, because afterwards the main loop will
1180    terminate.)
1181    
1182  Arguments:  Arguments:
1183    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1038  Arguments: Line 1185  Arguments:
1185    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1186    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1187    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1188      utf8         TRUE if we are in UTF-8 mode
1189    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1190    
1191  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
# Line 1045  Returns: the number of the named s Line 1193  Returns: the number of the named s
1193    
1194  static int  static int
1195  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,
1196    BOOL xmode, int *count)    BOOL xmode, BOOL utf8, int *count)
1197  {  {
1198  uschar *ptr = *ptrptr;  uschar *ptr = *ptrptr;
1199  int start_count = *count;  int start_count = *count;
# Line 1057  dealing with. The very first call may no Line 1205  dealing with. The very first call may no
1205    
1206  if (ptr[0] == CHAR_LEFT_PARENTHESIS)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1207    {    {
1208    if (ptr[1] == CHAR_QUESTION_MARK &&    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1209        ptr[2] == CHAR_VERTICAL_LINE)  
1210      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1211    
1212      /* Handle a normal, unnamed capturing parenthesis. */
1213    
1214      else if (ptr[1] != CHAR_QUESTION_MARK)
1215        {
1216        *count += 1;
1217        if (name == NULL && *count == lorn) return *count;
1218        ptr++;
1219        }
1220    
1221      /* All cases now have (? at the start. Remember when we are in a group
1222      where the parenthesis numbers are duplicated. */
1223    
1224      else if (ptr[2] == CHAR_VERTICAL_LINE)
1225      {      {
1226      ptr += 3;      ptr += 3;
1227      dup_parens = TRUE;      dup_parens = TRUE;
1228      }      }
1229    
1230    /* Handle a normal, unnamed capturing parenthesis */    /* Handle comments; all characters are allowed until a ket is reached. */
1231    
1232    else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)    else if (ptr[2] == CHAR_NUMBER_SIGN)
1233      {      {
1234      *count += 1;      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1235      if (name == NULL && *count == lorn) return *count;      goto FAIL_EXIT;
     ptr++;  
1236      }      }
1237    
1238    /* 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
1239    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
1240    condition (there can't be any nested parens. */    condition (there can't be any nested parens). */
1241    
1242    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1243      {      {
# Line 1087  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1249  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1249        }        }
1250      }      }
1251    
1252    /* We have either (? or (* and not a condition */    /* Start with (? but not a condition. */
1253    
1254    else    else
1255      {      {
# Line 1116  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1278  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1278    }    }
1279    
1280  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1281  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1282    first-pass call when this value is temporarily adjusted to stop at the current
1283    position. So DO NOT change this to a test for binary zero. */
1284    
1285  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1286    {    {
1287    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1288    
# Line 1192  for (; *ptr != 0; ptr++) Line 1356  for (; *ptr != 0; ptr++)
1356    
1357    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1358      {      {
1359      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1360        while (*ptr != 0)
1361          {
1362          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1363          ptr++;
1364    #ifdef SUPPORT_UTF8
1365          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1366    #endif
1367          }
1368      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1369      continue;      continue;
1370      }      }
# Line 1201  for (; *ptr != 0; ptr++) Line 1373  for (; *ptr != 0; ptr++)
1373    
1374    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1375      {      {
1376      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1377      if (rc > 0) return rc;      if (rc > 0) return rc;
1378      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1379      }      }
# Line 1209  for (; *ptr != 0; ptr++) Line 1381  for (; *ptr != 0; ptr++)
1381    else if (*ptr == CHAR_RIGHT_PARENTHESIS)    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1382      {      {
1383      if (dup_parens && *count < hwm_count) *count = hwm_count;      if (dup_parens && *count < hwm_count) *count = hwm_count;
1384      *ptrptr = ptr;      goto FAIL_EXIT;
     return -1;  
1385      }      }
1386    
1387    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
# Line 1248  Arguments: Line 1419  Arguments:
1419    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1420    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1421    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1422      utf8         TRUE if we are in UTF-8 mode
1423    
1424  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1425  */  */
1426    
1427  static int  static int
1428  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1429      BOOL utf8)
1430  {  {
1431  uschar *ptr = (uschar *)cd->start_pattern;  uschar *ptr = (uschar *)cd->start_pattern;
1432  int count = 0;  int count = 0;
# Line 1266  matching closing parens. That is why we Line 1439  matching closing parens. That is why we
1439    
1440  for (;;)  for (;;)
1441    {    {
1442    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1443    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1444    }    }
1445    
# Line 1282  return rc; Line 1455  return rc;
1455    
1456  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1457  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
1458  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
1459  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
1460  assertions, and also the \b assertion; for others it does not.  does not.
1461    
1462  Arguments:  Arguments:
1463    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  
1464    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1465    
1466  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1467  */  */
1468    
1469  static const uschar*  static const uschar*
1470  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1471  {  {
1472  for (;;)  for (;;)
1473    {    {
1474    switch ((int)*code)    switch ((int)*code)
1475      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1476      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1477      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1478      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1359  and doing the check at the end; a flag s Line 1522  and doing the check at the end; a flag s
1522    
1523  Arguments:  Arguments:
1524    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1525    options  the compiling options    utf8     TRUE in UTF-8 mode
1526    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1527    cd       the "compile data" structure    cd       the "compile data" structure
1528    
1529  Returns:   the fixed length,  Returns:   the fixed length,
1530               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1531               or -2 if \C was encountered               or -2 if \C was encountered (in UTF-8 mode only)
1532               or -3 if an OP_RECURSE item was encountered and atend is FALSE               or -3 if an OP_RECURSE item was encountered and atend is FALSE
1533                 or -4 if an unknown opcode was encountered (internal error)
1534  */  */
1535    
1536  static int  static int
1537  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1538  {  {
1539  int length = -1;  int length = -1;
1540    
# Line 1387  for (;;) Line 1551  for (;;)
1551    register int op = *cc;    register int op = *cc;
1552    switch (op)    switch (op)
1553      {      {
1554        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1555        OP_BRA (normal non-capturing bracket) because the other variants of these
1556        opcodes are all concerned with unlimited repeated groups, which of course
1557        are not of fixed length. */
1558    
1559      case OP_CBRA:      case OP_CBRA:
1560      case OP_BRA:      case OP_BRA:
1561      case OP_ONCE:      case OP_ONCE:
1562        case OP_ONCE_NC:
1563      case OP_COND:      case OP_COND:
1564      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1565      if (d < 0) return d;      if (d < 0) return d;
1566      branchlength += d;      branchlength += d;
1567      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1568      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1569      break;      break;
1570    
1571      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested call.
1572      call. If it's ALT it is an alternation in a nested call. If it is      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1573      END it's the end of the outer call. All can be handled by the same code. */      an ALT. If it is END it's the end of the outer call. All can be handled by
1574        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1575        because they all imply an unlimited repeat. */
1576    
1577      case OP_ALT:      case OP_ALT:
1578      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1579      case OP_END:      case OP_END:
1580        case OP_ACCEPT:
1581        case OP_ASSERT_ACCEPT:
1582      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1583        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1584      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1423  for (;;) Line 1595  for (;;)
1595      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1596      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1597      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1598      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1599      if (d < 0) return d;      if (d < 0) return d;
1600      branchlength += d;      branchlength += d;
1601      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1440  for (;;) Line 1612  for (;;)
1612    
1613      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1614    
1615      case OP_REVERSE:      case OP_MARK:
1616        case OP_PRUNE_ARG:
1617        case OP_SKIP_ARG:
1618        case OP_THEN_ARG:
1619        cc += cc[1] + _pcre_OP_lengths[*cc];
1620        break;
1621    
1622        case OP_CALLOUT:
1623        case OP_CIRC:
1624        case OP_CIRCM:
1625        case OP_CLOSE:
1626        case OP_COMMIT:
1627      case OP_CREF:      case OP_CREF:
     case OP_NCREF:  
     case OP_RREF:  
     case OP_NRREF:  
1628      case OP_DEF:      case OP_DEF:
1629      case OP_OPT:      case OP_DOLL:
1630      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
     case OP_SET_SOM:  
1631      case OP_EOD:      case OP_EOD:
1632      case OP_EODN:      case OP_EODN:
1633      case OP_CIRC:      case OP_FAIL:
1634      case OP_DOLL:      case OP_NCREF:
1635        case OP_NRREF:
1636      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1637        case OP_PRUNE:
1638        case OP_REVERSE:
1639        case OP_RREF:
1640        case OP_SET_SOM:
1641        case OP_SKIP:
1642        case OP_SOD:
1643        case OP_SOM:
1644        case OP_THEN:
1645      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1646      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
1647      break;      break;
# Line 1463  for (;;) Line 1649  for (;;)
1649      /* Handle literal characters */      /* Handle literal characters */
1650    
1651      case OP_CHAR:      case OP_CHAR:
1652      case OP_CHARNC:      case OP_CHARI:
1653      case OP_NOT:      case OP_NOT:
1654        case OP_NOTI:
1655      branchlength++;      branchlength++;
1656      cc += 2;      cc += 2;
1657  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1658      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];  
1659  #endif  #endif
1660      break;      break;
1661    
# Line 1477  for (;;) Line 1663  for (;;)
1663      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1664    
1665      case OP_EXACT:      case OP_EXACT:
1666        case OP_EXACTI:
1667        case OP_NOTEXACT:
1668        case OP_NOTEXACTI:
1669      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1670      cc += 4;      cc += 4;
1671  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1672      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];  
1673  #endif  #endif
1674      break;      break;
1675    
# Line 1498  for (;;) Line 1686  for (;;)
1686      cc += 2;      cc += 2;
1687      /* Fall through */      /* Fall through */
1688    
1689        case OP_HSPACE:
1690        case OP_VSPACE:
1691        case OP_NOT_HSPACE:
1692        case OP_NOT_VSPACE:
1693      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1694      case OP_DIGIT:      case OP_DIGIT:
1695      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1510  for (;;) Line 1702  for (;;)
1702      cc++;      cc++;
1703      break;      break;
1704    
1705      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1706        otherwise \C is coded as OP_ALLANY. */
1707    
1708      case OP_ANYBYTE:      case OP_ANYBYTE:
1709      return -2;      return -2;
