/[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 412 by ph10, Sat Apr 11 10:34:37 2009 UTC revision 751 by ph10, Fri Nov 18 11:13:30 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) 1997-2009 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 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 91  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 118  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 165  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 182  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 196  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 250  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 262  the number of relocations needed when a Line 320  the number of relocations needed when a
320  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
321  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
323  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329  static const char error_texts[] =  static const char error_texts[] =
330    "no error\0"    "no error\0"
# Line 309  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 331  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"
404    "subpattern name expected\0"    "subpattern name expected\0"
405    "digit expected after (?+\0"    "digit expected after (?+\0"
406    "] is an invalid data character in JavaScript compatibility mode";    "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "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 478  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 500  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--) while (*s++ != 0) {};  for (; n > 0; n--)
574      {
575      while (*s++ != 0) {};
576      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 571  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 594  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 722  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 772  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 782  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 809  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 909  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 1017  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 1028  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 1035  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 1047  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 1077  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 1100  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1272  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1272        if (name != NULL && lorn == ptr - thisname &&        if (name != NULL && lorn == ptr - thisname &&
1273            strncmp((const char *)name, (const char *)thisname, lorn) == 0)            strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1274          return *count;          return *count;
1275          term++;
1276        }        }
1277      }      }
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 1134  for (; *ptr != 0; ptr++) Line 1309  for (; *ptr != 0; ptr++)
1309      BOOL negate_class = FALSE;      BOOL negate_class = FALSE;
1310      for (;;)      for (;;)
1311        {        {
1312        int c = *(++ptr);        if (ptr[1] == CHAR_BACKSLASH)
       if (c == CHAR_BACKSLASH)  
1313          {          {
1314          if (ptr[1] == CHAR_E)          if (ptr[2] == CHAR_E)
1315            ptr++;            ptr+= 2;
1316          else if (strncmp((const char *)ptr+1,          else if (strncmp((const char *)ptr+2,
1317                   STR_Q STR_BACKSLASH STR_E, 3) == 0)                   STR_Q STR_BACKSLASH STR_E, 3) == 0)
1318            ptr += 3;            ptr += 4;
1319          else          else
1320            break;            break;
1321          }          }
1322        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1323            {
1324          negate_class = TRUE;          negate_class = TRUE;
1325            ptr++;
1326            }
1327        else break;        else break;
1328        }        }
1329    
# Line 1179  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 1188  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 1196  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 1235  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 1253  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 1269  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 1312  for (;;) Line 1488  for (;;)
1488    
1489      case OP_CALLOUT:      case OP_CALLOUT:
1490      case OP_CREF:      case OP_CREF:
1491        case OP_NCREF:
1492      case OP_RREF:      case OP_RREF:
1493        case OP_NRREF:
1494      case OP_DEF:      case OP_DEF:
1495      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1496      break;      break;
# Line 1328  for (;;) Line 1506  for (;;)
1506    
1507    
1508  /*************************************************  /*************************************************
1509  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1510  *************************************************/  *************************************************/
1511    
1512  /* Scan a pattern and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1513  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1514  In UTF8 mode, the result is in characters rather than bytes.  In UTF8 mode, the result is in characters rather than bytes. The branch is
1515    temporarily terminated with OP_END when this function is called.
1516    
1517    This function is called when a backward assertion is encountered, so that if it
1518    fails, the error message can point to the correct place in the pattern.
1519    However, we cannot do this when the assertion contains subroutine calls,
1520    because they can be forward references. We solve this by remembering this case
1521    and doing the check at the end; a flag specifies which mode we are running in.
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
1527      cd       the "compile data" structure
1528    
1529  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1530                 or -1 if there is no fixed length,
1531               or -2 if \C was encountered               or -2 if \C was encountered
1532                 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)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1538  {  {
1539  int length = -1;  int length = -1;
1540    
# Line 1357  branch, check the length against that of Line 1547  branch, check the length against that of
1547  for (;;)  for (;;)
1548    {    {
1549    int d;    int d;
1550      uschar *ce, *cs;
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);      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 1387  for (;;) Line 1586  for (;;)
1586      branchlength = 0;      branchlength = 0;
1587      break;      break;
1588    
1589        /* A true recursion implies not fixed length, but a subroutine call may
1590        be OK. If the subroutine is a forward reference, we can't deal with
1591        it until the end of the pattern, so return -3. */
1592    
1593        case OP_RECURSE:
1594        if (!atend) return -3;
1595        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1596        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1597        if (cc > cs && cc < ce) return -1;                /* Recursion */
1598        d = find_fixedlength(cs + 2, utf8, atend, cd);
1599        if (d < 0) return d;
1600        branchlength += d;
1601        cc += 1 + LINK_SIZE;
1602        break;
1603    
1604      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1605    
1606      case OP_ASSERT:      case OP_ASSERT:
# Line 1398  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_RREF:  
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:  
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 1418  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)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1659  #endif  #endif
1660      break;      break;
1661    
# Line 1434  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)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1673  #endif  #endif
1674      break;      break;
1675    
# Line 1457  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 1488  for (;;) Line 1721  for (;;)
1721    
1722      switch (*cc)      switch (*cc)
1723        {        {
1724          case OP_CRPLUS:
1725          case OP_CRMINPLUS:
1726        case OP_CRSTAR:        case OP_CRSTAR:
1727        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1728        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1508  for (;;) Line 1743  for (;;)
1743    
1744      /* Anything else is variable length */      /* Anything else is variable length */
1745    
1746      default:      case OP_ANYNL:
1747        case OP_BRAMINZERO:
1748        case OP_BRAPOS:
1749        case OP_BRAPOSZERO:
1750        case OP_BRAZERO:
1751        case OP_CBRAPOS:
1752        case OP_EXTUNI:
1753        case OP_KETRMAX:
1754        case OP_KETRMIN:
1755        case OP_KETRPOS:
1756        case OP_MINPLUS:
1757        case OP_MINPLUSI:
1758        case OP_MINQUERY:
1759        case OP_MINQUERYI:
1760        case OP_MINSTAR:
1761        case OP_MINSTARI:
1762        case OP_MINUPTO:
1763        case OP_MINUPTOI:
1764        case OP_NOTMINPLUS:
1765        case OP_NOTMINPLUSI:
1766        case OP_NOTMINQUERY:
1767        case OP_NOTMINQUERYI:
1768        case OP_NOTMINSTAR:
1769        case OP_NOTMINSTARI:
1770        case OP_NOTMINUPTO:
1771        case OP_NOTMINUPTOI:
1772        case OP_NOTPLUS:
1773        case OP_NOTPLUSI:
1774        case OP_NOTPOSPLUS:
1775        case OP_NOTPOSPLUSI:
1776        case OP_NOTPOSQUERY:
1777        case OP_NOTPOSQUERYI:
1778        case OP_NOTPOSSTAR:
1779        case OP_NOTPOSSTARI:
1780        case OP_NOTPOSUPTO:
1781        case OP_NOTPOSUPTOI:
1782        case OP_NOTQUERY:
1783        case OP_NOTQUERYI:
1784        case OP_NOTSTAR:
1785        case OP_NOTSTARI:
1786        case OP_NOTUPTO:
1787        case OP_NOTUPTOI:
1788        case OP_PLUS:
1789        case OP_PLUSI:
1790        case OP_POSPLUS:
1791        case OP_POSPLUSI:
1792        case OP_POSQUERY:
1793        case OP_POSQUERYI:
1794        case OP_POSSTAR:
1795        case OP_POSSTARI:
1796        case OP_POSUPTO:
1797        case OP_POSUPTOI:
1798        case OP_QUERY:
1799        case OP_QUERYI:
1800        case OP_REF:
1801        case OP_REFI:
1802        case OP_SBRA:
1803        case OP_SBRAPOS:
1804        case OP_SCBRA:
1805        case OP_SCBRAPOS:
1806        case OP_SCOND:
1807        case OP_SKIPZERO:
1808        case OP_STAR:
1809        case OP_STARI:
1810        case OP_TYPEMINPLUS:
1811        case OP_TYPEMINQUERY:
1812        case OP_TYPEMINSTAR:
1813        case OP_TYPEMINUPTO:
1814        case OP_TYPEPLUS:
1815        case OP_TYPEPOSPLUS:
1816        case OP_TYPEPOSQUERY:
1817        case OP_TYPEPOSSTAR:
1818        case OP_TYPEPOSUPTO:
1819        case OP_TYPEQUERY:
1820        case OP_TYPESTAR:
1821        case OP_TYPEUPTO:
1822        case OP_UPTO:
1823        case OP_UPTOI:
1824      return -1;      return -1;
1825    
1826        /* Catch unrecognized opcodes so that when new ones are added they
1827        are not forgotten, as has happened in the past. */
1828    
1829        default:
1830        return -4;
1831      }      }
1832    }    }
1833  /* Control never gets here */  /* Control never gets here */
# Line 1519  for (;;) Line 1837  for (;;)
1837    
1838    
1839  /*************************************************  /*************************************************
1840  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1841  *************************************************/  *************************************************/
1842    
1843  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1844  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1845    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1846    so that it can be called from pcre_study() when finding the minimum matching
1847    length.
