/[pcre]/code/branches/pcre16/pcre_compile.c
ViewVC logotype

Diff of /code/branches/pcre16/pcre_compile.c

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

revision 528 by ph10, Sat May 29 16:40:22 2010 UTC revision 749 by ph10, Fri Nov 18 10:36:45 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 261  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  /* 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  substitutes must be in the order of the names, defined above, and there are
266  both positive and negative cases. NULL means no substitute. */  both positive and negative cases. NULL means no substitute. */
267    
268  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 272  static const uschar *substitutes[] = { Line 272  static const uschar *substitutes[] = {
272    (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */    (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273    (uschar *)"\\p{Xsp}",   /* \s */    (uschar *)"\\p{Xsp}",   /* \s */
274    (uschar *)"\\P{Xwd}",   /* \W */    (uschar *)"\\P{Xwd}",   /* \W */
275    (uschar *)"\\p{Xwd}"    /* \w */    (uschar *)"\\p{Xwd}"    /* \w */
276  };  };
277    
278  static const uschar *posix_substitutes[] = {  static const uschar *posix_substitutes[] = {
279    (uschar *)"\\p{L}",     /* alpha */    (uschar *)"\\p{L}",     /* alpha */
280    (uschar *)"\\p{Ll}",    /* lower */    (uschar *)"\\p{Ll}",    /* lower */
281    (uschar *)"\\p{Lu}",    /* upper */    (uschar *)"\\p{Lu}",    /* upper */
282    (uschar *)"\\p{Xan}",   /* alnum */    (uschar *)"\\p{Xan}",   /* alnum */
283    NULL,                   /* ascii */    NULL,                   /* ascii */
284    (uschar *)"\\h",        /* blank */    (uschar *)"\\h",        /* blank */
285    NULL,                   /* cntrl */    NULL,                   /* cntrl */
# Line 289  static const uschar *posix_substitutes[] Line 289  static const uschar *posix_substitutes[]
289    NULL,                   /* punct */    NULL,                   /* punct */
290    (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */    (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291    (uschar *)"\\p{Xwd}",   /* word */    (uschar *)"\\p{Xwd}",   /* word */
292    NULL,                   /* xdigit */    NULL,                   /* xdigit */
293    /* Negated cases */    /* Negated cases */
294    (uschar *)"\\P{L}",     /* ^alpha */    (uschar *)"\\P{L}",     /* ^alpha */
295    (uschar *)"\\P{Ll}",    /* ^lower */    (uschar *)"\\P{Ll}",    /* ^lower */
296    (uschar *)"\\P{Lu}",    /* ^upper */    (uschar *)"\\P{Lu}",    /* ^upper */
297    (uschar *)"\\P{Xan}",   /* ^alnum */    (uschar *)"\\P{Xan}",   /* ^alnum */
298    NULL,                   /* ^ascii */    NULL,                   /* ^ascii */
299    (uschar *)"\\H",        /* ^blank */    (uschar *)"\\H",        /* ^blank */
300    NULL,                   /* ^cntrl */    NULL,                   /* ^cntrl */
# Line 304  static const uschar *posix_substitutes[] Line 304  static const uschar *posix_substitutes[]
304    NULL,                   /* ^punct */    NULL,                   /* ^punct */
305    (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */    (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306    (uschar *)"\\P{Xwd}",   /* ^word */    (uschar *)"\\P{Xwd}",   /* ^word */
307    NULL                    /* ^xdigit */    NULL                    /* ^xdigit */
308  };  };
309  #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310  #endif  #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 393  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"
# Line 407  static const char error_texts[] = Line 407  static const char error_texts[] =
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"    "(*MARK) must have an argument\0"
410    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413      /* 70 */
414      "internal error: unknown opcode in find_fixedlength()\0"
415    ;    ;
416    
417  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 544  static const unsigned char ebcdic_charta Line 548  static const unsigned char ebcdic_charta
548  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
549    
550  static BOOL  static BOOL
551    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
552      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
553    
554    
# Line 576  return s; Line 580  return s;
580    
581    
582  /*************************************************  /*************************************************
583    *            Check for counted repeat            *
584    *************************************************/
585    
586    /* This function is called when a '{' is encountered in a place where it might
587    start a quantifier. It looks ahead to see if it really is a quantifier or not.
588    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
589    where the ddds are digits.
590    
591    Arguments:
592      p         pointer to the first char after '{'
593    
594    Returns:    TRUE or FALSE
595    */
596    
597    static BOOL
598    is_counted_repeat(const uschar *p)
599    {
600    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
601    while ((digitab[*p] & ctype_digit) != 0) p++;
602    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
603    
604    if (*p++ != CHAR_COMMA) return FALSE;
605    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
606    
607    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
608    while ((digitab[*p] & ctype_digit) != 0) p++;
609    
610    return (*p == CHAR_RIGHT_CURLY_BRACKET);
611    }
612    
613    
614    
615    /*************************************************
616  *            Handle escapes                      *  *            Handle escapes                      *
617  *************************************************/  *************************************************/
618    
# Line 641  else Line 678  else
678    
679      case CHAR_l:      case CHAR_l:
680      case CHAR_L:      case CHAR_L:
681        *errorcodeptr = ERR37;
682        break;
683    
684      case CHAR_u:      case CHAR_u:
685        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
686          {
687          /* In JavaScript, \u must be followed by four hexadecimal numbers.
688          Otherwise it is a lowercase u letter. */
689          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
690               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
691            {
692            c = 0;
693            for (i = 0; i < 4; ++i)
694              {
695              register int cc = *(++ptr);
696    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
697              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
698              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
699    #else           /* EBCDIC coding */
700              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
701              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
702    #endif
703              }
704            }
705          }
706        else
707          *errorcodeptr = ERR37;
708        break;
709    
710      case CHAR_U:      case CHAR_U:
711      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
712        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
713      break;      break;
714    
715      /* \g must be followed by one of a number of specific things:      /* In a character class, \g is just a literal "g". Outside a character
716        class, \g must be followed by one of a number of specific things:
717    
718      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
719      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 663  else Line 730  else
730      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
731    
732      case CHAR_g:      case CHAR_g:
733        if (isclass) break;
734      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
735        {        {
736        c = -ESC_g;        c = -ESC_g;
# Line 791  else Line 859  else
859      treated as a data character. */      treated as a data character. */
860    
861      case CHAR_x:      case CHAR_x:
862        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
863          {
864          /* In JavaScript, \x must be followed by two hexadecimal numbers.
865          Otherwise it is a lowercase x letter. */
866          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
867            {
868            c = 0;
869            for (i = 0; i < 2; ++i)
870              {
871              register int cc = *(++ptr);
872    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
873              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
874              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
875    #else           /* EBCDIC coding */
876              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
877              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
878    #endif
879              }
880            }
881          break;
882          }
883    
884      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
885        {        {
886        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
# Line 841  else Line 931  else
931      break;      break;
932    
933      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
934      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
935        coding is ASCII-specific, but then the whole concept of \cx is
936      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
937    
938      case CHAR_c:      case CHAR_c:
# Line 851  else Line 942  else
942        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
943        break;        break;
944        }        }
945    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
946  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
947          {
948          *errorcodeptr = ERR68;
949          break;
950          }
951      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
952      c ^= 0x40;      c ^= 0x40;
953  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
954      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
955      c ^= 0xC0;      c ^= 0xC0;
956  #endif  #endif
# Line 879  else Line 974  else
974    }    }
975    
976  /* Perl supports \N{name} for character names, as well as plain \N for "not  /* Perl supports \N{name} for character names, as well as plain \N for "not
977  newline". PCRE does not support \N{name}. */  newline". PCRE does not support \N{name}. However, it does support
978    quantification such as \N{2,3}. */
979    
980  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
981         !is_counted_repeat(ptr+2))
982    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
983    
984  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
# Line 991  return -1; Line 1088  return -1;
1088    
1089    
1090  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if (*p++ != CHAR_COMMA) return FALSE;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == CHAR_RIGHT_CURLY_BRACKET);  
 }  
   
   
   