# Line 1529  for (;;) Line 1722  for (;;)
1722    
1723      switch (*cc)      switch (*cc)
1724        {        {
1725          case OP_CRPLUS:
1726          case OP_CRMINPLUS:
1727        case OP_CRSTAR:        case OP_CRSTAR:
1728        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1729        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1549  for (;;) Line 1744  for (;;)
1744    
1745      /* Anything else is variable length */      /* Anything else is variable length */
1746    
1747      default:      case OP_ANYNL:
1748        case OP_BRAMINZERO:
1749        case OP_BRAPOS:
1750        case OP_BRAPOSZERO:
1751        case OP_BRAZERO:
1752        case OP_CBRAPOS:
1753        case OP_EXTUNI:
1754        case OP_KETRMAX:
1755        case OP_KETRMIN:
1756        case OP_KETRPOS:
1757        case OP_MINPLUS:
1758        case OP_MINPLUSI:
1759        case OP_MINQUERY:
1760        case OP_MINQUERYI:
1761        case OP_MINSTAR:
1762        case OP_MINSTARI:
1763        case OP_MINUPTO:
1764        case OP_MINUPTOI:
1765        case OP_NOTMINPLUS:
1766        case OP_NOTMINPLUSI:
1767        case OP_NOTMINQUERY:
1768        case OP_NOTMINQUERYI:
1769        case OP_NOTMINSTAR:
1770        case OP_NOTMINSTARI:
1771        case OP_NOTMINUPTO:
1772        case OP_NOTMINUPTOI:
1773        case OP_NOTPLUS:
1774        case OP_NOTPLUSI:
1775        case OP_NOTPOSPLUS:
1776        case OP_NOTPOSPLUSI:
1777        case OP_NOTPOSQUERY:
1778        case OP_NOTPOSQUERYI:
1779        case OP_NOTPOSSTAR:
1780        case OP_NOTPOSSTARI:
1781        case OP_NOTPOSUPTO:
1782        case OP_NOTPOSUPTOI:
1783        case OP_NOTQUERY:
1784        case OP_NOTQUERYI:
1785        case OP_NOTSTAR:
1786        case OP_NOTSTARI:
1787        case OP_NOTUPTO:
1788        case OP_NOTUPTOI:
1789        case OP_PLUS:
1790        case OP_PLUSI:
1791        case OP_POSPLUS:
1792        case OP_POSPLUSI:
1793        case OP_POSQUERY:
1794        case OP_POSQUERYI:
1795        case OP_POSSTAR:
1796        case OP_POSSTARI:
1797        case OP_POSUPTO:
1798        case OP_POSUPTOI:
1799        case OP_QUERY:
1800        case OP_QUERYI:
1801        case OP_REF:
1802        case OP_REFI:
1803        case OP_SBRA:
1804        case OP_SBRAPOS:
1805        case OP_SCBRA:
1806        case OP_SCBRAPOS:
1807        case OP_SCOND:
1808        case OP_SKIPZERO:
1809        case OP_STAR:
1810        case OP_STARI:
1811        case OP_TYPEMINPLUS:
1812        case OP_TYPEMINQUERY:
1813        case OP_TYPEMINSTAR:
1814        case OP_TYPEMINUPTO:
1815        case OP_TYPEPLUS:
1816        case OP_TYPEPOSPLUS:
1817        case OP_TYPEPOSQUERY:
1818        case OP_TYPEPOSSTAR:
1819        case OP_TYPEPOSUPTO:
1820        case OP_TYPEQUERY:
1821        case OP_TYPESTAR:
1822        case OP_TYPEUPTO:
1823        case OP_UPTO:
1824        case OP_UPTOI:
1825      return -1;      return -1;
1826    
1827        /* Catch unrecognized opcodes so that when new ones are added they
1828        are not forgotten, as has happened in the past. */
1829    
1830        default:
1831        return -4;
1832      }      }
1833    }    }
1834  /* Control never gets here */  /* Control never gets here */
# Line 1583  _pcre_find_bracket(const uschar *code, B Line 1861  _pcre_find_bracket(const uschar *code, B
1861  for (;;)  for (;;)
1862    {    {
1863    register int c = *code;    register int c = *code;
1864    
1865    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1866    
1867    /* 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 1601  for (;;) Line 1880  for (;;)
1880    
1881    /* Handle capturing bracket */    /* Handle capturing bracket */
1882    
1883    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1884               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1885      {      {
1886      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1887      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1610  for (;;) Line 1890  for (;;)
1890    
1891    /* 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
1892    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
1893    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1894      must add in its length. */
1895    
1896    else    else
1897      {      {
# Line 1634  for (;;) Line 1915  for (;;)
1915        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1916        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1917        break;        break;
1918    
1919          case OP_MARK:
1920          case OP_PRUNE_ARG:
1921          case OP_SKIP_ARG:
1922          code += code[1];
1923          break;
1924    
1925          case OP_THEN_ARG:
1926          code += code[1];
1927          break;
1928        }        }
1929    
1930      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1648  for (;;) Line 1939  for (;;)
1939      if (utf8) switch(c)      if (utf8) switch(c)
1940        {        {
1941        case OP_CHAR:        case OP_CHAR:
1942        case OP_CHARNC:        case OP_CHARI:
1943        case OP_EXACT:        case OP_EXACT:
1944          case OP_EXACTI:
1945        case OP_UPTO:        case OP_UPTO:
1946          case OP_UPTOI:
1947        case OP_MINUPTO:        case OP_MINUPTO:
1948          case OP_MINUPTOI:
1949        case OP_POSUPTO:        case OP_POSUPTO:
1950          case OP_POSUPTOI:
1951        case OP_STAR:        case OP_STAR:
1952          case OP_STARI:
1953        case OP_MINSTAR:        case OP_MINSTAR:
1954          case OP_MINSTARI:
1955        case OP_POSSTAR:        case OP_POSSTAR:
1956          case OP_POSSTARI:
1957        case OP_PLUS:        case OP_PLUS:
1958          case OP_PLUSI:
1959        case OP_MINPLUS:        case OP_MINPLUS:
1960          case OP_MINPLUSI:
1961        case OP_POSPLUS:        case OP_POSPLUS:
1962          case OP_POSPLUSI:
1963        case OP_QUERY:        case OP_QUERY:
1964          case OP_QUERYI:
1965        case OP_MINQUERY:        case OP_MINQUERY:
1966          case OP_MINQUERYI:
1967        case OP_POSQUERY:        case OP_POSQUERY:
1968          case OP_POSQUERYI:
1969        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1970        break;        break;
1971        }        }
# Line 1705  for (;;) Line 2009  for (;;)
2009    
2010    /* 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
2011    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
2012    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2013      must add in its length. */
2014    
2015    else    else
2016      {      {
# Line 1729  for (;;) Line 2034  for (;;)
2034        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2035        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2036        break;        break;
2037    
2038          case OP_MARK:
2039          case OP_PRUNE_ARG:
2040          case OP_SKIP_ARG:
2041          code += code[1];
2042          break;
2043    
2044          case OP_THEN_ARG:
2045          code += code[1];
2046          break;
2047        }        }
2048    
2049      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1743  for (;;) Line 2058  for (;;)
2058      if (utf8) switch(c)      if (utf8) switch(c)
2059        {        {
2060        case OP_CHAR:        case OP_CHAR:
2061        case OP_CHARNC:        case OP_CHARI:
2062        case OP_EXACT:        case OP_EXACT:
2063          case OP_EXACTI:
2064        case OP_UPTO:        case OP_UPTO:
2065          case OP_UPTOI:
2066        case OP_MINUPTO:        case OP_MINUPTO:
2067          case OP_MINUPTOI:
2068        case OP_POSUPTO:        case OP_POSUPTO:
2069          case OP_POSUPTOI:
2070        case OP_STAR:        case OP_STAR:
2071          case OP_STARI:
2072        case OP_MINSTAR:        case OP_MINSTAR:
2073          case OP_MINSTARI:
2074        case OP_POSSTAR:        case OP_POSSTAR:
2075          case OP_POSSTARI:
2076        case OP_PLUS:        case OP_PLUS:
2077          case OP_PLUSI:
2078        case OP_MINPLUS:        case OP_MINPLUS:
2079          case OP_MINPLUSI:
2080        case OP_POSPLUS:        case OP_POSPLUS:
2081          case OP_POSPLUSI:
2082        case OP_QUERY:        case OP_QUERY:
2083          case OP_QUERYI:
2084        case OP_MINQUERY:        case OP_MINQUERY:
2085          case OP_MINQUERYI:
2086        case OP_POSQUERY:        case OP_POSQUERY:
2087          case OP_POSQUERYI:
2088        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
2089        break;        break;
2090        }        }
# Line 1795  could_be_empty_branch(const uschar *code Line 2123  could_be_empty_branch(const uschar *code
2123    compile_data *cd)    compile_data *cd)
2124  {  {
2125  register int c;  register int c;
2126  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2127       code < endcode;       code < endcode;
2128       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2129    {    {
2130    const uschar *ccode;    const uschar *ccode;
2131    
2132    c = *code;    c = *code;
2133    
2134    /* Skip over forward assertions; the other assertions are skipped by    /* Skip over forward assertions; the other assertions are skipped by
2135    first_significant_code() with a TRUE final argument. */    first_significant_code() with a TRUE final argument. */
2136    
# Line 1813  for (code = first_significant_code(code Line 2141  for (code = first_significant_code(code
2141      continue;      continue;
2142      }      }
2143    
   /* Groups with zero repeats can of course be empty; skip them. */  
   
   if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)  
     {  
     code += _pcre_OP_lengths[c];  
     do code += GET(code, 1); while (*code == OP_ALT);  
     c = *code;  
     continue;  
     }  
   
2144    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
2145    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
2146      forward reference subroutine call, we can't. To detect forward reference
2147      we have to scan up the list that is kept in the workspace. This function is
2148      called only when doing the real compile, not during the pre-compile that
2149      measures the size of the compiled pattern. */
2150    
2151    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2152      {      {
2153      const uschar *scode = cd->start_code + GET(code, 1);      const uschar *scode;
2154        BOOL empty_branch;
2155    
2156        /* Test for forward reference */
2157    
2158        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2159          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2160    
2161        /* Not a forward reference, test for completed backward reference */
2162    
2163        empty_branch = FALSE;
2164        scode = cd->start_code + GET(code, 1);
2165      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2166    
2167        /* Completed backwards reference */
2168    
2169      do      do
2170        {        {
2171        if (!could_be_empty_branch(scode, endcode, utf8, cd)) return FALSE;        if (could_be_empty_branch(scode, endcode, utf8, cd))
2172            {
2173            empty_branch = TRUE;
2174            break;
2175            }
2176        scode += GET(scode, 1);        scode += GET(scode, 1);
2177        }        }
2178      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2179    
2180        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2181      continue;      continue;
2182      }      }
2183    
2184      /* Groups with zero repeats can of course be empty; skip them. */
2185    
2186      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2187          c == OP_BRAPOSZERO)
2188        {
2189        code += _pcre_OP_lengths[c];
2190        do code += GET(code, 1); while (*code == OP_ALT);
2191        c = *code;
2192        continue;
2193        }
2194    
2195      /* A nested group that is already marked as "could be empty" can just be
2196      skipped. */
2197    
2198      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2199          c == OP_SCBRA || c == OP_SCBRAPOS)
2200        {
2201        do code += GET(code, 1); while (*code == OP_ALT);
2202        c = *code;
2203        continue;
2204        }
2205    
2206    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2207    
2208    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2209          c == OP_CBRA || c == OP_CBRAPOS ||
2210          c == OP_ONCE || c == OP_ONCE_NC ||
2211          c == OP_COND)
2212      {      {
2213      BOOL empty_branch;      BOOL empty_branch;
2214      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1928  for (code = first_significant_code(code Line 2295  for (code = first_significant_code(code
2295      case OP_ALLANY:      case OP_ALLANY:
2296      case OP_ANYBYTE:      case OP_ANYBYTE:
2297      case OP_CHAR:      case OP_CHAR:
2298      case OP_CHARNC:      case OP_CHARI:
2299      case OP_NOT:      case OP_NOT:
2300        case OP_NOTI:
2301      case OP_PLUS:      case OP_PLUS:
2302      case OP_MINPLUS:      case OP_MINPLUS:
2303      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1969  for (code = first_significant_code(code Line 2337  for (code = first_significant_code(code
2337      case OP_KET:      case OP_KET:
2338      case OP_KETRMAX:      case OP_KETRMAX:
2339      case OP_KETRMIN:      case OP_KETRMIN:
2340        case OP_KETRPOS:
2341      case OP_ALT:      case OP_ALT:
2342      return TRUE;      return TRUE;
2343    
# Line 1977  for (code = first_significant_code(code Line 2346  for (code = first_significant_code(code
2346    
2347  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2348      case OP_STAR:      case OP_STAR:
2349        case OP_STARI:
2350      case OP_MINSTAR:      case OP_MINSTAR:
2351        case OP_MINSTARI:
2352      case OP_POSSTAR:      case OP_POSSTAR:
2353        case OP_POSSTARI:
2354      case OP_QUERY:      case OP_QUERY:
2355        case OP_QUERYI:
2356      case OP_MINQUERY:      case OP_MINQUERY:
2357        case OP_MINQUERYI:
2358      case OP_POSQUERY:      case OP_POSQUERY:
2359        case OP_POSQUERYI:
2360      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2361      break;      break;
2362    
2363      case OP_UPTO:      case OP_UPTO:
2364        case OP_UPTOI:
2365      case OP_MINUPTO:      case OP_MINUPTO:
2366        case OP_MINUPTOI:
2367      case OP_POSUPTO:      case OP_POSUPTO:
2368        case OP_POSUPTOI:
2369      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2370      break;      break;
2371  #endif  #endif
2372    
2373        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2374        string. */
2375    
2376        case OP_MARK:
2377        case OP_PRUNE_ARG:
2378        case OP_SKIP_ARG:
2379        code += code[1];
2380        break;
2381    
2382        case OP_THEN_ARG:
2383        code += code[1];
2384        break;
2385    
2386      /* None of the remaining opcodes are required to match a character. */      /* None of the remaining opcodes are required to match a character. */
2387    
2388      default:      default:
2389      break;      break;
2390      }      }
2391    }    }
2392    
# Line 2012  return TRUE; Line 2403  return TRUE;
2403  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
2404  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,
2405  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.
2406    This function is called only during the real compile, not during the
2407    pre-compile.
2408    
2409  Arguments:  Arguments:
2410    code        points to start of the recursion    code        points to start of the recursion
2411    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2412    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2413    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2414    cd          pointers to tables etc    cd          pointers to tables etc
2415    
2416  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2417  */  */
# Line 2062  where Perl recognizes it as the POSIX cl Line 2455  where Perl recognizes it as the POSIX cl
2455  "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,
2456  I think.  I think.
2457    
2458    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2459    It seems that the appearance of a nested POSIX class supersedes an apparent
2460    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2461    a digit.
2462    
2463    In Perl, unescaped square brackets may also appear as part of class names. For
2464    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2465    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2466    seem right at all. PCRE does not allow closing square brackets in POSIX class
2467    names.
2468    
2469  Arguments:  Arguments:
2470    ptr      pointer to the initial [    ptr      pointer to the initial [
2471    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2076  int terminator; /* Don't combin Line 2480  int terminator; /* Don't combin
2480  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2481  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2482    {    {
2483    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2484        ptr++;
2485      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2486      else
2487      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2488      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2489        {        {
2490        *endptr = ptr;        *endptr = ptr;
2491        return TRUE;        return TRUE;
2492        }        }
2493        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2494             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2495              ptr[1] == CHAR_EQUALS_SIGN) &&
2496            check_posix_syntax(ptr, endptr))
2497          return FALSE;
2498      }      }
2499    }    }
2500  return FALSE;  return FALSE;
# Line 2212  auto_callout(uschar *code, const uschar Line 2623  auto_callout(uschar *code, const uschar
2623  {  {
2624  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2625  *code++ = 255;  *code++ = 255;
2626  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2627  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2628  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2629  }  }
2630    
# Line 2238  Returns: nothing Line 2649  Returns: nothing
2649  static void  static void
2650  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2651  {  {
2652  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2653  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2654  }  }
2655    
# Line 2288  for (++c; c <= d; c++) Line 2699  for (++c; c <= d; c++)
2699    
2700  return TRUE;  return TRUE;
2701  }  }
2702    
2703    
2704    
2705    /*************************************************
2706    *        Check a character and a property        *
2707    *************************************************/
2708    
2709    /* This function is called by check_auto_possessive() when a property item
2710    is adjacent to a fixed character.
2711    
2712    Arguments:
2713      c            the character
2714      ptype        the property type
2715      pdata        the data for the type
2716      negated      TRUE if it's a negated property (\P or \p{^)
2717    
2718    Returns:       TRUE if auto-possessifying is OK
2719    */
2720    
2721    static BOOL
2722    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2723    {
2724    const ucd_record *prop = GET_UCD(c);
2725    switch(ptype)
2726      {
2727      case PT_LAMP:
2728      return (prop->chartype == ucp_Lu ||
2729              prop->chartype == ucp_Ll ||
2730              prop->chartype == ucp_Lt) == negated;
2731    
2732      case PT_GC:
2733      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2734    
2735      case PT_PC:
2736      return (pdata == prop->chartype) == negated;
2737    
2738      case PT_SC:
2739      return (pdata == prop->script) == negated;
2740    
2741      /* These are specials */
2742    
2743      case PT_ALNUM:
2744      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2745              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2746    
2747      case PT_SPACE:    /* Perl space */
2748      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2749              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2750              == negated;
2751    
2752      case PT_PXSPACE:  /* POSIX space */
2753      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2754              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2755              c == CHAR_FF || c == CHAR_CR)
2756              == negated;
2757    
2758      case PT_WORD:
2759      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2760              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2761              c == CHAR_UNDERSCORE) == negated;
2762      }
2763    return FALSE;
2764    }
2765  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2766    
2767    
# Line 2301  whether the next thing could possibly ma Line 2775  whether the next thing could possibly ma
2775  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2776    
2777  Arguments:  Arguments:
2778    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2779    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2780    ptr           next character in pattern    ptr           next character in pattern
2781    options       options bits    options       options bits
2782    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2313  Returns: TRUE if possessifying is Line 2785  Returns: TRUE if possessifying is
2785  */  */
2786    
2787  static BOOL  static BOOL
2788  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2789    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2790  {  {
2791  int next;  int c, next;
2792    int op_code = *previous++;
2793    
2794  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2795    
# Line 2327  if ((options & PCRE_EXTENDED) != 0) Line 2800  if ((options & PCRE_EXTENDED) != 0)
2800      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2801      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2802        {        {
2803        while (*(++ptr) != 0)        ptr++;
2804          while (*ptr != 0)
2805            {
2806          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2807            ptr++;
2808    #ifdef SUPPORT_UTF8
2809            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2810    #endif
2811            }
2812        }        }
2813      else break;      else break;
2814      }      }
# Line 2364  if ((options & PCRE_EXTENDED) != 0) Line 2844  if ((options & PCRE_EXTENDED) != 0)
2844      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2845      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2846        {        {
2847        while (*(++ptr) != 0)        ptr++;
2848          while (*ptr != 0)
2849            {
2850          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2851            ptr++;
2852    #ifdef SUPPORT_UTF8
2853            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2854    #endif
2855            }
2856        }        }
2857      else break;      else break;
2858      }      }
# Line 2377  if (*ptr == CHAR_ASTERISK || *ptr == CHA Line 2864  if (*ptr == CHAR_ASTERISK || *ptr == CHA
2864    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)
2865      return FALSE;      return FALSE;
2866    
2867  /* 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
2868  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. */  
2869    
2870  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2871    {    {
2872    case OP_CHAR:    case OP_CHAR:
2873  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2874    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2875  #else  #else
2876    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2877  #endif  #endif
2878    return item != next;    return c != next;
2879    
2880    /* 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
2881    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
2882    high-valued characters. */    high-valued characters. */
2883    
2884    case OP_CHARNC:    case OP_CHARI:
2885  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2886    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2887    #else
2888      c = *previous;
2889  #endif  #endif
2890    if (item == next) return FALSE;    if (c == next) return FALSE;
2891  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2892    if (utf8)    if (utf8)
2893      {      {
# Line 2414  if (next >= 0) switch(op_code) Line 2898  if (next >= 0) switch(op_code)
2898  #else  #else
2899      othercase = NOTACHAR;      othercase = NOTACHAR;
2900  #endif  #endif
2901      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2902      }      }
2903    else    else
2904  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2905    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2906    
2907    /* 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
2908      opcodes are not used for multi-byte characters, because they are coded using
2909      an XCLASS instead. */
2910    
2911    case OP_NOT:    case OP_NOT:
2912    if (item == next) return TRUE;    return (c = *previous) == next;
2913    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2914      case OP_NOTI:
2915      if ((c = *previous) == next) return TRUE;
2916  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2917    if (utf8)    if (utf8)
2918      {      {
# Line 2435  if (next >= 0) switch(op_code) Line 2923  if (next >= 0) switch(op_code)
2923  #else  #else
2924      othercase = NOTACHAR;      othercase = NOTACHAR;
2925  #endif  #endif
2926      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2927      }      }
2928    else    else
2929  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2930    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2931    
2932      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2933      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2934    
2935    case OP_DIGIT:    case OP_DIGIT:
2936    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2482  if (next >= 0) switch(op_code) Line 2973  if (next >= 0) switch(op_code)
2973      case 0x202f:      case 0x202f:
2974      case 0x205f:      case 0x205f:
2975      case 0x3000:      case 0x3000:
2976      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2977      default:      default:
2978      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2979      }      }
2980    
2981      case OP_ANYNL:
2982    case OP_VSPACE:    case OP_VSPACE:
2983    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2984    switch(next)    switch(next)
# Line 2498  if (next >= 0) switch(op_code) Line 2990  if (next >= 0) switch(op_code)
2990      case 0x85:      case 0x85:
2991      case 0x2028:      case 0x2028:
2992      case 0x2029:      case 0x2029:
2993      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2994      default:      default:
2995      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2996      }      }
2997    
2998    #ifdef SUPPORT_UCP
2999      case OP_PROP:
3000      return check_char_prop(next, previous[0], previous[1], FALSE);
3001    
3002      case OP_NOTPROP:
3003      return check_char_prop(next, previous[0], previous[1], TRUE);
3004    #endif
3005    
3006    default:    default:
3007    return FALSE;    return FALSE;
3008    }    }
3009    
3010    
3011  /* 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