1848    
1849  Arguments:  Arguments:
1850    code        points to start of expression    code        points to start of expression
1851    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1852    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1853    
1854  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
1855  */  */
1856    
1857  static const uschar *  const uschar *
1858  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1859  {  {
1860  for (;;)  for (;;)
1861    {    {
1862    register int c = *code;    register int c = *code;
1863    
1864    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1865    
1866    /* 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 1547  for (;;) Line 1869  for (;;)
1869    
1870    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1871    
1872      /* Handle recursion */
1873    
1874      else if (c == OP_REVERSE)
1875        {
1876        if (number < 0) return (uschar *)code;
1877        code += _pcre_OP_lengths[c];
1878        }
1879    
1880    /* Handle capturing bracket */    /* Handle capturing bracket */
1881    
1882    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1883               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1884      {      {
1885      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1886      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1558  for (;;) Line 1889  for (;;)
1889    
1890    /* 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
1891    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
1892    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1893      must add in its length. */
1894    
1895    else    else
1896      {      {
# Line 1582  for (;;) Line 1914  for (;;)
1914        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1915        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1916        break;        break;
1917    
1918          case OP_MARK:
1919          case OP_PRUNE_ARG:
1920          case OP_SKIP_ARG:
1921          code += code[1];
1922          break;
1923    
1924          case OP_THEN_ARG:
1925          code += code[1];
1926          break;
1927        }        }
1928    
1929      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1596  for (;;) Line 1938  for (;;)
1938      if (utf8) switch(c)      if (utf8) switch(c)
1939        {        {
1940        case OP_CHAR:        case OP_CHAR:
1941        case OP_CHARNC:        case OP_CHARI:
1942        case OP_EXACT:        case OP_EXACT:
1943          case OP_EXACTI:
1944        case OP_UPTO:        case OP_UPTO:
1945          case OP_UPTOI:
1946        case OP_MINUPTO:        case OP_MINUPTO:
1947          case OP_MINUPTOI:
1948        case OP_POSUPTO:        case OP_POSUPTO:
1949          case OP_POSUPTOI:
1950        case OP_STAR:        case OP_STAR:
1951          case OP_STARI:
1952        case OP_MINSTAR:        case OP_MINSTAR:
1953          case OP_MINSTARI:
1954        case OP_POSSTAR:        case OP_POSSTAR:
1955          case OP_POSSTARI:
1956        case OP_PLUS:        case OP_PLUS:
1957          case OP_PLUSI:
1958        case OP_MINPLUS:        case OP_MINPLUS:
1959          case OP_MINPLUSI:
1960        case OP_POSPLUS:        case OP_POSPLUS:
1961          case OP_POSPLUSI:
1962        case OP_QUERY:        case OP_QUERY:
1963          case OP_QUERYI:
1964        case OP_MINQUERY:        case OP_MINQUERY:
1965          case OP_MINQUERYI:
1966        case OP_POSQUERY:        case OP_POSQUERY:
1967          case OP_POSQUERYI:
1968        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1969        break;        break;
1970        }        }
# Line 1653  for (;;) Line 2008  for (;;)
2008    
2009    /* 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
2010    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
2011    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2012      must add in its length. */
2013    
2014    else    else
2015      {      {
# Line 1677  for (;;) Line 2033  for (;;)
2033        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2034        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2035        break;        break;
2036    
2037          case OP_MARK:
2038          case OP_PRUNE_ARG:
2039          case OP_SKIP_ARG:
2040          code += code[1];
2041          break;
2042    
2043          case OP_THEN_ARG:
2044          code += code[1];
2045          break;
2046        }        }
2047    
2048      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1691  for (;;) Line 2057  for (;;)
2057      if (utf8) switch(c)      if (utf8) switch(c)
2058        {        {
2059        case OP_CHAR:        case OP_CHAR:
2060        case OP_CHARNC:        case OP_CHARI:
2061        case OP_EXACT:        case OP_EXACT:
2062          case OP_EXACTI:
2063        case OP_UPTO:        case OP_UPTO:
2064          case OP_UPTOI:
2065        case OP_MINUPTO:        case OP_MINUPTO:
2066          case OP_MINUPTOI:
2067        case OP_POSUPTO:        case OP_POSUPTO:
2068          case OP_POSUPTOI:
2069        case OP_STAR:        case OP_STAR:
2070          case OP_STARI:
2071        case OP_MINSTAR:        case OP_MINSTAR:
2072          case OP_MINSTARI:
2073        case OP_POSSTAR:        case OP_POSSTAR:
2074          case OP_POSSTARI:
2075        case OP_PLUS:        case OP_PLUS:
2076          case OP_PLUSI:
2077        case OP_MINPLUS:        case OP_MINPLUS:
2078          case OP_MINPLUSI:
2079        case OP_POSPLUS:        case OP_POSPLUS:
2080          case OP_POSPLUSI:
2081        case OP_QUERY:        case OP_QUERY:
2082          case OP_QUERYI:
2083        case OP_MINQUERY:        case OP_MINQUERY:
2084          case OP_MINQUERYI:
2085        case OP_POSQUERY:        case OP_POSQUERY:
2086          case OP_POSQUERYI:
2087        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
2088        break;        break;
2089        }        }
# Line 1733  Arguments: Line 2112  Arguments:
2112    code        points to start of search    code        points to start of search
2113    endcode     points to where to stop    endcode     points to where to stop
2114    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
2115      cd          contains pointers to tables etc.
2116    
2117  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2118  */  */
2119    
2120  static BOOL  static BOOL
2121  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
2122      compile_data *cd)
2123  {  {
2124  register int c;  register int c;
2125  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2126       code < endcode;       code < endcode;
2127       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2128    {    {
2129    const uschar *ccode;    const uschar *ccode;
2130    
# Line 1759  for (code = first_significant_code(code Line 2140  for (code = first_significant_code(code
2140      continue;      continue;
2141      }      }
2142    
2143      /* For a recursion/subroutine call, if its end has been reached, which
2144      implies a backward reference subroutine call, we can scan it. If it's a
2145      forward reference subroutine call, we can't. To detect forward reference
2146      we have to scan up the list that is kept in the workspace. This function is
2147      called only when doing the real compile, not during the pre-compile that
2148      measures the size of the compiled pattern. */
2149    
2150      if (c == OP_RECURSE)
2151        {
2152        const uschar *scode;
2153        BOOL empty_branch;
2154    
2155        /* Test for forward reference */
2156    
2157        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2158          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2159    
2160        /* Not a forward reference, test for completed backward reference */
2161    
2162        empty_branch = FALSE;
2163        scode = cd->start_code + GET(code, 1);
2164        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2165    
2166        /* Completed backwards reference */
2167    
2168        do
2169          {
2170          if (could_be_empty_branch(scode, endcode, utf8, cd))
2171            {
2172            empty_branch = TRUE;
2173            break;
2174            }
2175          scode += GET(scode, 1);
2176          }
2177        while (*scode == OP_ALT);
2178    
2179        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2180        continue;
2181        }
2182    
2183    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2184    
2185    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2186          c == OP_BRAPOSZERO)
2187      {      {
2188      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
2189      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1769  for (code = first_significant_code(code Line 2191  for (code = first_significant_code(code
2191      continue;      continue;
2192      }      }
2193    
2194      /* A nested group that is already marked as "could be empty" can just be
2195      skipped. */
2196    
2197      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2198          c == OP_SCBRA || c == OP_SCBRAPOS)
2199        {
2200        do code += GET(code, 1); while (*code == OP_ALT);
2201        c = *code;
2202        continue;
2203        }
2204    
2205    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2206    
2207    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2208          c == OP_CBRA || c == OP_CBRAPOS ||
2209          c == OP_ONCE || c == OP_ONCE_NC ||
2210          c == OP_COND)
2211      {      {
2212      BOOL empty_branch;      BOOL empty_branch;
2213      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1787  for (code = first_significant_code(code Line 2223  for (code = first_significant_code(code
2223        empty_branch = FALSE;        empty_branch = FALSE;
2224        do        do
2225          {          {
2226          if (!empty_branch && could_be_empty_branch(code, endcode, utf8))          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2227            empty_branch = TRUE;            empty_branch = TRUE;
2228          code += GET(code, 1);          code += GET(code, 1);
2229          }          }
# Line 1858  for (code = first_significant_code(code Line 2294  for (code = first_significant_code(code
2294      case OP_ALLANY:      case OP_ALLANY:
2295      case OP_ANYBYTE:      case OP_ANYBYTE:
2296      case OP_CHAR:      case OP_CHAR:
2297      case OP_CHARNC:      case OP_CHARI:
2298      case OP_NOT:      case OP_NOT:
2299        case OP_NOTI:
2300      case OP_PLUS:      case OP_PLUS:
2301      case OP_MINPLUS:      case OP_MINPLUS:
2302      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1899  for (code = first_significant_code(code Line 2336  for (code = first_significant_code(code
2336      case OP_KET:      case OP_KET:
2337      case OP_KETRMAX:      case OP_KETRMAX:
2338      case OP_KETRMIN:      case OP_KETRMIN:
2339        case OP_KETRPOS:
2340      case OP_ALT:      case OP_ALT:
2341      return TRUE;      return TRUE;
2342    
# Line 1907  for (code = first_significant_code(code Line 2345  for (code = first_significant_code(code
2345    
2346  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2347      case OP_STAR:      case OP_STAR:
2348        case OP_STARI:
2349      case OP_MINSTAR:      case OP_MINSTAR:
2350        case OP_MINSTARI:
2351      case OP_POSSTAR:      case OP_POSSTAR:
2352        case OP_POSSTARI:
2353      case OP_QUERY:      case OP_QUERY:
2354        case OP_QUERYI:
2355      case OP_MINQUERY:      case OP_MINQUERY:
2356        case OP_MINQUERYI:
2357      case OP_POSQUERY:      case OP_POSQUERY:
2358        case OP_POSQUERYI:
2359        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2360        break;
2361    
2362      case OP_UPTO:      case OP_UPTO:
2363        case OP_UPTOI:
2364      case OP_MINUPTO:      case OP_MINUPTO:
2365        case OP_MINUPTOI:
2366      case OP_POSUPTO:      case OP_POSUPTO:
2367      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2368        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2369      break;      break;
2370  #endif  #endif
2371    
2372        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2373        string. */
2374    
2375        case OP_MARK:
2376        case OP_PRUNE_ARG:
2377        case OP_SKIP_ARG:
2378        code += code[1];
2379        break;
2380    
2381        case OP_THEN_ARG:
2382        code += code[1];
2383        break;
2384    
2385        /* None of the remaining opcodes are required to match a character. */
2386    
2387        default:
2388        break;
2389      }      }
2390    }    }
2391    
# Line 1934  return TRUE; Line 2402  return TRUE;
2402  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
2403  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,
2404  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.
2405    This function is called only during the real compile, not during the
2406    pre-compile.