 /*************************************************  
1091  *         Read repeat counts                     *  *         Read repeat counts                     *
1092  *************************************************/  *************************************************/
1093    
# Line 1099  top-level call starts at the beginning o Line 1163  top-level call starts at the beginning o
1163  start at a parenthesis. It scans along a pattern's text looking for capturing  start at a parenthesis. It scans along a pattern's text looking for capturing
1164  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1165  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1166  returns when it reaches a given numbered subpattern. We know that if (?P< is  returns when it reaches a given numbered subpattern. Recursion is used to keep
1167  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1168  first pass. Recursion is used to keep track of subpatterns that reset the  
1169  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1170    that if (?< or (?' or (?P< is encountered, the name will be correctly
1171    terminated because that is checked in the first pass. There is now one call to
1172    this function in the first pass, to check for a recursive back reference by
1173    name (so that we can make the whole group atomic). In this case, we need check
1174    only up to the current position in the pattern, and that is still OK because
1175    and previous occurrences will have been checked. To make this work, the test
1176    for "end of pattern" is a check against cd->end_pattern in the main loop,
1177    instead of looking for a binary zero. This means that the special first-pass
1178    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1179    processing items within the loop are OK, because afterwards the main loop will
1180    terminate.)
1181    
1182  Arguments:  Arguments:
1183    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1110  Arguments: Line 1185  Arguments:
1185    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1186    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1187    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1188      utf8         TRUE if we are in UTF-8 mode
1189    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1190    
1191  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
# Line 1117  Returns:       the number of the named s Line 1193  Returns:       the number of the named s
1193    
1194  static int  static int
1195  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1196    BOOL xmode, int *count)    BOOL xmode, BOOL utf8, int *count)
1197  {  {
1198  uschar *ptr = *ptrptr;  uschar *ptr = *ptrptr;
1199  int start_count = *count;  int start_count = *count;
# Line 1129  dealing with. The very first call may no Line 1205  dealing with. The very first call may no
1205    
1206  if (ptr[0] == CHAR_LEFT_PARENTHESIS)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1207    {    {
1208    if (ptr[1] == CHAR_QUESTION_MARK &&    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1209        ptr[2] == CHAR_VERTICAL_LINE)  
1210      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1211    
1212      /* Handle a normal, unnamed capturing parenthesis. */
1213    
1214      else if (ptr[1] != CHAR_QUESTION_MARK)
1215        {
1216        *count += 1;
1217        if (name == NULL && *count == lorn) return *count;
1218        ptr++;
1219        }
1220    
1221      /* All cases now have (? at the start. Remember when we are in a group
1222      where the parenthesis numbers are duplicated. */
1223    
1224      else if (ptr[2] == CHAR_VERTICAL_LINE)
1225      {      {
1226      ptr += 3;      ptr += 3;
1227      dup_parens = TRUE;      dup_parens = TRUE;
1228      }      }
1229    
1230    /* Handle a normal, unnamed capturing parenthesis */    /* Handle comments; all characters are allowed until a ket is reached. */
1231    
1232    else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)    else if (ptr[2] == CHAR_NUMBER_SIGN)
1233      {      {
1234      *count += 1;      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1235      if (name == NULL && *count == lorn) return *count;      goto FAIL_EXIT;
     ptr++;  
1236      }      }
1237    
1238    /* Handle a condition. If it is an assertion, just carry on so that it    /* Handle a condition. If it is an assertion, just carry on so that it
1239    is processed as normal. If not, skip to the closing parenthesis of the    is processed as normal. If not, skip to the closing parenthesis of the
1240    condition (there can't be any nested parens. */    condition (there can't be any nested parens). */
1241    
1242    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1243      {      {
# Line 1159  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1249  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1249        }        }
1250      }      }
1251    
1252    /* We have either (? or (* and not a condition */    /* Start with (? but not a condition. */
1253    
1254    else    else
1255      {      {
# Line 1188  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1278  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1278    }    }
1279    
1280  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1281  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1282    first-pass call when this value is temporarily adjusted to stop at the current
1283    position. So DO NOT change this to a test for binary zero. */
1284    
1285  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1286    {    {
1287    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1288    
# Line 1264  for (; *ptr != 0; ptr++) Line 1356  for (; *ptr != 0; ptr++)
1356    
1357    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1358      {      {
1359      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1360        while (*ptr != 0)
1361          {
1362          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1363          ptr++;
1364    #ifdef SUPPORT_UTF8
1365          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1366    #endif
1367          }
1368      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1369      continue;      continue;
1370      }      }
# Line 1273  for (; *ptr != 0; ptr++) Line 1373  for (; *ptr != 0; ptr++)
1373    
1374    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1375      {      {
1376      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1377      if (rc > 0) return rc;      if (rc > 0) return rc;
1378      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1379      }      }
# Line 1281  for (; *ptr != 0; ptr++) Line 1381  for (; *ptr != 0; ptr++)
1381    else if (*ptr == CHAR_RIGHT_PARENTHESIS)    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1382      {      {
1383      if (dup_parens && *count < hwm_count) *count = hwm_count;      if (dup_parens && *count < hwm_count) *count = hwm_count;
1384      *ptrptr = ptr;      goto FAIL_EXIT;
     return -1;  
1385      }      }
1386    
1387    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
# Line 1320  Arguments: Line 1419  Arguments:
1419    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1420    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1421    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1422      utf8         TRUE if we are in UTF-8 mode
1423    
1424  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1425  */  */
1426    
1427  static int  static int
1428  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1429      BOOL utf8)
1430  {  {
1431  uschar *ptr = (uschar *)cd->start_pattern;  uschar *ptr = (uschar *)cd->start_pattern;
1432  int count = 0;  int count = 0;
# Line 1338  matching closing parens. That is why we Line 1439  matching closing parens. That is why we
1439    
1440  for (;;)  for (;;)
1441    {    {
1442    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1443    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1444    }    }
1445    
# Line 1354  return rc; Line 1455  return rc;
1455    
1456  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1457  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1458  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1459  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1460  assertions, and also the \b assertion; for others it does not.  does not.
1461    
1462  Arguments:  Arguments:
1463    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1464    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1465    
1466  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1467  */  */
1468    
1469  static const uschar*  static const uschar*
1470  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1471  {  {
1472  for (;;)  for (;;)
1473    {    {
1474    switch ((int)*code)    switch ((int)*code)
1475      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1476      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1477      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1478      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1431  and doing the check at the end; a flag s Line 1522  and doing the check at the end; a flag s
1522    
1523  Arguments:  Arguments:
1524    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1525    options  the compiling options    utf8     TRUE in UTF-8 mode
1526    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1527    cd       the "compile data" structure    cd       the "compile data" structure
1528    
# Line 1439  Returns:   the fixed length, Line 1530  Returns:   the fixed length,
1530               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1531               or -2 if \C was encountered               or -2 if \C was encountered
1532               or -3 if an OP_RECURSE item was encountered and atend is FALSE               or -3 if an OP_RECURSE item was encountered and atend is FALSE
1533                 or -4 if an unknown opcode was encountered (internal error)
1534  */  */
1535    
1536  static int  static int
1537  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1538  {  {
1539  int length = -1;  int length = -1;
1540    
# Line 1459  for (;;) Line 1551  for (;;)
1551    register int op = *cc;    register int op = *cc;
1552    switch (op)    switch (op)
1553      {      {
1554        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1555        OP_BRA (normal non-capturing bracket) because the other variants of these
1556        opcodes are all concerned with unlimited repeated groups, which of course
1557        are not of fixed length. */
1558    
1559      case OP_CBRA:      case OP_CBRA:
1560      case OP_BRA:      case OP_BRA:
1561      case OP_ONCE:      case OP_ONCE:
1562        case OP_ONCE_NC:
1563      case OP_COND:      case OP_COND:
1564      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1565      if (d < 0) return d;      if (d < 0) return d;
1566      branchlength += d;      branchlength += d;
1567      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1568      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1569      break;      break;
1570    
1571      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested call.
1572      call. If it's ALT it is an alternation in a nested call. If it is      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1573      END it's the end of the outer call. All can be handled by the same code. */      an ALT. If it is END it's the end of the outer call. All can be handled by
1574        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1575        because they all imply an unlimited repeat. */
1576    
1577      case OP_ALT:      case OP_ALT:
1578      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1579      case OP_END:      case OP_END:
1580        case OP_ACCEPT:
1581        case OP_ASSERT_ACCEPT:
1582      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1583        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1584      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1495  for (;;) Line 1595  for (;;)
1595      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1596      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1597      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1598      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1599      if (d < 0) return d;      if (d < 0) return d;
1600      branchlength += d;      branchlength += d;
1601      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1512  for (;;) Line 1612  for (;;)
1612    
1613      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1614    
1615      case OP_REVERSE:      case OP_MARK:
1616        case OP_PRUNE_ARG:
1617        case OP_SKIP_ARG:
1618        case OP_THEN_ARG:
1619        cc += cc[1] + _pcre_OP_lengths[*cc];
1620        break;
1621    
1622        case OP_CALLOUT:
1623        case OP_CIRC:
1624        case OP_CIRCM:
1625        case OP_CLOSE:
1626        case OP_COMMIT:
1627      case OP_CREF:      case OP_CREF:
     case OP_NCREF:  
     case OP_RREF:  
     case OP_NRREF:  
1628      case OP_DEF:      case OP_DEF:
1629      case OP_OPT:      case OP_DOLL:
1630      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
     case OP_SET_SOM:  
1631      case OP_EOD:      case OP_EOD:
1632      case OP_EODN:      case OP_EODN:
1633      case OP_CIRC:      case OP_FAIL:
1634      case OP_DOLL:      case OP_NCREF:
1635        case OP_NRREF:
1636      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1637        case OP_PRUNE:
1638        case OP_REVERSE:
1639        case OP_RREF:
1640        case OP_SET_SOM:
1641        case OP_SKIP:
1642        case OP_SOD:
1643        case OP_SOM:
1644        case OP_THEN:
1645      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1646      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
1647      break;      break;
# Line 1535  for (;;) Line 1649  for (;;)
1649      /* Handle literal characters */      /* Handle literal characters */
1650    
1651      case OP_CHAR:      case OP_CHAR:
1652      case OP_CHARNC:      case OP_CHARI:
1653      case OP_NOT:      case OP_NOT:
1654        case OP_NOTI:
1655      branchlength++;      branchlength++;
1656      cc += 2;      cc += 2;
1657  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1658      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1659  #endif  #endif
1660      break;      break;
1661    
# Line 1549  for (;;) Line 1663  for (;;)
1663      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1664    
1665      case OP_EXACT:      case OP_EXACT:
1666        case OP_EXACTI:
1667        case OP_NOTEXACT:
1668        case OP_NOTEXACTI:
1669      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1670      cc += 4;      cc += 4;
1671  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1672      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1673  #endif  #endif
1674      break;      break;
1675    
# Line 1570  for (;;) Line 1686  for (;;)
1686      cc += 2;      cc += 2;
1687      /* Fall through */      /* Fall through */
1688    
1689        case OP_HSPACE:
1690        case OP_VSPACE:
1691        case OP_NOT_HSPACE:
1692        case OP_NOT_VSPACE:
1693      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1694      case OP_DIGIT:      case OP_DIGIT:
1695      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1601  for (;;) Line 1721  for (;;)
1721    
1722      switch (*cc)      switch (*cc)
1723        {        {
1724          case OP_CRPLUS:
1725          case OP_CRMINPLUS:
1726        case OP_CRSTAR:        case OP_CRSTAR:
1727        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1728        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1621  for (;;) Line 1743  for (;;)
1743    
1744      /* Anything else is variable length */      /* Anything else is variable length */
1745    
1746      default:      case OP_ANYNL:
1747        case OP_BRAMINZERO:
1748        case OP_BRAPOS:
1749        case OP_BRAPOSZERO:
1750        case OP_BRAZERO:
1751        case OP_CBRAPOS:
1752        case OP_EXTUNI:
1753        case OP_KETRMAX:
1754        case OP_KETRMIN:
1755        case OP_KETRPOS:
1756        case OP_MINPLUS:
1757        case OP_MINPLUSI:
1758        case OP_MINQUERY:
1759        case OP_MINQUERYI:
1760        case OP_MINSTAR:
1761        case OP_MINSTARI:
1762        case OP_MINUPTO:
1763        case OP_MINUPTOI:
1764        case OP_NOTMINPLUS:
1765        case OP_NOTMINPLUSI:
1766        case OP_NOTMINQUERY:
1767        case OP_NOTMINQUERYI:
1768        case OP_NOTMINSTAR:
1769        case OP_NOTMINSTARI:
1770        case OP_NOTMINUPTO:
1771        case OP_NOTMINUPTOI:
1772        case OP_NOTPLUS:
1773        case OP_NOTPLUSI:
1774        case OP_NOTPOSPLUS:
1775        case OP_NOTPOSPLUSI:
1776        case OP_NOTPOSQUERY:
1777        case OP_NOTPOSQUERYI:
1778        case OP_NOTPOSSTAR:
1779        case OP_NOTPOSSTARI:
1780        case OP_NOTPOSUPTO:
1781        case OP_NOTPOSUPTOI:
1782        case OP_NOTQUERY:
1783        case OP_NOTQUERYI:
1784        case OP_NOTSTAR:
1785        case OP_NOTSTARI:
1786        case OP_NOTUPTO:
1787        case OP_NOTUPTOI:
1788        case OP_PLUS:
1789        case OP_PLUSI:
1790        case OP_POSPLUS:
1791        case OP_POSPLUSI:
1792        case OP_POSQUERY:
1793        case OP_POSQUERYI:
1794        case OP_POSSTAR:
1795        case OP_POSSTARI:
1796        case OP_POSUPTO:
1797        case OP_POSUPTOI:
1798        case OP_QUERY:
1799        case OP_QUERYI:
1800        case OP_REF:
1801        case OP_REFI:
1802        case OP_SBRA:
1803        case OP_SBRAPOS:
1804        case OP_SCBRA:
1805        case OP_SCBRAPOS:
1806        case OP_SCOND:
1807        case OP_SKIPZERO:
1808        case OP_STAR:
1809        case OP_STARI:
1810        case OP_TYPEMINPLUS:
1811        case OP_TYPEMINQUERY:
1812        case OP_TYPEMINSTAR:
1813        case OP_TYPEMINUPTO:
1814        case OP_TYPEPLUS:
1815        case OP_TYPEPOSPLUS:
1816        case OP_TYPEPOSQUERY:
1817        case OP_TYPEPOSSTAR:
1818        case OP_TYPEPOSUPTO:
1819        case OP_TYPEQUERY:
1820        case OP_TYPESTAR:
1821        case OP_TYPEUPTO:
1822        case OP_UPTO:
1823        case OP_UPTOI:
1824      return -1;      return -1;
1825    
1826        /* Catch unrecognized opcodes so that when new ones are added they
1827        are not forgotten, as has happened in the past. */
1828    
1829        default:
1830        return -4;
1831      }      }
1832    }    }
1833  /* Control never gets here */  /* Control never gets here */
# Line 1655  _pcre_find_bracket(const uschar *code, B Line 1860  _pcre_find_bracket(const uschar *code, B
1860  for (;;)  for (;;)
1861    {    {
1862    register int c = *code;    register int c = *code;
1863    
1864    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1865    
1866    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1673  for (;;) Line 1879  for (;;)
1879    
1880    /* Handle capturing bracket */    /* Handle capturing bracket */
1881    
1882    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1883               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1884      {      {
1885      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1886      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1711  for (;;) Line 1918  for (;;)
1918        case OP_MARK:        case OP_MARK:
1919        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
1920        case OP_SKIP_ARG:        case OP_SKIP_ARG:
1921          code += code[1];
1922          break;
1923    
1924        case OP_THEN_ARG:        case OP_THEN_ARG:
1925        code += code[1];        code += code[1];
1926        break;        break;
# Line 1728  for (;;) Line 1938  for (;;)
1938      if (utf8) switch(c)      if (utf8) switch(c)
1939        {        {
1940        case OP_CHAR:        case OP_CHAR:
1941        case OP_CHARNC:        case OP_CHARI:
1942        case OP_EXACT:        case OP_EXACT:
1943          case OP_EXACTI:
1944        case OP_UPTO:        case OP_UPTO:
1945          case OP_UPTOI:
1946        case OP_MINUPTO:        case OP_MINUPTO:
1947          case OP_MINUPTOI:
1948        case OP_POSUPTO:        case OP_POSUPTO:
1949          case OP_POSUPTOI:
1950        case OP_STAR:        case OP_STAR:
1951          case OP_STARI:
1952        case OP_MINSTAR:        case OP_MINSTAR:
1953          case OP_MINSTARI:
1954        case OP_POSSTAR:        case OP_POSSTAR:
1955          case OP_POSSTARI:
1956        case OP_PLUS:        case OP_PLUS:
1957          case OP_PLUSI:
1958        case OP_MINPLUS:        case OP_MINPLUS:
1959          case OP_MINPLUSI:
1960        case OP_POSPLUS:        case OP_POSPLUS:
1961          case OP_POSPLUSI:
1962        case OP_QUERY:        case OP_QUERY:
1963          case OP_QUERYI:
1964        case OP_MINQUERY:        case OP_MINQUERY:
1965          case OP_MINQUERYI:
1966        case OP_POSQUERY:        case OP_POSQUERY:
1967          case OP_POSQUERYI:
1968        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1969        break;        break;
1970        }        }
# Line 1814  for (;;) Line 2037  for (;;)
2037        case OP_MARK:        case OP_MARK:
2038        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2039        case OP_SKIP_ARG:        case OP_SKIP_ARG:
2040          code += code[1];
2041          break;
2042    
2043        case OP_THEN_ARG:        case OP_THEN_ARG:
2044        code += code[1];        code += code[1];
2045        break;        break;
# Line 1831  for (;;) Line 2057  for (;;)
2057      if (utf8) switch(c)      if (utf8) switch(c)
2058        {        {
2059        case OP_CHAR:        case OP_CHAR:
2060        case OP_CHARNC:        case OP_CHARI:
2061        case OP_EXACT:        case OP_EXACT:
2062          case OP_EXACTI:
2063        case OP_UPTO:        case OP_UPTO:
2064          case OP_UPTOI:
2065        case OP_MINUPTO:        case OP_MINUPTO:
2066          case OP_MINUPTOI:
2067        case OP_POSUPTO:        case OP_POSUPTO:
2068          case OP_POSUPTOI:
2069        case OP_STAR:        case OP_STAR:
2070          case OP_STARI:
2071        case OP_MINSTAR:        case OP_MINSTAR:
2072          case OP_MINSTARI:
2073        case OP_POSSTAR:        case OP_POSSTAR:
2074          case OP_POSSTARI:
2075        case OP_PLUS:        case OP_PLUS:
2076          case OP_PLUSI:
2077        case OP_MINPLUS:        case OP_MINPLUS:
2078          case OP_MINPLUSI:
2079        case OP_POSPLUS:        case OP_POSPLUS:
2080          case OP_POSPLUSI:
2081        case OP_QUERY:        case OP_QUERY:
2082          case OP_QUERYI:
2083        case OP_MINQUERY:        case OP_MINQUERY:
2084          case OP_MINQUERYI:
2085        case OP_POSQUERY:        case OP_POSQUERY:
2086          case OP_POSQUERYI:
2087        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
2088        break;        break;
2089        }        }
# Line 1883  could_be_empty_branch(const uschar *code Line 2122  could_be_empty_branch(const uschar *code
2122    compile_data *cd)    compile_data *cd)
2123  {  {
2124  register int c;  register int c;
2125  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2126       code < endcode;       code < endcode;
2127       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2128    {    {
2129    const uschar *ccode;    const uschar *ccode;
2130    
# Line 1901  for (code = first_significant_code(code Line 2140  for (code = first_significant_code(code
2140      continue;      continue;
2141      }      }
2142    
   /* Groups with zero repeats can of course be empty; skip them. */  
   