3012    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3013    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3014    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3015    replaced by OP_PROP codes when PCRE_UCP is set. */
3016    
3017  switch(op_code)  switch(op_code)
3018    {    {
3019    case OP_CHAR:    case OP_CHAR:
3020    case OP_CHARNC:    case OP_CHARI:
3021  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3022    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3023    #else
3024      c = *previous;
3025  #endif  #endif
3026    switch(-next)    switch(-next)
3027      {      {
3028      case ESC_d:      case ESC_d:
3029      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3030    
3031      case ESC_D:      case ESC_D:
3032      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3033    
3034      case ESC_s:      case ESC_s:
3035      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3036    
3037      case ESC_S:      case ESC_S:
3038      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3039    
3040      case ESC_w:      case ESC_w:
3041      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3042    
3043      case ESC_W:      case ESC_W:
3044      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3045    
3046      case ESC_h:      case ESC_h:
3047      case ESC_H:      case ESC_H:
3048      switch(item)      switch(c)
3049        {        {
3050        case 0x09:        case 0x09:
3051        case 0x20:        case 0x20:
# Line 2567  switch(op_code) Line 3073  switch(op_code)
3073    
3074      case ESC_v:      case ESC_v:
3075      case ESC_V:      case ESC_V:
3076      switch(item)      switch(c)
3077        {        {
3078        case 0x0a:        case 0x0a:
3079        case 0x0b:        case 0x0b:
# Line 2581  switch(op_code) Line 3087  switch(op_code)
3087        return -next == ESC_v;        return -next == ESC_v;
3088        }        }
3089    
3090        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3091        their substitutions and process them. The result will always be either
3092        -ESC_p or -ESC_P. Then fall through to process those values. */
3093    
3094    #ifdef SUPPORT_UCP
3095        case ESC_du:
3096        case ESC_DU:
3097        case ESC_wu:
3098        case ESC_WU:
3099        case ESC_su:
3100        case ESC_SU:
3101          {
3102          int temperrorcode = 0;
3103          ptr = substitutes[-next - ESC_DU];
3104          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3105          if (temperrorcode != 0) return FALSE;
3106          ptr++;    /* For compatibility */
3107          }
3108        /* Fall through */
3109    
3110        case ESC_p:
3111        case ESC_P:
3112          {
3113          int ptype, pdata, errorcodeptr;
3114          BOOL negated;
3115    
3116          ptr--;      /* Make ptr point at the p or P */
3117          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3118          if (ptype < 0) return FALSE;
3119          ptr++;      /* Point past the final curly ket */
3120    
3121          /* If the property item is optional, we have to give up. (When generated
3122          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3123          to the original \d etc. At this point, ptr will point to a zero byte. */
3124    
3125          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3126            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3127              return FALSE;
3128    
3129          /* Do the property check. */
3130    
3131          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3132          }
3133    #endif
3134    
3135      default:      default:
3136      return FALSE;      return FALSE;
3137      }      }
3138    
3139      /* In principle, support for Unicode properties should be integrated here as
3140      well. It means re-organizing the above code so as to get hold of the property
3141      values before switching on the op-code. However, I wonder how many patterns
3142      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3143      these op-codes are never generated.) */
3144    
3145    case OP_DIGIT:    case OP_DIGIT:
3146    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3147           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3148    
3149    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3150    return next == -ESC_d;    return next == -ESC_d;
3151    
3152    case OP_WHITESPACE:    case OP_WHITESPACE:
3153    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3154    
3155    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3156    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3157    
3158    case OP_HSPACE:    case OP_HSPACE:
3159    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3160             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3161    
3162    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3163    return next == -ESC_h;    return next == -ESC_h;
3164    
3165    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3166      case OP_ANYNL:
3167    case OP_VSPACE:    case OP_VSPACE:
3168    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3169    
3170    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3171    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3172    
3173    case OP_WORDCHAR:    case OP_WORDCHAR:
3174    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3175             next == -ESC_v || next == -ESC_R;
3176    
3177    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3178    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2644  Arguments: Line 3204  Arguments:
3204    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3205    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3206    bcptr          points to current branch chain    bcptr          points to current branch chain
3207      cond_depth     conditional nesting depth
3208    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3209    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3210                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2655  Returns: TRUE on success Line 3216  Returns: TRUE on success
3216  static BOOL  static BOOL
3217  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3218    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3219    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3220  {  {
3221  int repeat_type, op_type;  int repeat_type, op_type;
3222  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2664  int greedy_default, greedy_non_default; Line 3225  int greedy_default, greedy_non_default;
3225  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3226  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3227  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3228  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3229  int after_manual_callout = 0;  int after_manual_callout = 0;
3230  int length_prevgroup = 0;  int length_prevgroup = 0;
3231  register int c;  register int c;
# Line 2676  BOOL inescq = FALSE; Line 3237  BOOL inescq = FALSE;
3237  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3238  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3239  const uschar *tempptr;  const uschar *tempptr;
3240    const uschar *nestptr = NULL;
3241  uschar *previous = NULL;  uschar *previous = NULL;
3242  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3243  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3244  uschar classbits[32];  uschar classbits[32];
3245    
3246    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3247    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3248    dynamically as we process the pattern. */
3249    
3250  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3251  BOOL class_utf8;  BOOL class_utf8;
3252  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2689  uschar *class_utf8data_base; Line 3255  uschar *class_utf8data_base;
3255  uschar utf8_char[6];  uschar utf8_char[6];
3256  #else  #else
3257  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3258  #endif  #endif
3259    
3260  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 2740  for (;; ptr++) Line 3305  for (;; ptr++)
3305    int subfirstbyte;    int subfirstbyte;
3306    int terminator;    int terminator;
3307    int mclength;    int mclength;
3308      int tempbracount;
3309    uschar mcbuffer[8];    uschar mcbuffer[8];
3310    
3311    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3312    
3313    c = *ptr;    c = *ptr;
3314    
3315      /* If we are at the end of a nested substitution, revert to the outer level
3316      string. Nesting only happens one level deep. */
3317    
3318      if (c == 0 && nestptr != NULL)
3319        {
3320        ptr = nestptr;
3321        nestptr = NULL;
3322        c = *ptr;
3323        }
3324    
3325    /* 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
3326    previous cycle of this loop. */    previous cycle of this loop. */
3327    
# Line 2754  for (;; ptr++) Line 3330  for (;; ptr++)
3330  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
3331      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3332  #endif  #endif
3333      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3334        {        {
3335        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3336        goto FAILED;        goto FAILED;
# Line 2776  for (;; ptr++) Line 3352  for (;; ptr++)
3352        goto FAILED;        goto FAILED;
3353        }        }
3354    
3355      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3356      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, (int)(code - last_code),
3357          c));
3358    
3359      /* 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
3360      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
# Line 2803  for (;; ptr++) Line 3380  for (;; ptr++)
3380    /* 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
3381    reference list. */    reference list. */
3382    
3383    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3384      {      {
3385      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3386      goto FAILED;      goto FAILED;
# Line 2851  for (;; ptr++) Line 3428  for (;; ptr++)
3428      previous_callout = NULL;      previous_callout = NULL;
3429      }      }
3430    
3431    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3432    
3433    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3434      {      {
3435      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3436      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3437        {        {
3438        while (*(++ptr) != 0)        ptr++;
3439          while (*ptr != 0)
3440          {          {
3441          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3442            ptr++;
3443    #ifdef SUPPORT_UTF8
3444            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3445    #endif
3446          }          }
3447        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3448    
# Line 2894  for (;; ptr++) Line 3476  for (;; ptr++)
3476          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3477          goto FAILED;          goto FAILED;
3478          }          }
3479        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3480        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3481        }        }
3482      return TRUE;      return TRUE;
# Line 2905  for (;; ptr++) Line 3487  for (;; ptr++)
3487      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3488    
3489      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3490        previous = NULL;
3491      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3492        {        {
3493        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3494          *code++ = OP_CIRCM;
3495        }        }
3496      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3497      break;      break;
3498    
3499      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3500      previous = NULL;      previous = NULL;
3501      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3502      break;      break;
3503    
3504      /* 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 3099  for (;; ptr++) Line 3682  for (;; ptr++)
3682            ptr++;            ptr++;
3683            }            }
3684    
3685          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3686          if (posix_class < 0)          if (posix_class < 0)
3687            {            {
3688            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 3113  for (;; ptr++) Line 3696  for (;; ptr++)
3696          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3697            posix_class = 0;            posix_class = 0;
3698    
3699          /* 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
3700          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3701          subtract bits that may be in the main map already. At the end we or the  
3702          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3703            if ((options & PCRE_UCP) != 0)
3704              {
3705              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3706              if (posix_substitutes[pc] != NULL)
3707                {
3708                nestptr = tempptr + 1;
3709                ptr = posix_substitutes[pc] - 1;
3710                continue;
3711                }
3712              }
3713    #endif
3714            /* In the non-UCP case, we build the bit map for the POSIX class in a
3715            chunk of local store because we may be adding and subtracting from it,
3716            and we don't want to subtract bits that may be in the main map already.
