/[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 500 by ph10, Sat Mar 6 19:00:29 2010 UTC revision 762 by ph10, Tue Nov 22 13:36:51 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2010 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 92  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
# Line 119  static const short int escapes[] = { Line 124  static const short int escapes[] = {
124       -ESC_H,                  0,       -ESC_H,                  0,
125       0,                       -ESC_K,       0,                       -ESC_K,
126       0,                       0,       0,                       0,
127       0,                       0,       -ESC_N,                  0,
128       -ESC_P,                  -ESC_Q,       -ESC_P,                  -ESC_Q,
129       -ESC_R,                  -ESC_S,       -ESC_R,                  -ESC_S,
130       0,                       0,       0,                       0,
# Line 166  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 183  string is built from string macros so th Line 188  string is built from string macros so th
188  platforms. */  platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    int   len;    int   len;                 /* Length of verb name */
192    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static const char verbnames[] =  static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199    STRING_ACCEPT0    STRING_ACCEPT0
200    STRING_COMMIT0    STRING_COMMIT0
201    STRING_F0    STRING_F0
# Line 197  static const char verbnames[] = Line 205  static const char verbnames[] =
205    STRING_THEN;    STRING_THEN;
206    
207  static const verbitem verbs[] = {  static const verbitem verbs[] = {
208    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
209    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
210    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
212    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
213    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
214    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
# Line 251  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 263  the number of relocations needed when a Line 320  the number of relocations needed when a
320  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
321  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
323  because these strings are used only when there is a compilation error.  because these strings are used only when there is a compilation error.
324    
325  Each substring ends with \0 to insert a null character. This includes the final  Each substring ends with \0 to insert a null character. This includes the final
326  substring, so that the whole string ends with \0\0, which can be detected when  substring, so that the whole string ends with \0\0, which can be detected when
327  counting through. */  counting through. */
328    
329  static const char error_texts[] =  static const char error_texts[] =
# Line 314  static const char error_texts[] = Line 371  static const char error_texts[] =
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255\0"    "number after (?C is > 255\0"
376    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
# Line 336  static const char error_texts[] = Line 393  static const char error_texts[] =
393    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
400    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401    /* 60 */    /* 60 */
402    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
403    "number is too big\0"    "number is too big\0"
# Line 348  static const char error_texts[] = Line 405  static const char error_texts[] =
405    "digit expected after (?+\0"    "digit expected after (?+\0"
406    "] is an invalid data character in JavaScript compatibility mode\0"    "] is an invalid data character in JavaScript compatibility mode\0"
407    /* 65 */    /* 65 */
408    "different names for subpatterns of the same number are not allowed\0";    "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413      /* 70 */
414      "internal error: unknown opcode in find_fixedlength()\0"
415      "\\N is not supported in a class\0"
416      "too many forward references\0"
417      ;
418    
419  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
420  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 484  static const unsigned char ebcdic_charta Line 550  static const unsigned char ebcdic_charta
550  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
551    
552  static BOOL  static BOOL
553    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
554      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
555    
556    
# Line 506  static const char * Line 572  static const char *
572  find_error_text(int n)  find_error_text(int n)
573  {  {
574  const char *s = error_texts;  const char *s = error_texts;
575  for (; n > 0; n--)  for (; n > 0; n--)
576    {    {
577    while (*s++ != 0) {};    while (*s++ != 0) {};
578    if (*s == 0) return "Error text not found (please report)";    if (*s == 0) return "Error text not found (please report)";
579    }    }
580  return s;  return s;
581  }  }
582    
583    
584  /*************************************************  /*************************************************
585    *            Check for counted repeat            *
586    *************************************************/
587    
588    /* This function is called when a '{' is encountered in a place where it might
589    start a quantifier. It looks ahead to see if it really is a quantifier or not.
590    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
591    where the ddds are digits.
592    
593    Arguments:
594      p         pointer to the first char after '{'
595    
596    Returns:    TRUE or FALSE
597    */
598    
599    static BOOL
600    is_counted_repeat(const uschar *p)
601    {
602    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
603    while ((digitab[*p] & ctype_digit) != 0) p++;
604    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
605    
606    if (*p++ != CHAR_COMMA) return FALSE;
607    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
608    
609    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
610    while ((digitab[*p] & ctype_digit) != 0) p++;
611    
612    return (*p == CHAR_RIGHT_CURLY_BRACKET);
613    }
614    
615    
616    
617    /*************************************************
618  *            Handle escapes                      *  *            Handle escapes                      *
619  *************************************************/  *************************************************/
620    
# Line 581  else Line 680  else
680    
681      case CHAR_l:      case CHAR_l:
682      case CHAR_L:      case CHAR_L:
683      case CHAR_N:      *errorcodeptr = ERR37;
684        break;
685    
686      case CHAR_u:      case CHAR_u:
687        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
688          {
689          /* In JavaScript, \u must be followed by four hexadecimal numbers.
690          Otherwise it is a lowercase u letter. */
691          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
692               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
693            {
694            c = 0;
695            for (i = 0; i < 4; ++i)
696              {
697              register int cc = *(++ptr);
698    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
699              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
700              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
701    #else           /* EBCDIC coding */
702              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
703              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
704    #endif
705              }
706            }
707          }
708        else
709          *errorcodeptr = ERR37;
710        break;
711    
712      case CHAR_U:      case CHAR_U:
713      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
714        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
715      break;      break;
716    
717      /* \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
718        class, \g must be followed by one of a number of specific things:
719    
720      (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
721      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 604  else Line 732  else
732      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
733    
734      case CHAR_g:      case CHAR_g:
735        if (isclass) break;
736      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
737        {        {
738        c = -ESC_g;        c = -ESC_g;
# Line 732  else Line 861  else
861      treated as a data character. */      treated as a data character. */
862    
863      case CHAR_x:      case CHAR_x:
864        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
865          {
866          /* In JavaScript, \x must be followed by two hexadecimal numbers.
867          Otherwise it is a lowercase x letter. */
868          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
869            {
870            c = 0;
871            for (i = 0; i < 2; ++i)
872              {
873              register int cc = *(++ptr);
874    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
875              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
876              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
877    #else           /* EBCDIC coding */
878              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
879              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
880    #endif
881              }
882            }
883          break;
884          }
885    
886      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
887        {        {
888        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
# Line 782  else Line 933  else
933      break;      break;
934    
935      /* 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.
936      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
937        coding is ASCII-specific, but then the whole concept of \cx is
938      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
939    
940      case CHAR_c:      case CHAR_c:
# Line 792  else Line 944  else
944        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
945        break;        break;
946        }        }
947    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
948  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
949          {
950          *errorcodeptr = ERR68;
951          break;
952          }
953      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
954      c ^= 0x40;      c ^= 0x40;
955  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
956      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
957      c ^= 0xC0;      c ^= 0xC0;
958  #endif  #endif
# Line 819  else Line 975  else
975      }      }
976    }    }
977    
978    /* Perl supports \N{name} for character names, as well as plain \N for "not
979    newline". PCRE does not support \N{name}. However, it does support
980    quantification such as \N{2,3}. */
981    
982    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
983         !is_counted_repeat(ptr+2))
984      *errorcodeptr = ERR37;
985    
986    /* If PCRE_UCP is set, we change the values for \d etc. */
987    
988    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
989      c -= (ESC_DU - ESC_D);
990    
991    /* Set the pointer to the final character before returning. */
992    
993  *ptrptr = ptr;  *ptrptr = ptr;
994  return c;  return c;
995  }  }
# Line 919  return -1; Line 1090  return -1;
1090    
1091    
1092  /*************************************************  /*************************************************
 *            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);  
 }  
   
   
   
 /*************************************************  
1093  *         Read repeat counts                     *  *         Read repeat counts                     *
1094  *************************************************/  *************************************************/
1095    
# Line 1027  top-level call starts at the beginning o Line 1165  top-level call starts at the beginning o
1165  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
1166  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
1167  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
1168  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
1169  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1170  first pass. Recursion is used to keep track of subpatterns that reset the  
1171  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1172    that if (?< or (?' or (?P< is encountered, the name will be correctly
1173    terminated because that is checked in the first pass. There is now one call to
1174    this function in the first pass, to check for a recursive back reference by
1175    name (so that we can make the whole group atomic). In this case, we need check
1176    only up to the current position in the pattern, and that is still OK because
1177    and previous occurrences will have been checked. To make this work, the test
1178    for "end of pattern" is a check against cd->end_pattern in the main loop,
1179    instead of looking for a binary zero. This means that the special first-pass
1180    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1181    processing items within the loop are OK, because afterwards the main loop will
1182    terminate.)
1183    
1184  Arguments:  Arguments:
1185    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1038  Arguments: Line 1187  Arguments:
1187    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1188    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1189    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1190      utf8         TRUE if we are in UTF-8 mode
1191    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1192    
1193  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
# Line 1045  Returns:       the number of the named s Line 1195  Returns:       the number of the named s
1195    
1196  static int  static int
1197  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,
1198    BOOL xmode, int *count)    BOOL xmode, BOOL utf8, int *count)
1199  {  {
1200  uschar *ptr = *ptrptr;  uschar *ptr = *ptrptr;
1201  int start_count = *count;  int start_count = *count;
# Line 1057  dealing with. The very first call may no Line 1207  dealing with. The very first call may no
1207    
1208  if (ptr[0] == CHAR_LEFT_PARENTHESIS)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1209    {    {
1210    if (ptr[1] == CHAR_QUESTION_MARK &&    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1211        ptr[2] == CHAR_VERTICAL_LINE)  
1212      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1213    
1214      /* Handle a normal, unnamed capturing parenthesis. */
1215    
1216      else if (ptr[1] != CHAR_QUESTION_MARK)
1217        {
1218        *count += 1;
1219        if (name == NULL && *count == lorn) return *count;
1220        ptr++;
1221        }
1222    
1223      /* All cases now have (? at the start. Remember when we are in a group
1224      where the parenthesis numbers are duplicated. */
1225    
1226      else if (ptr[2] == CHAR_VERTICAL_LINE)
1227      {      {
1228      ptr += 3;      ptr += 3;
1229      dup_parens = TRUE;      dup_parens = TRUE;
1230      }      }
1231    
1232    /* Handle a normal, unnamed capturing parenthesis */    /* Handle comments; all characters are allowed until a ket is reached. */
1233    
1234    else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)    else if (ptr[2] == CHAR_NUMBER_SIGN)
1235      {      {
1236      *count += 1;      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1237      if (name == NULL && *count == lorn) return *count;      goto FAIL_EXIT;
     ptr++;  
1238      }      }
1239    
1240    /* 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
1241    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
1242    condition (there can't be any nested parens. */    condition (there can't be any nested parens). */
1243    
1244    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1245      {      {
# Line 1087  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1251  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1251        }        }
1252      }      }
1253    
1254    /* We have either (? or (* and not a condition */    /* Start with (? but not a condition. */
1255    
1256    else    else
1257      {      {
# Line 1116  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1280  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1280    }    }
1281    
1282  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1283  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1284    first-pass call when this value is temporarily adjusted to stop at the current
1285    position. So DO NOT change this to a test for binary zero. */
1286    
1287  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1288    {    {
1289    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1290    
# Line 1192  for (; *ptr != 0; ptr++) Line 1358  for (; *ptr != 0; ptr++)
1358    
1359    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1360      {      {
1361      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1362        while (*ptr != 0)
1363          {
1364          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1365          ptr++;
1366    #ifdef SUPPORT_UTF8
1367          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1368    #endif
1369          }
1370      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1371      continue;      continue;
1372      }      }
# Line 1201  for (; *ptr != 0; ptr++) Line 1375  for (; *ptr != 0; ptr++)
1375    
1376    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1377      {      {
1378      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1379      if (rc > 0) return rc;      if (rc > 0) return rc;
1380      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1381      }      }
# Line 1209  for (; *ptr != 0; ptr++) Line 1383  for (; *ptr != 0; ptr++)
1383    else if (*ptr == CHAR_RIGHT_PARENTHESIS)    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1384      {      {
1385      if (dup_parens && *count < hwm_count) *count = hwm_count;      if (dup_parens && *count < hwm_count) *count = hwm_count;
1386      *ptrptr = ptr;      goto FAIL_EXIT;
     return -1;  
1387      }      }
1388    
1389    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
# Line 1248  Arguments: Line 1421  Arguments:
1421    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1422    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1423    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1424      utf8         TRUE if we are in UTF-8 mode
1425    
1426  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1427  */  */
1428    
1429  static int  static int
1430  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1431      BOOL utf8)
1432  {  {
1433  uschar *ptr = (uschar *)cd->start_pattern;  uschar *ptr = (uschar *)cd->start_pattern;
1434  int count = 0;  int count = 0;
# Line 1266  matching closing parens. That is why we Line 1441  matching closing parens. That is why we
1441    
1442  for (;;)  for (;;)
1443    {    {
1444    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1445    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1446    }    }
1447    
# Line 1282  return rc; Line 1457  return rc;
1457    
1458  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1459  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
1460  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
1461  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
1462  assertions, and also the \b assertion; for others it does not.  does not.
