/[pcre]/code/trunk/pcre_compile.c
ViewVC logotype

Diff of /code/trunk/pcre_compile.c

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

revision 512 by ph10, Tue Mar 30 11:11:52 2010 UTC revision 745 by ph10, Mon Nov 14 11:41:03 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 124  static const short int escapes[] = { Line 124  static const short int escapes[] = {
124       -ESC_H,                  0,       -ESC_H,                  0,
125       0,                       -ESC_K,       0,                       -ESC_K,
126       0,                       0,       0,                       0,
127       0,                       0,       -ESC_N,                  0,
128       -ESC_P,                  -ESC_Q,       -ESC_P,                  -ESC_Q,
129       -ESC_R,                  -ESC_S,       -ESC_R,                  -ESC_S,
130       0,                       0,       0,                       0,
# Line 171  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 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
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 324  static const char error_texts[] = Line 371  static const char error_texts[] =
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255\0"    "number after (?C is > 255\0"
376    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
# Line 346  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 360  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"
411      "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413    ;    ;
414    
415  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 496  static const unsigned char ebcdic_charta Line 546  static const unsigned char ebcdic_charta
546  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
547    
548  static BOOL  static BOOL
549    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
550      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
551    
552    
# Line 528  return s; Line 578  return s;
578    
579    
580  /*************************************************  /*************************************************
581    *            Check for counted repeat            *
582    *************************************************/
583    
584    /* This function is called when a '{' is encountered in a place where it might
585    start a quantifier. It looks ahead to see if it really is a quantifier or not.
586    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
587    where the ddds are digits.
588    
589    Arguments:
590      p         pointer to the first char after '{'
591    
592    Returns:    TRUE or FALSE
593    */
594    
595    static BOOL
596    is_counted_repeat(const uschar *p)
597    {
598    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
599    while ((digitab[*p] & ctype_digit) != 0) p++;
600    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
601    
602    if (*p++ != CHAR_COMMA) return FALSE;
603    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
604    
605    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
606    while ((digitab[*p] & ctype_digit) != 0) p++;
607    
608    return (*p == CHAR_RIGHT_CURLY_BRACKET);
609    }
610    
611    
612    
613    /*************************************************
614  *            Handle escapes                      *  *            Handle escapes                      *
615  *************************************************/  *************************************************/
616    
# Line 593  else Line 676  else
676    
677      case CHAR_l:      case CHAR_l:
678      case CHAR_L:      case CHAR_L:
679      case CHAR_N:      *errorcodeptr = ERR37;
680        break;
681    
682      case CHAR_u:      case CHAR_u:
683        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
684          {
685          /* In JavaScript, \u must be followed by four hexadecimal numbers.
686          Otherwise it is a lowercase u letter. */
687          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
688               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
689            {
690            c = 0;
691            for (i = 0; i < 4; ++i)
692              {
693              register int cc = *(++ptr);
694    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
695              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
696              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
697    #else           /* EBCDIC coding */
698              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
699              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
700    #endif
701              }
702            }
703          }
704        else
705          *errorcodeptr = ERR37;
706        break;
707    
708      case CHAR_U:      case CHAR_U:
709      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
710        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
711      break;      break;
712    
713      /* \g must be followed by one of a number of specific things:      /* In a character class, \g is just a literal "g". Outside a character
714        class, \g must be followed by one of a number of specific things:
715    
716      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
717      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 616  else Line 728  else
728      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
729    
730      case CHAR_g:      case CHAR_g:
731        if (isclass) break;
732      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
733        {        {
734        c = -ESC_g;        c = -ESC_g;
# Line 744  else Line 857  else
857      treated as a data character. */      treated as a data character. */
858    
859      case CHAR_x:      case CHAR_x:
860        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
861          {
862          /* In JavaScript, \x must be followed by two hexadecimal numbers.