2407    
2408  Arguments:  Arguments:
2409    code        points to start of the recursion    code        points to start of the recursion
2410    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2411    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2412    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2413      cd          pointers to tables etc
2414    
2415  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2416  */  */
2417    
2418  static BOOL  static BOOL
2419  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2420    BOOL utf8)    BOOL utf8, compile_data *cd)
2421  {  {
2422  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2423    {    {
2424    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2425        return FALSE;
2426    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2427    }    }
2428  return TRUE;  return TRUE;
# Line 1982  where Perl recognizes it as the POSIX cl Line 2454  where Perl recognizes it as the POSIX cl
2454  "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,
2455  I think.  I think.
2456    
2457    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2458    It seems that the appearance of a nested POSIX class supersedes an apparent
2459    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2460    a digit.
2461    
2462    In Perl, unescaped square brackets may also appear as part of class names. For
2463    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2464    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2465    seem right at all. PCRE does not allow closing square brackets in POSIX class
2466    names.
2467    
2468  Arguments:  Arguments:
2469    ptr      pointer to the initial [    ptr      pointer to the initial [
2470    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 1996  int terminator;          /* Don't combin Line 2479  int terminator;          /* Don't combin
2479  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2480  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2481    {    {
2482    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2483        ptr++;
2484      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2485      else
2486      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2487      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2488        {        {
2489        *endptr = ptr;        *endptr = ptr;
2490        return TRUE;        return TRUE;
2491        }        }
2492        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2493             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2494              ptr[1] == CHAR_EQUALS_SIGN) &&
2495            check_posix_syntax(ptr, endptr))
2496          return FALSE;
2497      }      }
2498    }    }
2499  return FALSE;  return FALSE;
# Line 2132  auto_callout(uschar *code, const uschar Line 2622  auto_callout(uschar *code, const uschar
2622  {  {
2623  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2624  *code++ = 255;  *code++ = 255;
2625  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2626  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2627  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2628  }  }
2629    
# Line 2158  Returns:             nothing Line 2648  Returns:             nothing
2648  static void  static void
2649  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2650  {  {
2651  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2652  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2653  }  }
2654    
# Line 2208  for (++c; c <= d; c++) Line 2698  for (++c; c <= d; c++)
2698    
2699  return TRUE;  return TRUE;
2700  }  }
2701    
2702    
2703    
2704    /*************************************************
2705    *        Check a character and a property        *
2706    *************************************************/
2707    
2708    /* This function is called by check_auto_possessive() when a property item
2709    is adjacent to a fixed character.
2710    
2711    Arguments:
2712      c            the character
2713      ptype        the property type
2714      pdata        the data for the type
2715      negated      TRUE if it's a negated property (\P or \p{^)
2716    
2717    Returns:       TRUE if auto-possessifying is OK
2718    */
2719    
2720    static BOOL
2721    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2722    {
2723    const ucd_record *prop = GET_UCD(c);
2724    switch(ptype)
2725      {
2726      case PT_LAMP:
2727      return (prop->chartype == ucp_Lu ||
2728              prop->chartype == ucp_Ll ||
2729              prop->chartype == ucp_Lt) == negated;
2730    
2731      case PT_GC:
2732      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2733    
2734      case PT_PC:
2735      return (pdata == prop->chartype) == negated;
2736    
2737      case PT_SC:
2738      return (pdata == prop->script) == negated;
2739    
2740      /* These are specials */
2741    
2742      case PT_ALNUM:
2743      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2744              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2745    
2746      case PT_SPACE:    /* Perl space */
2747      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2748              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2749              == negated;
2750    
2751      case PT_PXSPACE:  /* POSIX space */
2752      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2753              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2754              c == CHAR_FF || c == CHAR_CR)
2755              == negated;
2756    
2757      case PT_WORD:
2758      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2759              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2760              c == CHAR_UNDERSCORE) == negated;
2761      }
2762    return FALSE;
2763    }
2764  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2765    
2766    
# Line 2221  whether the next thing could possibly ma Line 2774  whether the next thing could possibly ma
2774  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2775    
2776  Arguments:  Arguments:
2777    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2778    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2779    ptr           next character in pattern    ptr           next character in pattern
2780    options       options bits    options       options bits
2781    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2233  Returns:        TRUE if possessifying is Line 2784  Returns:        TRUE if possessifying is
2784  */  */
2785    
2786  static BOOL  static BOOL
2787  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2788    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2789  {  {
2790  int next;  int c, next;
2791    int op_code = *previous++;
2792    
2793  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2794    
# Line 2247  if ((options & PCRE_EXTENDED) != 0) Line 2799  if ((options & PCRE_EXTENDED) != 0)
2799      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2800      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2801        {        {
2802        while (*(++ptr) != 0)        ptr++;
2803          while (*ptr != 0)
2804            {
2805          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2806            ptr++;
2807    #ifdef SUPPORT_UTF8
2808            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2809    #endif
2810            }
2811        }        }
2812      else break;      else break;
2813      }      }
# Line 2284  if ((options & PCRE_EXTENDED) != 0) Line 2843  if ((options & PCRE_EXTENDED) != 0)
2843      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2844      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2845        {        {
2846        while (*(++ptr) != 0)        ptr++;
2847          while (*ptr != 0)
2848            {
2849          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2850            ptr++;
2851    #ifdef SUPPORT_UTF8
2852            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2853    #endif
2854            }
2855        }        }
2856      else break;      else break;
2857      }      }
# Line 2297  if (*ptr == CHAR_ASTERISK || *ptr == CHA Line 2863  if (*ptr == CHAR_ASTERISK || *ptr == CHA
2863    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)
2864      return FALSE;      return FALSE;
2865    
2866  /* 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
2867  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. */  
2868    
2869  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2870    {    {
2871    case OP_CHAR:    case OP_CHAR:
2872  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2873    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2874  #else  #else
2875    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2876  #endif  #endif
2877    return item != next;    return c != next;
2878    
2879    /* 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
2880    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
2881    high-valued characters. */    high-valued characters. */
2882    
2883    case OP_CHARNC:    case OP_CHARI:
2884  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2885    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2886    #else
2887      c = *previous;
2888  #endif  #endif
2889    if (item == next) return FALSE;    if (c == next) return FALSE;
2890  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2891    if (utf8)    if (utf8)
2892      {      {
# Line 2334  if (next >= 0) switch(op_code) Line 2897  if (next >= 0) switch(op_code)
2897  #else  #else
2898      othercase = NOTACHAR;      othercase = NOTACHAR;
2899  #endif  #endif
2900      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2901      }      }
2902    else    else
2903  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2904    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2905    
2906    /* 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
2907      opcodes are not used for multi-byte characters, because they are coded using
2908      an XCLASS instead. */
2909    
2910    case OP_NOT:    case OP_NOT:
2911    if (item == next) return TRUE;    return (c = *previous) == next;
2912    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2913      case OP_NOTI:
2914      if ((c = *previous) == next) return TRUE;
2915  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2916    if (utf8)    if (utf8)
2917      {      {
# Line 2355  if (next >= 0) switch(op_code) Line 2922  if (next >= 0) switch(op_code)
2922  #else  #else
2923      othercase = NOTACHAR;      othercase = NOTACHAR;
2924  #endif  #endif
2925      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2926      }      }
2927    else    else
2928  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2929    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2930    
2931      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2932      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2933    
2934    case OP_DIGIT:    case OP_DIGIT:
2935    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2402  if (next >= 0) switch(op_code) Line 2972  if (next >= 0) switch(op_code)
2972      case 0x202f:      case 0x202f:
2973      case 0x205f:      case 0x205f:
2974      case 0x3000:      case 0x3000:
2975      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2976      default:      default:
2977      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2978      }      }
2979    
2980      case OP_ANYNL:
2981    case OP_VSPACE:    case OP_VSPACE:
2982    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2983    switch(next)    switch(next)
# Line 2418  if (next >= 0) switch(op_code) Line 2989  if (next >= 0) switch(op_code)
2989      case 0x85:      case 0x85:
2990      case 0x2028:      case 0x2028:
2991      case 0x2029:      case 0x2029:
2992      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2993      default:      default:
2994      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2995      }      }
2996    
2997    #ifdef SUPPORT_UCP
2998      case OP_PROP:
2999      return check_char_prop(next, previous[0], previous[1], FALSE);
3000    
3001      case OP_NOTPROP:
3002      return check_char_prop(next, previous[0], previous[1], TRUE);
3003    #endif
3004    
3005    default:    default:
3006    return FALSE;    return FALSE;
3007    }    }
3008    
3009    
3010  /* 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
3011    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3012    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3013    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3014    replaced by OP_PROP codes when PCRE_UCP is set. */
3015    
3016  switch(op_code)  switch(op_code)
3017    {    {
3018    case OP_CHAR:    case OP_CHAR:
3019    case OP_CHARNC:    case OP_CHARI:
3020  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3021    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3022    #else
3023      c = *previous;
3024  #endif  #endif
3025    switch(-next)    switch(-next)
3026      {      {
3027      case ESC_d:      case ESC_d:
3028      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3029    
3030      case ESC_D:      case ESC_D:
3031      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3032    
3033      case ESC_s:      case ESC_s:
3034      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3035    
3036      case ESC_S:      case ESC_S:
3037      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3038    
3039      case ESC_w:      case ESC_w:
3040      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3041    
3042      case ESC_W:      case ESC_W:
3043      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3044    
3045      case ESC_h:      case ESC_h:
3046      case ESC_H:      case ESC_H:
3047      switch(item)      switch(c)
3048        {        {
3049        case 0x09:        case 0x09:
3050        case 0x20:        case 0x20:
# Line 2487  switch(op_code) Line 3072  switch(op_code)
3072    
3073      case ESC_v:      case ESC_v:
3074      case ESC_V:      case ESC_V:
3075      switch(item)      switch(c)
3076        {        {
3077        case 0x0a:        case 0x0a:
3078        case 0x0b:        case 0x0b:
# Line 2501  switch(op_code) Line 3086  switch(op_code)
3086        return -next == ESC_v;        return -next == ESC_v;
3087        }        }
3088    
3089        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3090        their substitutions and process them. The result will always be either
3091        -ESC_p or -ESC_P. Then fall through to process those values. */
3092    
3093    #ifdef SUPPORT_UCP
3094        case ESC_du:
3095        case ESC_DU:
3096        case ESC_wu:
3097        case ESC_WU:
3098        case ESC_su:
3099        case ESC_SU:
3100          {
3101          int temperrorcode = 0;
3102          ptr = substitutes[-next - ESC_DU];
3103          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3104          if (temperrorcode != 0) return FALSE;
3105          ptr++;    /* For compatibility */
3106          }
3107        /* Fall through */
3108    
3109        case ESC_p:
3110        case ESC_P:
3111          {
3112          int ptype, pdata, errorcodeptr;
3113          BOOL negated;
3114    
3115          ptr--;      /* Make ptr point at the p or P */
3116          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3117          if (ptype < 0) return FALSE;
3118          ptr++;      /* Point past the final curly ket */
3119    
3120          /* If the property item is optional, we have to give up. (When generated
3121          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3122          to the original \d etc. At this point, ptr will point to a zero byte. */
3123    
3124          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3125            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3126              return FALSE;
3127    
3128          /* Do the property check. */
3129    
3130          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3131          }
3132    #endif
3133    
3134      default:      default:
3135      return FALSE;      return FALSE;
3136      }      }
3137    
3138      /* In principle, support for Unicode properties should be integrated here as
3139      well. It means re-organizing the above code so as to get hold of the property
3140      values before switching on the op-code. However, I wonder how many patterns
3141      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3142      these op-codes are never generated.) */
3143    
3144    case OP_DIGIT:    case OP_DIGIT:
3145    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3146           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3147    
3148    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3149    return next == -ESC_d;    return next == -ESC_d;
3150    
3151    case OP_WHITESPACE:    case OP_WHITESPACE:
3152    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3153    