   if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)  
     {  
     code += _pcre_OP_lengths[c];  
     do code += GET(code, 1); while (*code == OP_ALT);  
     c = *code;  
     continue;  
     }  
   
2143    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
2144    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
2145      forward reference subroutine call, we can't. To detect forward reference
2146      we have to scan up the list that is kept in the workspace. This function is
2147      called only when doing the real compile, not during the pre-compile that
2148      measures the size of the compiled pattern. */
2149    
2150    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2151      {      {
2152      BOOL empty_branch = FALSE;      const uschar *scode;
2153      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
2154    
2155        /* Test for forward reference */
2156    
2157        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2158          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2159    
2160        /* Not a forward reference, test for completed backward reference */
2161    
2162        empty_branch = FALSE;
2163        scode = cd->start_code + GET(code, 1);
2164      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2165    
2166        /* Completed backwards reference */
2167    
2168      do      do
2169        {        {
2170        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf8, cd))
# Line 1929  for (code = first_significant_code(code Line 2175  for (code = first_significant_code(code
2175        scode += GET(scode, 1);        scode += GET(scode, 1);
2176        }        }
2177      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2178    
2179      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2180      continue;      continue;
2181      }      }
2182    
2183      /* Groups with zero repeats can of course be empty; skip them. */
2184    
2185      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2186          c == OP_BRAPOSZERO)
2187        {
2188        code += _pcre_OP_lengths[c];
2189        do code += GET(code, 1); while (*code == OP_ALT);
2190        c = *code;
2191        continue;
2192        }
2193    
2194      /* A nested group that is already marked as "could be empty" can just be
2195      skipped. */
2196    
2197      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2198          c == OP_SCBRA || c == OP_SCBRAPOS)
2199        {
2200        do code += GET(code, 1); while (*code == OP_ALT);
2201        c = *code;
2202        continue;
2203        }
2204    
2205    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2206    
2207    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2208          c == OP_CBRA || c == OP_CBRAPOS ||
2209          c == OP_ONCE || c == OP_ONCE_NC ||
2210          c == OP_COND)
2211      {      {
2212      BOOL empty_branch;      BOOL empty_branch;
2213      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2022  for (code = first_significant_code(code Line 2294  for (code = first_significant_code(code
2294      case OP_ALLANY:      case OP_ALLANY:
2295      case OP_ANYBYTE:      case OP_ANYBYTE:
2296      case OP_CHAR:      case OP_CHAR:
2297      case OP_CHARNC:      case OP_CHARI:
2298      case OP_NOT:      case OP_NOT:
2299        case OP_NOTI:
2300      case OP_PLUS:      case OP_PLUS:
2301      case OP_MINPLUS:      case OP_MINPLUS:
2302      case OP_POSPLUS:      case OP_POSPLUS:
# Line 2063  for (code = first_significant_code(code Line 2336  for (code = first_significant_code(code
2336      case OP_KET:      case OP_KET:
2337      case OP_KETRMAX:      case OP_KETRMAX:
2338      case OP_KETRMIN:      case OP_KETRMIN:
2339        case OP_KETRPOS:
2340      case OP_ALT:      case OP_ALT:
2341      return TRUE;      return TRUE;
2342    
# Line 2071  for (code = first_significant_code(code Line 2345  for (code = first_significant_code(code
2345    
2346  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2347      case OP_STAR:      case OP_STAR:
2348        case OP_STARI:
2349      case OP_MINSTAR:      case OP_MINSTAR:
2350        case OP_MINSTARI:
2351      case OP_POSSTAR:      case OP_POSSTAR:
2352        case OP_POSSTARI:
2353      case OP_QUERY:      case OP_QUERY:
2354        case OP_QUERYI:
2355      case OP_MINQUERY:      case OP_MINQUERY:
2356        case OP_MINQUERYI:
2357      case OP_POSQUERY:      case OP_POSQUERY:
2358        case OP_POSQUERYI:
2359      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2360      break;      break;
2361    
2362      case OP_UPTO:      case OP_UPTO:
2363        case OP_UPTOI:
2364      case OP_MINUPTO:      case OP_MINUPTO:
2365        case OP_MINUPTOI:
2366      case OP_POSUPTO:      case OP_POSUPTO:
2367        case OP_POSUPTOI:
2368      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2369      break;      break;
2370  #endif  #endif
# Line 2092  for (code = first_significant_code(code Line 2375  for (code = first_significant_code(code
2375      case OP_MARK:      case OP_MARK:
2376      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2377      case OP_SKIP_ARG:      case OP_SKIP_ARG:
2378        code += code[1];
2379        break;
2380    
2381      case OP_THEN_ARG:      case OP_THEN_ARG:
2382      code += code[1];      code += code[1];
2383      break;      break;
# Line 2116  return TRUE; Line 2402  return TRUE;
2402  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2403  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2404  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2405    This function is called only during the real compile, not during the
2406    pre-compile.
2407    
2408  Arguments:  Arguments:
2409    code        points to start of the recursion    code        points to start of the recursion
# Line 2166  where Perl recognizes it as the POSIX cl Line 2454  where Perl recognizes it as the POSIX cl
2454  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2455  I think.  I think.
2456    
2457    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2458    It seems that the appearance of a nested POSIX class supersedes an apparent
2459    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2460    a digit.
2461    
2462    In Perl, unescaped square brackets may also appear as part of class names. For
2463    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2464    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2465    seem right at all. PCRE does not allow closing square brackets in POSIX class
2466    names.
2467    
2468  Arguments:  Arguments:
2469    ptr      pointer to the initial [    ptr      pointer to the initial [
2470    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2180  int terminator;          /* Don't combin Line 2479  int terminator;          /* Don't combin
2479  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2480  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2481    {    {
2482    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2483        ptr++;
2484      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2485      else
2486      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2487      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2488        {        {
2489        *endptr = ptr;        *endptr = ptr;
2490        return TRUE;        return TRUE;
2491        }        }
2492        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2493             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2494              ptr[1] == CHAR_EQUALS_SIGN) &&
2495            check_posix_syntax(ptr, endptr))
2496          return FALSE;
2497      }      }
2498    }    }
2499  return FALSE;  return FALSE;
# Line 2316  auto_callout(uschar *code, const uschar Line 2622  auto_callout(uschar *code, const uschar
2622  {  {
2623  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2624  *code++ = 255;  *code++ = 255;
2625  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2626  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2627  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2628  }  }
2629    
# Line 2342  Returns:             nothing Line 2648  Returns:             nothing
2648  static void  static void
2649  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2650  {  {
2651  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2652  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2653  }  }
2654    
# Line 2392  for (++c; c <= d; c++) Line 2698  for (++c; c <= d; c++)
2698    
2699  return TRUE;  return TRUE;
2700  }  }
2701    
2702    
2703    
2704    /*************************************************
2705    *        Check a character and a property        *
2706    *************************************************/
2707    
2708    /* This function is called by check_auto_possessive() when a property item
2709    is adjacent to a fixed character.
2710    
2711    Arguments:
2712      c            the character
2713      ptype        the property type
2714      pdata        the data for the type
2715      negated      TRUE if it's a negated property (\P or \p{^)
2716    
2717    Returns:       TRUE if auto-possessifying is OK
2718    */
2719    
2720    static BOOL
2721    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2722    {
2723    const ucd_record *prop = GET_UCD(c);
2724    switch(ptype)
2725      {
2726      case PT_LAMP:
2727      return (prop->chartype == ucp_Lu ||
2728              prop->chartype == ucp_Ll ||
2729              prop->chartype == ucp_Lt) == negated;
2730    
2731      case PT_GC:
2732      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2733    
2734      case PT_PC:
2735      return (pdata == prop->chartype) == negated;
2736    
2737      case PT_SC:
2738      return (pdata == prop->script) == negated;
2739    
2740      /* These are specials */
2741    
2742      case PT_ALNUM:
2743      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2744              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2745    
2746      case PT_SPACE:    /* Perl space */
2747      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2748              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2749              == negated;
2750    
2751      case PT_PXSPACE:  /* POSIX space */
2752      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2753              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2754              c == CHAR_FF || c == CHAR_CR)
2755              == negated;
2756    
2757      case PT_WORD:
2758      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2759              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2760              c == CHAR_UNDERSCORE) == negated;
2761      }
2762    return FALSE;
2763    }
2764  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2765    
2766    
# Line 2405  whether the next thing could possibly ma Line 2774  whether the next thing could possibly ma
2774  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2775    
2776  Arguments:  Arguments:
2777    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2778    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2779    ptr           next character in pattern    ptr           next character in pattern
2780    options       options bits    options       options bits
2781    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2417  Returns:        TRUE if possessifying is Line 2784  Returns:        TRUE if possessifying is
2784  */  */
2785    
2786  static BOOL  static BOOL
2787  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2788    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2789  {  {
2790  int next;  int c, next;
2791    int op_code = *previous++;
2792    
2793  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2794    
# Line 2431  if ((options & PCRE_EXTENDED) != 0) Line 2799  if ((options & PCRE_EXTENDED) != 0)
2799      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2800      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2801        {        {
2802        while (*(++ptr) != 0)        ptr++;
2803          while (*ptr != 0)
2804            {
2805          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2806            ptr++;
2807    #ifdef SUPPORT_UTF8
2808            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2809    #endif
2810            }
2811        }        }
2812      else break;      else break;
2813      }      }
# Line 2468  if ((options & PCRE_EXTENDED) != 0) Line 2843  if ((options & PCRE_EXTENDED) != 0)
2843      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2844      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2845        {        {
2846        while (*(++ptr) != 0)        ptr++;
2847          while (*ptr != 0)
2848            {
2849          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2850            ptr++;
2851    #ifdef SUPPORT_UTF8
2852            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2853    #endif
2854            }
2855        }        }
2856      else break;      else break;
2857      }      }
# Line 2481  if (*ptr == CHAR_ASTERISK || *ptr == CHA Line 2863  if (*ptr == CHAR_ASTERISK || *ptr == CHA
2863    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2864      return FALSE;      return FALSE;
2865    
2866  /* Now compare the next item with the previous opcode. If the previous is a  /* Now compare the next item with the previous opcode. First, handle cases when
2867  positive single character match, "item" either contains the character or, if  the next item is a character. */
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
   