863          Otherwise it is a lowercase x letter. */
864          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
865            {
866            c = 0;
867            for (i = 0; i < 2; ++i)
868              {
869              register int cc = *(++ptr);
870    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
871              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
872              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
873    #else           /* EBCDIC coding */
874              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
875              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
876    #endif
877              }
878            }
879          break;
880          }
881    
882      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
883        {        {
884        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
# Line 794  else Line 929  else
929      break;      break;
930    
931      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
932      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
933        coding is ASCII-specific, but then the whole concept of \cx is
934      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
935    
936      case CHAR_c:      case CHAR_c:
# Line 804  else Line 940  else
940        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
941        break;        break;
942        }        }
943    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
944  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
945          {
946          *errorcodeptr = ERR68;
947          break;
948          }
949      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
950      c ^= 0x40;      c ^= 0x40;
951  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
952      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
953      c ^= 0xC0;      c ^= 0xC0;
954  #endif  #endif
# Line 831  else Line 971  else
971      }      }
972    }    }
973    
974    /* Perl supports \N{name} for character names, as well as plain \N for "not
975    newline". PCRE does not support \N{name}. However, it does support
976    quantification such as \N{2,3}. */
977    
978    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
979         !is_counted_repeat(ptr+2))
980      *errorcodeptr = ERR37;
981    
982    /* If PCRE_UCP is set, we change the values for \d etc. */
983    
984    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
985      c -= (ESC_DU - ESC_D);
986    
987    /* Set the pointer to the final character before returning. */
988    
989  *ptrptr = ptr;  *ptrptr = ptr;
990  return c;  return c;
991  }  }
# Line 931  return -1; Line 1086  return -1;
1086    
1087    
1088  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if (*p++ != CHAR_COMMA) return FALSE;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == CHAR_RIGHT_CURLY_BRACKET);  
 }  
   
   
   
 /*************************************************  
1089  *         Read repeat counts                     *  *         Read repeat counts                     *
1090  *************************************************/  *************************************************/
1091    
# Line 1039  top-level call starts at the beginning o Line 1161  top-level call starts at the beginning o
1161  start at a parenthesis. It scans along a pattern's text looking for capturing  start at a parenthesis. It scans along a pattern's text looking for capturing
1162  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1163  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1164  returns when it reaches a given numbered subpattern. We know that if (?P< is  returns when it reaches a given numbered subpattern. Recursion is used to keep
1165  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1166  first pass. Recursion is used to keep track of subpatterns that reset the  
1167  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1168    that if (?< or (?' or (?P< is encountered, the name will be correctly
1169    terminated because that is checked in the first pass. There is now one call to
1170    this function in the first pass, to check for a recursive back reference by
1171    name (so that we can make the whole group atomic). In this case, we need check
1172    only up to the current position in the pattern, and that is still OK because
1173    and previous occurrences will have been checked. To make this work, the test
1174    for "end of pattern" is a check against cd->end_pattern in the main loop,
1175    instead of looking for a binary zero. This means that the special first-pass
1176    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1177    processing items within the loop are OK, because afterwards the main loop will
1178    terminate.)