1463    
1464  Arguments:  Arguments:
1465    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  
1466    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1467    
1468  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1469  */  */
1470    
1471  static const uschar*  static const uschar*
1472  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1473  {  {
1474  for (;;)  for (;;)
1475    {    {
1476    switch ((int)*code)    switch ((int)*code)
1477      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1478      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1479      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1480      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1359  and doing the check at the end; a flag s Line 1524  and doing the check at the end; a flag s
1524    
1525  Arguments:  Arguments:
1526    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1527    options  the compiling options    utf8     TRUE in UTF-8 mode
1528    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1529    cd       the "compile data" structure    cd       the "compile data" structure
1530    
1531  Returns:   the fixed length,  Returns:   the fixed length,
1532               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1533               or -2 if \C was encountered               or -2 if \C was encountered (in UTF-8 mode only)
1534               or -3 if an OP_RECURSE item was encountered and atend is FALSE               or -3 if an OP_RECURSE item was encountered and atend is FALSE
1535                 or -4 if an unknown opcode was encountered (internal error)
1536  */  */
1537    
1538  static int  static int
1539  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1540  {  {
1541  int length = -1;  int length = -1;
1542    
# Line 1387  for (;;) Line 1553  for (;;)
1553    register int op = *cc;    register int op = *cc;
1554    switch (op)    switch (op)
1555      {      {
1556        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1557        OP_BRA (normal non-capturing bracket) because the other variants of these
1558        opcodes are all concerned with unlimited repeated groups, which of course
1559        are not of fixed length. */
1560    
1561      case OP_CBRA:      case OP_CBRA:
1562      case OP_BRA:      case OP_BRA:
1563      case OP_ONCE:      case OP_ONCE:
1564        case OP_ONCE_NC:
1565      case OP_COND:      case OP_COND:
1566      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1567      if (d < 0) return d;      if (d < 0) return d;
1568      branchlength += d;      branchlength += d;
1569      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1570      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1571      break;      break;
1572    
1573      /* 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.
1574      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
1575      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
1576        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1577        because they all imply an unlimited repeat. */
1578    
1579      case OP_ALT:      case OP_ALT:
1580      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1581      case OP_END:      case OP_END:
1582        case OP_ACCEPT:
1583        case OP_ASSERT_ACCEPT:
1584      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1585        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1586      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1423  for (;;) Line 1597  for (;;)
1597      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1598      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1599      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1600      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1601      if (d < 0) return d;      if (d < 0) return d;
1602      branchlength += d;      branchlength += d;
1603      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1440  for (;;) Line 1614  for (;;)
1614    
1615      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1616    
1617      case OP_REVERSE:      case OP_MARK:
1618        case OP_PRUNE_ARG:
1619        case OP_SKIP_ARG:
1620        case OP_THEN_ARG:
1621        cc += cc[1] + _pcre_OP_lengths[*cc];
1622        break;
1623    
1624        case OP_CALLOUT:
1625        case OP_CIRC:
1626        case OP_CIRCM:
1627        case OP_CLOSE:
1628        case OP_COMMIT:
1629      case OP_CREF:      case OP_CREF:
     case OP_NCREF:  
     case OP_RREF:  
     case OP_NRREF:  
1630      case OP_DEF:      case OP_DEF:
1631      case OP_OPT:      case OP_DOLL:
1632      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
     case OP_SET_SOM:  
1633      case OP_EOD:      case OP_EOD:
1634      case OP_EODN:      case OP_EODN:
1635      case OP_CIRC:      case OP_FAIL:
1636      case OP_DOLL:      case OP_NCREF:
1637        case OP_NRREF:
1638      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1639        case OP_PRUNE:
1640        case OP_REVERSE:
1641        case OP_RREF:
1642        case OP_SET_SOM:
1643        case OP_SKIP:
1644        case OP_SOD:
1645        case OP_SOM:
1646        case OP_THEN:
1647      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1648      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
1649      break;      break;
# Line 1463  for (;;) Line 1651  for (;;)
1651      /* Handle literal characters */      /* Handle literal characters */
1652    
1653      case OP_CHAR:      case OP_CHAR:
1654      case OP_CHARNC:      case OP_CHARI:
1655      case OP_NOT:      case OP_NOT:
1656        case OP_NOTI:
1657      branchlength++;      branchlength++;
1658      cc += 2;      cc += 2;
1659  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1660      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1661  #endif  #endif
1662      break;      break;
1663    
# Line 1477  for (;;) Line 1665  for (;;)
1665      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1666    
1667      case OP_EXACT:      case OP_EXACT:
1668        case OP_EXACTI:
1669        case OP_NOTEXACT:
1670        case OP_NOTEXACTI:
1671      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1672      cc += 4;      cc += 4;
1673  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1674      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1675  #endif  #endif
1676      break;      break;
1677    
# Line 1498  for (;;) Line 1688  for (;;)
1688      cc += 2;      cc += 2;
1689      /* Fall through */      /* Fall through */
1690    
1691        case OP_HSPACE:
1692        case OP_VSPACE:
1693        case OP_NOT_HSPACE:
1694        case OP_NOT_VSPACE:
1695      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1696      case OP_DIGIT:      case OP_DIGIT:
1697      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1510  for (;;) Line 1704  for (;;)
1704      cc++;      cc++;
1705      break;      break;
1706    
1707      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1708        otherwise \C is coded as OP_ALLANY. */
1709    
1710      case OP_ANYBYTE:      case OP_ANYBYTE:
1711      return -2;      return -2;
# Line 1529  for (;;) Line 1724  for (;;)
1724    
1725      switch (*cc)      switch (*cc)
1726        {        {
1727          case OP_CRPLUS:
1728          case OP_CRMINPLUS:
1729        case OP_CRSTAR:        case OP_CRSTAR:
1730        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1731        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1549  for (;;) Line 1746  for (;;)
1746    
1747      /* Anything else is variable length */      /* Anything else is variable length */
1748    
1749      default:      case OP_ANYNL:
1750        case OP_BRAMINZERO:
1751        case OP_BRAPOS:
1752        case OP_BRAPOSZERO:
1753        case OP_BRAZERO:
1754        case OP_CBRAPOS:
1755        case OP_EXTUNI:
1756        case OP_KETRMAX:
1757        case OP_KETRMIN:
1758        case OP_KETRPOS:
1759        case OP_MINPLUS:
1760        case OP_MINPLUSI:
1761        case OP_MINQUERY:
1762        case OP_MINQUERYI:
1763        case OP_MINSTAR:
1764        case OP_MINSTARI:
1765        case OP_MINUPTO:
1766        case OP_MINUPTOI:
1767        case OP_NOTMINPLUS:
1768        case OP_NOTMINPLUSI:
1769        case OP_NOTMINQUERY:
1770        case OP_NOTMINQUERYI:
1771        case OP_NOTMINSTAR:
1772        case OP_NOTMINSTARI:
1773        case OP_NOTMINUPTO:
1774        case OP_NOTMINUPTOI:
1775        case OP_NOTPLUS:
1776        case OP_NOTPLUSI:
1777        case OP_NOTPOSPLUS:
1778        case OP_NOTPOSPLUSI:
1779        case OP_NOTPOSQUERY:
1780        case OP_NOTPOSQUERYI:
1781        case OP_NOTPOSSTAR:
1782        case OP_NOTPOSSTARI:
1783        case OP_NOTPOSUPTO:
1784        case OP_NOTPOSUPTOI:
1785        case OP_NOTQUERY:
1786        case OP_NOTQUERYI:
1787        case OP_NOTSTAR:
1788        case OP_NOTSTARI:
1789        case OP_NOTUPTO:
1790        case OP_NOTUPTOI:
1791        case OP_PLUS:
1792        case OP_PLUSI:
1793        case OP_POSPLUS:
1794        case OP_POSPLUSI:
1795        case OP_POSQUERY:
1796        case OP_POSQUERYI:
1797        case OP_POSSTAR:
1798        case OP_POSSTARI:
1799        case OP_POSUPTO:
1800        case OP_POSUPTOI:
1801        case OP_QUERY:
1802        case OP_QUERYI:
1803        case OP_REF:
1804        case OP_REFI:
1805        case OP_SBRA:
1806        case OP_SBRAPOS:
1807        case OP_SCBRA:
1808        case OP_SCBRAPOS:
1809        case OP_SCOND:
1810        case OP_SKIPZERO:
1811        case OP_STAR:
1812        case OP_STARI:
1813        case OP_TYPEMINPLUS:
1814        case OP_TYPEMINQUERY:
1815        case OP_TYPEMINSTAR:
1816        case OP_TYPEMINUPTO:
1817        case OP_TYPEPLUS:
1818        case OP_TYPEPOSPLUS:
1819        case OP_TYPEPOSQUERY:
1820        case OP_TYPEPOSSTAR:
1821        case OP_TYPEPOSUPTO:
1822        case OP_TYPEQUERY:
1823        case OP_TYPESTAR:
1824        case OP_TYPEUPTO:
1825        case OP_UPTO:
1826        case OP_UPTOI:
1827      return -1;      return -1;
1828    
1829        /* Catch unrecognized opcodes so that when new ones are added they
1830        are not forgotten, as has happened in the past. */
1831    
1832        default:
1833        return -4;
1834      }      }
1835    }    }
1836  /* Control never gets here */  /* Control never gets here */
# Line 1583  _pcre_find_bracket(const uschar *code, B Line 1863  _pcre_find_bracket(const uschar *code, B
1863  for (;;)  for (;;)
1864    {    {
1865    register int c = *code;    register int c = *code;
1866    
1867    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1868    
1869    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1601  for (;;) Line 1882  for (;;)
1882    
1883    /* Handle capturing bracket */    /* Handle capturing bracket */
1884    
1885    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1886               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1887      {      {
1888      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1889      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1610  for (;;) Line 1892  for (;;)
1892    
1893    /* 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
1894    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
1895    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1896      must add in its length. */
1897    
1898    else    else
1899      {      {
# Line 1634  for (;;) Line 1917  for (;;)
1917        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1918        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1919        break;        break;
1920    
1921          case OP_MARK:
1922          case OP_PRUNE_ARG:
1923          case OP_SKIP_ARG:
1924          code += code[1];
1925          break;
1926    
1927          case OP_THEN_ARG:
1928          code += code[1];
1929          break;
1930        }        }
1931    
1932      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1648  for (;;) Line 1941  for (;;)
1941      if (utf8) switch(c)      if (utf8) switch(c)
1942        {        {
1943        case OP_CHAR:        case OP_CHAR:
1944        case OP_CHARNC:        case OP_CHARI:
1945        case OP_EXACT:        case OP_EXACT:
1946          case OP_EXACTI:
1947        case OP_UPTO:        case OP_UPTO:
1948          case OP_UPTOI:
1949        case OP_MINUPTO:        case OP_MINUPTO:
1950          case OP_MINUPTOI:
1951        case OP_POSUPTO:        case OP_POSUPTO:
1952          case OP_POSUPTOI:
1953        case OP_STAR:        case OP_STAR:
1954          case OP_STARI:
1955        case OP_MINSTAR:        case OP_MINSTAR:
1956          case OP_MINSTARI:
1957        case OP_POSSTAR:        case OP_POSSTAR:
1958          case OP_POSSTARI:
1959        case OP_PLUS:        case OP_PLUS:
1960          case OP_PLUSI:
1961        case OP_MINPLUS:        case OP_MINPLUS:
1962          case OP_MINPLUSI:
1963        case OP_POSPLUS:        case OP_POSPLUS:
1964          case OP_POSPLUSI:
1965        case OP_QUERY:        case OP_QUERY:
1966          case OP_QUERYI:
1967        case OP_MINQUERY:        case OP_MINQUERY:
1968          case OP_MINQUERYI:
1969        case OP_POSQUERY:        case OP_POSQUERY:
1970          case OP_POSQUERYI:
1971        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1972        break;        break;
1973        }        }
# Line 1705  for (;;) Line 2011  for (;;)
2011    
2012    /* 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
2013    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
2014    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2015      must add in its length. */
2016    
2017    else    else
2018      {      {
# Line 1729  for (;;) Line 2036  for (;;)
2036        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2037        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2038        break;        break;
2039    
2040          case OP_MARK:
2041          case OP_PRUNE_ARG:
2042          case OP_SKIP_ARG:
2043          code += code[1];
2044          break;
2045    
2046          case OP_THEN_ARG:
2047          code += code[1];
2048          break;
2049        }        }
2050    
2051      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1743  for (;;) Line 2060  for (;;)
2060      if (utf8) switch(c)      if (utf8) switch(c)
2061        {        {
2062        case OP_CHAR:        case OP_CHAR:
2063        case OP_CHARNC:        case OP_CHARI:
2064        case OP_EXACT:        case OP_EXACT:
2065          case OP_EXACTI:
2066        case OP_UPTO:        case OP_UPTO:
2067          case OP_UPTOI:
2068        case OP_MINUPTO:        case OP_MINUPTO:
2069          case OP_MINUPTOI:
2070        case OP_POSUPTO:        case OP_POSUPTO:
2071          case OP_POSUPTOI:
2072        case OP_STAR:        case OP_STAR:
2073          case OP_STARI:
2074        case OP_MINSTAR:        case OP_MINSTAR:
2075          case OP_MINSTARI:
2076        case OP_POSSTAR:        case OP_POSSTAR:
2077          case OP_POSSTARI:
2078        case OP_PLUS:        case OP_PLUS:
2079          case OP_PLUSI:
2080        case OP_MINPLUS:        case OP_MINPLUS:
2081          case OP_MINPLUSI:
2082        case OP_POSPLUS:        case OP_POSPLUS:
2083          case OP_POSPLUSI:
2084        case OP_QUERY:        case OP_QUERY:
2085          case OP_QUERYI:
2086        case OP_MINQUERY:        case OP_MINQUERY:
2087          case OP_MINQUERYI:
2088        case OP_POSQUERY:        case OP_POSQUERY:
2089          case OP_POSQUERYI:
2090        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
2091        break;        break;
2092        }        }
# Line 1785  Arguments: Line 2115  Arguments:
2115    code        points to start of search    code        points to start of search
2116    endcode     points to where to stop    endcode     points to where to stop
2117    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
2118      cd          contains pointers to tables etc.
2119    
2120  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2121  */  */
2122    
2123  static BOOL  static BOOL
2124  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
2125      compile_data *cd)
2126  {  {
2127  register int c;  register int c;
2128  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2129       code < endcode;       code < endcode;
2130       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2131    {    {
2132    const uschar *ccode;    const uschar *ccode;
2133    
# Line 1811  for (code = first_significant_code(code Line 2143  for (code = first_significant_code(code
2143      continue;      continue;
2144      }      }
2145    
2146      /* For a recursion/subroutine call, if its end has been reached, which
2147      implies a backward reference subroutine call, we can scan it. If it's a
2148      forward reference subroutine call, we can't. To detect forward reference
2149      we have to scan up the list that is kept in the workspace. This function is
2150      called only when doing the real compile, not during the pre-compile that
2151      measures the size of the compiled pattern. */
2152    
2153      if (c == OP_RECURSE)
2154        {
2155        const uschar *scode;
2156        BOOL empty_branch;
2157    
2158        /* Test for forward reference */
2159    
2160        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2161          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2162    
2163        /* Not a forward reference, test for completed backward reference */
2164    
2165        empty_branch = FALSE;
2166        scode = cd->start_code + GET(code, 1);
2167        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2168    
2169        /* Completed backwards reference */
2170    
2171        do
2172          {
2173          if (could_be_empty_branch(scode, endcode, utf8, cd))
2174            {
2175            empty_branch = TRUE;
2176            break;
2177            }
2178          scode += GET(scode, 1);
2179          }
2180        while (*scode == OP_ALT);
2181    
2182        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2183        continue;
2184        }
2185    
2186    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2187    
2188    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2189          c == OP_BRAPOSZERO)
2190      {      {
2191      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
2192      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1821  for (code = first_significant_code(code Line 2194  for (code = first_significant_code(code
2194      continue;      continue;
2195      }      }
2196    
2197      /* A nested group that is already marked as "could be empty" can just be
2198      skipped. */
2199    
2200      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2201          c == OP_SCBRA || c == OP_SCBRAPOS)
2202        {
2203        do code += GET(code, 1); while (*code == OP_ALT);
2204        c = *code;
2205        continue;
2206        }
2207    
2208    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2209    
2210    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2211          c == OP_CBRA || c == OP_CBRAPOS ||
2212          c == OP_ONCE || c == OP_ONCE_NC ||
2213          c == OP_COND)
2214      {      {
2215      BOOL empty_branch;      BOOL empty_branch;
2216      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1839  for (code = first_significant_code(code Line 2226  for (code = first_significant_code(code
2226        empty_branch = FALSE;        empty_branch = FALSE;
2227        do        do
2228          {          {
2229          if (!empty_branch && could_be_empty_branch(code, endcode, utf8))          if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2230            empty_branch = TRUE;            empty_branch = TRUE;
2231          code += GET(code, 1);          code += GET(code, 1);
2232          }          }
# Line 1910  for (code = first_significant_code(code Line 2297  for (code = first_significant_code(code
2297      case OP_ALLANY:      case OP_ALLANY:
2298      case OP_ANYBYTE:      case OP_ANYBYTE:
2299      case OP_CHAR:      case OP_CHAR:
2300      case OP_CHARNC:      case OP_CHARI:
2301      case OP_NOT:      case OP_NOT:
2302        case OP_NOTI:
2303      case OP_PLUS:      case OP_PLUS:
2304      case OP_MINPLUS:      case OP_MINPLUS:
2305      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1951  for (code = first_significant_code(code Line 2339  for (code = first_significant_code(code
2339      case OP_KET:      case OP_KET:
2340      case OP_KETRMAX:      case OP_KETRMAX:
2341      case OP_KETRMIN:      case OP_KETRMIN:
2342        case OP_KETRPOS:
2343      case OP_ALT:      case OP_ALT:
2344      return TRUE;      return TRUE;
2345    
# Line 1959  for (code = first_significant_code(code Line 2348  for (code = first_significant_code(code
2348    
2349  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2350      case OP_STAR:      case OP_STAR:
2351        case OP_STARI:
2352      case OP_MINSTAR:      case OP_MINSTAR:
2353        case OP_MINSTARI:
2354      case OP_POSSTAR:      case OP_POSSTAR:
2355        case OP_POSSTARI:
2356      case OP_QUERY:      case OP_QUERY:
2357        case OP_QUERYI:
2358      case OP_MINQUERY:      case OP_MINQUERY:
2359        case OP_MINQUERYI:
2360      case OP_POSQUERY:      case OP_POSQUERY:
2361        case OP_POSQUERYI:
2362      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2363      break;      break;
2364    
2365      case OP_UPTO:      case OP_UPTO:
2366        case OP_UPTOI:
2367      case OP_MINUPTO:      case OP_MINUPTO:
2368        case OP_MINUPTOI:
2369      case OP_POSUPTO:      case OP_POSUPTO:
2370        case OP_POSUPTOI:
2371      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2372      break;      break;
2373  #endif  #endif
2374    
2375        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2376        string. */
2377    
2378        case OP_MARK:
2379        case OP_PRUNE_ARG:
2380        case OP_SKIP_ARG:
2381        code += code[1];
2382        break;
2383    
2384        case OP_THEN_ARG:
2385        code += code[1];
2386        break;
2387    
2388        /* None of the remaining opcodes are required to match a character. */
2389    
2390        default:
2391        break;
2392      }      }
2393    }    }
2394    
# Line 1989  return TRUE; Line 2405  return TRUE;
2405  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
2406  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,
2407  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.
2408    This function is called only during the real compile, not during the
2409    pre-compile.
2410    
2411  Arguments:  Arguments:
2412    code        points to start of the recursion    code        points to start of the recursion
2413    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2414    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2415    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2416      cd          pointers to tables etc
2417    
2418  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2419  */  */
2420    
2421  static BOOL  static BOOL
2422  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2423    BOOL utf8)    BOOL utf8, compile_data *cd)
2424  {  {
2425  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2426    {    {
2427    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2428      return FALSE;      return FALSE;
2429    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2430    }    }
# Line 2038  where Perl recognizes it as the POSIX cl Line 2457  where Perl recognizes it as the POSIX cl
2457  "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,
2458  I think.  I think.
2459    
2460    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2461    It seems that the appearance of a nested POSIX class supersedes an apparent
2462    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2463    a digit.
2464    
2465    In Perl, unescaped square brackets may also appear as part of class names. For
2466    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2467    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2468    seem right at all. PCRE does not allow closing square brackets in POSIX class
2469    names.