3154    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3155    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3156    
3157    case OP_HSPACE:    case OP_HSPACE:
3158    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3159             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3160    
3161    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3162    return next == -ESC_h;    return next == -ESC_h;
3163    
3164    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3165      case OP_ANYNL:
3166    case OP_VSPACE:    case OP_VSPACE:
3167    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3168    
3169    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3170    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3171    
3172    case OP_WORDCHAR:    case OP_WORDCHAR:
3173    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3174             next == -ESC_v || next == -ESC_R;
3175    
3176    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3177    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2564  Arguments: Line 3203  Arguments:
3203    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3204    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3205    bcptr          points to current branch chain    bcptr          points to current branch chain
3206      cond_depth     conditional nesting depth
3207    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3208    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3209                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2575  Returns:         TRUE on success Line 3215  Returns:         TRUE on success
3215  static BOOL  static BOOL
3216  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3217    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3218    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3219  {  {
3220  int repeat_type, op_type;  int repeat_type, op_type;
3221  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2584  int greedy_default, greedy_non_default; Line 3224  int greedy_default, greedy_non_default;
3224  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3225  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3226  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3227  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3228  int after_manual_callout = 0;  int after_manual_callout = 0;
3229  int length_prevgroup = 0;  int length_prevgroup = 0;
3230  register int c;  register int c;
# Line 2596  BOOL inescq = FALSE; Line 3236  BOOL inescq = FALSE;
3236  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3237  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3238  const uschar *tempptr;  const uschar *tempptr;
3239    const uschar *nestptr = NULL;
3240  uschar *previous = NULL;  uschar *previous = NULL;
3241  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3242  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3243  uschar classbits[32];  uschar classbits[32];
3244    
3245    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3246    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3247    dynamically as we process the pattern. */
3248    
3249  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3250  BOOL class_utf8;  BOOL class_utf8;
3251  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2609  uschar *class_utf8data_base; Line 3254  uschar *class_utf8data_base;
3254  uschar utf8_char[6];  uschar utf8_char[6];
3255  #else  #else
3256  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3257  #endif  #endif
3258    
3259  #ifdef DEBUG  #ifdef PCRE_DEBUG
3260  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3261  #endif  #endif
3262    
# Line 2660  for (;; ptr++) Line 3304  for (;; ptr++)
3304    int subfirstbyte;    int subfirstbyte;
3305    int terminator;    int terminator;
3306    int mclength;    int mclength;
3307      int tempbracount;
3308    uschar mcbuffer[8];    uschar mcbuffer[8];
3309    
3310    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3311    
3312    c = *ptr;    c = *ptr;
3313    
3314      /* If we are at the end of a nested substitution, revert to the outer level
3315      string. Nesting only happens one level deep. */
3316    
3317      if (c == 0 && nestptr != NULL)
3318        {
3319        ptr = nestptr;
3320        nestptr = NULL;
3321        c = *ptr;
3322        }
3323    
3324    /* 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
3325    previous cycle of this loop. */    previous cycle of this loop. */
3326    
3327    if (lengthptr != NULL)    if (lengthptr != NULL)
3328      {      {
3329  #ifdef DEBUG  #ifdef PCRE_DEBUG
3330      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3331  #endif  #endif
3332      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3333        {        {
3334        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3335        goto FAILED;        goto FAILED;
# Line 2696  for (;; ptr++) Line 3351  for (;; ptr++)
3351        goto FAILED;        goto FAILED;
3352        }        }
3353    
3354      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3355      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),
3356          c));
3357    
3358      /* 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
3359      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 2723  for (;; ptr++) Line 3379  for (;; ptr++)
3379    /* 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
3380    reference list. */    reference list. */
3381    
3382    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3383      {      {
3384      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3385      goto FAILED;      goto FAILED;
# Line 2771  for (;; ptr++) Line 3427  for (;; ptr++)
3427      previous_callout = NULL;      previous_callout = NULL;
3428      }      }
3429    
3430    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3431    
3432    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3433      {      {
3434      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3435      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3436        {        {
3437        while (*(++ptr) != 0)        ptr++;
3438          while (*ptr != 0)
3439          {          {
3440          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3441            ptr++;
3442    #ifdef SUPPORT_UTF8
3443            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3444    #endif
3445          }          }
3446        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3447    
# Line 2814  for (;; ptr++) Line 3475  for (;; ptr++)
3475          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3476          goto FAILED;          goto FAILED;
3477          }          }
3478        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3479        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3480        }        }
3481      return TRUE;      return TRUE;
# Line 2825  for (;; ptr++) Line 3486  for (;; ptr++)
3486      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3487    
3488      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3489        previous = NULL;
3490      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3491        {        {
3492        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3493          *code++ = OP_CIRCM;
3494        }        }
3495      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3496      break;      break;
3497    
3498      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3499      previous = NULL;      previous = NULL;
3500      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3501      break;      break;
3502    
3503      /* 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 3019  for (;; ptr++) Line 3681  for (;; ptr++)
3681            ptr++;            ptr++;
3682            }            }
3683    
3684          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3685          if (posix_class < 0)          if (posix_class < 0)
3686            {            {
3687            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 3033  for (;; ptr++) Line 3695  for (;; ptr++)
3695          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3696            posix_class = 0;            posix_class = 0;
3697    
3698          /* 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
3699          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3700          subtract bits that may be in the main map already. At the end we or the  
3701          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3702            if ((options & PCRE_UCP) != 0)
3703              {
3704              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3705              if (posix_substitutes[pc] != NULL)
3706                {
3707                nestptr = tempptr + 1;
3708                ptr = posix_substitutes[pc] - 1;
3709                continue;
3710                }
3711              }
3712    #endif
3713            /* In the non-UCP case, we build the bit map for the POSIX class in a
3714            chunk of local store because we may be adding and subtracting from it,
3715            and we don't want to subtract bits that may be in the main map already.
3716            At the end we or the result into the bit map that is being built. */
3717    
3718          posix_class *= 3;          posix_class *= 3;
3719    
# Line 3080  for (;; ptr++) Line 3757  for (;; ptr++)
3757    
3758        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3759        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
3760        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
3761        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
3762        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
3763        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3764          PCRE_EXTRA is set. */
3765    
3766        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3767          {          {
3768          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3769          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3770    
3771          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 */  
3772          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3773            {            {
3774            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3109  for (;; ptr++) Line 3785  for (;; ptr++)
3785            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3786            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3787    
3788            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3789              {              {
3790    #ifdef SUPPORT_UCP
3791                case ESC_du:     /* These are the values given for \d etc */
3792                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3793                case ESC_wu:     /* escape sequence with an appropriate \p */
3794                case ESC_WU:     /* or \P to test Unicode properties instead */
3795                case ESC_su:     /* of the default ASCII testing. */
3796                case ESC_SU:
3797                nestptr = ptr;
3798                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3799                class_charcount -= 2;                /* Undo! */
3800                continue;
3801    #endif
3802              case ESC_d:              case ESC_d:
3803              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3804              continue;              continue;
# Line 3131  for (;; ptr++) Line 3817  for (;; ptr++)
3817              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3818              continue;              continue;
3819    
3820                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3821                if it was previously set by something earlier in the character
3822                class. */
3823    
3824              case ESC_s:              case ESC_s:
3825              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3826              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3827                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3828              continue;              continue;
3829    
3830              case ESC_S:              case ESC_S:
# Line 3142  for (;; ptr++) Line 3833  for (;; ptr++)
3833              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3834              continue;              continue;
3835    
3836              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)  
             {  
3837              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3838              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3839              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 3179  for (;; ptr++) Line 3857  for (;; ptr++)
3857                }                }
3858  #endif  #endif
3859              continue;              continue;
             }  
3860    
3861            if (-c == ESC_H)              case ESC_H:
             {  
3862              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3863                {                {
3864                int x = 0xff;                int x = 0xff;
# Line 3224  for (;; ptr++) Line 3900  for (;; ptr++)
3900                }                }
3901  #endif  #endif
3902              continue;              continue;
             }  
3903    
3904            if (-c == ESC_v)              case ESC_v:
             {  
3905              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3906              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3907              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3243  for (;; ptr++) Line 3917  for (;; ptr++)
3917                }                }
3918  #endif  #endif
3919              continue;              continue;
             }  
3920    
3921            if (-c == ESC_V)              case ESC_V:
             {  
3922              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3923                {                {
3924                int x = 0xff;                int x = 0xff;
# Line 3276  for (;; ptr++) Line 3948  for (;; ptr++)
3948                }                }
3949  #endif  #endif
3950              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3951    
3952  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3953            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3954              {              case ESC_P:
3955              BOOL negated;                {
3956              int pdata;                BOOL negated;
3957              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3958              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3959              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3960              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3961                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3962              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3963              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3964              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3965              continue;                class_charcount -= 2;   /* Not a < 256 character */
3966              }                continue;
3967                  }
3968  #endif  #endif
3969            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3970            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3971            treated as literals. */              treated as literals. */
3972    
3973            if ((options & PCRE_EXTRA) != 0)              default:
3974              {              if ((options & PCRE_EXTRA) != 0)
3975              *errorcodeptr = ERR7;                {
3976              goto FAILED;                *errorcodeptr = ERR7;
3977                  goto FAILED;
3978                  }
3979                class_charcount -= 2;  /* Undo the default count from above */
3980                c = *ptr;              /* Get the final character and fall through */
3981                break;
3982              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3983            }            }
3984    
3985          /* 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 3377  for (;; ptr++) Line 4049  for (;; ptr++)
4049            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
4050            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
4051    
4052            /* \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 */  
4053    
4054            if (d < 0)            if (d < 0)
4055              {              {
4056              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  
4057                {                {
4058                ptr = oldptr;                ptr = oldptr;
4059                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3550  for (;; ptr++) Line 4219  for (;; ptr++)
4219          }          }
4220        }        }
4221    
4222      /* 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.