1179    
1180  Arguments:  Arguments:
1181    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1050  Arguments: Line 1183  Arguments:
1183    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1184    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1185    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1186      utf8         TRUE if we are in UTF-8 mode
1187    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1188    
1189  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
# Line 1057  Returns:       the number of the named s Line 1191  Returns:       the number of the named s
1191    
1192  static int  static int
1193  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1194    BOOL xmode, int *count)    BOOL xmode, BOOL utf8, int *count)
1195  {  {
1196  uschar *ptr = *ptrptr;  uschar *ptr = *ptrptr;
1197  int start_count = *count;  int start_count = *count;
# Line 1069  dealing with. The very first call may no Line 1203  dealing with. The very first call may no
1203    
1204  if (ptr[0] == CHAR_LEFT_PARENTHESIS)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1205    {    {
1206    if (ptr[1] == CHAR_QUESTION_MARK &&    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1207        ptr[2] == CHAR_VERTICAL_LINE)  
1208      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1209    
1210      /* Handle a normal, unnamed capturing parenthesis. */
1211    
1212      else if (ptr[1] != CHAR_QUESTION_MARK)
1213        {
1214        *count += 1;
1215        if (name == NULL && *count == lorn) return *count;
1216        ptr++;
1217        }
1218    
1219      /* All cases now have (? at the start. Remember when we are in a group
1220      where the parenthesis numbers are duplicated. */
1221    
1222      else if (ptr[2] == CHAR_VERTICAL_LINE)
1223      {      {
1224      ptr += 3;      ptr += 3;
1225      dup_parens = TRUE;      dup_parens = TRUE;
1226      }      }
1227    
1228    /* Handle a normal, unnamed capturing parenthesis */    /* Handle comments; all characters are allowed until a ket is reached. */
1229    
1230    else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)    else if (ptr[2] == CHAR_NUMBER_SIGN)
1231      {      {
1232      *count += 1;      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1233      if (name == NULL && *count == lorn) return *count;      goto FAIL_EXIT;
     ptr++;  
1234      }      }
1235    
1236    /* Handle a condition. If it is an assertion, just carry on so that it    /* Handle a condition. If it is an assertion, just carry on so that it
1237    is processed as normal. If not, skip to the closing parenthesis of the    is processed as normal. If not, skip to the closing parenthesis of the
1238    condition (there can't be any nested parens. */    condition (there can't be any nested parens). */
1239    
1240    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1241      {      {
# Line 1099  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1247  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1247        }        }
1248      }      }
1249    
1250    /* We have either (? or (* and not a condition */    /* Start with (? but not a condition. */
1251    
1252    else    else
1253      {      {
# Line 1128  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1276  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1276    }    }
1277    
1278  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1279  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1280    first-pass call when this value is temporarily adjusted to stop at the current
1281    position. So DO NOT change this to a test for binary zero. */
1282    
1283  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1284    {    {
1285    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1286    
# Line 1204  for (; *ptr != 0; ptr++) Line 1354  for (; *ptr != 0; ptr++)
1354    
1355    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1356      {      {
1357      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1358        while (*ptr != 0)
1359          {
1360          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1361          ptr++;
1362    #ifdef SUPPORT_UTF8
1363          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1364    #endif
1365          }
1366      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1367      continue;      continue;
1368      }      }
# Line 1213  for (; *ptr != 0; ptr++) Line 1371  for (; *ptr != 0; ptr++)
1371    
1372    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1373      {      {
1374      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1375      if (rc > 0) return rc;      if (rc > 0) return rc;
1376      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1377      }      }
# Line 1221  for (; *ptr != 0; ptr++) Line 1379  for (; *ptr != 0; ptr++)
1379    else if (*ptr == CHAR_RIGHT_PARENTHESIS)    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1380      {      {
1381      if (dup_parens && *count < hwm_count) *count = hwm_count;      if (dup_parens && *count < hwm_count) *count = hwm_count;
1382      *ptrptr = ptr;      goto FAIL_EXIT;
     return -1;  
1383      }      }
1384    
1385    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
# Line 1260  Arguments: Line 1417  Arguments:
1417    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1418    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1419    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1420      utf8         TRUE if we are in UTF-8 mode
1421    
1422  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1423  */  */
1424    
1425  static int  static int
1426  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1427      BOOL utf8)
1428  {  {
1429  uschar *ptr = (uschar *)cd->start_pattern;  uschar *ptr = (uschar *)cd->start_pattern;
1430  int count = 0;  int count = 0;
# Line 1278  matching closing parens. That is why we Line 1437  matching closing parens. That is why we
1437    
1438  for (;;)  for (;;)
1439    {    {
1440    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1441    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1442    }    }
1443    
# Line 1294  return rc; Line 1453  return rc;
1453    
1454  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1455  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1456  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1457  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1458  assertions, and also the \b assertion; for others it does not.  does not.