3717            At the end we or the result into the bit map that is being built. */
3718    
3719          posix_class *= 3;          posix_class *= 3;
3720    
# Line 3160  for (;; ptr++) Line 3758  for (;; ptr++)
3758    
3759        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3760        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
3761        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
3762        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
3763        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
3764        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3765          PCRE_EXTRA is set. */
3766    
3767        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3768          {          {
3769          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3770          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3771    
3772          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 */  
3773          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3774            {            {
3775            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3189  for (;; ptr++) Line 3786  for (;; ptr++)
3786            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3787            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3788    
3789            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3790              {              {
3791    #ifdef SUPPORT_UCP
3792                case ESC_du:     /* These are the values given for \d etc */
3793                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3794                case ESC_wu:     /* escape sequence with an appropriate \p */
3795                case ESC_WU:     /* or \P to test Unicode properties instead */
3796                case ESC_su:     /* of the default ASCII testing. */
3797                case ESC_SU:
3798                nestptr = ptr;
3799                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3800                class_charcount -= 2;                /* Undo! */
3801                continue;
3802    #endif
3803              case ESC_d:              case ESC_d:
3804              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3805              continue;              continue;
# Line 3211  for (;; ptr++) Line 3818  for (;; ptr++)
3818              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3819              continue;              continue;
3820    
3821                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3822                if it was previously set by something earlier in the character
3823                class. */
3824    
3825              case ESC_s:              case ESC_s:
3826              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3827              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3828                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3829              continue;              continue;
3830    
3831              case ESC_S:              case ESC_S:
# Line 3222  for (;; ptr++) Line 3834  for (;; ptr++)
3834              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3835              continue;              continue;
3836    
3837              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)  
             {  
3838              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3839              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3840              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 3259  for (;; ptr++) Line 3858  for (;; ptr++)
3858                }                }
3859  #endif  #endif
3860              continue;              continue;
             }  
3861    
3862            if (-c == ESC_H)              case ESC_H:
             {  
3863              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3864                {                {
3865                int x = 0xff;                int x = 0xff;
# Line 3304  for (;; ptr++) Line 3901  for (;; ptr++)
3901                }                }
3902  #endif  #endif
3903              continue;              continue;
             }  
3904    
3905            if (-c == ESC_v)              case ESC_v:
             {  
3906              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3907              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3908              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3323  for (;; ptr++) Line 3918  for (;; ptr++)
3918                }                }
3919  #endif  #endif
3920              continue;              continue;
             }  
3921    
3922            if (-c == ESC_V)              case ESC_V:
             {  
3923              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3924                {                {
3925                int x = 0xff;                int x = 0xff;
# Line 3356  for (;; ptr++) Line 3949  for (;; ptr++)
3949                }                }
3950  #endif  #endif
3951              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3952    
3953  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3954            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3955              {              case ESC_P:
3956              BOOL negated;                {
3957              int pdata;                BOOL negated;
3958              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3959              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3960              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3961              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3962                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3963              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3964              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3965              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3966              continue;                class_charcount -= 2;   /* Not a < 256 character */
3967              }                continue;
3968                  }
3969  #endif  #endif
3970            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3971            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3972            treated as literals. */              treated as literals. */
3973    
3974            if ((options & PCRE_EXTRA) != 0)              default:
3975              {              if ((options & PCRE_EXTRA) != 0)
3976              *errorcodeptr = ERR7;                {
3977              goto FAILED;                *errorcodeptr = ERR7;
3978                  goto FAILED;
3979                  }
3980                class_charcount -= 2;  /* Undo the default count from above */
3981                c = *ptr;              /* Get the final character and fall through */
3982                break;
3983              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3984            }            }
3985    
3986          /* 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 3457  for (;; ptr++) Line 4050  for (;; ptr++)
4050            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
4051            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
4052    
4053            /* \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 */  
4054    
4055            if (d < 0)            if (d < 0)
4056              {              {
4057              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  
4058                {                {
4059                ptr = oldptr;                ptr = oldptr;
4060                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3630  for (;; ptr++) Line 4220  for (;; ptr++)
4220          }          }
4221        }        }
4222    
4223      /* 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.
4224        If we are at the end of an internal nested string, revert to the outer
4225        string. */
4226    
4227        while (((c = *(++ptr)) != 0 ||
4228               (nestptr != NULL &&
4229                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4230               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4231    
4232      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
4233    
4234      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4235        {        {
4236        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4237        goto FAILED;        goto FAILED;
4238        }        }
4239    
   
 /* 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  
   
   
4240      /* 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
4241      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
4242      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 3666  we set the flag only if there is a liter Line 4244  we set the flag only if there is a liter
4244    
4245      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
4246      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4247      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4248      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4249    
4250      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
4251      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.
4252      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
4253      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
4254      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
4255      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4256    
4257  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4258      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3685  we set the flag only if there is a liter Line 4263  we set the flag only if there is a liter
4263        {        {
4264        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4265    
4266        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4267    
4268        if (negate_class)        if (negate_class)
4269          {          {
4270          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4271          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4272          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4273          *code++ = class_lastchar;          *code++ = class_lastchar;
4274          break;          break;
4275          }          }
# Line 3722  we set the flag only if there is a liter Line 4300  we set the flag only if there is a liter
4300    
4301      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4302      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4303      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
4304      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
4305      (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
4306      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
4307        actual compiled code. */
4308    
4309  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4310      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4311        {        {
4312        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4313        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3754  we set the flag only if there is a liter Line 4333  we set the flag only if there is a liter
4333        }        }
4334  #endif  #endif
4335    
4336      /* 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
4337      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
4338      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
4339      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4340        negating it if necessary. */
4341    
4342      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4343      if (negate_class)      if (negate_class)
# Line 3817  we set the flag only if there is a liter Line 4397  we set the flag only if there is a liter
4397      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4398      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4399    
4400      /* 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
4401      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4402    
4403      tempcode = previous;      tempcode = previous;
4404    
# Line 3841  we set the flag only if there is a liter Line 4421  we set the flag only if there is a liter
4421        }        }
4422      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4423    
4424        /* If previous was a recursion call, wrap it in atomic brackets so that
4425        previous becomes the atomic group. All recursions were so wrapped in the
4426        past, but it no longer happens for non-repeated recursions. In fact, the
4427        repeated ones could be re-implemented independently so as not to need this,
4428        but for the moment we rely on the code for repeating groups. */
4429    
4430        if (*previous == OP_RECURSE)
4431          {
4432          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4433          *previous = OP_ONCE;
4434          PUT(previous, 1, 2 + 2*LINK_SIZE);
4435          previous[2 + 2*LINK_SIZE] = OP_KET;
4436          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4437          code += 2 + 2 * LINK_SIZE;
4438          length_prevgroup = 3 + 3*LINK_SIZE;
4439    
4440          /* When actually compiling, we need to check whether this was a forward
4441          reference, and if so, adjust the offset. */
4442    
4443          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4444            {
4445            int offset = GET(cd->hwm, -LINK_SIZE);
4446            if (offset == previous + 1 - cd->start_code)
4447              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4448            }
4449          }
4450    
4451        /* Now handle repetition for the different types of item. */
4452    
4453      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4454      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
4455      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
4456      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
4457      instead.  */      instead.  */
4458    
4459      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4460        {        {
4461          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4462    
4463        /* 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
4464        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
4465        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 3881  we set the flag only if there is a liter Line 4492  we set the flag only if there is a liter
4492    
4493        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4494            repeat_max < 0 &&            repeat_max < 0 &&
4495            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4496          {          {
4497          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4498          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3894  we set the flag only if there is a liter Line 4504  we set the flag only if there is a liter
4504      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4505      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-
4506      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4507      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
4508      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4509    
4510      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4511        {        {
4512        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4513        c = previous[1];        c = previous[1];
4514        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4515            repeat_max < 0 &&            repeat_max < 0 &&
4516            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4517          {          {
4518          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4519          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3927  we set the flag only if there is a liter Line 4537  we set the flag only if there is a liter
4537    
4538        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4539            repeat_max < 0 &&            repeat_max < 0 &&
4540            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4541          {          {
4542          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4543          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4096  we set the flag only if there is a liter Line 4706  we set the flag only if there is a liter
4706  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4707               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4708  #endif  #endif
4709               *previous == OP_REF)               *previous == OP_REF ||
4710                 *previous == OP_REFI)
4711        {        {
4712        if (repeat_max == 0)        if (repeat_max == 0)
4713          {          {
# Line 4130  we set the flag only if there is a liter Line 4741  we set the flag only if there is a liter
4741        }        }
4742    
4743      /* 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
4744      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4745        opcodes such as BRA and CBRA, as this is the place where they get converted
4746        into the more special varieties such as BRAPOS and SBRA. A test for >=
4747        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4748        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4749        repetition of assertions, but now it does, for Perl compatibility. */
4750    
4751      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4752        {        {
4753        register int i;        register int i;
4754        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4755        uschar *bralink = NULL;        uschar *bralink = NULL;
4756          uschar *brazeroptr = NULL;
4757    
4758        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4759          we just ignore the repeat. */
4760    
4761        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4762          {          goto END_REPEAT;
         *errorcodeptr = ERR55;  
         goto FAILED;  
         }  
4763    
4764        /* If the maximum repeat count is unlimited, find the end of the bracket        /* There is no sense in actually repeating assertions. The only potential
4765        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,
4766        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
4767        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. */
4768        pointer. */  
4769          if (*previous < OP_ONCE)    /* Assertion */
4770        if (repeat_max == -1)          {
4771          {          if (repeat_min > 0) goto END_REPEAT;
4772          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;  
4773          }          }
4774    
4775        /* 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 4179  we set the flag only if there is a liter Line 4790  we set the flag only if there is a liter
4790          **   goto END_REPEAT;          **   goto END_REPEAT;
4791          **   }          **   }
4792    
4793          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
4794          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
4795          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
4796          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4797            selectively.