4223        If we are at the end of an internal nested string, revert to the outer
4224        string. */
4225    
4226        while (((c = *(++ptr)) != 0 ||
4227               (nestptr != NULL &&
4228                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4229               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4230    
4231      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
4232    
4233      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4234        {        {
4235        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4236        goto FAILED;        goto FAILED;
4237        }        }
4238    
   
 /* 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  
   
   
4239      /* 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
4240      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
4241      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 3586  we set the flag only if there is a liter Line 4243  we set the flag only if there is a liter
4243    
4244      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
4245      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4246      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4247      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4248    
4249      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
4250      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.
4251      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
4252      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
4253      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
4254      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4255    
4256  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4257      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3605  we set the flag only if there is a liter Line 4262  we set the flag only if there is a liter
4262        {        {
4263        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4264    
4265        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4266    
4267        if (negate_class)        if (negate_class)
4268          {          {
4269          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4270          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4271          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4272          *code++ = class_lastchar;          *code++ = class_lastchar;
4273          break;          break;
4274          }          }
# Line 3642  we set the flag only if there is a liter Line 4299  we set the flag only if there is a liter
4299    
4300      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4301      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4302      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
4303      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
4304      (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
4305      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
4306        actual compiled code. */
4307    
4308  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4309      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4310        {        {
4311        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4312        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3674  we set the flag only if there is a liter Line 4332  we set the flag only if there is a liter
4332        }        }
4333  #endif  #endif
4334    
4335      /* 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
4336      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
4337      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
4338      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4339        negating it if necessary. */
4340    
4341      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4342      if (negate_class)      if (negate_class)
# Line 3737  we set the flag only if there is a liter Line 4396  we set the flag only if there is a liter
4396      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4397      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4398    
4399      /* 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
4400      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4401    
4402      tempcode = previous;      tempcode = previous;
4403    
# Line 3761  we set the flag only if there is a liter Line 4420  we set the flag only if there is a liter
4420        }        }
4421      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4422    
4423        /* If previous was a recursion call, wrap it in atomic brackets so that
4424        previous becomes the atomic group. All recursions were so wrapped in the
4425        past, but it no longer happens for non-repeated recursions. In fact, the
4426        repeated ones could be re-implemented independently so as not to need this,
4427        but for the moment we rely on the code for repeating groups. */
4428    
4429        if (*previous == OP_RECURSE)
4430          {
4431          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4432          *previous = OP_ONCE;
4433          PUT(previous, 1, 2 + 2*LINK_SIZE);
4434          previous[2 + 2*LINK_SIZE] = OP_KET;
4435          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4436          code += 2 + 2 * LINK_SIZE;
4437          length_prevgroup = 3 + 3*LINK_SIZE;
4438    
4439          /* When actually compiling, we need to check whether this was a forward
4440          reference, and if so, adjust the offset. */
4441    
4442          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4443            {
4444            int offset = GET(cd->hwm, -LINK_SIZE);
4445            if (offset == previous + 1 - cd->start_code)
4446              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4447            }
4448          }
4449    
4450        /* Now handle repetition for the different types of item. */
4451    
4452      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4453      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
4454      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
4455      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
4456      instead.  */      instead.  */
4457    
4458      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4459        {        {
4460          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4461    
4462        /* 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
4463        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
4464        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 3801  we set the flag only if there is a liter Line 4491  we set the flag only if there is a liter
4491    
4492        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4493            repeat_max < 0 &&            repeat_max < 0 &&
4494            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4495          {          {
4496          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4497          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3814  we set the flag only if there is a liter Line 4503  we set the flag only if there is a liter
4503      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4504      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-
4505      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4506      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
4507      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4508    
4509      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4510        {        {
4511        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4512        c = previous[1];        c = previous[1];
4513        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4514            repeat_max < 0 &&            repeat_max < 0 &&
4515            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4516          {          {
4517          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4518          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3847  we set the flag only if there is a liter Line 4536  we set the flag only if there is a liter
4536    
4537        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4538            repeat_max < 0 &&            repeat_max < 0 &&
4539            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4540          {          {
4541          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4542          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3869  we set the flag only if there is a liter Line 4558  we set the flag only if there is a liter
4558    
4559        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4560    
4561          /*--------------------------------------------------------------------*/
4562          /* This code is obsolete from release 8.00; the restriction was finally
4563          removed: */
4564    
4565        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4566        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4567    
4568        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4569          /*--------------------------------------------------------------------*/
4570    
4571        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4572    
# Line 4011  we set the flag only if there is a liter Line 4705  we set the flag only if there is a liter
4705  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4706               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4707  #endif  #endif
4708               *previous == OP_REF)               *previous == OP_REF ||
4709                 *previous == OP_REFI)
4710        {        {
4711        if (repeat_max == 0)        if (repeat_max == 0)
4712          {          {
# Line 4019  we set the flag only if there is a liter Line 4714  we set the flag only if there is a liter
4714          goto END_REPEAT;          goto END_REPEAT;
4715          }          }
4716    
4717          /*--------------------------------------------------------------------*/
4718          /* This code is obsolete from release 8.00; the restriction was finally
4719          removed: */
4720    
4721        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4722        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4723    
4724        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4725          /*--------------------------------------------------------------------*/
4726    
4727        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4728          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 4040  we set the flag only if there is a liter Line 4740  we set the flag only if there is a liter
4740        }        }
4741    
4742      /* 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
4743      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4744        opcodes such as BRA and CBRA, as this is the place where they get converted
4745        into the more special varieties such as BRAPOS and SBRA. A test for >=
4746        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4747        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4748        repetition of assertions, but now it does, for Perl compatibility. */
4749    
4750      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4751        {        {
4752        register int i;        register int i;
4753        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4754        uschar *bralink = NULL;        uschar *bralink = NULL;
4755          uschar *brazeroptr = NULL;
4756    
4757        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4758          we just ignore the repeat. */
4759    
4760        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4761          {          goto END_REPEAT;
4762          *errorcodeptr = ERR55;  
4763          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4764          }        use of repetition is in cases when the assertion is optional. Therefore,
4765          if the minimum is greater than zero, just ignore the repeat. If the
4766          maximum is not not zero or one, set it to 1. */
4767    
4768        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4769        by scanning through from the start, and compute the offset back to it          {
4770        from the current code pointer. There may be an OP_OPT setting following          if (repeat_min > 0) goto END_REPEAT;
4771        the final KET, so we can't find the end just by going back from the code          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
       pointer. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = code - ket;  
4772          }          }
4773    
4774        /* 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 4089  we set the flag only if there is a liter Line 4789  we set the flag only if there is a liter
4789          **   goto END_REPEAT;          **   goto END_REPEAT;
4790          **   }          **   }
4791    
4792          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
4793          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
4794          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
4795          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4796            selectively.
4797    
4798          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
4799          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 4112  we set the flag only if there is a liter Line 4813  we set the flag only if there is a liter
4813              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4814              goto END_REPEAT;              goto END_REPEAT;
4815              }              }
4816              brazeroptr = previous;    /* Save for possessive optimizing */
4817            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4818            }            }
4819    
# Line 4136  we set the flag only if there is a liter Line 4838  we set the flag only if there is a liter
4838            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4839            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4840    
4841            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4842            bralink = previous;            bralink = previous;
4843            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4844            }            }
# Line 4157  we set the flag only if there is a liter Line 4859  we set the flag only if there is a liter
4859            {            {
4860            /* In the pre-compile phase, we don't actually do the replication. We            /* In the pre-compile phase, we don't actually do the replication. We
4861            just adjust the length as if we had. Do some paranoid checks for            just adjust the length as if we had. Do some paranoid checks for
4862            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4863              integer type when available, otherwise double. */
4864    
4865            if (lengthptr != NULL)            if (lengthptr != NULL)
4866              {              {
4867              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4868              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4869                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4870                        (INT64_OR_DOUBLE)INT_MAX ||
4871                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4872                {                {
4873                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 4209  we set the flag only if there is a liter Line 4913  we set the flag only if there is a liter
4913          just adjust the length as if we had. For each repetition we must add 1          just adjust the length as if we had. For each repetition we must add 1
4914          to the length for BRAZERO and for all but the last repetition we must          to the length for BRAZERO and for all but the last repetition we must
4915          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4916          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4917            a 64-bit integer type when available, otherwise double. */
4918    
4919          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4920            {            {
4921            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4922                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4923            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4924                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4925                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4926                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4927              {              {
4928              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4242  we set the flag only if there is a liter Line 4947  we set the flag only if there is a liter
4947              {              {
4948              int offset;              int offset;
4949              *code++ = OP_BRA;              *code++ = OP_BRA;
4950              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4951              bralink = code;              bralink = code;
4952              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4953              }              }
# Line 4263  we set the flag only if there is a liter Line 4968  we set the flag only if there is a liter
4968          while (bralink != NULL)          while (bralink != NULL)
4969            {            {
4970            int oldlinkoffset;            int oldlinkoffset;
4971            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4972            uschar *bra = code - offset;            uschar *bra = code - offset;
4973            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4974            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4273  we set the flag only if there is a liter Line 4978  we set the flag only if there is a liter
4978            }            }
4979          }          }
4980    
4981        /* 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
4982        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4983        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
4984        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4985          deal with possessive ONCEs specially.