1459    
1460  Arguments:  Arguments:
1461    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1462    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1463    
1464  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1465  */  */
1466    
1467  static const uschar*  static const uschar*
1468  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1469  {  {
1470  for (;;)  for (;;)
1471    {    {
1472    switch ((int)*code)    switch ((int)*code)
1473      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1474      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1475      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1476      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1371  and doing the check at the end; a flag s Line 1520  and doing the check at the end; a flag s
1520    
1521  Arguments:  Arguments:
1522    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1523    options  the compiling options    utf8     TRUE in UTF-8 mode
1524    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1525    cd       the "compile data" structure    cd       the "compile data" structure
1526    
# Line 1382  Returns:   the fixed length, Line 1531  Returns:   the fixed length,
1531  */  */
1532    
1533  static int  static int
1534  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1535  {  {
1536  int length = -1;  int length = -1;
1537    
# Line 1399  for (;;) Line 1548  for (;;)
1548    register int op = *cc;    register int op = *cc;
1549    switch (op)    switch (op)
1550      {      {
1551        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1552        OP_BRA (normal non-capturing bracket) because the other variants of these
1553        opcodes are all concerned with unlimited repeated groups, which of course
1554        are not of fixed length. They will cause a -1 response from the default
1555        case of this switch. */
1556    
1557      case OP_CBRA:      case OP_CBRA:
1558      case OP_BRA:      case OP_BRA:
1559      case OP_ONCE:      case OP_ONCE:
1560        case OP_ONCE_NC:
1561      case OP_COND:      case OP_COND:
1562      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1563      if (d < 0) return d;      if (d < 0) return d;
1564      branchlength += d;      branchlength += d;
1565      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1412  for (;;) Line 1568  for (;;)
1568    
1569      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested
1570      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is
1571      END it's the end of the outer call. All can be handled by the same code. */      END it's the end of the outer call. All can be handled by the same code.
1572        Note that we must not include the OP_KETRxxx opcodes here, because they
1573        all imply an unlimited repeat. */
1574    
1575      case OP_ALT:      case OP_ALT:
1576      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1577      case OP_END:      case OP_END:
1578      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1579        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1435  for (;;) Line 1591  for (;;)
1591      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1592      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1593      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1594      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1595      if (d < 0) return d;      if (d < 0) return d;
1596      branchlength += d;      branchlength += d;
1597      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1458  for (;;) Line 1614  for (;;)
1614      case OP_RREF:      case OP_RREF:
1615      case OP_NRREF:      case OP_NRREF:
1616      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1617      case OP_CALLOUT:      case OP_CALLOUT:
1618      case OP_SOD:      case OP_SOD:
1619      case OP_SOM:      case OP_SOM:
# Line 1466  for (;;) Line 1621  for (;;)
1621      case OP_EOD:      case OP_EOD:
1622      case OP_EODN:      case OP_EODN:
1623      case OP_CIRC:      case OP_CIRC:
1624        case OP_CIRCM:
1625      case OP_DOLL:      case OP_DOLL:
1626        case OP_DOLLM:
1627      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1628      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1629      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1475  for (;;) Line 1632  for (;;)
1632      /* Handle literal characters */      /* Handle literal characters */
1633    
1634      case OP_CHAR:      case OP_CHAR:
1635      case OP_CHARNC:      case OP_CHARI:
1636      case OP_NOT:      case OP_NOT:
1637        case OP_NOTI:
1638      branchlength++;      branchlength++;
1639      cc += 2;      cc += 2;
1640  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1641      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1642  #endif  #endif
1643      break;      break;
1644    
# Line 1492  for (;;) Line 1649  for (;;)
1649      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1650      cc += 4;      cc += 4;
1651  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1652      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];  
1653  #endif  #endif
1654      break;      break;
1655    