4798    
4799          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
4800          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 4202  we set the flag only if there is a liter Line 4814  we set the flag only if there is a liter
4814              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4815              goto END_REPEAT;              goto END_REPEAT;
4816              }              }
4817              brazeroptr = previous;    /* Save for possessive optimizing */
4818            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4819            }            }
4820    
# Line 4226  we set the flag only if there is a liter Line 4839  we set the flag only if there is a liter
4839            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4840            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4841    
4842            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4843            bralink = previous;            bralink = previous;
4844            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4845            }            }
# Line 4335  we set the flag only if there is a liter Line 4948  we set the flag only if there is a liter
4948              {              {
4949              int offset;              int offset;
4950              *code++ = OP_BRA;              *code++ = OP_BRA;
4951              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4952              bralink = code;              bralink = code;
4953              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4954              }              }
# Line 4356  we set the flag only if there is a liter Line 4969  we set the flag only if there is a liter
4969          while (bralink != NULL)          while (bralink != NULL)
4970            {            {
4971            int oldlinkoffset;            int oldlinkoffset;
4972            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4973            uschar *bra = code - offset;            uschar *bra = code - offset;
4974            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4975            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4366  we set the flag only if there is a liter Line 4979  we set the flag only if there is a liter
4979            }            }
4980          }          }
4981    
4982        /* 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
4983        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4984        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
4985        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4986          deal with possessive ONCEs specially.
4987        Then, when we are doing the actual compile phase, check to see whether  
4988        this group is a non-atomic one that could match an empty string. If so,        Otherwise, when we are doing the actual compile phase, check to see
4989          whether this group is one that could match an empty string. If so,
4990        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
4991        that runtime checking can be done. [This check is also applied to        that runtime checking can be done. [This check is also applied to ONCE
4992        atomic groups at runtime, but in a different way.] */        groups at runtime, but in a different way.]
4993    
4994          Then, if the quantifier was possessive and the bracket is not a
4995          conditional, we convert the BRA code to the POS form, and the KET code to
4996          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
4997          subpattern at both the start and at the end.) The use of special opcodes
4998          makes it possible to reduce greatly the stack usage in pcre_exec(). If
4999          the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
5000    
5001          Then, if the minimum number of matches is 1 or 0, cancel the possessive
5002          flag so that the default action below, of wrapping everything inside
5003          atomic brackets, does not happen. When the minimum is greater than 1,
5004          there will be earlier copies of the group, and so we still have to wrap
5005          the whole thing. */
5006    
5007        else        else
5008          {          {
5009          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
5010          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
5011          *ketcode = OP_KETRMAX + repeat_type;  
5012          if (lengthptr == NULL && *bracode != OP_ONCE)          /* Convert possessive ONCE brackets to non-capturing */
5013    
5014            if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5015                possessive_quantifier) *bracode = OP_BRA;
5016    
5017            /* For non-possessive ONCE brackets, all we need to do is to
5018            set the KET. */
5019    
5020            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5021              *ketcode = OP_KETRMAX + repeat_type;
5022    
5023            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5024            converted to non-capturing above). */
5025    
5026            else
5027            {            {
5028            uschar *scode = bracode;            /* In the compile phase, check for empty string matching. */
5029            do  
5030              if (lengthptr == NULL)
5031              {              {
5032              if (could_be_empty_branch(scode, ketcode, utf8, cd))              uschar *scode = bracode;
5033                do
5034                {                {
5035                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
5036                break;                  {
5037                    *bracode += OP_SBRA - OP_BRA;
5038                    break;
5039                    }
5040                  scode += GET(scode, 1);
5041                }                }
5042              scode += GET(scode, 1);              while (*scode == OP_ALT);
5043              }              }
5044            while (*scode == OP_ALT);  
5045              /* Handle possessive quantifiers. */
5046    
5047              if (possessive_quantifier)
5048                {
5049                /* For COND brackets, we wrap the whole thing in a possessively
5050                repeated non-capturing bracket, because we have not invented POS
5051                versions of the COND opcodes. Because we are moving code along, we
5052                must ensure that any pending recursive references are updated. */
5053    
5054                if (*bracode == OP_COND || *bracode == OP_SCOND)
5055                  {
5056                  int nlen = (int)(code - bracode);
5057                  *code = OP_END;
5058                  adjust_recurse(bracode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5059                  memmove(bracode + 1+LINK_SIZE, bracode, nlen);
5060                  code += 1 + LINK_SIZE;
5061                  nlen += 1 + LINK_SIZE;
5062                  *bracode = OP_BRAPOS;
5063                  *code++ = OP_KETRPOS;
5064                  PUTINC(code, 0, nlen);
5065                  PUT(bracode, 1, nlen);
5066                  }
5067    
5068                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5069    
5070                else
5071                  {
5072                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5073                  *ketcode = OP_KETRPOS;
5074                  }
5075    
5076                /* If the minimum is zero, mark it as possessive, then unset the
5077                possessive flag when the minimum is 0 or 1. */
5078    
5079                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5080                if (repeat_min < 2) possessive_quantifier = FALSE;
5081                }
5082    
5083              /* Non-possessive quantifier */
5084    
5085              else *ketcode = OP_KETRMAX + repeat_type;
5086            }            }
5087          }          }
5088        }        }
# Line 4415  we set the flag only if there is a liter Line 5103  we set the flag only if there is a liter
5103        }        }
5104    
5105      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
5106      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
5107      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
5108      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
5109      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
5110      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
5111      tempcode, not at previous, which might be the first part of a string whose  
5112      (former) last char we repeated.      Some (but not all) possessively repeated subpatterns have already been
5113        completely handled in the code just above. For them, possessive_quantifier
5114        is always FALSE at this stage.
5115    
5116        Note that the repeated item starts at tempcode, not at previous, which
5117        might be the first part of a string whose (former) last char we repeated.
5118    
5119      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
5120      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 4444  we set the flag only if there is a liter Line 5137  we set the flag only if there is a liter
5137  #endif  #endif
5138          }          }
5139    
5140        len = code - tempcode;        len = (int)(code - tempcode);
5141        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
5142          {          {
5143          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4452  we set the flag only if there is a liter Line 5145  we set the flag only if there is a liter
5145          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
5146          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
5147    
5148          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
5149          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
5150          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
5151          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
5152    
5153          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
5154          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
5155          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
5156          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
5157    
5158            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
5159            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
5160            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
5161            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
5162    
5163            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
5164            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
5165            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
5166            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
5167    
5168          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
5169          pending recursive references are updated. */          pending recursive references are updated. */
5170    
5171          default:          default:
5172          *code = OP_END;          *code = OP_END;
5173          adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);          adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
# Line 4503  we set the flag only if there is a liter Line 5206  we set the flag only if there is a liter
5206    
5207      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
5208    
5209      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
5210             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
5211        {        {
5212        int i, namelen;        int i, namelen;
5213          int arglen = 0;
5214        const char *vn = verbnames;        const char *vn = verbnames;
5215        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
5216          const uschar *arg = NULL;
5217        previous = NULL;        previous = NULL;
5218        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5219          namelen = (int)(ptr - name);
5220    
5221          /* It appears that Perl allows any characters whatsoever, other than
5222          a closing parenthesis, to appear in arguments, so we no longer insist on
5223          letters, digits, and underscores. */
5224    
5225        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5226          {          {
5227          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
5228          goto FAILED;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5229            arglen = (int)(ptr - arg);
5230          }          }
5231    
5232        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
5233          {          {
5234          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
5235          goto FAILED;          goto FAILED;
5236          }          }
5237        namelen = ptr - name;  
5238          /* Scan the table of verb names */
5239    
5240        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
5241          {          {
5242          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5243              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5244            {            {
5245            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5246              ASSERT_ACCEPT if in an assertion. */
5247    
5248            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5249              {              {
5250              open_capitem *oc;              open_capitem *oc;
5251                if (arglen != 0)
5252                  {
5253                  *errorcodeptr = ERR59;
5254                  goto FAILED;
5255                  }
5256              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5257              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5258                {                {
5259                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5260                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5261                }                }
5262                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5263    
5264                /* Do not set firstbyte after *ACCEPT */
5265                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5266              }              }
5267            *code++ = verbs[i].op;  
5268            break;            /* Handle other cases with/without an argument */
5269    
5270              else if (arglen == 0)
5271                {
5272                if (verbs[i].op < 0)   /* Argument is mandatory */
5273                  {
5274                  *errorcodeptr = ERR66;
5275                  goto FAILED;
5276                  }
5277                *code = verbs[i].op;
5278                if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
5279                }
5280    
5281              else
5282                {
5283                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
5284                  {
5285                  *errorcodeptr = ERR59;
5286                  goto FAILED;
5287                  }
5288                *code = verbs[i].op_arg;
5289                if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
5290                *code++ = arglen;
5291                memcpy(code, arg, arglen);
5292                code += arglen;
5293                *code++ = 0;
5294                }
5295    
5296              break;  /* Found verb, exit loop */
5297            }            }
5298    
5299          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
5300          }          }
5301        if (i < verbcount) continue;  
5302        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
5303          *errorcodeptr = ERR60;          /* Verb not recognized */
5304        goto FAILED;        goto FAILED;
5305        }        }
5306    
# Line 4663  we set the flag only if there is a liter Line 5419  we set the flag only if there is a liter
5419                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5420            ptr++;            ptr++;
5421            }            }
5422          namelen = ptr - name;          namelen = (int)(ptr - name);
5423    
5424          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
5425              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
# Line 4724  we set the flag only if there is a liter Line 5480  we set the flag only if there is a liter
5480          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5481    
5482          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5483                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5484            {            {
5485            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5486            code[1+LINK_SIZE]++;            code[1+LINK_SIZE]++;
# Line 4792  we set the flag only if there is a liter Line 5548  we set the flag only if there is a liter
5548          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5549          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5550          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5551            cd->assert_depth += 1;
5552          ptr++;          ptr++;
5553          break;          break;
5554    
# Line 4806  we set the flag only if there is a liter Line 5563  we set the flag only if there is a liter
5563            continue;            continue;
5564            }            }
5565          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5566            cd->assert_depth += 1;
5567          break;          break;
5568    
5569    
# Line 4815  we set the flag only if there is a liter Line 5573  we set the flag only if there is a liter
5573            {            {
5574            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5575            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5576              cd->assert_depth += 1;
5577            ptr += 2;            ptr += 2;
5578            break;            break;
5579    
5580            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5581            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5582              cd->assert_depth += 1;
5583            ptr += 2;            ptr += 2;
5584            break;            break;
5585    
# Line 4841  we set the flag only if there is a liter Line 5601  we set the flag only if there is a liter
5601    
5602          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5603          case CHAR_C:                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
5604          previous_callout = code;  /* Save for later completion */          previous_callout = code;     /* Save for later completion */
5605          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1;    /* Skip one item before completing */
5606          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
5607            {            {
5608            int n = 0;            int n = 0;
# Line 4859  we set the flag only if there is a liter Line 5619  we set the flag only if there is a liter
5619              goto FAILED;              goto FAILED;
5620              }              }
5621            *code++ = n;            *code++ = n;
5622            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5623            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5624            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5625            }            }
5626          previous = NULL;          previous = NULL;
# Line 4893  we set the flag only if there is a liter Line 5653  we set the flag only if there is a liter
5653            name = ++ptr;            name = ++ptr;
5654    
5655            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5656            namelen = ptr - name;            namelen = (int)(ptr - name);
5657    
5658            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5659    
# Line 5023  we set the flag only if there is a liter Line 5783  we set the flag only if there is a liter
5783          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5784          name = ++ptr;          name = ++ptr;
5785          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5786          namelen = ptr - name;          namelen = (int)(ptr - name);
5787    
5788          /* 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
5789          reference number. */          a dummy reference number, because it was not used in the first pass.