 /* Handle cases when the next item is a character. */  
2868    
2869  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2870    {    {
2871    case OP_CHAR:    case OP_CHAR:
2872  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2873    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2874  #else  #else
2875    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2876  #endif  #endif
2877    return item != next;    return c != next;
2878    
2879    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
2880    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
2881    high-valued characters. */    high-valued characters. */
2882    
2883    case OP_CHARNC:    case OP_CHARI:
2884  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2885    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2886    #else
2887      c = *previous;
2888  #endif  #endif
2889    if (item == next) return FALSE;    if (c == next) return FALSE;
2890  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2891    if (utf8)    if (utf8)
2892      {      {
# Line 2518  if (next >= 0) switch(op_code) Line 2897  if (next >= 0) switch(op_code)
2897  #else  #else
2898      othercase = NOTACHAR;      othercase = NOTACHAR;
2899  #endif  #endif
2900      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2901      }      }
2902    else    else
2903  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2904    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2905    
2906    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2907      opcodes are not used for multi-byte characters, because they are coded using
2908      an XCLASS instead. */
2909    
2910    case OP_NOT:    case OP_NOT:
2911    if (item == next) return TRUE;    return (c = *previous) == next;
2912    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2913      case OP_NOTI:
2914      if ((c = *previous) == next) return TRUE;
2915  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2916    if (utf8)    if (utf8)
2917      {      {
# Line 2539  if (next >= 0) switch(op_code) Line 2922  if (next >= 0) switch(op_code)
2922  #else  #else
2923      othercase = NOTACHAR;      othercase = NOTACHAR;
2924  #endif  #endif
2925      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2926      }      }
2927    else    else
2928  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2929    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2930    
2931    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2932    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2933    
2934    case OP_DIGIT:    case OP_DIGIT:
2935    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2611  if (next >= 0) switch(op_code) Line 2994  if (next >= 0) switch(op_code)
2994      return op_code != OP_NOT_VSPACE;      return op_code != OP_NOT_VSPACE;
2995      }      }
2996    
2997    #ifdef SUPPORT_UCP
2998      case OP_PROP:
2999      return check_char_prop(next, previous[0], previous[1], FALSE);
3000    
3001      case OP_NOTPROP:
3002      return check_char_prop(next, previous[0], previous[1], TRUE);
3003    #endif
3004    
3005    default:    default:
3006    return FALSE;    return FALSE;
3007    }    }
3008    
3009    
3010  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3011  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3012  generated only when PCRE_UCP is *not* set, that is, when only ASCII  generated only when PCRE_UCP is *not* set, that is, when only ASCII
3013  characteristics are recognized. */  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3014    replaced by OP_PROP codes when PCRE_UCP is set. */
3015    
3016  switch(op_code)  switch(op_code)
3017    {    {
3018    case OP_CHAR:    case OP_CHAR:
3019    case OP_CHARNC:    case OP_CHARI:
3020  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3021    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3022    #else
3023      c = *previous;
3024  #endif  #endif
3025    switch(-next)    switch(-next)
3026      {      {
3027      case ESC_d:      case ESC_d:
3028      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3029    
3030      case ESC_D:      case ESC_D:
3031      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3032    
3033      case ESC_s:      case ESC_s:
3034      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3035    
3036      case ESC_S:      case ESC_S:
3037      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3038    
3039      case ESC_w:      case ESC_w:
3040      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3041    
3042      case ESC_W:      case ESC_W:
3043      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3044    
3045      case ESC_h:      case ESC_h:
3046      case ESC_H:      case ESC_H:
3047      switch(item)      switch(c)
3048        {        {
3049        case 0x09:        case 0x09:
3050        case 0x20:        case 0x20:
# Line 2678  switch(op_code) Line 3072  switch(op_code)
3072    
3073      case ESC_v:      case ESC_v:
3074      case ESC_V:      case ESC_V:
3075      switch(item)      switch(c)
3076        {        {
3077        case 0x0a:        case 0x0a:
3078        case 0x0b:        case 0x0b:
# Line 2692  switch(op_code) Line 3086  switch(op_code)
3086        return -next == ESC_v;        return -next == ESC_v;
3087        }        }
3088    
3089        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3090        their substitutions and process them. The result will always be either
3091        -ESC_p or -ESC_P. Then fall through to process those values. */
3092    
3093    #ifdef SUPPORT_UCP
3094        case ESC_du:
3095        case ESC_DU:
3096        case ESC_wu:
3097        case ESC_WU:
3098        case ESC_su:
3099        case ESC_SU:
3100          {
3101          int temperrorcode = 0;
3102          ptr = substitutes[-next - ESC_DU];
3103          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3104          if (temperrorcode != 0) return FALSE;
3105          ptr++;    /* For compatibility */
3106          }
3107        /* Fall through */
3108    
3109        case ESC_p:
3110        case ESC_P:
3111          {
3112          int ptype, pdata, errorcodeptr;
3113          BOOL negated;
3114    
3115          ptr--;      /* Make ptr point at the p or P */
3116          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3117          if (ptype < 0) return FALSE;
3118          ptr++;      /* Point past the final curly ket */
3119    
3120          /* If the property item is optional, we have to give up. (When generated
3121          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3122          to the original \d etc. At this point, ptr will point to a zero byte. */
3123    
3124          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3125            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3126              return FALSE;
3127    
3128          /* Do the property check. */
3129    
3130          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3131          }
3132    #endif
3133    
3134      default:      default:
3135      return FALSE;      return FALSE;
3136      }      }
3137    
3138      /* In principle, support for Unicode properties should be integrated here as
3139      well. It means re-organizing the above code so as to get hold of the property
3140      values before switching on the op-code. However, I wonder how many patterns
3141      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3142      these op-codes are never generated.) */
3143    
3144    case OP_DIGIT:    case OP_DIGIT:
3145    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3146           next == -ESC_h || next == -ESC_v || next == -ESC_R;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
# Line 2710  switch(op_code) Line 3155  switch(op_code)
3155    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3156    
3157    case OP_HSPACE:    case OP_HSPACE:
3158    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3159           next == -ESC_w || next == -ESC_v || next == -ESC_R;           next == -ESC_w || next == -ESC_v || next == -ESC_R;
3160    
3161    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3162    return next == -ESC_h;    return next == -ESC_h;
3163    
3164    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3165    case OP_ANYNL:    case OP_ANYNL:
3166    case OP_VSPACE:    case OP_VSPACE:
3167    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3168    
# Line 2725  switch(op_code) Line 3170  switch(op_code)
3170    return next == -ESC_v || next == -ESC_R;    return next == -ESC_v || next == -ESC_R;
3171    
3172    case OP_WORDCHAR:    case OP_WORDCHAR:
3173    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3174           next == -ESC_v || next == -ESC_R;           next == -ESC_v || next == -ESC_R;
3175    
3176    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
# Line 2758  Arguments: Line 3203  Arguments:
3203    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3204    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3205    bcptr          points to current branch chain    bcptr          points to current branch chain
3206      cond_depth     conditional nesting depth
3207    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3208    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3209                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2769  Returns:         TRUE on success Line 3215  Returns:         TRUE on success
3215  static BOOL  static BOOL
3216  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3217    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3218    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3219  {  {
3220  int repeat_type, op_type;  int repeat_type, op_type;
3221  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2778  int greedy_default, greedy_non_default; Line 3224  int greedy_default, greedy_non_default;
3224  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3225  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3226  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3227  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3228  int after_manual_callout = 0;  int after_manual_callout = 0;
3229  int length_prevgroup = 0;  int length_prevgroup = 0;
3230  register int c;  register int c;
# Line 2796  uschar *previous_callout = NULL; Line 3242  uschar *previous_callout = NULL;
3242  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3243  uschar classbits[32];  uschar classbits[32];
3244    
3245    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3246    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3247    dynamically as we process the pattern. */
3248    
3249  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3250  BOOL class_utf8;  BOOL class_utf8;
3251  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2804  uschar *class_utf8data_base; Line 3254  uschar *class_utf8data_base;
3254  uschar utf8_char[6];  uschar utf8_char[6];
3255  #else  #else
3256  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3257  #endif  #endif
3258    
3259  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 2855  for (;; ptr++) Line 3304  for (;; ptr++)
3304    int subfirstbyte;    int subfirstbyte;
3305    int terminator;    int terminator;
3306    int mclength;    int mclength;
3307      int tempbracount;
3308    uschar mcbuffer[8];    uschar mcbuffer[8];
3309    
3310    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3311    
3312    c = *ptr;    c = *ptr;
3313    
3314    /* If we are at the end of a nested substitution, revert to the outer level    /* If we are at the end of a nested substitution, revert to the outer level
3315    string. Nesting only happens one level deep. */    string. Nesting only happens one level deep. */
3316    
3317    if (c == 0 && nestptr != NULL)    if (c == 0 && nestptr != NULL)
# Line 2901  for (;; ptr++) Line 3351  for (;; ptr++)
3351        goto FAILED;        goto FAILED;
3352        }        }
3353    
3354      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3355      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3356    
3357      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
# Line 2976  for (;; ptr++) Line 3426  for (;; ptr++)
3426      previous_callout = NULL;      previous_callout = NULL;
3427      }      }
3428    
3429    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3430    
3431    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3432      {      {
3433      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3434      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3435        {        {
3436        while (*(++ptr) != 0)        ptr++;
3437          while (*ptr != 0)
3438          {          {
3439          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3440            ptr++;
3441    #ifdef SUPPORT_UTF8
3442            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3443    #endif
3444          }          }
3445        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3446    
# Line 3019  for (;; ptr++) Line 3474  for (;; ptr++)
3474          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3475          goto FAILED;          goto FAILED;
3476          }          }
3477        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3478        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3479        }        }
3480      return TRUE;      return TRUE;
# Line 3030  for (;; ptr++) Line 3485  for (;; ptr++)
3485      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3486    
3487      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3488        previous = NULL;
3489      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3490        {        {
3491        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3492          *code++ = OP_CIRCM;
3493        }        }
3494      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3495      break;      break;
3496    
3497      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3498      previous = NULL;      previous = NULL;
3499      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3500      break;      break;
3501    
3502      /* 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 3168  for (;; ptr++) Line 3624  for (;; ptr++)
3624          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3625          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3626          }          }
3627    
3628        /* In the pre-compile phase, accumulate the length of any UTF-8 extra        /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3629        data and reset the pointer. This is so that very large classes that        data and reset the pointer. This is so that very large classes that
3630        contain a zillion UTF-8 characters no longer overwrite the work space        contain a zillion UTF-8 characters no longer overwrite the work space
# Line 3224  for (;; ptr++) Line 3680  for (;; ptr++)
3680            ptr++;            ptr++;
3681            }            }
3682    
3683          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3684          if (posix_class < 0)          if (posix_class < 0)
3685            {            {
3686            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 3237  for (;; ptr++) Line 3693  for (;; ptr++)
3693    
3694          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3695            posix_class = 0;            posix_class = 0;
3696    
3697          /* When PCRE_UCP is set, some of the POSIX classes are converted to          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3698          different escape sequences that use Unicode properties. */          different escape sequences that use Unicode properties. */
3699    
3700  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3701          if ((options & PCRE_UCP) != 0)          if ((options & PCRE_UCP) != 0)
3702            {            {
3703            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3704            if (posix_substitutes[pc] != NULL)            if (posix_substitutes[pc] != NULL)
3705              {              {
3706              nestptr = tempptr + 1;              nestptr = tempptr + 1;
3707              ptr = posix_substitutes[pc] - 1;              ptr = posix_substitutes[pc] - 1;
3708              continue;              continue;
3709              }              }
3710            }            }
3711  #endif  #endif
3712          /* In the non-UCP case, we build the bit map for the POSIX class in a          /* In the non-UCP case, we build the bit map for the POSIX class in a
3713          chunk of local store because we may be adding and subtracting from it,          chunk of local store because we may be adding and subtracting from it,
3714          and we don't want to subtract bits that may be in the main map already.          and we don't want to subtract bits that may be in the main map already.
# Line 3339  for (;; ptr++) Line 3795  for (;; ptr++)
3795              case ESC_SU:              case ESC_SU:
3796              nestptr = ptr;              nestptr = ptr;
3797              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3798              class_charcount -= 2;                /* Undo! */              class_charcount -= 2;                /* Undo! */
3799              continue;              continue;
3800  #endif  #endif
3801              case ESC_d:              case ESC_d:
# Line 3360  for (;; ptr++) Line 3816  for (;; ptr++)
3816              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3817              continue;              continue;
3818    
3819                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3820                if it was previously set by something earlier in the character
3821                class. */
3822    
3823              case ESC_s:              case ESC_s:
3824              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3825              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3826                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3827              continue;              continue;
3828    
3829              case ESC_S:              case ESC_S:
# Line 3781  for (;; ptr++) Line 4242  for (;; ptr++)
4242    
4243      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
4244      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4245      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4246      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4247    
4248      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
4249      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.
4250      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
4251      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
4252      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
4253      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4254    
4255  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4256      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
4257        (!utf8 || !negate_class || class_lastchar < 128))        (!utf8 || !negate_class || class_lastchar < 128))
# Line 3800  for (;; ptr++) Line 4261  for (;; ptr++)
4261        {        {
4262        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4263    
4264        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4265    
4266        if (negate_class)        if (negate_class)
4267          {          {
4268          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4269          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4270          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4271          *code++ = class_lastchar;          *code++ = class_lastchar;
4272          break;          break;
4273          }          }
# Line 3870  for (;; ptr++) Line 4331  for (;; ptr++)
4331        }        }
4332  #endif  #endif
4333    
4334      /* If there are no characters > 255, or they are all to be included or      /* If there are no characters > 255, or they are all to be included or
4335      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
4336      whole class was negated and whether there were negative specials such as \S      whole class was negated and whether there were negative specials such as \S
4337      (non-UCP) in the class. Then copy the 32-byte map into the code vector,      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
# Line 3934  for (;; ptr++) Line 4395  for (;; ptr++)
4395      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4396      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4397    
4398      /* 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
4399      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4400    
4401      tempcode = previous;      tempcode = previous;
4402    
# Line 3958  for (;; ptr++) Line 4419  for (;; ptr++)
4419        }        }
4420      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4421    
4422        /* If previous was a recursion call, wrap it in atomic brackets so that
4423        previous becomes the atomic group. All recursions were so wrapped in the
4424        past, but it no longer happens for non-repeated recursions. In fact, the
4425        repeated ones could be re-implemented independently so as not to need this,
4426        but for the moment we rely on the code for repeating groups. */
4427    
4428        if (*previous == OP_RECURSE)
4429          {
4430          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4431          *previous = OP_ONCE;
4432          PUT(previous, 1, 2 + 2*LINK_SIZE);
4433          previous[2 + 2*LINK_SIZE] = OP_KET;
4434          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4435          code += 2 + 2 * LINK_SIZE;
4436          length_prevgroup = 3 + 3*LINK_SIZE;
4437    
4438          /* When actually compiling, we need to check whether this was a forward
4439          reference, and if so, adjust the offset. */
4440    
4441          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4442            {
4443            int offset = GET(cd->hwm, -LINK_SIZE);
4444            if (offset == previous + 1 - cd->start_code)
4445              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4446            }
4447          }
4448    
4449        /* Now handle repetition for the different types of item. */
4450    
4451      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4452      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
4453      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
4454      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
4455      instead.  */      instead.  */
4456    
4457      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4458        {        {
4459          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4460    
4461        /* 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
4462        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
4463        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 3998  for (;; ptr++) Line 4490  for (;; ptr++)
4490    
4491        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4492            repeat_max < 0 &&            repeat_max < 0 &&
4493            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4494          {          {
4495          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4496          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4011  for (;; ptr++) Line 4502  for (;; ptr++)
4502      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4503      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-
4504      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4505      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
4506      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4507    
4508      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4509        {        {
4510        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4511        c = previous[1];        c = previous[1];
4512        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4513            repeat_max < 0 &&            repeat_max < 0 &&
4514            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4515          {          {
4516          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4517          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4044  for (;; ptr++) Line 4535  for (;; ptr++)
4535    
4536        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4537            repeat_max < 0 &&            repeat_max < 0 &&
4538            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4539          {          {
4540          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4541          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4213  for (;; ptr++) Line 4704  for (;; ptr++)