# Line 1595  _pcre_find_bracket(const uschar *code, B Line 1751  _pcre_find_bracket(const uschar *code, B
1751  for (;;)  for (;;)
1752    {    {
1753    register int c = *code;    register int c = *code;
1754    
1755    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1756    
1757    /* 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 1613  for (;;) Line 1770  for (;;)
1770    
1771    /* Handle capturing bracket */    /* Handle capturing bracket */
1772    
1773    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1774               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1775      {      {
1776      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1777      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1651  for (;;) Line 1809  for (;;)
1809        case OP_MARK:        case OP_MARK:
1810        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
1811        case OP_SKIP_ARG:        case OP_SKIP_ARG:
1812          code += code[1];
1813          break;
1814    
1815        case OP_THEN_ARG:        case OP_THEN_ARG:
1816        code += code[1];        code += code[1];
1817        break;        break;
# Line 1668  for (;;) Line 1829  for (;;)
1829      if (utf8) switch(c)      if (utf8) switch(c)
1830        {        {
1831        case OP_CHAR:        case OP_CHAR:
1832        case OP_CHARNC:        case OP_CHARI:
1833        case OP_EXACT:        case OP_EXACT:
1834          case OP_EXACTI:
1835        case OP_UPTO:        case OP_UPTO:
1836          case OP_UPTOI:
1837        case OP_MINUPTO:        case OP_MINUPTO:
1838          case OP_MINUPTOI:
1839        case OP_POSUPTO:        case OP_POSUPTO:
1840          case OP_POSUPTOI:
1841        case OP_STAR:        case OP_STAR:
1842          case OP_STARI:
1843        case OP_MINSTAR:        case OP_MINSTAR:
1844          case OP_MINSTARI:
1845        case OP_POSSTAR:        case OP_POSSTAR:
1846          case OP_POSSTARI:
1847        case OP_PLUS:        case OP_PLUS:
1848          case OP_PLUSI:
1849        case OP_MINPLUS:        case OP_MINPLUS:
1850          case OP_MINPLUSI:
1851        case OP_POSPLUS:        case OP_POSPLUS:
1852          case OP_POSPLUSI:
1853        case OP_QUERY:        case OP_QUERY:
1854          case OP_QUERYI:
1855        case OP_MINQUERY:        case OP_MINQUERY:
1856          case OP_MINQUERYI:
1857        case OP_POSQUERY:        case OP_POSQUERY:
1858          case OP_POSQUERYI:
1859        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1860        break;        break;
1861        }        }
# Line 1754  for (;;) Line 1928  for (;;)
1928        case OP_MARK:        case OP_MARK:
1929        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
1930        case OP_SKIP_ARG:        case OP_SKIP_ARG:
1931          code += code[1];
1932          break;
1933    
1934        case OP_THEN_ARG:        case OP_THEN_ARG:
1935        code += code[1];        code += code[1];
1936        break;        break;
# Line 1771  for (;;) Line 1948  for (;;)
1948      if (utf8) switch(c)      if (utf8) switch(c)
1949        {        {
1950        case OP_CHAR:        case OP_CHAR:
1951        case OP_CHARNC:        case OP_CHARI:
1952        case OP_EXACT:        case OP_EXACT:
1953          case OP_EXACTI:
1954        case OP_UPTO:        case OP_UPTO:
1955          case OP_UPTOI:
1956        case OP_MINUPTO:        case OP_MINUPTO:
1957          case OP_MINUPTOI:
1958        case OP_POSUPTO:        case OP_POSUPTO:
1959          case OP_POSUPTOI:
1960        case OP_STAR:        case OP_STAR:
1961          case OP_STARI:
1962        case OP_MINSTAR:        case OP_MINSTAR:
1963          case OP_MINSTARI:
1964        case OP_POSSTAR:        case OP_POSSTAR:
1965          case OP_POSSTARI:
1966        case OP_PLUS:        case OP_PLUS:
1967          case OP_PLUSI:
1968        case OP_MINPLUS:        case OP_MINPLUS:
1969          case OP_MINPLUSI:
1970        case OP_POSPLUS:        case OP_POSPLUS:
1971          case OP_POSPLUSI:
1972        case OP_QUERY:        case OP_QUERY:
1973          case OP_QUERYI:
1974        case OP_MINQUERY:        case OP_MINQUERY:
1975          case OP_MINQUERYI:
1976        case OP_POSQUERY:        case OP_POSQUERY:
1977          case OP_POSQUERYI:
1978        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1979        break;        break;
1980        }        }
# Line 1823  could_be_empty_branch(const uschar *code Line 2013  could_be_empty_branch(const uschar *code
2013    compile_data *cd)    compile_data *cd)
2014  {  {
2015  register int c;  register int c;
2016  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2017       code < endcode;       code < endcode;
2018       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2019    {    {
2020    const uschar *ccode;    const uschar *ccode;
2021    
# Line 1841  for (code = first_significant_code(code Line 2031  for (code = first_significant_code(code
2031      continue;      continue;
2032      }      }
2033    
   /* 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;  
     }  
   
2034    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
2035    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
2036      forward reference subroutine call, we can't. To detect forward reference