4986        Then, when we are doing the actual compile phase, check to see whether  
4987        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
4988          whether this group is one that could match an empty string. If so,
4989        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
4990        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
4991        atomic groups at runtime, but in a different way.] */        groups at runtime, but in a different way.]
4992    
4993          Then, if the quantifier was possessive and the bracket is not a
4994          conditional, we convert the BRA code to the POS form, and the KET code to
4995          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
4996          subpattern at both the start and at the end.) The use of special opcodes
4997          makes it possible to reduce greatly the stack usage in pcre_exec(). If
4998          the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
4999    
5000          Then, if the minimum number of matches is 1 or 0, cancel the possessive
5001          flag so that the default action below, of wrapping everything inside
5002          atomic brackets, does not happen. When the minimum is greater than 1,
5003          there will be earlier copies of the group, and so we still have to wrap
5004          the whole thing. */
5005    
5006        else        else
5007          {          {
5008          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
5009          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
5010          *ketcode = OP_KETRMAX + repeat_type;  
5011          if (lengthptr == NULL && *bracode != OP_ONCE)          /* Convert possessive ONCE brackets to non-capturing */
5012    
5013            if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5014                possessive_quantifier) *bracode = OP_BRA;
5015    
5016            /* For non-possessive ONCE brackets, all we need to do is to
5017            set the KET. */
5018    
5019            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5020              *ketcode = OP_KETRMAX + repeat_type;
5021    
5022            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5023            converted to non-capturing above). */
5024    
5025            else
5026            {            {
5027            uschar *scode = bracode;            /* In the compile phase, check for empty string matching. */
5028            do  
5029              if (lengthptr == NULL)
5030              {              {
5031              if (could_be_empty_branch(scode, ketcode, utf8))              uschar *scode = bracode;
5032                do
5033                {                {
5034                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
5035                break;                  {
5036                    *bracode += OP_SBRA - OP_BRA;
5037                    break;
5038                    }
5039                  scode += GET(scode, 1);
5040                }                }
5041              scode += GET(scode, 1);              while (*scode == OP_ALT);
5042              }              }
5043            while (*scode == OP_ALT);  
5044              /* Handle possessive quantifiers. */
5045    
5046              if (possessive_quantifier)
5047                {
5048                /* For COND brackets, we wrap the whole thing in a possessively
5049                repeated non-capturing bracket, because we have not invented POS
5050                versions of the COND opcodes. Because we are moving code along, we
5051                must ensure that any pending recursive references are updated. */
5052    
5053                if (*bracode == OP_COND || *bracode == OP_SCOND)
5054                  {
5055                  int nlen = (int)(code - bracode);
5056                  *code = OP_END;
5057                  adjust_recurse(bracode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5058                  memmove(bracode + 1+LINK_SIZE, bracode, nlen);
5059                  code += 1 + LINK_SIZE;
5060                  nlen += 1 + LINK_SIZE;
5061                  *bracode = OP_BRAPOS;
5062                  *code++ = OP_KETRPOS;
5063                  PUTINC(code, 0, nlen);
5064                  PUT(bracode, 1, nlen);
5065                  }
5066    
5067                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5068    
5069                else
5070                  {
5071                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5072                  *ketcode = OP_KETRPOS;
5073                  }
5074    
5075                /* If the minimum is zero, mark it as possessive, then unset the
5076                possessive flag when the minimum is 0 or 1. */
5077    
5078                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5079                if (repeat_min < 2) possessive_quantifier = FALSE;
5080                }
5081    
5082              /* Non-possessive quantifier */
5083    
5084              else *ketcode = OP_KETRMAX + repeat_type;
5085            }            }
5086          }          }
5087        }        }
# Line 4322  we set the flag only if there is a liter Line 5102  we set the flag only if there is a liter
5102        }        }
5103    
5104      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
5105      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
5106      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
5107      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
5108      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
5109      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
5110      tempcode, not at previous, which might be the first part of a string whose  
5111      (former) last char we repeated.      Some (but not all) possessively repeated subpatterns have already been
5112        completely handled in the code just above. For them, possessive_quantifier
5113        is always FALSE at this stage.
5114    
5115        Note that the repeated item starts at tempcode, not at previous, which
5116        might be the first part of a string whose (former) last char we repeated.
5117    
5118      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
5119      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 4337  we set the flag only if there is a liter Line 5122  we set the flag only if there is a liter
5122      if (possessive_quantifier)      if (possessive_quantifier)
5123        {        {
5124        int len;        int len;
5125        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
5126            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
5127          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
5128            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
5129               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
5130        len = code - tempcode;        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
5131            {
5132            tempcode += _pcre_OP_lengths[*tempcode];
5133    #ifdef SUPPORT_UTF8
5134            if (utf8 && tempcode[-1] >= 0xc0)
5135              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
5136    #endif
5137            }
5138    
5139          len = (int)(code - tempcode);
5140        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
5141          {          {
5142          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4350  we set the flag only if there is a liter Line 5144  we set the flag only if there is a liter
5144          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
5145          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
5146    
5147          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
5148          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
5149          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
5150          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
5151    
5152          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
5153          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
5154          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
5155          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
5156    
5157            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
5158            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
5159            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
5160            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
5161    
5162            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
5163            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
5164            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
5165            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
5166    
5167            /* Because we are moving code along, we must ensure that any
5168            pending recursive references are updated. */
5169    
5170          default:          default:
5171            *code = OP_END;
5172            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5173          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
5174          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
5175          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4396  we set the flag only if there is a liter Line 5205  we set the flag only if there is a liter
5205    
5206      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
5207    
5208      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
5209             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
5210        {        {
5211        int i, namelen;        int i, namelen;
5212          int arglen = 0;
5213        const char *vn = verbnames;        const char *vn = verbnames;
5214        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
5215          const uschar *arg = NULL;
5216        previous = NULL;        previous = NULL;
5217        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5218          namelen = (int)(ptr - name);
5219    
5220          /* It appears that Perl allows any characters whatsoever, other than
5221          a closing parenthesis, to appear in arguments, so we no longer insist on
5222          letters, digits, and underscores. */
5223    
5224        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5225          {          {
5226          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
5227          goto FAILED;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5228            arglen = (int)(ptr - arg);
5229          }          }
5230    
5231        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
5232          {          {
5233          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
5234          goto FAILED;          goto FAILED;
5235          }          }
5236        namelen = ptr - name;  
5237          /* Scan the table of verb names */
5238    
5239        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
5240          {          {
5241          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5242              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5243            {            {
5244            *code = verbs[i].op;            /* Check for open captures before ACCEPT and convert it to
5245            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;            ASSERT_ACCEPT if in an assertion. */
5246            break;  
5247              if (verbs[i].op == OP_ACCEPT)
5248                {
5249                open_capitem *oc;
5250                if (arglen != 0)
5251                  {
5252                  *errorcodeptr = ERR59;
5253                  goto FAILED;
5254                  }
5255                cd->had_accept = TRUE;
5256                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5257                  {
5258                  *code++ = OP_CLOSE;
5259                  PUT2INC(code, 0, oc->number);
5260                  }
5261                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5262    
5263                /* Do not set firstbyte after *ACCEPT */
5264                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5265                }
5266    
5267              /* Handle other cases with/without an argument */
5268    
5269              else if (arglen == 0)
5270                {
5271                if (verbs[i].op < 0)   /* Argument is mandatory */
5272                  {
5273                  *errorcodeptr = ERR66;
5274                  goto FAILED;
5275                  }
5276                *code = verbs[i].op;
5277                if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
5278                }
5279    
5280              else
5281                {
5282                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
5283                  {
5284                  *errorcodeptr = ERR59;
5285                  goto FAILED;
5286                  }
5287                *code = verbs[i].op_arg;
5288                if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
5289                *code++ = arglen;
5290                memcpy(code, arg, arglen);
5291                code += arglen;
5292                *code++ = 0;
5293                }
5294    
5295              break;  /* Found verb, exit loop */
5296            }            }
5297    
5298          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
5299          }          }
5300        if (i < verbcount) continue;  
5301        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
5302          *errorcodeptr = ERR60;          /* Verb not recognized */
5303        goto FAILED;        goto FAILED;
5304        }        }
5305    
# Line 4545  we set the flag only if there is a liter Line 5418  we set the flag only if there is a liter
5418                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5419            ptr++;            ptr++;
5420            }            }
5421          namelen = ptr - name;          namelen = (int)(ptr - name);
5422    
5423          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
5424              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
# Line 4582  we set the flag only if there is a liter Line 5455  we set the flag only if there is a liter
5455            }            }
5456    
5457          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
5458          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
5459            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
5460            except they record that the reference was originally to a name. The
5461            information is used to check duplicate names. */
5462    
5463          slot = cd->name_table;          slot = cd->name_table;
5464          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4597  we set the flag only if there is a liter Line 5473  we set the flag only if there is a liter
5473            {            {
5474            recno = GET2(slot, 0);            recno = GET2(slot, 0);
5475            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5476              code[1+LINK_SIZE]++;
5477            }            }
5478    
5479          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5480    
5481          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5482                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5483            {            {
5484            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5485              code[1+LINK_SIZE]++;
5486            }            }
5487    
5488          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4669  we set the flag only if there is a liter Line 5547  we set the flag only if there is a liter
5547          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5548          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5549          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5550            cd->assert_depth += 1;
5551          ptr++;          ptr++;
5552          break;          break;
5553    
# Line 4683  we set the flag only if there is a liter Line 5562  we set the flag only if there is a liter
5562            continue;            continue;
5563            }            }
5564          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5565            cd->assert_depth += 1;
5566          break;          break;
5567    
5568    
# Line 4692  we set the flag only if there is a liter Line 5572  we set the flag only if there is a liter
5572            {            {
5573            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5574            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5575              cd->assert_depth += 1;
5576            ptr += 2;            ptr += 2;
5577            break;            break;
5578    
5579            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5580            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5581              cd->assert_depth += 1;