2470    
2471  Arguments:  Arguments:
2472    ptr      pointer to the initial [    ptr      pointer to the initial [
2473    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2052  int terminator;          /* Don't combin Line 2482  int terminator;          /* Don't combin
2482  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2483  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2484    {    {
2485    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2486        ptr++;
2487      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2488      else
2489      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2490      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2491        {        {
2492        *endptr = ptr;        *endptr = ptr;
2493        return TRUE;        return TRUE;
2494        }        }
2495        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2496             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2497              ptr[1] == CHAR_EQUALS_SIGN) &&
2498            check_posix_syntax(ptr, endptr))
2499          return FALSE;
2500      }      }
2501    }    }
2502  return FALSE;  return FALSE;
# Line 2188  auto_callout(uschar *code, const uschar Line 2625  auto_callout(uschar *code, const uschar
2625  {  {
2626  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2627  *code++ = 255;  *code++ = 255;
2628  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2629  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2630  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2631  }  }
2632    
# Line 2214  Returns:             nothing Line 2651  Returns:             nothing
2651  static void  static void
2652  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2653  {  {
2654  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2655  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2656  }  }
2657    
# Line 2264  for (++c; c <= d; c++) Line 2701  for (++c; c <= d; c++)
2701    
2702  return TRUE;  return TRUE;
2703  }  }
2704    
2705    
2706    
2707    /*************************************************
2708    *        Check a character and a property        *
2709    *************************************************/
2710    
2711    /* This function is called by check_auto_possessive() when a property item
2712    is adjacent to a fixed character.
2713    
2714    Arguments:
2715      c            the character
2716      ptype        the property type
2717      pdata        the data for the type
2718      negated      TRUE if it's a negated property (\P or \p{^)
2719    
2720    Returns:       TRUE if auto-possessifying is OK
2721    */
2722    
2723    static BOOL
2724    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2725    {
2726    const ucd_record *prop = GET_UCD(c);
2727    switch(ptype)
2728      {
2729      case PT_LAMP:
2730      return (prop->chartype == ucp_Lu ||
2731              prop->chartype == ucp_Ll ||
2732              prop->chartype == ucp_Lt) == negated;
2733    
2734      case PT_GC:
2735      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2736    
2737      case PT_PC:
2738      return (pdata == prop->chartype) == negated;
2739    
2740      case PT_SC:
2741      return (pdata == prop->script) == negated;
2742    
2743      /* These are specials */
2744    
2745      case PT_ALNUM:
2746      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2747              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2748    
2749      case PT_SPACE:    /* Perl space */
2750      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2751              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2752              == negated;
2753    
2754      case PT_PXSPACE:  /* POSIX space */
2755      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2756              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2757              c == CHAR_FF || c == CHAR_CR)
2758              == negated;
2759    
2760      case PT_WORD:
2761      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2762              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2763              c == CHAR_UNDERSCORE) == negated;
2764      }
2765    return FALSE;
2766    }
2767  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2768    
2769    
# Line 2277  whether the next thing could possibly ma Line 2777  whether the next thing could possibly ma
2777  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2778    
2779  Arguments:  Arguments:
2780    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2781    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2782    ptr           next character in pattern    ptr           next character in pattern
2783    options       options bits    options       options bits
2784    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2289  Returns:        TRUE if possessifying is Line 2787  Returns:        TRUE if possessifying is
2787  */  */
2788    
2789  static BOOL  static BOOL
2790  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2791    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2792  {  {
2793  int next;  int c, next;
2794    int op_code = *previous++;
2795    
2796  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2797    
# Line 2303  if ((options & PCRE_EXTENDED) != 0) Line 2802  if ((options & PCRE_EXTENDED) != 0)
2802      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2803      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2804        {        {
2805        while (*(++ptr) != 0)        ptr++;
2806          while (*ptr != 0)
2807            {
2808          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2809            ptr++;
2810    #ifdef SUPPORT_UTF8
2811            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2812    #endif
2813            }
2814        }        }
2815      else break;      else break;
2816      }      }
# Line 2340  if ((options & PCRE_EXTENDED) != 0) Line 2846  if ((options & PCRE_EXTENDED) != 0)
2846      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2847      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2848        {        {
2849        while (*(++ptr) != 0)        ptr++;
2850          while (*ptr != 0)
2851            {
2852          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2853            ptr++;
2854    #ifdef SUPPORT_UTF8
2855            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2856    #endif
2857            }
2858        }        }
2859      else break;      else break;
2860      }      }
# Line 2353  if (*ptr == CHAR_ASTERISK || *ptr == CHA Line 2866  if (*ptr == CHAR_ASTERISK || *ptr == CHA
2866    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)
2867      return FALSE;      return FALSE;
2868    
2869  /* 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
2870  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. */  
2871    
2872  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2873    {    {
2874    case OP_CHAR:    case OP_CHAR:
2875  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2876    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2877  #else  #else
2878    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2879  #endif  #endif
2880    return item != next;    return c != next;
2881    
2882    /* 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
2883    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
2884    high-valued characters. */    high-valued characters. */
2885    
2886    case OP_CHARNC:    case OP_CHARI:
2887  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2888    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2889    #else
2890      c = *previous;
2891  #endif  #endif
2892    if (item == next) return FALSE;    if (c == next) return FALSE;
2893  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2894    if (utf8)    if (utf8)
2895      {      {
# Line 2390  if (next >= 0) switch(op_code) Line 2900  if (next >= 0) switch(op_code)
2900  #else  #else
2901      othercase = NOTACHAR;      othercase = NOTACHAR;
2902  #endif  #endif
2903      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2904      }      }
2905    else    else
2906  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2907    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2908    
2909    /* 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
2910      opcodes are not used for multi-byte characters, because they are coded using
2911      an XCLASS instead. */
2912    
2913    case OP_NOT:    case OP_NOT:
2914    if (item == next) return TRUE;    return (c = *previous) == next;
2915    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2916      case OP_NOTI:
2917      if ((c = *previous) == next) return TRUE;
2918  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2919    if (utf8)    if (utf8)
2920      {      {
# Line 2411  if (next >= 0) switch(op_code) Line 2925  if (next >= 0) switch(op_code)
2925  #else  #else
2926      othercase = NOTACHAR;      othercase = NOTACHAR;
2927  #endif  #endif
2928      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2929      }      }
2930    else    else
2931  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2932    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2933    
2934      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2935      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2936    
2937    case OP_DIGIT:    case OP_DIGIT:
2938    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2458  if (next >= 0) switch(op_code) Line 2975  if (next >= 0) switch(op_code)
2975      case 0x202f:      case 0x202f:
2976      case 0x205f:      case 0x205f:
2977      case 0x3000:      case 0x3000:
2978      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2979      default:      default:
2980      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2981      }      }
2982    
2983      case OP_ANYNL:
2984    case OP_VSPACE:    case OP_VSPACE:
2985    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2986    switch(next)    switch(next)
# Line 2474  if (next >= 0) switch(op_code) Line 2992  if (next >= 0) switch(op_code)
2992      case 0x85:      case 0x85:
2993      case 0x2028:      case 0x2028:
2994      case 0x2029:      case 0x2029:
2995      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2996      default:      default:
2997      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2998      }      }
2999    
3000    #ifdef SUPPORT_UCP
3001      case OP_PROP:
3002      return check_char_prop(next, previous[0], previous[1], FALSE);
3003    
3004      case OP_NOTPROP:
3005      return check_char_prop(next, previous[0], previous[1], TRUE);
3006    #endif
3007    
3008    default:    default:
3009    return FALSE;    return FALSE;
3010    }    }
3011    
3012    
3013  /* 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
3014    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3015    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3016    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3017    replaced by OP_PROP codes when PCRE_UCP is set. */
3018    
3019  switch(op_code)  switch(op_code)
3020    {    {
3021    case OP_CHAR:    case OP_CHAR:
3022    case OP_CHARNC:    case OP_CHARI:
3023  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3024    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3025    #else
3026      c = *previous;
3027  #endif  #endif
3028    switch(-next)    switch(-next)
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_D:      case ESC_D:
3034      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 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_S:      case ESC_S:
3040      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 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_W:      case ESC_W:
3046      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3047    
3048      case ESC_h:      case ESC_h:
3049      case ESC_H:      case ESC_H:
3050      switch(item)      switch(c)
3051        {        {
3052        case 0x09:        case 0x09:
3053        case 0x20:        case 0x20:
# Line 2543  switch(op_code) Line 3075  switch(op_code)
3075    
3076      case ESC_v:      case ESC_v:
3077      case ESC_V:      case ESC_V:
3078      switch(item)      switch(c)
3079        {        {
3080        case 0x0a:        case 0x0a:
3081        case 0x0b:        case 0x0b:
# Line 2557  switch(op_code) Line 3089  switch(op_code)
3089        return -next == ESC_v;        return -next == ESC_v;
3090        }        }
3091    
3092        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3093        their substitutions and process them. The result will always be either
3094        -ESC_p or -ESC_P. Then fall through to process those values. */
3095    
3096    #ifdef SUPPORT_UCP
3097        case ESC_du:
3098        case ESC_DU:
3099        case ESC_wu:
3100        case ESC_WU:
3101        case ESC_su:
3102        case ESC_SU:
3103          {
3104          int temperrorcode = 0;
3105          ptr = substitutes[-next - ESC_DU];
3106          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3107          if (temperrorcode != 0) return FALSE;
3108          ptr++;    /* For compatibility */
3109          }
3110        /* Fall through */
3111    
3112        case ESC_p:
3113        case ESC_P:
3114          {
3115          int ptype, pdata, errorcodeptr;
3116          BOOL negated;
3117    
3118          ptr--;      /* Make ptr point at the p or P */
3119          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3120          if (ptype < 0) return FALSE;
3121          ptr++;      /* Point past the final curly ket */
3122    
3123          /* If the property item is optional, we have to give up. (When generated
3124          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3125          to the original \d etc. At this point, ptr will point to a zero byte. */
3126    
3127          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3128            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3129              return FALSE;
3130    
3131          /* Do the property check. */
3132    
3133          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3134          }
3135    #endif
3136    
3137      default:      default:
3138      return FALSE;      return FALSE;
3139      }      }
3140    
3141      /* In principle, support for Unicode properties should be integrated here as
3142      well. It means re-organizing the above code so as to get hold of the property
3143      values before switching on the op-code. However, I wonder how many patterns
3144      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3145      these op-codes are never generated.) */
3146    
3147    case OP_DIGIT:    case OP_DIGIT:
3148    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3149           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3150    
3151    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3152    return next == -ESC_d;    return next == -ESC_d;
3153    
3154    case OP_WHITESPACE:    case OP_WHITESPACE:
3155    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3156    
3157    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3158    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3159    
3160    case OP_HSPACE:    case OP_HSPACE:
3161    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3162             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3163    
3164    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3165    return next == -ESC_h;    return next == -ESC_h;
3166    
3167    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3168      case OP_ANYNL:
3169    case OP_VSPACE:    case OP_VSPACE:
3170    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3171    
3172    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3173    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3174    
3175    case OP_WORDCHAR:    case OP_WORDCHAR:
3176    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3177             next == -ESC_v || next == -ESC_R;
3178    
3179    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3180    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2620  Arguments: Line 3206  Arguments:
3206    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3207    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3208    bcptr          points to current branch chain    bcptr          points to current branch chain
3209      cond_depth     conditional nesting depth
3210    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3211    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3212                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2631  Returns:         TRUE on success Line 3218  Returns:         TRUE on success
3218  static BOOL  static BOOL