2037      we have to scan up the list that is kept in the workspace. This function is
2038      called only when doing the real compile, not during the pre-compile that
2039      measures the size of the compiled pattern. */
2040    
2041    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2042      {      {
2043      BOOL empty_branch = FALSE;      const uschar *scode;
2044      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
2045    
2046        /* Test for forward reference */
2047    
2048        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2049          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2050    
2051        /* Not a forward reference, test for completed backward reference */
2052    
2053        empty_branch = FALSE;
2054        scode = cd->start_code + GET(code, 1);
2055      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2056    
2057        /* Completed backwards reference */
2058    
2059      do      do
2060        {        {
2061        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf8, cd))
# Line 1869  for (code = first_significant_code(code Line 2066  for (code = first_significant_code(code
2066        scode += GET(scode, 1);        scode += GET(scode, 1);
2067        }        }
2068      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2069    
2070      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2071      continue;      continue;
2072      }      }
2073    
2074      /* Groups with zero repeats can of course be empty; skip them. */
2075    
2076      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2077          c == OP_BRAPOSZERO)
2078        {
2079        code += _pcre_OP_lengths[c];
2080        do code += GET(code, 1); while (*code == OP_ALT);
2081        c = *code;
2082        continue;
2083        }
2084    
2085      /* A nested group that is already marked as "could be empty" can just be
2086      skipped. */
2087    
2088      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2089          c == OP_SCBRA || c == OP_SCBRAPOS)
2090        {
2091        do code += GET(code, 1); while (*code == OP_ALT);
2092        c = *code;
2093        continue;
2094        }
2095    
2096    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2097    
2098    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2099          c == OP_CBRA || c == OP_CBRAPOS ||
2100          c == OP_ONCE || c == OP_ONCE_NC ||
2101          c == OP_COND)
2102      {      {
2103      BOOL empty_branch;      BOOL empty_branch;
2104      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1962  for (code = first_significant_code(code Line 2185  for (code = first_significant_code(code
2185      case OP_ALLANY:      case OP_ALLANY:
2186      case OP_ANYBYTE:      case OP_ANYBYTE:
2187      case OP_CHAR:      case OP_CHAR:
2188      case OP_CHARNC:      case OP_CHARI:
2189      case OP_NOT:      case OP_NOT:
2190        case OP_NOTI:
2191      case OP_PLUS:      case OP_PLUS:
2192      case OP_MINPLUS:      case OP_MINPLUS:
2193      case OP_POSPLUS:      case OP_POSPLUS:
# Line 2003  for (code = first_significant_code(code Line 2227  for (code = first_significant_code(code
2227      case OP_KET:      case OP_KET:
2228      case OP_KETRMAX:      case OP_KETRMAX:
2229      case OP_KETRMIN:      case OP_KETRMIN:
2230        case OP_KETRPOS:
2231      case OP_ALT:      case OP_ALT:
2232      return TRUE;      return TRUE;
2233    
# Line 2011  for (code = first_significant_code(code Line 2236  for (code = first_significant_code(code
2236    
2237  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2238      case OP_STAR:      case OP_STAR:
2239        case OP_STARI:
2240      case OP_MINSTAR:      case OP_MINSTAR:
2241        case OP_MINSTARI:
2242      case OP_POSSTAR:      case OP_POSSTAR:
2243        case OP_POSSTARI:
2244      case OP_QUERY:      case OP_QUERY:
2245        case OP_QUERYI:
2246      case OP_MINQUERY:      case OP_MINQUERY:
2247        case OP_MINQUERYI:
2248      case OP_POSQUERY:      case OP_POSQUERY:
2249        case OP_POSQUERYI:
2250      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2251      break;      break;
2252    
2253      case OP_UPTO:      case OP_UPTO:
2254        case OP_UPTOI:
2255      case OP_MINUPTO:      case OP_MINUPTO:
2256        case OP_MINUPTOI:
2257      case OP_POSUPTO:      case OP_POSUPTO:
2258        case OP_POSUPTOI:
2259      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2260      break;      break;
2261  #endif  #endif
# Line 2032  for (code = first_significant_code(code Line 2266  for (code = first_significant_code(code
2266      case OP_MARK:      case OP_MARK:
2267      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2268      case OP_SKIP_ARG:      case OP_SKIP_ARG:
2269        code += code[1];
2270        break;
2271    
2272      case OP_THEN_ARG:      case OP_THEN_ARG:
2273      code += code[1];      code += code[1];
2274      break;      break;
# Line 2056  return TRUE; Line 2293  return TRUE;
2293  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
2294  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,
2295  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.