5582            ptr += 2;            ptr += 2;
5583            break;            break;
5584    
# Line 4736  we set the flag only if there is a liter Line 5618  we set the flag only if there is a liter
5618              goto FAILED;              goto FAILED;
5619              }              }
5620            *code++ = n;            *code++ = n;
5621            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5622            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5623            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5624            }            }
5625          previous = NULL;          previous = NULL;
# Line 4770  we set the flag only if there is a liter Line 5652  we set the flag only if there is a liter
5652            name = ++ptr;            name = ++ptr;
5653    
5654            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5655            namelen = ptr - name;            namelen = (int)(ptr - name);
5656    
5657            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5658    
# Line 4797  we set the flag only if there is a liter Line 5679  we set the flag only if there is a liter
5679                }                }
5680              }              }
5681    
5682            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5683              alphabetical order. Duplicate names for different numbers are
5684              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5685              number are always OK. (An existing number can be re-used if (?|
5686              appears in the pattern.) In either event, a duplicate name results in
5687              a duplicate entry in the table, even if the number is the same. This
5688              is because the number of names, and hence the table size, is computed
5689              in the pre-compile, and it affects various numbers and pointers which
5690              would all have to be modified, and the compiled code moved down, if
5691              duplicates with the same number were omitted from the table. This
5692              doesn't seem worth the hassle. However, *different* names for the
5693              same number are not permitted. */
5694    
5695            else            else
5696              {              {
5697                BOOL dupname = FALSE;
5698              slot = cd->name_table;              slot = cd->name_table;
5699    
5700              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5701                {                {
5702                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4809  we set the flag only if there is a liter Line 5704  we set the flag only if there is a liter
5704                  {                  {
5705                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5706                    {                    {
5707                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5708                          (options & PCRE_DUPNAMES) == 0)
5709                      {                      {
5710                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5711                      goto FAILED;                      goto FAILED;
5712                      }                      }
5713                      else dupname = TRUE;
5714                    }                    }
5715                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5716                  }                  }
5717    
5718                  /* Make space in the table and break the loop for an earlier
5719                  name. For a duplicate or later name, carry on. We do this for
5720                  duplicates so that in the simple case (when ?(| is not used) they
5721                  are in order of their numbers. */
5722    
5723                if (crc < 0)                if (crc < 0)
5724                  {                  {
5725                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5726                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5727                  break;                  break;
5728                  }                  }
5729    
5730                  /* Continue the loop for a later or duplicate name */
5731    
5732                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5733                }                }
5734    
5735                /* For non-duplicate names, check for a duplicate number before
5736                adding the new name. */
5737    
5738                if (!dupname)
5739                  {
5740                  uschar *cslot = cd->name_table;
5741                  for (i = 0; i < cd->names_found; i++)
5742                    {
5743                    if (cslot != slot)
5744                      {
5745                      if (GET2(cslot, 0) == cd->bracount + 1)
5746                        {
5747                        *errorcodeptr = ERR65;
5748                        goto FAILED;
5749                        }
5750                      }
5751                    else i--;
5752                    cslot += cd->name_entry_size;
5753                    }
5754                  }
5755    
5756              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5757              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5758              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5759              }              }
5760            }            }
5761    
5762          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5763            encountered. */
5764    
         ptr++;                    /* Move past > or ' */  
5765          cd->names_found++;          cd->names_found++;
5766            ptr++;                    /* Move past > or ' */
5767          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5768    
5769    
# Line 4854  we set the flag only if there is a liter Line 5782  we set the flag only if there is a liter
5782          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5783          name = ++ptr;          name = ++ptr;
5784          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5785          namelen = ptr - name;          namelen = (int)(ptr - name);
5786    
5787          /* 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
5788          reference number. */          a dummy reference number, because it was not used in the first pass.
5789            However, with the change of recursive back references to be atomic,
5790            we have to look for the number so that this state can be identified, as
5791            otherwise the incorrect length is computed. If it's not a backwards
5792            reference, the dummy number will do. */
5793    
5794          if (lengthptr != NULL)          if (lengthptr != NULL)
5795            {            {
5796              const uschar *temp;
5797    
5798            if (namelen == 0)            if (namelen == 0)
5799              {              {
5800              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 4876  we set the flag only if there is a liter Line 5810  we set the flag only if there is a liter
5810              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5811              goto FAILED;              goto FAILED;
5812              }              }
5813            recno = 0;  
5814              /* The name table does not exist in the first pass, so we cannot
5815              do a simple search as in the code below. Instead, we have to scan the
5816              pattern to find the number. It is important that we scan it only as
5817              far as we have got because the syntax of named subpatterns has not
5818              been checked for the rest of the pattern, and find_parens() assumes
5819              correct syntax. In any case, it's a waste of resources to scan
5820              further. We stop the scan at the current point by temporarily
5821              adjusting the value of cd->endpattern. */
5822    
5823              temp = cd->end_pattern;
5824              cd->end_pattern = ptr;
5825              recno = find_parens(cd, name, namelen,
5826                (options & PCRE_EXTENDED) != 0, utf8);
5827              cd->end_pattern = temp;
5828              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5829            }            }
5830    
5831          /* 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 4901  we set the flag only if there is a liter Line 5850  we set the flag only if there is a liter
5850              }              }
5851            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5852                      find_parens(cd, name, namelen,                      find_parens(cd, name, namelen,
5853                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5854              {              {
5855              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5856              goto FAILED;              goto FAILED;
# Line 5004  we set the flag only if there is a liter Line 5953  we set the flag only if there is a liter
5953            if (lengthptr == NULL)            if (lengthptr == NULL)
5954              {              {
5955              *code = OP_END;              *code = OP_END;
5956              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5957                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5958    
5959              /* Forward reference */              /* Forward reference */
5960    
5961              if (called == NULL)              if (called == NULL)
5962                {                {
5963                if (find_parens(cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5964                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5965                  {                  {
5966                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5967                  goto FAILED;                  goto FAILED;
5968                  }                  }
5969    
5970                  /* Fudge the value of "called" so that when it is inserted as an
5971                  offset below, what it actually inserted is the reference number
5972                  of the group. Then remember the forward reference. */
5973    
5974                called = cd->start_code + recno;                called = cd->start_code + recno;
5975                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5976                }                }
5977    
5978              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
5979              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
5980              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
5981                must not, however, do this check if we are in a conditional
5982                subpattern because the condition might be testing for recursion in
5983                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
5984                Forever loops are also detected at runtime, so those that occur in
5985                conditional subpatterns will be picked up then. */
5986    
5987              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
5988                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5989                {                {
5990                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5991                goto FAILED;                goto FAILED;
5992                }                }
5993              }              }
5994    
5995            /* 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;  
5996    
5997            *code = OP_RECURSE;            *code = OP_RECURSE;
5998            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);  
5999            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
   
           length_prevgroup = 3 + 3*LINK_SIZE;  
6000            }            }
6001    
6002          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5109  we set the flag only if there is a liter Line 6057  we set the flag only if there is a liter
6057          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
6058          both phases.          both phases.
6059    
6060          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
6061          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. */  
6062    
6063          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
6064            {            {
# Line 5120  we set the flag only if there is a liter Line 6067  we set the flag only if there is a liter
6067              {              {
6068              cd->external_options = newoptions;              cd->external_options = newoptions;
6069              }              }
6070           else            else
6071              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
6072              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
6073              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
6074              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
6075              }              }
6076    
6077            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
6078            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). */  
6079    
6080            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
6081            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5152  we set the flag only if there is a liter Line 6092  we set the flag only if there is a liter
6092          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
6093        }       /* End of (? handling */        }       /* End of (? handling */
6094    
6095      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
6096      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
6097      brackets. */      brackets. */
6098    
6099      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 5171  we set the flag only if there is a liter Line 6111  we set the flag only if there is a liter
6111        skipbytes = 2;        skipbytes = 2;
6112        }        }
6113    
6114      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
6115      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
6116      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
6117      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. */  
6118    
6119      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                      /* For handling repetition */
6120      *code = bravalue;      *code = bravalue;
6121      tempcode = code;      tempcode = code;
6122      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
6123      length_prevgroup = 0;              /* Initialize for pre-compile phase */      tempbracount = cd->bracount;          /* Save value before bracket */
6124        length_prevgroup = 0;                 /* Initialize for pre-compile phase */
6125    
6126      if (!compile_regex(      if (!compile_regex(
6127           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
6128           options & PCRE_IMS,           /* The previous ims option state */           &tempcode,                       /* Where to put code (updated) */
6129           &tempcode,                    /* Where to put code (updated) */           &ptr,                            /* Input pointer (updated) */
6130           &ptr,                         /* Input pointer (updated) */           errorcodeptr,                    /* Where to put an error message */
          errorcodeptr,                 /* Where to put an error message */  
6131           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
6132            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
6133           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
6134           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
6135           &subfirstbyte,                /* For possible first char */           cond_depth +
6136           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
6137           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
6138           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
6139           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
6140             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
6141             (lengthptr == NULL)? NULL :      /* Actual compile phase */
6142               &length_prevgroup              /* Pre-compile phase */
6143           ))           ))
6144        goto FAILED;        goto FAILED;
6145    
6146        /* If this was an atomic group and there are no capturing groups within it,
6147        generate OP_ONCE_NC instead of OP_ONCE. */
6148    
6149        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
6150          *code = OP_ONCE_NC;
6151    
6152        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
6153          cd->assert_depth -= 1;
6154    
6155      /* 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
6156      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.