3219  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3220    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3221    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3222  {  {
3223  int repeat_type, op_type;  int repeat_type, op_type;
3224  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2640  int greedy_default, greedy_non_default; Line 3227  int greedy_default, greedy_non_default;
3227  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3228  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3229  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3230  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3231  int after_manual_callout = 0;  int after_manual_callout = 0;
3232  int length_prevgroup = 0;  int length_prevgroup = 0;
3233  register int c;  register int c;
# Line 2652  BOOL inescq = FALSE; Line 3239  BOOL inescq = FALSE;
3239  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3240  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3241  const uschar *tempptr;  const uschar *tempptr;
3242    const uschar *nestptr = NULL;
3243  uschar *previous = NULL;  uschar *previous = NULL;
3244  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3245  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3246  uschar classbits[32];  uschar classbits[32];
3247    
3248    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3249    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3250    dynamically as we process the pattern. */
3251    
3252  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3253  BOOL class_utf8;  BOOL class_utf8;
3254  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2665  uschar *class_utf8data_base; Line 3257  uschar *class_utf8data_base;
3257  uschar utf8_char[6];  uschar utf8_char[6];
3258  #else  #else
3259  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3260  #endif  #endif
3261    
3262  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 2716  for (;; ptr++) Line 3307  for (;; ptr++)
3307    int subfirstbyte;    int subfirstbyte;
3308    int terminator;    int terminator;
3309    int mclength;    int mclength;
3310      int tempbracount;
3311    uschar mcbuffer[8];    uschar mcbuffer[8];
3312    
3313    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3314    
3315    c = *ptr;    c = *ptr;
3316    
3317      /* If we are at the end of a nested substitution, revert to the outer level
3318      string. Nesting only happens one level deep. */
3319    
3320      if (c == 0 && nestptr != NULL)
3321        {
3322        ptr = nestptr;
3323        nestptr = NULL;
3324        c = *ptr;
3325        }
3326    
3327    /* 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
3328    previous cycle of this loop. */    previous cycle of this loop. */
3329    
# Line 2730  for (;; ptr++) Line 3332  for (;; ptr++)
3332  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
3333      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3334  #endif  #endif
3335      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3336        {        {
3337        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3338        goto FAILED;        goto FAILED;
# Line 2752  for (;; ptr++) Line 3354  for (;; ptr++)
3354        goto FAILED;        goto FAILED;
3355        }        }
3356    
3357      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3358      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),
3359          c));
3360    
3361      /* 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
3362      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 2779  for (;; ptr++) Line 3382  for (;; ptr++)
3382    /* 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
3383    reference list. */    reference list. */
3384    
3385    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3386      {      {
3387      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3388      goto FAILED;      goto FAILED;
# Line 2827  for (;; ptr++) Line 3430  for (;; ptr++)
3430      previous_callout = NULL;      previous_callout = NULL;
3431      }      }
3432    
3433    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3434    
3435    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3436      {      {
3437      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3438      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3439        {        {
3440        while (*(++ptr) != 0)        ptr++;
3441          while (*ptr != 0)
3442          {          {
3443          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3444            ptr++;
3445    #ifdef SUPPORT_UTF8
3446            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3447    #endif
3448          }          }
3449        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3450    
# Line 2870  for (;; ptr++) Line 3478  for (;; ptr++)
3478          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3479          goto FAILED;          goto FAILED;
3480          }          }
3481        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3482        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3483        }        }
3484      return TRUE;      return TRUE;
# Line 2881  for (;; ptr++) Line 3489  for (;; ptr++)
3489      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3490    
3491      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3492        previous = NULL;
3493      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3494        {        {
3495        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3496          *code++ = OP_CIRCM;
3497        }        }
3498      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3499      break;      break;
3500    
3501      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3502      previous = NULL;      previous = NULL;
3503      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3504      break;      break;
3505    
3506      /* 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 3075  for (;; ptr++) Line 3684  for (;; ptr++)
3684            ptr++;            ptr++;
3685            }            }
3686    
3687          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3688          if (posix_class < 0)          if (posix_class < 0)
3689            {            {
3690            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 3089  for (;; ptr++) Line 3698  for (;; ptr++)
3698          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3699            posix_class = 0;            posix_class = 0;
3700    
3701          /* 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
3702          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3703          subtract bits that may be in the main map already. At the end we or the  
3704          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3705            if ((options & PCRE_UCP) != 0)
3706              {
3707              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3708              if (posix_substitutes[pc] != NULL)
3709                {
3710                nestptr = tempptr + 1;
3711                ptr = posix_substitutes[pc] - 1;
3712                continue;
3713                }
3714              }
3715    #endif
3716            /* In the non-UCP case, we build the bit map for the POSIX class in a
3717            chunk of local store because we may be adding and subtracting from it,
3718            and we don't want to subtract bits that may be in the main map already.
3719            At the end we or the result into the bit map that is being built. */
3720    
3721          posix_class *= 3;          posix_class *= 3;
3722    
# Line 3136  for (;; ptr++) Line 3760  for (;; ptr++)
3760    
3761        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3762        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
3763        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
3764        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
3765        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
3766        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3767          PCRE_EXTRA is set. */
3768    
3769        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3770          {          {
3771          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3772          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3773    
3774          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 */
3775          else if (-c == ESC_X) c = CHAR_X;   /* \X is literal X in a class */          else if (-c == ESC_N)            /* \N is not supported in a class */
3776          else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */            {
3777              *errorcodeptr = ERR71;
3778              goto FAILED;
3779              }
3780          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3781            {            {
3782            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3165  for (;; ptr++) Line 3793  for (;; ptr++)
3793            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3794            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3795    
3796            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3797              {              {
3798    #ifdef SUPPORT_UCP
3799                case ESC_du:     /* These are the values given for \d etc */
3800                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3801                case ESC_wu:     /* escape sequence with an appropriate \p */
3802                case ESC_WU:     /* or \P to test Unicode properties instead */
3803                case ESC_su:     /* of the default ASCII testing. */
3804                case ESC_SU:
3805                nestptr = ptr;
3806                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3807                class_charcount -= 2;                /* Undo! */
3808                continue;
3809    #endif
3810              case ESC_d:              case ESC_d:
3811              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3812              continue;              continue;
# Line 3187  for (;; ptr++) Line 3825  for (;; ptr++)
3825              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3826              continue;              continue;
3827    
3828                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3829                if it was previously set by something earlier in the character
3830                class. */
3831    
3832              case ESC_s:              case ESC_s:
3833              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3834              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3835                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3836              continue;              continue;
3837    
3838              case ESC_S:              case ESC_S:
# Line 3198  for (;; ptr++) Line 3841  for (;; ptr++)
3841              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3842              continue;              continue;
3843    
3844              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)  
             {  
3845              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3846              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3847              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 3235  for (;; ptr++) Line 3865  for (;; ptr++)
3865                }                }
3866  #endif  #endif
3867              continue;              continue;
             }  
3868    
3869            if (-c == ESC_H)              case ESC_H:
             {  
3870              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3871                {                {
3872                int x = 0xff;                int x = 0xff;
# Line 3280  for (;; ptr++) Line 3908  for (;; ptr++)
3908                }                }
3909  #endif  #endif
3910              continue;              continue;
             }  
3911    
3912            if (-c == ESC_v)              case ESC_v:
             {  
3913              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3914              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3915              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3299  for (;; ptr++) Line 3925  for (;; ptr++)
3925                }                }
3926  #endif  #endif
3927              continue;              continue;
             }  
3928    
3929            if (-c == ESC_V)              case ESC_V:
             {  
3930              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3931                {                {
3932                int x = 0xff;                int x = 0xff;
# Line 3332  for (;; ptr++) Line 3956  for (;; ptr++)
3956                }                }
3957  #endif  #endif
3958              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3959    
3960  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3961            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3962              {              case ESC_P:
3963              BOOL negated;                {
3964              int pdata;                BOOL negated;
3965              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3966              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3967              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3968              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3969                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3970              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3971              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3972              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3973              continue;                class_charcount -= 2;   /* Not a < 256 character */
3974              }                continue;
3975                  }
3976  #endif  #endif
3977            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3978            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3979            treated as literals. */              treated as literals. */
3980    
3981            if ((options & PCRE_EXTRA) != 0)              default:
3982              {              if ((options & PCRE_EXTRA) != 0)
3983              *errorcodeptr = ERR7;                {
3984              goto FAILED;                *errorcodeptr = ERR7;
3985                  goto FAILED;
3986                  }
3987                class_charcount -= 2;  /* Undo the default count from above */
3988                c = *ptr;              /* Get the final character and fall through */
3989                break;
3990              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3991            }            }
3992    
3993          /* 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 3433  for (;; ptr++) Line 4057  for (;; ptr++)
4057            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
4058            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
4059    
4060            /* \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 */  
4061    
4062            if (d < 0)            if (d < 0)
4063              {              {
4064              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  
4065                {                {
4066                ptr = oldptr;                ptr = oldptr;
4067                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3606  for (;; ptr++) Line 4227  for (;; ptr++)
4227          }          }
4228        }        }
4229    
4230      /* 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.
4231        If we are at the end of an internal nested string, revert to the outer
4232        string. */
4233    
4234        while (((c = *(++ptr)) != 0 ||
4235               (nestptr != NULL &&
4236                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4237               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4238    
4239      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
4240    
4241      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4242        {        {
4243        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4244        goto FAILED;        goto FAILED;
4245        }        }
4246    
   
 /* 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  
   
   
4247      /* 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
4248      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
4249      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 3642  we set the flag only if there is a liter Line 4251  we set the flag only if there is a liter
4251    
4252      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
4253      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4254      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4255      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4256    
4257      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
4258      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.