4704  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4705               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4706  #endif  #endif
4707               *previous == OP_REF)               *previous == OP_REF ||
4708                 *previous == OP_REFI)
4709        {        {
4710        if (repeat_max == 0)        if (repeat_max == 0)
4711          {          {
# Line 4247  for (;; ptr++) Line 4739  for (;; ptr++)
4739        }        }
4740    
4741      /* 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
4742      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4743        opcodes such as BRA and CBRA, as this is the place where they get converted
4744        into the more special varieties such as BRAPOS and SBRA. A test for >=
4745        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4746        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4747        repetition of assertions, but now it does, for Perl compatibility. */
4748    
4749      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4750        {        {
4751        register int i;        register int i;
4752        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4753        uschar *bralink = NULL;        uschar *bralink = NULL;
4754          uschar *brazeroptr = NULL;
4755    
4756        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4757          we just ignore the repeat. */
4758    
4759        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4760          {          goto END_REPEAT;
4761          *errorcodeptr = ERR55;  
4762          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4763          }        use of repetition is in cases when the assertion is optional. Therefore,
4764          if the minimum is greater than zero, just ignore the repeat. If the
4765          maximum is not not zero or one, set it to 1. */
4766    
4767        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4768        by scanning through from the start, and compute the offset back to it          {
4769        from the current code pointer. There may be an OP_OPT setting following          if (repeat_min > 0) goto END_REPEAT;
4770        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;  
4771          }          }
4772    
4773        /* 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 4296  for (;; ptr++) Line 4788  for (;; ptr++)
4788          **   goto END_REPEAT;          **   goto END_REPEAT;
4789          **   }          **   }
4790    
4791          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
4792          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
4793          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
4794          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4795            selectively.
4796    
4797          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
4798          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 4319  for (;; ptr++) Line 4812  for (;; ptr++)
4812              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4813              goto END_REPEAT;              goto END_REPEAT;
4814              }              }
4815              brazeroptr = previous;    /* Save for possessive optimizing */
4816            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4817            }            }
4818    
# Line 4343  for (;; ptr++) Line 4837  for (;; ptr++)
4837            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4838            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4839    
4840            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4841            bralink = previous;            bralink = previous;
4842            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4843            }            }
# Line 4452  for (;; ptr++) Line 4946  for (;; ptr++)
4946              {              {
4947              int offset;              int offset;
4948              *code++ = OP_BRA;              *code++ = OP_BRA;
4949              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4950              bralink = code;              bralink = code;
4951              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4952              }              }
# Line 4473  for (;; ptr++) Line 4967  for (;; ptr++)
4967          while (bralink != NULL)          while (bralink != NULL)
4968            {            {
4969            int oldlinkoffset;            int oldlinkoffset;
4970            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4971            uschar *bra = code - offset;            uschar *bra = code - offset;
4972            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4973            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4483  for (;; ptr++) Line 4977  for (;; ptr++)
4977            }            }
4978          }          }
4979    
4980        /* 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
4981        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4982        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
4983        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4984          deal with possessive ONCEs specially.
4985        Then, when we are doing the actual compile phase, check to see whether  
4986        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
4987          whether this group is one that could match an empty string. If so,
4988        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
4989        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
4990        atomic groups at runtime, but in a different way.] */        groups at runtime, but in a different way.]
4991    
4992          Then, if the quantifier was possessive and the bracket is not a
4993          conditional, we convert the BRA code to the POS form, and the KET code to
4994          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
4995          subpattern at both the start and at the end.) The use of special opcodes
4996          makes it possible to reduce greatly the stack usage in pcre_exec(). If
4997          the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO. Then
4998          cancel the possessive flag so that the default action below, of wrapping
4999          everything inside atomic brackets, does not happen. */
5000    
5001        else        else
5002          {          {
5003          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
5004          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
5005          *ketcode = OP_KETRMAX + repeat_type;  
5006          if (lengthptr == NULL && *bracode != OP_ONCE)          /* Convert possessive ONCE brackets to non-capturing */
5007    
5008            if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5009                possessive_quantifier) *bracode = OP_BRA;
5010    
5011            /* For non-possessive ONCE brackets, all we need to do is to
5012            set the KET. */
5013    
5014            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5015              *ketcode = OP_KETRMAX + repeat_type;
5016    
5017            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5018            converted to non-capturing above). */
5019    
5020            else
5021            {            {
5022            uschar *scode = bracode;            /* In the compile phase, check for empty string matching. */
5023            do  
5024              if (lengthptr == NULL)
5025              {              {
5026              if (could_be_empty_branch(scode, ketcode, utf8, cd))              uschar *scode = bracode;
5027                do
5028                {                {
5029                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
5030                break;                  {
5031                    *bracode += OP_SBRA - OP_BRA;
5032                    break;
5033                    }
5034                  scode += GET(scode, 1);
5035                }                }
5036              scode += GET(scode, 1);              while (*scode == OP_ALT);
5037              }              }
5038            while (*scode == OP_ALT);  
5039              /* Handle possessive quantifiers. */
5040    
5041              if (possessive_quantifier)
5042                {
5043                /* For COND brackets, we wrap the whole thing in a possessively
5044                repeated non-capturing bracket, because we have not invented POS
5045                versions of the COND opcodes. Because we are moving code along, we
5046                must ensure that any pending recursive references are updated. */
5047    
5048                if (*bracode == OP_COND || *bracode == OP_SCOND)
5049                  {
5050                  int nlen = (int)(code - bracode);
5051                  *code = OP_END;
5052                  adjust_recurse(bracode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5053                  memmove(bracode + 1+LINK_SIZE, bracode, nlen);
5054                  code += 1 + LINK_SIZE;
5055                  nlen += 1 + LINK_SIZE;
5056                  *bracode = OP_BRAPOS;
5057                  *code++ = OP_KETRPOS;
5058                  PUTINC(code, 0, nlen);
5059                  PUT(bracode, 1, nlen);
5060                  }
5061    
5062                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5063    
5064                else
5065                  {
5066                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5067                  *ketcode = OP_KETRPOS;
5068                  }
5069    
5070                /* If the minimum is zero, mark it as possessive, then unset the
5071                possessive flag. */
5072    
5073                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5074                possessive_quantifier = FALSE;
5075                }
5076    
5077              /* Non-possessive quantifier */
5078    
5079              else *ketcode = OP_KETRMAX + repeat_type;
5080            }            }
5081          }          }
5082        }        }
# Line 4532  for (;; ptr++) Line 5097  for (;; ptr++)
5097        }        }
5098    
5099      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
5100      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
5101      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
5102      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
5103      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
5104      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
5105      tempcode, not at previous, which might be the first part of a string whose  
5106      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
5107        just above, so possessive_quantifier is always FALSE for them at this
5108        stage.
5109    
5110        Note that the repeated item starts at tempcode, not at previous, which
5111        might be the first part of a string whose (former) last char we repeated.
5112    
5113      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
5114      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 4561  for (;; ptr++) Line 5131  for (;; ptr++)
5131  #endif  #endif
5132          }          }
5133    
5134        len = code - tempcode;        len = (int)(code - tempcode);
5135        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
5136          {          {
5137          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4569  for (;; ptr++) Line 5139  for (;; ptr++)
5139          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
5140          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
5141    
5142          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
5143          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
5144          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
5145          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
5146    
5147          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
5148          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
5149          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
5150          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
5151    
5152            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
5153            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
5154            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
5155            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
5156    
5157            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
5158            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
5159            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
5160            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
5161    
5162          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
5163          pending recursive references are updated. */          pending recursive references are updated. */
5164    
# Line 4630  for (;; ptr++) Line 5210  for (;; ptr++)
5210        const uschar *arg = NULL;        const uschar *arg = NULL;
5211        previous = NULL;        previous = NULL;
5212        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5213        namelen = ptr - name;        namelen = (int)(ptr - name);
5214    
5215          /* It appears that Perl allows any characters whatsoever, other than
5216          a closing parenthesis, to appear in arguments, so we no longer insist on
5217          letters, digits, and underscores. */
5218    
5219        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5220          {          {
5221          arg = ++ptr;          arg = ++ptr;
5222          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5223            || *ptr == '_') ptr++;          arglen = (int)(ptr - arg);
         arglen = ptr - arg;  
5224          }          }
5225    
5226        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
# Line 4653  for (;; ptr++) Line 5236  for (;; ptr++)
5236          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5237              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5238            {            {
5239            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5240              ASSERT_ACCEPT if in an assertion. */
5241    
5242            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5243              {              {
5244              open_capitem *oc;              open_capitem *oc;
5245                if (arglen != 0)
5246                  {
5247                  *errorcodeptr = ERR59;
5248                  goto FAILED;
5249                  }
5250              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5251              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5252                {                {
5253                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5254                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5255                }                }
5256                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5257    
5258                /* Do not set firstbyte after *ACCEPT */
5259                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5260              }              }
5261    
5262            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
5263    
5264            if (arglen == 0)            else if (arglen == 0)
5265              {              {
5266              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
5267                {                {
5268                *errorcodeptr = ERR66;                *errorcodeptr = ERR66;
5269                goto FAILED;                goto FAILED;
5270                }                }
5271              *code++ = verbs[i].op;              *code = verbs[i].op;
5272                if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
5273              }              }
5274    
5275            else            else
# Line 4685  for (;; ptr++) Line 5279  for (;; ptr++)
5279                *errorcodeptr = ERR59;                *errorcodeptr = ERR59;
5280                goto FAILED;                goto FAILED;
5281                }                }
5282              *code++ = verbs[i].op_arg;              *code = verbs[i].op_arg;
5283                if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
5284              *code++ = arglen;              *code++ = arglen;
5285              memcpy(code, arg, arglen);              memcpy(code, arg, arglen);
5286              code += arglen;              code += arglen;
# Line 4818  for (;; ptr++) Line 5413  for (;; ptr++)
5413                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5414            ptr++;            ptr++;
5415            }            }
5416          namelen = ptr - name;          namelen = (int)(ptr - name);
5417    
5418          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
5419              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
# Line 4879  for (;; ptr++) Line 5474  for (;; ptr++)
5474          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5475    
5476          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5477                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5478            {            {
5479            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5480            code[1+LINK_SIZE]++;            code[1+LINK_SIZE]++;
# Line 4947  for (;; ptr++) Line 5542  for (;; ptr++)
5542          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5543          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5544          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5545            cd->assert_depth += 1;
5546          ptr++;          ptr++;
5547          break;          break;
5548    
# Line 4961  for (;; ptr++) Line 5557  for (;; ptr++)
5557            continue;            continue;
5558            }            }
5559          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5560            cd->assert_depth += 1;
5561          break;          break;
5562    
5563    
# Line 4970  for (;; ptr++) Line 5567  for (;; ptr++)
5567            {            {
5568            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5569            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5570              cd->assert_depth += 1;
5571            ptr += 2;            ptr += 2;
5572            break;            break;
5573    
5574            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5575            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5576              cd->assert_depth += 1;
5577            ptr += 2;            ptr += 2;
5578            break;            break;
5579    
# Line 5014  for (;; ptr++) Line 5613  for (;; ptr++)
5613              goto FAILED;              goto FAILED;
5614              }              }
5615            *code++ = n;            *code++ = n;
5616            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5617            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5618            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5619            }            }
5620          previous = NULL;          previous = NULL;
# Line 5048  for (;; ptr++) Line 5647  for (;; ptr++)
5647            name = ++ptr;            name = ++ptr;
5648    
5649            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5650            namelen = ptr - name;            namelen = (int)(ptr - name);
5651    
5652            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5653    
# Line 5178  for (;; ptr++) Line 5777  for (;; ptr++)
5777          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5778          name = ++ptr;          name = ++ptr;
5779          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5780          namelen = ptr - name;          namelen = (int)(ptr - name);
5781    
5782          /* 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
5783          reference number. */          a dummy reference number, because it was not used in the first pass.
5784            However, with the change of recursive back references to be atomic,
5785            we have to look for the number so that this state can be identified, as
5786            otherwise the incorrect length is computed. If it's not a backwards
5787            reference, the dummy number will do. */
5788    
5789          if (lengthptr != NULL)          if (lengthptr != NULL)
5790            {            {
5791              const uschar *temp;
5792    
5793            if (namelen == 0)            if (namelen == 0)
5794              {              {
5795              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5200  for (;; ptr++) Line 5805  for (;; ptr++)
5805              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5806              goto FAILED;              goto FAILED;
5807              }              }
5808            recno = 0;  
5809              /* The name table does not exist in the first pass, so we cannot
5810              do a simple search as in the code below. Instead, we have to scan the
5811              pattern to find the number. It is important that we scan it only as
5812              far as we have got because the syntax of named subpatterns has not
5813              been checked for the rest of the pattern, and find_parens() assumes
5814              correct syntax. In any case, it's a waste of resources to scan
5815              further. We stop the scan at the current point by temporarily
5816              adjusting the value of cd->endpattern. */
5817    
5818              temp = cd->end_pattern;
5819              cd->end_pattern = ptr;
5820              recno = find_parens(cd, name, namelen,
5821                (options & PCRE_EXTENDED) != 0, utf8);
5822              cd->end_pattern = temp;
5823              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5824            }            }
5825    
5826          /* 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 5225  for (;; ptr++) Line 5845  for (;; ptr++)
5845              }              }
5846            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5847                      find_parens(cd, name, namelen,                      find_parens(cd, name, namelen,
5848                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5849              {              {
5850              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5851              goto FAILED;              goto FAILED;
# Line 5336  for (;; ptr++) Line 5956  for (;; ptr++)
5956              if (called == NULL)              if (called == NULL)
5957                {                {
5958                if (find_parens(cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5959                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5960                  {                  {
5961                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5962                  goto FAILED;                  goto FAILED;
# Line 5344  for (;; ptr++) Line 5964  for (;; ptr++)
5964    
5965                /* 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
5966                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5967                of the group. */                of the group. Then remember the forward reference. */
5968    
5969                called = cd->start_code + recno;                called = cd->start_code + recno;
5970                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5971                }                }
5972    
5973              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
5974              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
5975              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
5976                must not, however, do this check if we are in a conditional
5977                subpattern because the condition might be testing for recursion in
5978                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
5979                Forever loops are also detected at runtime, so those that occur in
5980                conditional subpatterns will be picked up then. */
5981    
5982              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
5983                       could_be_empty(called, code, bcptr, utf8, cd))                       could_be_empty(called, code, bcptr, utf8, cd))
5984                {                {
5985                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
# Line 5362  for (;; ptr++) Line 5987  for (;; ptr++)
5987                }                }
5988              }              }
5989    
5990            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
           "once" brackets. Set up a "previous group" length so that a  
           subsequent quantifier will work. */  
   