6157      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
6158    
6159      /* 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
6160      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
6161      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
6162      not be available. */      not be available. */
# Line 5272  we set the flag only if there is a liter Line 6221  we set the flag only if there is a liter
6221          goto FAILED;          goto FAILED;
6222          }          }
6223        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
6224        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
6225        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
6226        *code++ = OP_KET;        *code++ = OP_KET;
6227        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5345  we set the flag only if there is a liter Line 6294  we set the flag only if there is a liter
6294    
6295      /* ===================================================================*/      /* ===================================================================*/
6296      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
6297      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
6298      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
6299      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
6300      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
6301      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
6302        ever created. */
6303    
6304      case CHAR_BACKSLASH:      case CHAR_BACKSLASH:
6305      tempptr = ptr;      tempptr = ptr;
# Line 5439  we set the flag only if there is a liter Line 6389  we set the flag only if there is a liter
6389          }          }
6390    
6391        /* \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).
6392        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6393    
6394        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6395          {          {
6396            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6397              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6398              {
6399              *errorcodeptr = ERR69;
6400              break;
6401              }
6402          is_recurse = FALSE;          is_recurse = FALSE;
6403          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6404            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
# Line 5457  we set the flag only if there is a liter Line 6412  we set the flag only if there is a liter
6412    
6413        if (-c >= ESC_REF)        if (-c >= ESC_REF)
6414          {          {
6415            open_capitem *oc;
6416          recno = -c - ESC_REF;          recno = -c - ESC_REF;
6417    
6418          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6419          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6420          previous = code;          previous = code;
6421          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6422          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6423          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6424          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
6425    
6426            /* Check to see if this back reference is recursive, that it, it
6427            is inside the group that it references. A flag is set so that the
6428            group can be made atomic. */
6429    
6430            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6431              {
6432              if (oc->number == recno)
6433                {
6434                oc->flag = TRUE;
6435                break;
6436                }
6437              }
6438          }          }
6439    
6440        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5495  we set the flag only if there is a liter Line 6464  we set the flag only if there is a liter
6464  #endif  #endif
6465    
6466        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
6467        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
6468          situation when PCRE_UCP is not set. When it *is* set, we substitute
6469          Unicode property tests. */
6470    
6471        else        else
6472          {          {
6473          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
6474          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
6475              {
6476              nestptr = ptr + 1;                   /* Where to resume */
6477              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
6478              }
6479            else
6480    #endif
6481              {
6482              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6483              *code++ = -c;
6484              }
6485          }          }
6486        continue;        continue;
6487        }        }
# Line 5545  we set the flag only if there is a liter Line 6526  we set the flag only if there is a liter
6526    
6527      ONE_CHAR:      ONE_CHAR:
6528      previous = code;      previous = code;
6529      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6530      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6531    
6532      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 5574  we set the flag only if there is a liter Line 6555  we set the flag only if there is a liter
6555        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6556        }        }
6557    
6558      /* 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
6559      1 or the matching is caseful. */      1 or the matching is caseful. */
6560    
6561      else      else
# Line 5609  return FALSE; Line 6590  return FALSE;
6590  /* 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
6591  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
6592  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.  
   
6593  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
6594  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
6595  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6596    
6597  Arguments:  Arguments:
6598    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  
6599    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6600    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6601    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
6602    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6603    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6604    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6605      cond_depth     depth of nesting for conditional subpatterns
6606    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6607    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6608    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 5638  Returns:         TRUE on success Line 6614  Returns:         TRUE on success
6614  */  */
6615    
6616  static BOOL  static BOOL
6617  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6618    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6619    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6620    int *lengthptr)    compile_data *cd, int *lengthptr)
6621  {  {
6622  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6623  uschar *code = *codeptr;  uschar *code = *codeptr;
6624  uschar *last_branch = code;  uschar *last_branch = code;
6625  uschar *start_bracket = code;  uschar *start_bracket = code;
6626  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
6627    open_capitem capitem;
6628    int capnumber = 0;
6629  int firstbyte, reqbyte;  int firstbyte, reqbyte;
6630  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
6631  int length;  int length;
# Line 5656  int max_bracount; Line 6634  int max_bracount;
6634  branch_chain bc;  branch_chain bc;
6635    
6636  bc.outer = bcptr;  bc.outer = bcptr;
6637  bc.current = code;  bc.current_branch = code;
6638    
6639  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
6640    
# Line 5674  the code that abstracts option settings Line 6652  the code that abstracts option settings
6652  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the
6653  pre-compile phase to find out whether anything has yet been compiled or not. */  pre-compile phase to find out whether anything has yet been compiled or not. */
6654    
6655    /* If this is a capturing subpattern, add to the chain of open capturing items
6656    so that we can detect them if (*ACCEPT) is encountered. This is also used to
6657    detect groups that contain recursive back references to themselves. Note that
6658    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6659    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6660    
6661    if (*code == OP_CBRA)
6662      {
6663      capnumber = GET2(code, 1 + LINK_SIZE);
6664      capitem.number = capnumber;
6665      capitem.next = cd->open_caps;
6666      capitem.flag = FALSE;
6667      cd->open_caps = &capitem;
6668      }
6669    
6670  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
6671    
6672  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5689  for (;;) Line 6682  for (;;)
6682    
6683    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6684    
   /* 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;  
     }  
   
6685    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6686    
6687    if (lookbehind)    if (lookbehind)
# Line 5712  for (;;) Line 6696  for (;;)
6696    into the length. */    into the length. */
6697    
6698    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6699          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6700            (lengthptr == NULL)? NULL : &length))
6701      {      {
6702      *ptrptr = ptr;      *ptrptr = ptr;
6703      return FALSE;      return FALSE;
# Line 5768  for (;;) Line 6753  for (;;)
6753    
6754      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
6755      put the length into the OP_REVERSE item. Temporarily mark the end of the      put the length into the OP_REVERSE item. Temporarily mark the end of the
6756      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
6757        because there may be forward references that we can't check here. Set a
6758        flag to cause another lookbehind check at the end. Why not do it all at the
6759        end? Because common, erroneous checks are picked up here and the offset of
6760        the problem can be shown. */
6761    
6762      if (lookbehind)      if (lookbehind)
6763        {        {
6764        int fixed_length;        int fixed_length;
6765        *code = OP_END;        *code = OP_END;
6766        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6767            FALSE, cd);
6768        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6769        if (fixed_length < 0)        if (fixed_length == -3)
6770          {          {
6771          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          cd->check_lookbehind = TRUE;
6772            }
6773          else if (fixed_length < 0)
6774            {
6775            *errorcodeptr = (fixed_length == -2)? ERR36 :
6776                            (fixed_length == -4)? ERR70: ERR25;
6777          *ptrptr = ptr;          *ptrptr = ptr;
6778          return FALSE;          return FALSE;
6779          }          }
6780        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
6781        }        }
6782      }      }
6783    
# Line 5791  for (;;) Line 6786  for (;;)
6786    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
6787    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
6788    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
6789    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. */  
6790    
6791    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6792      {      {
6793      if (lengthptr == NULL)      if (lengthptr == NULL)
6794        {        {
6795        int branch_length = code - last_branch;        int branch_length = (int)(code - last_branch);
6796        do        do
6797          {          {
6798          int prev_length = GET(last_branch, 1);          int prev_length = GET(last_branch, 1);
# Line 5813  for (;;) Line 6806  for (;;)
6806      /* Fill in the ket */      /* Fill in the ket */
6807    
6808      *code = OP_KET;      *code = OP_KET;
6809      PUT(code, 1, code - start_bracket);      PUT(code, 1, (int)(code - start_bracket));
6810      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6811    
6812      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6813        recursive back references. If so, we must wrap it in atomic brackets.
6814        In any event, remove the block from the chain. */
6815    
6816      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)      if (capnumber > 0)
6817        {        {
6818        *code++ = OP_OPT;        if (cd->open_caps->flag)
6819        *code++ = oldims;          {
6820        length += 2;          memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6821              code - start_bracket);
6822            *start_bracket = OP_ONCE;
6823            code += 1 + LINK_SIZE;
6824            PUT(start_bracket, 1, (int)(code - start_bracket));
6825            *code = OP_KET;
6826            PUT(code, 1, (int)(code - start_bracket));
6827            code += 1 + LINK_SIZE;
6828            length += 2 + 2*LINK_SIZE;
6829            }
6830          cd->open_caps = cd->open_caps->next;
6831        }        }
6832    
6833      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
# Line 5864  for (;;) Line 6869  for (;;)
6869    else    else
6870      {      {
6871      *code = OP_ALT;      *code = OP_ALT;
6872      PUT(code, 1, code - last_branch);      PUT(code, 1, (int)(code - last_branch));
6873      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6874      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6875      }      }
6876    
# Line 5884  for (;;) Line 6889  for (;;)
6889  /* 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
6890  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
6891  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
6892  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
6893  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6894    
6895  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.
6896  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 5906  of the more common cases more precisely. Line 6911  of the more common cases more precisely.
6911    
6912  Arguments:  Arguments:
6913    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6914    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
6915                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6916                    the less precise approach                    the less precise approach
# Line 5916  Returns:     TRUE or FALSE Line 6920  Returns:     TRUE or FALSE
6920  */  */
6921    
6922  static BOOL  static BOOL
6923  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6924    unsigned int backref_map)    unsigned int backref_map)
6925  {  {
6926  do {  do {
6927     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6928       options, PCRE_MULTILINE, FALSE);       FALSE);
6929     register int op = *scode;     register int op = *scode;
6930