5790            However, with the change of recursive back references to be atomic,
5791            we have to look for the number so that this state can be identified, as
5792            otherwise the incorrect length is computed. If it's not a backwards
5793            reference, the dummy number will do. */
5794    
5795          if (lengthptr != NULL)          if (lengthptr != NULL)
5796            {            {
5797              const uschar *temp;
5798    
5799            if (namelen == 0)            if (namelen == 0)
5800              {              {
5801              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5045  we set the flag only if there is a liter Line 5811  we set the flag only if there is a liter
5811              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5812              goto FAILED;              goto FAILED;
5813              }              }
5814            recno = 0;  
5815              /* The name table does not exist in the first pass, so we cannot
5816              do a simple search as in the code below. Instead, we have to scan the
5817              pattern to find the number. It is important that we scan it only as
5818              far as we have got because the syntax of named subpatterns has not
5819              been checked for the rest of the pattern, and find_parens() assumes
5820              correct syntax. In any case, it's a waste of resources to scan
5821              further. We stop the scan at the current point by temporarily
5822              adjusting the value of cd->endpattern. */
5823    
5824              temp = cd->end_pattern;
5825              cd->end_pattern = ptr;
5826              recno = find_parens(cd, name, namelen,
5827                (options & PCRE_EXTENDED) != 0, utf8);
5828              cd->end_pattern = temp;
5829              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5830            }            }
5831    
5832          /* 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 5070  we set the flag only if there is a liter Line 5851  we set the flag only if there is a liter
5851              }              }
5852            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5853                      find_parens(cd, name, namelen,                      find_parens(cd, name, namelen,
5854                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5855              {              {
5856              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5857              goto FAILED;              goto FAILED;
# Line 5181  we set the flag only if there is a liter Line 5962  we set the flag only if there is a liter
5962              if (called == NULL)              if (called == NULL)
5963                {                {
5964                if (find_parens(cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5965                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5966                  {                  {
5967                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5968                  goto FAILED;                  goto FAILED;
5969                  }                  }
5970    
5971                /* Fudge the value of "called" so that when it is inserted as an                /* Fudge the value of "called" so that when it is inserted as an
5972                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5973                of the group. */                of the group. Then remember the forward reference. */
5974    
5975                called = cd->start_code + recno;                called = cd->start_code + recno;
5976                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5977                }                }
5978    
5979              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
5980              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
5981              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
5982                must not, however, do this check if we are in a conditional
5983                subpattern because the condition might be testing for recursion in
5984                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
5985                Forever loops are also detected at runtime, so those that occur in
5986                conditional subpatterns will be picked up then. */
5987    
5988              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
5989                       could_be_empty(called, code, bcptr, utf8, cd))                       could_be_empty(called, code, bcptr, utf8, cd))
5990                {                {
5991                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
# Line 5207  we set the flag only if there is a liter Line 5993  we set the flag only if there is a liter
5993                }                }
5994              }              }
5995    
5996            /* 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;  
5997    
5998            *code = OP_RECURSE;            *code = OP_RECURSE;
5999            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);  
6000            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
   
           length_prevgroup = 3 + 3*LINK_SIZE;  
6001            }            }
6002    
6003          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5284  we set the flag only if there is a liter Line 6058  we set the flag only if there is a liter
6058          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
6059          both phases.          both phases.
6060    
6061          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
6062          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. */  
6063    
6064          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
6065            {            {
# Line 5297  we set the flag only if there is a liter Line 6070  we set the flag only if there is a liter
6070              }              }
6071            else            else
6072              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
6073              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
6074              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
6075              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
6076              }              }
6077    
6078            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
6079            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). */  
6080    
6081            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
6082            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5327  we set the flag only if there is a liter Line 6093  we set the flag only if there is a liter
6093          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
6094        }       /* End of (? handling */        }       /* End of (? handling */
6095    
6096      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
6097      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
6098      brackets. */      brackets. */
6099    
6100      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 5346  we set the flag only if there is a liter Line 6112  we set the flag only if there is a liter
6112        skipbytes = 2;        skipbytes = 2;
6113        }        }
6114    
6115      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
6116      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
6117      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
6118      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. */  
6119    
6120      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                      /* For handling repetition */
6121      *code = bravalue;      *code = bravalue;
6122      tempcode = code;      tempcode = code;
6123      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
6124      length_prevgroup = 0;              /* Initialize for pre-compile phase */      tempbracount = cd->bracount;          /* Save value before bracket */
6125        length_prevgroup = 0;                 /* Initialize for pre-compile phase */
6126    
6127      if (!compile_regex(      if (!compile_regex(
6128           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
6129           options & PCRE_IMS,           /* The previous ims option state */           &tempcode,                       /* Where to put code (updated) */
6130           &tempcode,                    /* Where to put code (updated) */           &ptr,                            /* Input pointer (updated) */
6131           &ptr,                         /* Input pointer (updated) */           errorcodeptr,                    /* Where to put an error message */
          errorcodeptr,                 /* Where to put an error message */  
6132           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
6133            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
6134           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
6135           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
6136           &subfirstbyte,                /* For possible first char */           cond_depth +
6137           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
6138           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
6139           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
6140           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
6141             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
6142             (lengthptr == NULL)? NULL :      /* Actual compile phase */
6143               &length_prevgroup              /* Pre-compile phase */
6144           ))           ))
6145        goto FAILED;        goto FAILED;
6146    
6147        /* If this was an atomic group and there are no capturing groups within it,
6148        generate OP_ONCE_NC instead of OP_ONCE. */
6149    
6150        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
6151          *code = OP_ONCE_NC;
6152    
6153        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
6154          cd->assert_depth -= 1;
6155    
6156      /* 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
6157      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.
6158      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
6159    
6160      /* 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
6161      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
6162      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
6163      not be available. */      not be available. */
# Line 5447  we set the flag only if there is a liter Line 6222  we set the flag only if there is a liter
6222          goto FAILED;          goto FAILED;
6223          }          }
6224        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
6225        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
6226        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
6227        *code++ = OP_KET;        *code++ = OP_KET;
6228        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5520  we set the flag only if there is a liter Line 6295  we set the flag only if there is a liter
6295    
6296      /* ===================================================================*/      /* ===================================================================*/
6297      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
6298      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
6299      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
6300      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
6301      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
6302      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
6303        ever created. */
6304    
6305      case CHAR_BACKSLASH:      case CHAR_BACKSLASH:
6306      tempptr = ptr;      tempptr = ptr;
# Line 5614  we set the flag only if there is a liter Line 6390  we set the flag only if there is a liter
6390          }          }
6391    
6392        /* \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).
6393        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6394    
6395        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6396          {          {
6397            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6398              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6399              {
6400              *errorcodeptr = ERR69;
6401              break;
6402              }
6403          is_recurse = FALSE;          is_recurse = FALSE;
6404          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6405            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
# Line 5638  we set the flag only if there is a liter Line 6419  we set the flag only if there is a liter
6419          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6420          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6421          previous = code;          previous = code;
6422          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6423          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6424          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6425          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
# Line 5684  we set the flag only if there is a liter Line 6465  we set the flag only if there is a liter
6465  #endif  #endif
6466    
6467        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
6468        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
6469          situation when PCRE_UCP is not set. When it *is* set, we substitute
6470          Unicode property tests. */
6471    
6472        else        else
6473          {          {
6474          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
6475          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
6476              {
6477              nestptr = ptr + 1;                   /* Where to resume */
6478              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
6479              }
6480            else
6481    #endif
6482            /* In non-UTF-8 mode, we turn \C into OP_ALLANY instead of OP_ANYBYTE
6483            so that it works in DFA mode and in lookbehinds. */
6484    
6485              {
6486              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6487              *code++ = (!utf8 && c == -ESC_C)? OP_ALLANY : -c;
6488              }
6489          }          }
6490        continue;        continue;
6491        }        }
# Line 5734  we set the flag only if there is a liter Line 6530  we set the flag only if there is a liter
6530    
6531      ONE_CHAR:      ONE_CHAR:
6532      previous = code;      previous = code;
6533      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6534      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6535    
6536      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 5763  we set the flag only if there is a liter Line 6559  we set the flag only if there is a liter
6559        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6560        }        }
6561    
6562      /* 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
6563      1 or the matching is caseful. */      1 or the matching is caseful. */
6564    
6565      else      else
# Line 5798  return FALSE; Line 6594  return FALSE;
6594  /* 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
6595  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
6596  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.  