           *code = OP_ONCE;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
5991    
5992            *code = OP_RECURSE;            *code = OP_RECURSE;
5993            PUT(code, 1, called - cd->start_code);            PUT(code, 1, (int)(called - cd->start_code));
5994            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
   
           *code = OP_KET;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
   
           length_prevgroup = 3 + 3*LINK_SIZE;  
5995            }            }
5996    
5997          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5439  for (;; ptr++) Line 6052  for (;; ptr++)
6052          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
6053          both phases.          both phases.
6054    
6055          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
6056          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. */  
6057    
6058          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
6059            {            {
# Line 5452  for (;; ptr++) Line 6064  for (;; ptr++)
6064              }              }
6065            else            else
6066              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
6067              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
6068              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
6069              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
6070              }              }
6071    
6072            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
6073            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). */  
6074    
6075            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
6076            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5501  for (;; ptr++) Line 6106  for (;; ptr++)
6106        skipbytes = 2;        skipbytes = 2;
6107        }        }
6108    
6109      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
6110      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
6111      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
6112      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. */  
6113    
6114      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                      /* For handling repetition */
6115      *code = bravalue;      *code = bravalue;
6116      tempcode = code;      tempcode = code;
6117      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
6118      length_prevgroup = 0;              /* Initialize for pre-compile phase */      tempbracount = cd->bracount;          /* Save value before bracket */
6119        length_prevgroup = 0;                 /* Initialize for pre-compile phase */
6120    
6121      if (!compile_regex(      if (!compile_regex(
6122           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
6123           options & PCRE_IMS,           /* The previous ims option state */           &tempcode,                       /* Where to put code (updated) */
6124           &tempcode,                    /* Where to put code (updated) */           &ptr,                            /* Input pointer (updated) */
6125           &ptr,                         /* Input pointer (updated) */           errorcodeptr,                    /* Where to put an error message */
          errorcodeptr,                 /* Where to put an error message */  
6126           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
6127            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
6128           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
6129           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
6130           &subfirstbyte,                /* For possible first char */           cond_depth +
6131           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
6132           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
6133           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
6134           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
6135             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
6136             (lengthptr == NULL)? NULL :      /* Actual compile phase */
6137               &length_prevgroup              /* Pre-compile phase */
6138           ))           ))
6139        goto FAILED;        goto FAILED;
6140    
6141        /* If this was an atomic group and there are no capturing groups within it,
6142        generate OP_ONCE_NC instead of OP_ONCE. */
6143    
6144        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
6145          *code = OP_ONCE_NC;
6146    
6147        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
6148          cd->assert_depth -= 1;
6149    
6150      /* 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
6151      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.
6152      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
6153    
6154      /* 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
6155      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
6156      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
6157      not be available. */      not be available. */
# Line 5602  for (;; ptr++) Line 6216  for (;; ptr++)
6216          goto FAILED;          goto FAILED;
6217          }          }
6218        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
6219        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
6220        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
6221        *code++ = OP_KET;        *code++ = OP_KET;
6222        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5675  for (;; ptr++) Line 6289  for (;; ptr++)
6289    
6290      /* ===================================================================*/      /* ===================================================================*/
6291      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
6292      are arranged to be the negation of the corresponding OP_values in the      are arranged to be the negation of the corresponding OP_values in the
6293      default case when PCRE_UCP is not set. For the back references, the values      default case when PCRE_UCP is not set. For the back references, the values
6294      are ESC_REF plus the reference number. Only back references and those types      are ESC_REF plus the reference number. Only back references and those types
6295      that consume a character may be repeated. We can test for values between      that consume a character may be repeated. We can test for values between
# Line 5770  for (;; ptr++) Line 6384  for (;; ptr++)
6384          }          }
6385    
6386        /* \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).
6387        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6388    
6389        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6390          {          {
6391            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6392              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6393              {
6394              *errorcodeptr = ERR69;
6395              break;
6396              }
6397          is_recurse = FALSE;          is_recurse = FALSE;
6398          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6399            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
# Line 5794  for (;; ptr++) Line 6413  for (;; ptr++)
6413          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6414          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6415          previous = code;          previous = code;
6416          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6417          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6418          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6419          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
# Line 5853  for (;; ptr++) Line 6472  for (;; ptr++)
6472            ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */            ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
6473            }            }
6474          else          else
6475  #endif  #endif
6476            {            {
6477            previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;            previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6478            *code++ = -c;            *code++ = -c;
6479            }            }
6480          }          }
6481        continue;        continue;
6482        }        }
# Line 5902  for (;; ptr++) Line 6521  for (;; ptr++)
6521    
6522      ONE_CHAR:      ONE_CHAR:
6523      previous = code;      previous = code;
6524      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6525      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6526    
6527      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 5931  for (;; ptr++) Line 6550  for (;; ptr++)
6550        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6551        }        }
6552    
6553      /* 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
6554      1 or the matching is caseful. */      1 or the matching is caseful. */
6555    
6556      else      else
# Line 5966  return FALSE; Line 6585  return FALSE;
6585  /* 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
6586  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
6587  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.  
   
6588  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
6589  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
6590  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6591    
6592  Arguments:  Arguments:
6593    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  
6594    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6595    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6596    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
6597    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6598    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6599    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6600      cond_depth     depth of nesting for conditional subpatterns
6601    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6602    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6603    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 5995  Returns:         TRUE on success Line 6609  Returns:         TRUE on success
6609  */  */
6610    
6611  static BOOL  static BOOL
6612  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6613    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6614    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6615    int *lengthptr)    compile_data *cd, int *lengthptr)
6616  {  {
6617  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6618  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 6012  int branchfirstbyte, branchreqbyte; Line 6626  int branchfirstbyte, branchreqbyte;
6626  int length;  int length;
6627  int orig_bracount;  int orig_bracount;
6628  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6629  branch_chain bc;  branch_chain bc;
6630    
6631  bc.outer = bcptr;  bc.outer = bcptr;
# Line 6036  pre-compile phase to find out whether an Line 6649  pre-compile phase to find out whether an
6649    
6650  /* 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
6651  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
6652  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6653    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6654    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6655    
6656  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6657    {    {
# Line 6062  for (;;) Line 6677  for (;;)
6677    
6678    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6679    
   /* 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;  
     }  
   
6680    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6681    
6682    if (lookbehind)    if (lookbehind)
# Line 6085  for (;;) Line 6691  for (;;)
6691    into the length. */    into the length. */
6692    
6693    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6694          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6695            (lengthptr == NULL)? NULL : &length))
6696      {      {
6697      *ptrptr = ptr;      *ptrptr = ptr;
6698      return FALSE;      return FALSE;
6699      }      }
6700    
   /* 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;  
   
6701    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6702    has fewer than the rest. */    has fewer than the rest. */
6703    
# Line 6160  for (;;) Line 6758  for (;;)
6758        {        {
6759        int fixed_length;        int fixed_length;
6760        *code = OP_END;        *code = OP_END;
6761        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6762            FALSE, cd);
6763        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6764        if (fixed_length == -3)        if (fixed_length == -3)
6765          {          {
# Line 6168  for (;;) Line 6767  for (;;)
6767          }          }
6768        else if (fixed_length < 0)        else if (fixed_length < 0)
6769          {          {
6770          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 :
6771                            (fixed_length == -4)? ERR70: ERR25;
6772          *ptrptr = ptr;          *ptrptr = ptr;
6773          return FALSE;          return FALSE;
6774          }          }
# Line 6181  for (;;) Line 6781  for (;;)
6781    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
6782    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
6783    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
6784    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. */  
6785    
6786    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6787      {      {
6788      if (lengthptr == NULL)      if (lengthptr == NULL)
6789        {        {
6790        int branch_length = code - last_branch;        int branch_length = (int)(code - last_branch);
6791        do        do
6792          {          {
6793          int prev_length = GET(last_branch, 1);          int prev_length = GET(last_branch, 1);
# Line 6203  for (;;) Line 6801  for (;;)
6801      /* Fill in the ket */      /* Fill in the ket */
6802    
6803      *code = OP_KET;      *code = OP_KET;
6804      PUT(code, 1, code - start_bracket);      PUT(code, 1, (int)(code - start_bracket));
6805      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6806    
6807      /* 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 6218  for (;;) Line 6816  for (;;)
6816            code - start_bracket);            code - start_bracket);
6817          *start_bracket = OP_ONCE;          *start_bracket = OP_ONCE;
6818          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
6819          PUT(start_bracket, 1, code - start_bracket);          PUT(start_bracket, 1, (int)(code - start_bracket));
6820          *code = OP_KET;          *code = OP_KET;
6821          PUT(code, 1, code - start_bracket);          PUT(code, 1, (int)(code - start_bracket));
6822          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
6823          length += 2 + 2*LINK_SIZE;          length += 2 + 2*LINK_SIZE;
6824          }          }
6825        cd->open_caps = cd->open_caps->next;        cd->open_caps = cd->open_caps->next;
6826        }        }
6827    
     /* Reset options if needed. */  
   
     if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)  
       {  
       *code++ = OP_OPT;  
       *code++ = oldims;  
       length += 2;  
       }  
   