4259      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
4260      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
4261      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
4262      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4263    
4264  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4265      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3661  we set the flag only if there is a liter Line 4270  we set the flag only if there is a liter
4270        {        {
4271        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4272    
4273        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4274    
4275        if (negate_class)        if (negate_class)
4276          {          {
4277          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4278          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4279          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4280          *code++ = class_lastchar;          *code++ = class_lastchar;
4281          break;          break;
4282          }          }
# Line 3698  we set the flag only if there is a liter Line 4307  we set the flag only if there is a liter
4307    
4308      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4309      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4310      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
4311      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
4312      (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
4313      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
4314        actual compiled code. */
4315    
4316  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4317      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4318        {        {
4319        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4320        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3730  we set the flag only if there is a liter Line 4340  we set the flag only if there is a liter
4340        }        }
4341  #endif  #endif
4342    
4343      /* 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
4344      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
4345      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
4346      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4347        negating it if necessary. */
4348    
4349      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4350      if (negate_class)      if (negate_class)
# Line 3793  we set the flag only if there is a liter Line 4404  we set the flag only if there is a liter
4404      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4405      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4406    
4407      /* 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
4408      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4409    
4410      tempcode = previous;      tempcode = previous;
4411    
# Line 3817  we set the flag only if there is a liter Line 4428  we set the flag only if there is a liter
4428        }        }
4429      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4430    
4431        /* If previous was a recursion call, wrap it in atomic brackets so that
4432        previous becomes the atomic group. All recursions were so wrapped in the
4433        past, but it no longer happens for non-repeated recursions. In fact, the
4434        repeated ones could be re-implemented independently so as not to need this,
4435        but for the moment we rely on the code for repeating groups. */
4436    
4437        if (*previous == OP_RECURSE)
4438          {
4439          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4440          *previous = OP_ONCE;
4441          PUT(previous, 1, 2 + 2*LINK_SIZE);
4442          previous[2 + 2*LINK_SIZE] = OP_KET;
4443          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4444          code += 2 + 2 * LINK_SIZE;
4445          length_prevgroup = 3 + 3*LINK_SIZE;
4446    
4447          /* When actually compiling, we need to check whether this was a forward
4448          reference, and if so, adjust the offset. */
4449    
4450          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4451            {
4452            int offset = GET(cd->hwm, -LINK_SIZE);
4453            if (offset == previous + 1 - cd->start_code)
4454              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4455            }
4456          }
4457    
4458        /* Now handle repetition for the different types of item. */
4459    
4460      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4461      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
4462      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
4463      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
4464      instead.  */      instead.  */
4465    
4466      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4467        {        {
4468          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4469    
4470        /* 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
4471        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
4472        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 3857  we set the flag only if there is a liter Line 4499  we set the flag only if there is a liter
4499    
4500        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4501            repeat_max < 0 &&            repeat_max < 0 &&
4502            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4503          {          {
4504          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4505          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3870  we set the flag only if there is a liter Line 4511  we set the flag only if there is a liter
4511      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4512      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-
4513      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4514      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
4515      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4516    
4517      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4518        {        {
4519        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4520        c = previous[1];        c = previous[1];
4521        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4522            repeat_max < 0 &&            repeat_max < 0 &&
4523            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4524          {          {
4525          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4526          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3903  we set the flag only if there is a liter Line 4544  we set the flag only if there is a liter
4544    
4545        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4546            repeat_max < 0 &&            repeat_max < 0 &&
4547            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4548          {          {
4549          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4550          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4072  we set the flag only if there is a liter Line 4713  we set the flag only if there is a liter
4713  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4714               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4715  #endif  #endif
4716               *previous == OP_REF)               *previous == OP_REF ||
4717                 *previous == OP_REFI)
4718        {        {
4719        if (repeat_max == 0)        if (repeat_max == 0)
4720          {          {
# Line 4106  we set the flag only if there is a liter Line 4748  we set the flag only if there is a liter
4748        }        }
4749    
4750      /* 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
4751      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4752        opcodes such as BRA and CBRA, as this is the place where they get converted
4753        into the more special varieties such as BRAPOS and SBRA. A test for >=
4754        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4755        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4756        repetition of assertions, but now it does, for Perl compatibility. */
4757    
4758      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4759        {        {
4760        register int i;        register int i;
4761        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4762        uschar *bralink = NULL;        uschar *bralink = NULL;
4763          uschar *brazeroptr = NULL;
4764    
4765        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4766          we just ignore the repeat. */
4767    
4768        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4769          {          goto END_REPEAT;
4770          *errorcodeptr = ERR55;  
4771          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4772          }        use of repetition is in cases when the assertion is optional. Therefore,
4773          if the minimum is greater than zero, just ignore the repeat. If the
4774          maximum is not not zero or one, set it to 1. */
4775    
4776        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4777        by scanning through from the start, and compute the offset back to it          {
4778        from the current code pointer. There may be an OP_OPT setting following          if (repeat_min > 0) goto END_REPEAT;
4779        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;  
4780          }          }
4781    
4782        /* 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 4155  we set the flag only if there is a liter Line 4797  we set the flag only if there is a liter
4797          **   goto END_REPEAT;          **   goto END_REPEAT;
4798          **   }          **   }
4799    
4800          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
4801          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
4802          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
4803          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4804            selectively.
4805    
4806          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
4807          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 4178  we set the flag only if there is a liter Line 4821  we set the flag only if there is a liter
4821              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4822              goto END_REPEAT;              goto END_REPEAT;
4823              }              }
4824              brazeroptr = previous;    /* Save for possessive optimizing */
4825            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4826            }            }
4827    
# Line 4202  we set the flag only if there is a liter Line 4846  we set the flag only if there is a liter
4846            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4847            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4848    
4849            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4850            bralink = previous;            bralink = previous;
4851            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4852            }            }
# Line 4240  we set the flag only if there is a liter Line 4884  we set the flag only if there is a liter
4884              *lengthptr += delta;              *lengthptr += delta;
4885              }              }
4886    
4887            /* This is compiling for real */            /* This is compiling for real. If there is a set first byte for
4888              the group, and we have not yet set a "required byte", set it. */
4889    
4890            else            else
4891              {              {
# Line 4252  we set the flag only if there is a liter Line 4897  we set the flag only if there is a liter
4897                memcpy(code, previous, len);                memcpy(code, previous, len);
4898                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
4899                  {                  {
4900                    if (cd->hwm >= cd->start_workspace + WORK_SIZE_CHECK)
4901                      {
4902                      *errorcodeptr = ERR72;
4903                      goto FAILED;
4904                      }
4905                  PUT(cd->hwm, 0, GET(hc, 0) + len);                  PUT(cd->hwm, 0, GET(hc, 0) + len);
4906                  cd->hwm += LINK_SIZE;                  cd->hwm += LINK_SIZE;
4907                  }                  }
# Line 4279  we set the flag only if there is a liter Line 4929  we set the flag only if there is a liter
4929          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
4930          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4931          a 64-bit integer type when available, otherwise double. */          a 64-bit integer type when available, otherwise double. */
4932    
4933          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4934            {            {
4935            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
# Line 4311  we set the flag only if there is a liter Line 4961  we set the flag only if there is a liter
4961              {              {
4962              int offset;              int offset;
4963              *code++ = OP_BRA;              *code++ = OP_BRA;
4964              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4965              bralink = code;              bralink = code;
4966              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4967              }              }
# Line 4319  we set the flag only if there is a liter Line 4969  we set the flag only if there is a liter
4969            memcpy(code, previous, len);            memcpy(code, previous, len);
4970            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
4971              {              {
4972                if (cd->hwm >= cd->start_workspace + WORK_SIZE_CHECK)
4973                  {
4974                  *errorcodeptr = ERR72;
4975                  goto FAILED;
4976                  }
4977              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
4978              cd->hwm += LINK_SIZE;              cd->hwm += LINK_SIZE;
4979              }              }
# Line 4332  we set the flag only if there is a liter Line 4987  we set the flag only if there is a liter
4987          while (bralink != NULL)          while (bralink != NULL)
4988            {            {
4989            int oldlinkoffset;            int oldlinkoffset;
4990            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4991            uschar *bra = code - offset;            uschar *bra = code - offset;
4992            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4993            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4342  we set the flag only if there is a liter Line 4997  we set the flag only if there is a liter
4997            }            }
4998          }          }
4999    
5000        /* 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
5001        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
5002        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
5003        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
5004          deal with possessive ONCEs specially.
5005        Then, when we are doing the actual compile phase, check to see whether  
5006        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
5007          whether this group is one that could match an empty string. If so,
5008        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
5009        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
5010        atomic groups at runtime, but in a different way.] */        groups at runtime, but in a different way.]
5011    
5012          Then, if the quantifier was possessive and the bracket is not a
5013          conditional, we convert the BRA code to the POS form, and the KET code to
5014          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
5015          subpattern at both the start and at the end.) The use of special opcodes
5016          makes it possible to reduce greatly the stack usage in pcre_exec(). If
5017          the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
5018    
5019          Then, if the minimum number of matches is 1 or 0, cancel the possessive
5020          flag so that the default action below, of wrapping everything inside
5021          atomic brackets, does not happen. When the minimum is greater than 1,
5022          there will be earlier copies of the group, and so we still have to wrap
5023          the whole thing. */
5024    
5025        else        else
5026          {          {
5027          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
5028          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
5029          *ketcode = OP_KETRMAX + repeat_type;  
5030          if (lengthptr == NULL && *bracode != OP_ONCE)          /* Convert possessive ONCE brackets to non-capturing */
5031    
5032            if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5033                possessive_quantifier) *bracode = OP_BRA;
5034    
5035            /* For non-possessive ONCE brackets, all we need to do is to
5036            set the KET. */
5037    
5038            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5039              *ketcode = OP_KETRMAX + repeat_type;
5040    
5041            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5042            converted to non-capturing above). */
5043    
5044            else
5045            {            {
5046            uschar *scode = bracode;            /* In the compile phase, check for empty string matching. */
5047            do  
5048              if (lengthptr == NULL)
5049              {              {
5050              if (could_be_empty_branch(scode, ketcode, utf8))              uschar *scode = bracode;
5051                do
5052                {                {
5053                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
5054                break;                  {
5055                    *bracode += OP_SBRA - OP_BRA;
5056                    break;
5057                    }
5058                  scode += GET(scode, 1);
5059                }                }
5060              scode += GET(scode, 1);              while (*scode == OP_ALT);
5061              }              }
5062            while (*scode == OP_ALT);  
5063              /* Handle possessive quantifiers. */
5064    
5065              if (possessive_quantifier)
5066                {
5067                /* For COND brackets, we wrap the whole thing in a possessively
5068                repeated non-capturing bracket, because we have not invented POS
5069                versions of the COND opcodes. Because we are moving code along, we
5070                must ensure that any pending recursive references are updated. */
5071    
5072                if (*bracode == OP_COND || *bracode == OP_SCOND)
5073                  {
5074                  int nlen = (int)(code - bracode);
5075                  *code = OP_END;
5076                  adjust_recurse(bracode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5077                  memmove(bracode + 1+LINK_SIZE, bracode, nlen);
5078                  code += 1 + LINK_SIZE;
5079                  nlen += 1 + LINK_SIZE;
5080                  *bracode = OP_BRAPOS;
5081                  *code++ = OP_KETRPOS;
5082                  PUTINC(code, 0, nlen);
5083                  PUT(bracode, 1, nlen);
5084                  }
5085    
5086                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5087    
5088                else
5089                  {
5090                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5091                  *ketcode = OP_KETRPOS;
5092                  }
5093    
5094                /* If the minimum is zero, mark it as possessive, then unset the
5095                possessive flag when the minimum is 0 or 1. */
5096    
5097                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5098                if (repeat_min < 2) possessive_quantifier = FALSE;
5099                }
5100    
5101              /* Non-possessive quantifier */
5102    
5103              else *ketcode = OP_KETRMAX + repeat_type;
5104            }            }
5105          }          }
5106        }        }
# Line 4391  we set the flag only if there is a liter Line 5121  we set the flag only if there is a liter
5121        }        }
5122    
5123      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
5124      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
5125      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
5126      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
5127      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
5128      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
5129      tempcode, not at previous, which might be the first part of a string whose  
5130      (former) last char we repeated.      Some (but not all) possessively repeated subpatterns have already been
5131        completely handled in the code just above. For them, possessive_quantifier
5132        is always FALSE at this stage.
5133    
5134        Note that the repeated item starts at tempcode, not at previous, which
5135        might be the first part of a string whose (former) last char we repeated.