   
6597  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
6598  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
6599  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6600    
6601  Arguments:  Arguments:
6602    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  
6603    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6604    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6605    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
6606    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6607    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6608    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6609      cond_depth     depth of nesting for conditional subpatterns
6610    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6611    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6612    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 5827  Returns: TRUE on success Line 6618  Returns: TRUE on success
6618  */  */
6619    
6620  static BOOL  static BOOL
6621  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6622    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6623    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6624    int *lengthptr)    compile_data *cd, int *lengthptr)
6625  {  {
6626  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6627  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 5844  int branchfirstbyte, branchreqbyte; Line 6635  int branchfirstbyte, branchreqbyte;
6635  int length;  int length;
6636  int orig_bracount;  int orig_bracount;
6637  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6638  branch_chain bc;  branch_chain bc;
6639    
6640  bc.outer = bcptr;  bc.outer = bcptr;
# Line 5868  pre-compile phase to find out whether an Line 6658  pre-compile phase to find out whether an
6658    
6659  /* 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
6660  so that we can detect them if (*ACCEPT) is encountered. This is also used to  so that we can detect them if (*ACCEPT) is encountered. This is also used to
6661  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6662    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6663    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6664    
6665  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6666    {    {
# Line 5894  for (;;) Line 6686  for (;;)
6686    
6687    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6688    
   /* 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;  
     }  
   
6689    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6690    
6691    if (lookbehind)    if (lookbehind)
# Line 5917  for (;;) Line 6700  for (;;)
6700    into the length. */    into the length. */
6701    
6702    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6703          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6704            (lengthptr == NULL)? NULL : &length))
6705      {      {
6706      *ptrptr = ptr;      *ptrptr = ptr;
6707      return FALSE;      return FALSE;
6708      }      }
6709    
   /* 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;  
   
6710    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6711    has fewer than the rest. */    has fewer than the rest. */
6712    
# Line 5992  for (;;) Line 6767  for (;;)
6767        {        {
6768        int fixed_length;        int fixed_length;
6769        *code = OP_END;        *code = OP_END;
6770        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6771            FALSE, cd);
6772        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6773        if (fixed_length == -3)        if (fixed_length == -3)
6774          {          {
# Line 6000  for (;;) Line 6776  for (;;)
6776          }          }
6777        else if (fixed_length < 0)        else if (fixed_length < 0)
6778          {          {
6779          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 :
6780                            (fixed_length == -4)? ERR70: ERR25;
6781          *ptrptr = ptr;          *ptrptr = ptr;
6782          return FALSE;          return FALSE;
6783          }          }
# Line 6013  for (;;) Line 6790  for (;;)
6790    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
6791    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
6792    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
6793    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. */  
6794    
6795    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6796      {      {
6797      if (lengthptr == NULL)      if (lengthptr == NULL)
6798        {        {
6799        int branch_length = code - last_branch;        int branch_length = (int)(code - last_branch);
6800        do        do
6801          {          {
6802          int prev_length = GET(last_branch, 1);          int prev_length = GET(last_branch, 1);
# Line 6035  for (;;) Line 6810  for (;;)
6810      /* Fill in the ket */      /* Fill in the ket */
6811    
6812      *code = OP_KET;      *code = OP_KET;
6813      PUT(code, 1, code - start_bracket);      PUT(code, 1, (int)(code - start_bracket));
6814      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6815    
6816      /* If it was a capturing subpattern, check to see if it contained any      /* If it was a capturing subpattern, check to see if it contained any
# Line 6050  for (;;) Line 6825  for (;;)
6825            code - start_bracket);            code - start_bracket);
6826          *start_bracket = OP_ONCE;          *start_bracket = OP_ONCE;
6827          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
6828          PUT(start_bracket, 1, code - start_bracket);          PUT(start_bracket, 1, (int)(code - start_bracket));
6829          *code = OP_KET;          *code = OP_KET;
6830          PUT(code, 1, code - start_bracket);          PUT(code, 1, (int)(code - start_bracket));
6831          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
6832          length += 2 + 2*LINK_SIZE;          length += 2 + 2*LINK_SIZE;
6833          }          }
6834        cd->open_caps = cd->open_caps->next;        cd->open_caps = cd->open_caps->next;
6835        }        }
6836    
     /* Reset options if needed. */  
   
     if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)  
       {  
       *code++ = OP_OPT;  
       *code++ = oldims;  
       length += 2;  
       }  
   
6837      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
6838    
6839      cd->bracount = max_bracount;      cd->bracount = max_bracount;
# Line 6107  for (;;) Line 6873  for (;;)
6873    else    else
6874      {      {
6875      *code = OP_ALT;      *code = OP_ALT;
6876      PUT(code, 1, code - last_branch);      PUT(code, 1, (int)(code - last_branch));
6877      bc.current_branch = last_branch = code;      bc.current_branch = last_branch = code;
6878      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6879      }      }
# Line 6127  for (;;) Line 6893  for (;;)
6893  /* 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
6894  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
6895  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
6896  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
6897  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6898    
6899  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.
6900  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 6149  of the more common cases more precisely. Line 6915  of the more common cases more precisely.
6915    
6916  Arguments:  Arguments:
6917    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6918    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
6919                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6920                    the less precise approach                    the less precise approach
# Line 6159  Returns: TRUE or FALSE Line 6924  Returns: TRUE or FALSE
6924  */  */
6925    
6926  static BOOL  static BOOL
6927  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6928    unsigned int backref_map)    unsigned int backref_map)
6929  {  {
6930  do {  do {
6931     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6932       options, PCRE_MULTILINE, FALSE);       FALSE);
6933     register int op = *scode;     register int op = *scode;
6934    
6935     /* Non-capturing brackets */     /* Non-capturing brackets */
6936    
6937     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6938           op == OP_SBRA || op == OP_SBRAPOS)
6939       {       {
6940       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6941       }       }
6942    
6943     /* Capturing brackets */     /* Capturing brackets */
6944    
6945     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6946                op == OP_SCBRA || op == OP_SCBRAPOS)
6947       {       {
6948       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6949       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6950       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6951       }       }
6952    
6953     /* Other brackets */     /* Other brackets */
6954    
6955     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC ||
6956                op == OP_COND)
6957       {       {
6958       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6959       }       }
6960    
6961     /* .* 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 6202  do { Line 6970  do {
6970    
6971     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6972    
6973     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;  
6974     code += GET(code, 1);     code += GET(code, 1);
6975     }     }
6976  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6240  is_startline(const uschar *code, unsigne Line 7006  is_startline(const uschar *code, unsigne
7006  {  {
7007  do {  do {
7008     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
7009       NULL, 0, FALSE);       FALSE);
7010     register int op = *scode;     register int op = *scode;
7011    
7012     /* 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 6267  do { Line 7033  do {
7033         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
7034         break;         break;
7035         }         }
7036       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
7037       op = *scode;       op = *scode;
7038       }       }
7039    
7040     /* Non-capturing brackets */     /* Non-capturing brackets */
7041    
7042     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
7043           op == OP_SBRA || op == OP_SBRAPOS)
7044       {       {
7045       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
7046       }       }
7047    
7048     /* Capturing brackets */     /* Capturing brackets */
7049    
7050     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
7051                op == OP_SCBRA || op == OP_SCBRAPOS)
7052       {       {
7053       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
7054       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6289  do { Line 7057  do {
7057    
7058     /* Other brackets */     /* Other brackets */
7059    
7060     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC)
7061       {       {
7062       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
7063       }       }
# Line 6304  do { Line 7072  do {
7072    
7073     /* Check for explicit circumflex */     /* Check for explicit circumflex */
7074    
7075     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
7076    
7077     /* Move on to the next alternative */     /* Move on to the next alternative */
7078    
# Line 6330  we return that char, otherwise -1. Line 7098  we return that char, otherwise -1.
7098    
7099  Arguments:  Arguments:
7100    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)  
7101    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
7102    
7103  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
7104  */  */
7105    
7106  static int  static int
7107  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
7108  {  {
7109  register int c = -1;  register int c = -1;
7110  do {  do {
7111     int d;     int d;
7112     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
7113       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
7114       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
7115     register int op = *scode;     register int op = *scode;
7116    
7117     switch(op)     switch(op)
# Line 6352  do { Line 7120  do {
7120       return -1;       return -1;
7121    
7122       case OP_BRA:       case OP_BRA:
7123         case OP_BRAPOS:
7124       case OP_CBRA:       case OP_CBRA:
7125         case OP_SCBRA:
7126         case OP_CBRAPOS:
7127         case OP_SCBRAPOS:
7128       case OP_ASSERT:       case OP_ASSERT:
7129       case OP_ONCE:       case OP_ONCE:
7130         case OP_ONCE_NC:
7131       case OP_COND:       case OP_COND:
7132       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
7133         return -1;         return -1;
7134       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
7135       break;       break;
7136    
7137       case OP_EXACT:       /* Fall through */       case OP_EXACT:
7138       scode += 2;       scode += 2;
7139         /* Fall through */
7140    
7141       case OP_CHAR:       case OP_CHAR:
      case OP_CHARNC:  
7142       case OP_PLUS:       case OP_PLUS:
7143       case OP_MINPLUS:       case OP_MINPLUS:
7144       case OP_POSPLUS:       case OP_POSPLUS:
7145       if (!inassert) return -1;       if (!inassert) return -1;
7146       if (c < 0)       if (c < 0) c = scode[1];
7147         {         else if (c != scode[1]) return -1;
7148         c = scode[1];       break;
7149         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
7150         }       case OP_EXACTI:
7151       else if (c != scode[1]) return -1;       scode += 2;
7152         /* Fall through */
7153    
7154         case OP_CHARI:
7155         case OP_PLUSI:
7156         case OP_MINPLUSI:
7157         case OP_POSPLUSI:
7158         if (!inassert) return -1;
7159         if (c < 0) c = scode[1] | REQ_CASELESS;
7160           else if (c != scode[1]) return -1;
7161       break;       break;
7162       }       }
7163    
# Line 6426  int length = 1; /* For final END opcode Line 7208  int length = 1; /* For final END opcode
7208  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
7209  int errorcode = 0;  int errorcode = 0;
7210  int skipatstart = 0;  int skipatstart = 0;
7211  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8;
7212  size_t size;  size_t size;
7213  uschar *code;  uschar *code;
7214  const uschar *codestart;  const uschar *codestart;
# Line 6496  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7278  while (ptr[skipatstart] == CHAR_LEFT_PAR
7278    
7279    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
7280      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
7281      else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
7282        { skipatstart += 6; options |= PCRE_UCP; continue; }
7283      else if (strncmp((char *)(ptr+skipatstart+2), STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
7284        { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
7285    
7286    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)