6828      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
6829    
6830      cd->bracount = max_bracount;      cd->bracount = max_bracount;
# Line 6275  for (;;) Line 6864  for (;;)
6864    else    else
6865      {      {
6866      *code = OP_ALT;      *code = OP_ALT;
6867      PUT(code, 1, code - last_branch);      PUT(code, 1, (int)(code - last_branch));
6868      bc.current_branch = last_branch = code;      bc.current_branch = last_branch = code;
6869      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6870      }      }
# Line 6295  for (;;) Line 6884  for (;;)
6884  /* 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
6885  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
6886  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
6887  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
6888  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6889    
6890  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.
6891  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 6317  of the more common cases more precisely. Line 6906  of the more common cases more precisely.
6906    
6907  Arguments:  Arguments:
6908    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6909    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
6910                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6911                    the less precise approach                    the less precise approach
# Line 6327  Returns:     TRUE or FALSE Line 6915  Returns:     TRUE or FALSE
6915  */  */
6916    
6917  static BOOL  static BOOL
6918  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6919    unsigned int backref_map)    unsigned int backref_map)
6920  {  {
6921  do {  do {
6922     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6923       options, PCRE_MULTILINE, FALSE);       FALSE);
6924     register int op = *scode;     register int op = *scode;
6925    
6926     /* Non-capturing brackets */     /* Non-capturing brackets */
6927    
6928     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6929           op == OP_SBRA || op == OP_SBRAPOS)
6930       {       {
6931       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6932       }       }
6933    
6934     /* Capturing brackets */     /* Capturing brackets */
6935    
6936     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6937                op == OP_SCBRA || op == OP_SCBRAPOS)
6938       {       {
6939       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6940       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6941       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6942       }       }
6943    
6944     /* Other brackets */     /* Other brackets */
6945    
6946     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC ||
6947                op == OP_COND)
6948       {       {
6949       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6950       }       }
6951    
6952     /* .* 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 6370  do { Line 6961  do {
6961    
6962     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6963    
6964     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;  
6965     code += GET(code, 1);     code += GET(code, 1);
6966     }     }
6967  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6408  is_startline(const uschar *code, unsigne Line 6997  is_startline(const uschar *code, unsigne
6997  {  {
6998  do {  do {
6999     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
7000       NULL, 0, FALSE);       FALSE);
7001     register int op = *scode;     register int op = *scode;
7002    
7003     /* 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 6435  do { Line 7024  do {
7024         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
7025         break;         break;
7026         }         }
7027       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
7028       op = *scode;       op = *scode;
7029       }       }
7030    
7031     /* Non-capturing brackets */     /* Non-capturing brackets */
7032    
7033     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
7034           op == OP_SBRA || op == OP_SBRAPOS)
7035       {       {
7036       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
7037       }       }
7038    
7039     /* Capturing brackets */     /* Capturing brackets */
7040    
7041     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
7042                op == OP_SCBRA || op == OP_SCBRAPOS)
7043       {       {
7044       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
7045       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6457  do { Line 7048  do {
7048    
7049     /* Other brackets */     /* Other brackets */
7050    
7051     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC)
7052       {       {
7053       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
7054       }       }
# Line 6472  do { Line 7063  do {
7063    
7064     /* Check for explicit circumflex */     /* Check for explicit circumflex */
7065    
7066     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
7067    
7068     /* Move on to the next alternative */     /* Move on to the next alternative */
7069    
# Line 6498  we return that char, otherwise -1. Line 7089  we return that char, otherwise -1.
7089    
7090  Arguments:  Arguments:
7091    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)  
7092    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
7093    
7094  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
7095  */  */
7096    
7097  static int  static int
7098  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
7099  {  {
7100  register int c = -1;  register int c = -1;
7101  do {  do {
7102     int d;     int d;
7103     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
7104       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
7105       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
7106     register int op = *scode;     register int op = *scode;
7107    
7108     switch(op)     switch(op)
# Line 6520  do { Line 7111  do {
7111       return -1;       return -1;
7112    
7113       case OP_BRA:       case OP_BRA:
7114         case OP_BRAPOS:
7115       case OP_CBRA:       case OP_CBRA:
7116         case OP_SCBRA:
7117         case OP_CBRAPOS:
7118         case OP_SCBRAPOS:
7119       case OP_ASSERT:       case OP_ASSERT:
7120       case OP_ONCE:       case OP_ONCE:
7121         case OP_ONCE_NC:
7122       case OP_COND:       case OP_COND:
7123       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
7124         return -1;         return -1;
7125       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
7126       break;       break;
7127    
7128       case OP_EXACT:       /* Fall through */       case OP_EXACT:
7129       scode += 2;       scode += 2;
7130         /* Fall through */
7131    
7132       case OP_CHAR:       case OP_CHAR:
      case OP_CHARNC:  
7133       case OP_PLUS:       case OP_PLUS:
7134       case OP_MINPLUS:       case OP_MINPLUS:
7135       case OP_POSPLUS:       case OP_POSPLUS:
7136       if (!inassert) return -1;       if (!inassert) return -1;
7137       if (c < 0)       if (c < 0) c = scode[1];
7138         {         else if (c != scode[1]) return -1;
7139         c = scode[1];       break;
7140         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
7141         }       case OP_EXACTI:
7142       else if (c != scode[1]) return -1;       scode += 2;
7143         /* Fall through */
7144    
7145         case OP_CHARI:
7146         case OP_PLUSI:
7147         case OP_MINPLUSI:
7148         case OP_POSPLUSI:
7149         if (!inassert) return -1;
7150         if (c < 0) c = scode[1] | REQ_CASELESS;
7151           else if (c != scode[1]) return -1;
7152       break;       break;
7153       }       }
7154    
# Line 6666  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7271  while (ptr[skipatstart] == CHAR_LEFT_PAR
7271      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
7272    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
7273      { skipatstart += 6; options |= PCRE_UCP; continue; }      { skipatstart += 6; options |= PCRE_UCP; continue; }
7274      else if (strncmp((char *)(ptr+skipatstart+2), STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
7275        { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
7276    
7277    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
7278      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6689  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7296  while (ptr[skipatstart] == CHAR_LEFT_PAR
7296      options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;      options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
7297    else break;    else break;
7298    }    }
7299    
7300  utf8 = (options & PCRE_UTF8) != 0;  utf8 = (options & PCRE_UTF8) != 0;
7301    
7302  /* Can't support UTF8 unless PCRE has been compiled to include the code. */  /* Can't support UTF8 unless PCRE has been compiled to include the code. The
7303    return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7304    release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7305    not used here. */
7306    
7307  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7308  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7309       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7310    {    {
7311    errorcode = ERR44;    errorcode = ERR44;
7312    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6715  if (utf8) Line 7325  if (utf8)
7325  if ((options & PCRE_UCP) != 0)  if ((options & PCRE_UCP) != 0)
7326    {    {
7327    errorcode = ERR67;    errorcode = ERR67;
7328    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
7329    }    }
7330  #endif  #endif
7331    
7332  /* Check validity of \R options. */  /* Check validity of \R options. */
7333    
7334  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
7335         (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
7336    {    {
7337    case 0:    errorcode = ERR56;
7338    case PCRE_BSR_ANYCRLF:    goto PCRE_EARLY_ERROR_RETURN;
   case PCRE_BSR_UNICODE:  
   break;  
   default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;  
7339    }    }
7340    
7341  /* Handle different types of newline. The three bits give seven cases. The  /* Handle different types of newline. The three bits give seven cases. The
# Line 6812  outside can help speed up starting point Line 7420  outside can help speed up starting point
7420  ptr += skipatstart;  ptr += skipatstart;
7421  code = cworkspace;  code = cworkspace;
7422  *code = OP_BRA;  *code = OP_BRA;
7423  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7424    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7425  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7426    
7427  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 6847  regex compiled on a system with 4-byte p Line 7454  regex compiled on a system with 4-byte p
7454  pointers. */  pointers. */
7455    
7456  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
7457  re->size = size;  re->size = (int)size;
7458  re->options = cd->external_options;  re->options = cd->external_options;
7459  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7460  re->dummy1 = 0;  re->dummy1 = 0;
# Line 6868  field; this time it's used for rememberi Line 7475  field; this time it's used for rememberi
7475  */  */
7476    
7477  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7478    cd->assert_depth = 0;
7479  cd->bracount = 0;  cd->bracount = 0;
7480  cd->names_found = 0;  cd->names_found = 0;
7481  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 6886  of the function here. */ Line 7494  of the function here. */
7494  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7495  code = (uschar *)codestart;  code = (uschar *)codestart;
7496  *code = OP_BRA;  *code = OP_BRA;
7497  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0, 0,
7498    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7499  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7500  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7501  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7502    
7503  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */  if (cd->had_accept) reqbyte = REQ_NONE;   /* Must disable after (*ACCEPT) */
7504    
7505  /* If not reached end of pattern on success, there's an excess bracket. */  /* If not reached end of pattern on success, there's an excess bracket. */
7506    
# Line 6918  while (errorcode == 0 && cd->hwm > cwork Line 7526  while (errorcode == 0 && cd->hwm > cwork
7526    recno = GET(codestart, offset);    recno = GET(codestart, offset);
7527    groupptr = _pcre_find_bracket(codestart, utf8, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
7528    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
7529      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, (int)(groupptr - codestart));
7530    }    }
7531    
7532  /* Give an error if there's back reference to a non-existent capturing  /* Give an error if there's back reference to a non-existent capturing
# Line 6953  if (cd->check_lookbehind) Line 7561  if (cd->check_lookbehind)
7561        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7562        int end_op = *be;        int end_op = *be;
7563        *be = OP_END;        *be = OP_END;
7564        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7565            cd);
7566        *be = end_op;        *be = end_op;
7567        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7568        if (fixed_length < 0)        if (fixed_length < 0)
7569          {          {
7570          errorcode = (fixed_length == -2)? ERR36 : ERR25;          errorcode = (fixed_length == -2)? ERR36 :
7571                        (fixed_length == -4)? ERR70 : ERR25;
7572          break;          break;
7573          }          }
7574        PUT(cc, 1, fixed_length);        PUT(cc, 1, fixed_length);
# Line 6973  if (errorcode != 0) Line 7583  if (errorcode != 0)
7583    {    {
7584    (pcre_free)(re);    (pcre_free)(re);
7585    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
7586    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = (int)(ptr - (const uschar *)pattern);
7587    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
7588    *errorptr = find_error_text(errorcode);    *errorptr = find_error_text(errorcode);
7589    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
# Line 6992  start with ^. and also when all branches Line 7602  start with ^. and also when all branches
7602    
7603  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7604    {    {
7605    int temp_options = re->options;   /* May get changed during these scans */    if (is_anchored(codestart, 0, cd->backref_map))
   if (is_anchored(codestart, &temp_options, 0, cd->backref_map))  
7606      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7607    else    else
7608      {      {
7609      if (firstbyte < 0)      if (firstbyte < 0)
7610        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7611      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7612        {        {
7613        int ch = firstbyte & 255;        int ch = firstbyte & 255;

Legend:
Removed from v.528  
changed lines
  Added in v.749

  ViewVC Help
Powered by ViewVC 1.1.5