5136    
5137      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
5138      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 4420  we set the flag only if there is a liter Line 5155  we set the flag only if there is a liter
5155  #endif  #endif
5156          }          }
5157    
5158        len = code - tempcode;        len = (int)(code - tempcode);
5159        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
5160          {          {
5161          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4428  we set the flag only if there is a liter Line 5163  we set the flag only if there is a liter
5163          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
5164          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
5165    
5166          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
5167          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
5168          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
5169          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
5170    
5171          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
5172          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
5173          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
5174          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
5175    
5176            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
5177            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
5178            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
5179            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
5180    
5181            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
5182            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
5183            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
5184            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
5185    
5186          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
5187          pending recursive references are updated. */          pending recursive references are updated. */
5188    
5189          default:          default:
5190          *code = OP_END;          *code = OP_END;
5191          adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);          adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
# Line 4479  we set the flag only if there is a liter Line 5224  we set the flag only if there is a liter
5224    
5225      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
5226    
5227      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
5228             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
5229        {        {
5230        int i, namelen;        int i, namelen;
5231          int arglen = 0;
5232        const char *vn = verbnames;        const char *vn = verbnames;
5233        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
5234          const uschar *arg = NULL;
5235        previous = NULL;        previous = NULL;
5236        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5237          namelen = (int)(ptr - name);
5238    
5239          /* It appears that Perl allows any characters whatsoever, other than
5240          a closing parenthesis, to appear in arguments, so we no longer insist on
5241          letters, digits, and underscores. */
5242    
5243        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5244          {          {
5245          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
5246          goto FAILED;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5247            arglen = (int)(ptr - arg);
5248          }          }
5249    
5250        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
5251          {          {
5252          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
5253          goto FAILED;          goto FAILED;
5254          }          }
5255        namelen = ptr - name;  
5256          /* Scan the table of verb names */
5257    
5258        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
5259          {          {
5260          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5261              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5262            {            {
5263            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5264              ASSERT_ACCEPT if in an assertion. */
5265    
5266            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5267              {              {
5268              open_capitem *oc;              open_capitem *oc;
5269                if (arglen != 0)
5270                  {
5271                  *errorcodeptr = ERR59;
5272                  goto FAILED;
5273                  }
5274              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5275              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5276                {                {
5277                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5278                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5279                }                }
5280                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5281    
5282                /* Do not set firstbyte after *ACCEPT */
5283                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5284              }              }
5285            *code++ = verbs[i].op;  
5286            break;            /* Handle other cases with/without an argument */
5287    
5288              else if (arglen == 0)
5289                {
5290                if (verbs[i].op < 0)   /* Argument is mandatory */
5291                  {
5292                  *errorcodeptr = ERR66;
5293                  goto FAILED;
5294                  }
5295                *code = verbs[i].op;
5296                if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
5297                }
5298    
5299              else
5300                {
5301                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
5302                  {
5303                  *errorcodeptr = ERR59;
5304                  goto FAILED;
5305                  }
5306                *code = verbs[i].op_arg;
5307                if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
5308                *code++ = arglen;
5309                memcpy(code, arg, arglen);
5310                code += arglen;
5311                *code++ = 0;
5312                }
5313    
5314              break;  /* Found verb, exit loop */
5315            }            }
5316    
5317          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
5318          }          }
5319        if (i < verbcount) continue;  
5320        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
5321          *errorcodeptr = ERR60;          /* Verb not recognized */
5322        goto FAILED;        goto FAILED;
5323        }        }
5324    
# Line 4639  we set the flag only if there is a liter Line 5437  we set the flag only if there is a liter
5437                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5438            ptr++;            ptr++;
5439            }            }
5440          namelen = ptr - name;          namelen = (int)(ptr - name);
5441    
5442          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
5443              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
# Line 4700  we set the flag only if there is a liter Line 5498  we set the flag only if there is a liter
5498          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5499    
5500          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5501                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5502            {            {
5503            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5504            code[1+LINK_SIZE]++;            code[1+LINK_SIZE]++;
# Line 4768  we set the flag only if there is a liter Line 5566  we set the flag only if there is a liter
5566          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5567          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5568          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5569            cd->assert_depth += 1;
5570          ptr++;          ptr++;
5571          break;          break;
5572    
# Line 4782  we set the flag only if there is a liter Line 5581  we set the flag only if there is a liter
5581            continue;            continue;
5582            }            }
5583          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5584            cd->assert_depth += 1;
5585          break;          break;
5586    
5587    
# Line 4791  we set the flag only if there is a liter Line 5591  we set the flag only if there is a liter
5591            {            {
5592            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5593            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5594              cd->assert_depth += 1;
5595            ptr += 2;            ptr += 2;
5596            break;            break;
5597    
5598            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5599            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5600              cd->assert_depth += 1;
5601            ptr += 2;            ptr += 2;
5602            break;            break;
5603    
# Line 4817  we set the flag only if there is a liter Line 5619  we set the flag only if there is a liter
5619    
5620          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5621          case CHAR_C:                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
5622          previous_callout = code;  /* Save for later completion */          previous_callout = code;     /* Save for later completion */
5623          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1;    /* Skip one item before completing */
5624          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
5625            {            {
5626            int n = 0;            int n = 0;
# Line 4835  we set the flag only if there is a liter Line 5637  we set the flag only if there is a liter
5637              goto FAILED;              goto FAILED;
5638              }              }
5639            *code++ = n;            *code++ = n;
5640            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5641            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5642            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5643            }            }
5644          previous = NULL;          previous = NULL;
# Line 4869  we set the flag only if there is a liter Line 5671  we set the flag only if there is a liter
5671            name = ++ptr;            name = ++ptr;
5672    
5673            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5674            namelen = ptr - name;            namelen = (int)(ptr - name);
5675    
5676            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5677    
# Line 4999  we set the flag only if there is a liter Line 5801  we set the flag only if there is a liter
5801          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5802          name = ++ptr;          name = ++ptr;
5803          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5804          namelen = ptr - name;          namelen = (int)(ptr - name);
5805    
5806          /* 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
5807          reference number. */          a dummy reference number, because it was not used in the first pass.
5808            However, with the change of recursive back references to be atomic,
5809            we have to look for the number so that this state can be identified, as
5810            otherwise the incorrect length is computed. If it's not a backwards
5811            reference, the dummy number will do. */
5812    
5813          if (lengthptr != NULL)          if (lengthptr != NULL)
5814            {            {
5815              const uschar *temp;
5816    
5817            if (namelen == 0)            if (namelen == 0)
5818              {              {
5819              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5021  we set the flag only if there is a liter Line 5829  we set the flag only if there is a liter
5829              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5830              goto FAILED;              goto FAILED;
5831              }              }
5832            recno = 0;  
5833              /* The name table does not exist in the first pass, so we cannot
5834              do a simple search as in the code below. Instead, we have to scan the
5835              pattern to find the number. It is important that we scan it only as
5836              far as we have got because the syntax of named subpatterns has not
5837              been checked for the rest of the pattern, and find_parens() assumes
5838              correct syntax. In any case, it's a waste of resources to scan
5839              further. We stop the scan at the current point by temporarily
5840              adjusting the value of cd->endpattern. */
5841    
5842              temp = cd->end_pattern;
5843              cd->end_pattern = ptr;
5844              recno = find_parens(cd, name, namelen,
5845                (options & PCRE_EXTENDED) != 0, utf8);
5846              cd->end_pattern = temp;
5847              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5848            }            }
5849    
5850          /* 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 5046  we set the flag only if there is a liter Line 5869  we set the flag only if there is a liter
5869              }              }
5870            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5871                      find_parens(cd, name, namelen,                      find_parens(cd, name, namelen,
5872                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5873              {              {
5874              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5875              goto FAILED;              goto FAILED;
# Line 5157  we set the flag only if there is a liter Line 5980  we set the flag only if there is a liter
5980              if (called == NULL)              if (called == NULL)
5981                {                {
5982                if (find_parens(cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5983                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5984                  {                  {
5985                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5986                  goto FAILED;                  goto FAILED;
5987                  }                  }
5988    
5989                /* Fudge the value of "called" so that when it is inserted as an                /* Fudge the value of "called" so that when it is inserted as an
5990                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5991                of the group. */                of the group. Then remember the forward reference. */
5992    
5993                called = cd->start_code + recno;                called = cd->start_code + recno;
5994                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                if (cd->hwm >= cd->start_workspace + WORK_SIZE_CHECK)
5995                    {
5996                    *errorcodeptr = ERR72;
5997                    goto FAILED;
5998                    }
5999                  PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
6000                }                }
6001    
6002              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
6003              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
6004              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
6005                must not, however, do this check if we are in a conditional
6006                subpattern because the condition might be testing for recursion in
6007                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
6008                Forever loops are also detected at runtime, so those that occur in
6009                conditional subpatterns will be picked up then. */
6010    
6011              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
6012                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
6013                {                {
6014                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
6015                goto FAILED;                goto FAILED;
6016                }                }
6017              }              }
6018    
6019            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. It does not have a set first
6020            "once" brackets. Set up a "previous group" length so that a            byte (relevant if it is repeated, because it will then be wrapped
6021            subsequent quantifier will work. */            with ONCE brackets). */
   
           *code = OP_ONCE;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
6022    
6023            *code = OP_RECURSE;            *code = OP_RECURSE;
6024            PUT(code, 1, called - cd->start_code);            PUT(code, 1, (int)(called - cd->start_code));
6025            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
6026              groupsetfirstbyte = FALSE;
           *code = OP_KET;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
   
           length_prevgroup = 3 + 3*LINK_SIZE;  
6027            }            }
6028    
6029          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5260  we set the flag only if there is a liter Line 6084  we set the flag only if there is a liter
6084          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
6085          both phases.          both phases.
6086    
6087          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
6088          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. */  
6089    
6090          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
6091            {            {
# Line 5273  we set the flag only if there is a liter Line 6096  we set the flag only if there is a liter
6096              }              }
6097            else            else
6098              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
6099              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
6100              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
6101              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
6102              }              }
6103    
6104            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
6105            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). */  
6106    
6107            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
6108            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5303  we set the flag only if there is a liter Line 6119  we set the flag only if there is a liter
6119          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
6120        }       /* End of (? handling */        }       /* End of (? handling */
6121    
6122      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
6123      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
6124      brackets. */      brackets. */
6125    
6126      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 5322  we set the flag only if there is a liter Line 6138  we set the flag only if there is a liter
6138        skipbytes = 2;        skipbytes = 2;
6139        }        }
6140    
6141      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
6142      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
6143      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
6144      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. */  
6145    
6146      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                      /* For handling repetition */
6147      *code = bravalue;      *code = bravalue;
6148      tempcode = code;      tempcode = code;
6149      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
6150      length_prevgroup = 0;              /* Initialize for pre-compile phase */      tempbracount = cd->bracount;          /* Save value before bracket */
6151        length_prevgroup = 0;                 /* Initialize for pre-compile phase */
6152    
6153      if (!compile_regex(      if (!compile_regex(
6154           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
6155           options & PCRE_IMS,           /* The previous ims option state */           &tempcode,                       /* Where to put code (updated) */
6156           &tempcode,                    /* Where to put code (updated) */           &ptr,                            /* Input pointer (updated) */
6157           &ptr,                         /* Input pointer (updated) */           errorcodeptr,                    /* Where to put an error message */
          errorcodeptr,                 /* Where to put an error message */  
6158           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
6159            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
6160           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
6161           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
6162           &subfirstbyte,                /* For possible first char */           cond_depth +
6163           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
6164           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
6165           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
6166           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
6167             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
6168             (lengthptr == NULL)? NULL :      /* Actual compile phase */
6169               &length_prevgroup              /* Pre-compile phase */
6170           ))           ))
6171        goto FAILED;        goto FAILED;
6172    
6173        /* If this was an atomic group and there are no capturing groups within it,
6174        generate OP_ONCE_NC instead of OP_ONCE. */
6175    
6176        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
6177          *code = OP_ONCE_NC;
6178    
6179        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
6180          cd->assert_depth -= 1;
6181    
6182      /* 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
6183      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.
6184      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
6185    
6186      /* 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
6187      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
6188      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
6189      not be available. */      not be available. */
# Line 5423  we set the flag only if there is a liter Line 6248  we set the flag only if there is a liter
6248          goto FAILED;          goto FAILED;
6249          }          }
6250        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
6251        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
6252        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
6253        *code++ = OP_KET;        *code++ = OP_KET;
6254        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5496  we set the flag only if there is a liter Line 6321  we set the flag only if there is a liter
6321    
6322      /* ===================================================================*/      /* ===================================================================*/
6323      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
6324      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
6325      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
6326      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
6327      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
6328      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
6329        ever created. */
6330    
6331      case CHAR_BACKSLASH:      case CHAR_BACKSLASH:
6332      tempptr = ptr;      tempptr = ptr;
# Line 5590  we set the flag only if there is a liter Line 6416  we set the flag only if there is a liter
6416          }          }
6417    
6418        /* \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).
6419        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6420    
6421        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6422          {          {
6423            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6424              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6425              {
6426              *errorcodeptr = ERR69;
6427              break;
6428              }
6429          is_recurse = FALSE;          is_recurse = FALSE;
6430          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6431            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
# Line 5614  we set the flag only if there is a liter Line 6445  we set the flag only if there is a liter
6445          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6446          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6447          previous = code;          previous = code;
6448          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6449          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6450          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6451          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
# Line 5660  we set the flag only if there is a liter Line 6491  we set the flag only if there is a liter
6491  #endif  #endif
6492    
6493        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
6494        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
6495          situation when PCRE_UCP is not set. When it *is* set, we substitute
6496          Unicode property tests. */
6497    
6498        else        else
6499          {          {
6500          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
6501          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
6502              {
6503              nestptr = ptr + 1;                   /* Where to resume */
6504              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
6505              }
6506            else
6507    #endif
6508            /* In non-UTF-8 mode, we turn \C into OP_ALLANY instead of OP_ANYBYTE
6509            so that it works in DFA mode and in lookbehinds. */
6510    
6511              {
6512              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6513              *code++ = (!utf8 && c == -ESC_C)? OP_ALLANY : -c;
6514              }
6515          }          }
6516        continue;        continue;
6517        }        }
# Line 5710  we set the flag only if there is a liter Line 6556  we set the flag only if there is a liter
6556    
6557      ONE_CHAR:      ONE_CHAR:
6558      previous = code;      previous = code;
6559      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6560      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6561    
6562      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 5739  we set the flag only if there is a liter Line 6585  we set the flag only if there is a liter
6585        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6586        }        }
6587    
6588      /* 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
6589      1 or the matching is caseful. */      1 or the matching is caseful. */
6590    
6591      else      else
# Line 5774  return FALSE; Line 6620  return FALSE;
6620  /* 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
6621  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
6622  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.  
   
6623  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
6624  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
6625  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6626    
6627  Arguments:  Arguments:
6628    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  
6629    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6630    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6631    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
6632    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6633    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6634    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6635      cond_depth     depth of nesting for conditional subpatterns
6636    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6637    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6638    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 5803  Returns:         TRUE on success Line 6644  Returns:         TRUE on success
6644  */  */
6645    
6646  static BOOL  static BOOL
6647  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6648    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6649    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6650    int *lengthptr)    compile_data *cd, int *lengthptr)
6651  {  {
6652  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6653  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 5820  int branchfirstbyte, branchreqbyte; Line 6661  int branchfirstbyte, branchreqbyte;
6661  int length;  int length;
6662  int orig_bracount;  int orig_bracount;
6663  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6664  branch_chain bc;  branch_chain bc;
6665    
6666  bc.outer = bcptr;  bc.outer = bcptr;
# Line 5844  pre-compile phase to find out whether an Line 6684  pre-compile phase to find out whether an
6684    
6685  /* If this is a capturing subpattern, add to the chain of open capturing items  /* If this is a capturing subpattern, add to the chain of open capturing items
6686  so that we can detect them if (*ACCEPT) is encountered. This is also used to  so that we can detect them if (*ACCEPT) is encountered. This is also used to
6687  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6688    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6689    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6690    
6691  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6692    {    {
# Line 5870  for (;;) Line 6712  for (;;)
6712    
6713    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6714    
   /* 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;  
     }  
   
6715    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6716    
6717    if (lookbehind)    if (lookbehind)
# Line 5893  for (;;) Line 6726  for (;;)
6726    into the length. */    into the length. */
6727    
6728    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6729          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6730            (lengthptr == NULL)? NULL : &length))
6731      {      {
6732      *ptrptr = ptr;      *ptrptr = ptr;
6733      return FALSE;      return FALSE;
6734      }      }
6735    
   /* If the external options have changed during this branch, it means that we  
   are at the top level, and a leading option setting has been encountered. We  
   need to re-set the original option values to take account of this so that,  
   during the pre-compile phase, we know to allow for a re-set at the start of  
   subsequent branches. */  
   
   if (old_external_options != cd->external_options)  
     oldims = cd->external_options & PCRE_IMS;  
   
6736    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6737    has fewer than the rest. */    has fewer than the rest. */
6738    
# Line 5968  for (;;) Line 6793  for (;;)
6793        {        {
6794        int fixed_length;        int fixed_length;
6795        *code = OP_END;        *code = OP_END;
6796        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6797            FALSE, cd);
6798        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6799        if (fixed_length == -3)        if (fixed_length == -3)
6800          {          {
# Line 5976  for (;;) Line 6802  for (;;)
6802          }          }
6803        else if (fixed_length < 0)        else if (fixed_length < 0)
6804          {          {
6805          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 :
6806                            (fixed_length == -4)? ERR70: ERR25;
6807          *ptrptr = ptr;          *ptrptr = ptr;
6808          return FALSE;          return FALSE;
6809          }          }
# Line 5989  for (;;) Line 6816  for (;;)
6816    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
6817    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
6818    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
6819    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. */  
6820    
6821    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6822      {      {
6823      if (lengthptr == NULL)      if (lengthptr == NULL)
6824        {        {
6825        int branch_length = code - last_branch;        int branch_length = (int)(code - last_branch);
6826        do        do
6827          {          {
6828          int prev_length = GET(last_branch, 1);          int prev_length = GET(last_branch, 1);
# Line 6011  for (;;) Line 6836  for (;;)
6836      /* Fill in the ket */      /* Fill in the ket */
6837    
6838      *code = OP_KET;      *code = OP_KET;
6839      PUT(code, 1, code - start_bracket);      PUT(code, 1, (int)(code - start_bracket));
6840      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6841    
6842      /* If it was a capturing subpattern, check to see if it contained any      /* If it was a capturing subpattern, check to see if it contained any
# Line 6026  for (;;) Line 6851  for (;;)
6851            code - start_bracket);            code - start_bracket);
6852          *start_bracket = OP_ONCE;          *start_bracket = OP_ONCE;
6853          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
6854          PUT(start_bracket, 1, code - start_bracket);          PUT(start_bracket, 1, (int)(code - start_bracket));
6855          *code = OP_KET;          *code = OP_KET;
6856          PUT(code, 1, code - start_bracket);          PUT(code, 1, (int)(code - start_bracket));
6857          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
6858          length += 2 + 2*LINK_SIZE;          length += 2 + 2*LINK_SIZE;
6859          }          }
6860        cd->open_caps = cd->open_caps->next;        cd->open_caps = cd->open_caps->next;
6861        }        }
6862    
     /* Reset options if needed. */  
   
     if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)  
       {  
       *code++ = OP_OPT;  
       *code++ = oldims;  
       length += 2;  
       }  
   
6863      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
6864    
6865      cd->bracount = max_bracount;      cd->bracount = max_bracount;
# Line 6083  for (;;) Line 6899  for (;;)
6899    else    else
6900      {      {
6901      *code = OP_ALT;      *code = OP_ALT;
6902      PUT(code, 1, code - last_branch);      PUT(code, 1, (int)(code - last_branch));
6903      bc.current_branch = last_branch = code;      bc.current_branch = last_branch = code;
6904      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6905      }      }
# Line 6103  for (;;) Line 6919  for (;;)
6919  /* 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
6920  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
6921  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
6922  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
6923  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6924    
6925  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.
6926  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 6125  of the more common cases more precisely. Line 6941  of the more common cases more precisely.
6941    
6942  Arguments:  Arguments:
6943    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6944    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
6945                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6946                    the less precise approach                    the less precise approach
# Line 6135  Returns:     TRUE or FALSE Line 6950  Returns:     TRUE or FALSE
6950  */  */
6951    
6952  static BOOL  static BOOL
6953  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6954    unsigned int backref_map)    unsigned int backref_map)
6955  {  {
6956  do {  do {
6957     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6958       options, PCRE_MULTILINE, FALSE);       FALSE);
6959     register int op = *scode;     register int op = *scode;
6960    
6961     /* Non-capturing brackets */     /* Non-capturing brackets */
6962    
6963     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6964           op == OP_SBRA || op == OP_SBRAPOS)
6965       {       {
6966       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6967       }       }
6968    
6969     /* Capturing brackets */     /* Capturing brackets */
6970    
6971     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6972                op == OP_SCBRA || op == OP_SCBRAPOS)
6973       {       {
6974       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6975       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6976       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6977       }       }
6978    
6979     /* Other brackets */     /* Other brackets */
6980    
6981     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC ||
6982                op == OP_COND)
6983       {       {
6984       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6985       }       }
6986    
6987     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
# Line 6178  do { Line 6996  do {
6996    
6997     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6998    
6999     else if (op != OP_SOD && op != OP_SOM &&     else if (op != OP_SOD && op != OP_SOM && op != OP_CIRC) return FALSE;
            ((*options & PCRE_MULTILINE) != 0 || op != OP_CIRC))  
      return FALSE;  
7000     code += GET(code, 1);     code += GET(code, 1);
7001     }     }
7002  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6216  is_startline(const uschar *code, unsigne Line 7032  is_startline(const uschar *code, unsigne
7032  {  {
7033  do {  do {
7034     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
7035       NULL, 0, FALSE);       FALSE);
7036     register int op = *scode;     register int op = *scode;
7037    
7038     /* If we are at the start of a conditional assertion group, *both* the     /* If we are at the start of a conditional assertion group, *both* the
# Line 6243  do { Line 7059  do {
7059         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
7060         break;         break;
7061         }         }
7062       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
7063       op = *scode;       op = *scode;
7064       }       }
7065    
7066     /* Non-capturing brackets */     /* Non-capturing brackets */
7067    
7068     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
7069           op == OP_SBRA || op == OP_SBRAPOS)
7070       {       {
7071       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
7072       }       }
7073    
7074     /* Capturing brackets */     /* Capturing brackets */
7075    
7076     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
7077                op == OP_SCBRA || op == OP_SCBRAPOS)
7078       {       {
7079       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
7080       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6265  do { Line 7083  do {
7083    
7084     /* Other brackets */     /* Other brackets */
7085    
7086     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC)
7087       {       {
7088       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
7089       }       }
# Line 6280  do { Line 7098  do {
7098    
7099     /* Check for explicit circumflex */     /* Check for explicit circumflex */
7100    
7101     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
7102    
7103     /* Move on to the next alternative */     /* Move on to the next alternative */
7104    
# Line 6306  we return that char, otherwise -1. Line 7124  we return that char, otherwise -1.
7124    
7125  Arguments:  Arguments:
7126    code       points to start of expression (the bracket)    code       points to start of expression (the bracket)
   options    pointer to the options (used to check casing changes)  
7127    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
7128    
7129  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
7130  */  */
7131    
7132  static int  static int
7133  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
7134  {  {
7135  register int c = -1;  register int c = -1;
7136  do {  do {
7137     int d;     int d;
7138     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
7139       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
7140       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
7141     register int op = *scode;     register int op = *scode;
7142    
7143     switch(op)     switch(op)
# Line 6328  do { Line 7146  do {
7146       return -1;       return -1;
7147    
7148       case OP_BRA:       case OP_BRA:
7149         case OP_BRAPOS:
7150       case OP_CBRA:       case OP_CBRA:
7151         case OP_SCBRA:
7152         case OP_CBRAPOS:
7153         case OP_SCBRAPOS:
7154       case OP_ASSERT:       case OP_ASSERT:
7155       case OP_ONCE:       case OP_ONCE:
7156         case OP_ONCE_NC:
7157       case OP_COND:       cas