2296    This function is called only during the real compile, not during the
2297    pre-compile.
2298    
2299  Arguments:  Arguments:
2300    code        points to start of the recursion    code        points to start of the recursion
# Line 2106  where Perl recognizes it as the POSIX cl Line 2345  where Perl recognizes it as the POSIX cl
2345  "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,
2346  I think.  I think.
2347    
2348    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2349    It seems that the appearance of a nested POSIX class supersedes an apparent
2350    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2351    a digit.
2352    
2353    In Perl, unescaped square brackets may also appear as part of class names. For
2354    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2355    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2356    seem right at all. PCRE does not allow closing square brackets in POSIX class
2357    names.
2358    
2359  Arguments:  Arguments:
2360    ptr      pointer to the initial [    ptr      pointer to the initial [
2361    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2120  int terminator;          /* Don't combin Line 2370  int terminator;          /* Don't combin
2370  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2371  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2372    {    {
2373    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2374        ptr++;
2375      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2376      else
2377      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2378      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2379        {        {
2380        *endptr = ptr;        *endptr = ptr;
2381        return TRUE;        return TRUE;
2382        }        }
2383        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2384             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2385              ptr[1] == CHAR_EQUALS_SIGN) &&
2386            check_posix_syntax(ptr, endptr))
2387          return FALSE;
2388      }      }
2389    }    }
2390  return FALSE;  return FALSE;
# Line 2256  auto_callout(uschar *code, const uschar Line 2513  auto_callout(uschar *code, const uschar
2513  {  {
2514  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2515  *code++ = 255;  *code++ = 255;
2516  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2517  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2518  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2519  }  }
2520    
# Line 2282  Returns:             nothing Line 2539  Returns:             nothing
2539  static void  static void
2540  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2541  {  {
2542  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2543  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2544  }  }
2545    
# Line 2332  for (++c; c <= d; c++) Line 2589  for (++c; c <= d; c++)
2589    
2590  return TRUE;  return TRUE;
2591  }  }
2592    
2593    
2594    
2595    /*************************************************
2596    *        Check a character and a property        *
2597    *************************************************/
2598    
2599    /* This function is called by check_auto_possessive() when a property item
2600    is adjacent to a fixed character.
2601    
2602    Arguments:
2603      c            the character
2604      ptype        the property type
2605      pdata        the data for the type
2606      negated      TRUE if it's a negated property (\P or \p{^)
2607    
2608    Returns:       TRUE if auto-possessifying is OK
2609    */
2610    
2611    static BOOL
2612    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2613    {
2614    const ucd_record *prop = GET_UCD(c);
2615    switch(ptype)
2616      {
2617      case PT_LAMP:
2618      return (prop->chartype == ucp_Lu ||
2619              prop->chartype == ucp_Ll ||
2620              prop->chartype == ucp_Lt) == negated;
2621    
2622      case PT_GC:
2623      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2624    
2625      case PT_PC:
2626      return (pdata == prop->chartype) == negated;
2627    
2628      case PT_SC:
2629      return (pdata == prop->script) == negated;
2630    
2631      /* These are specials */
2632    
2633      case PT_ALNUM:
2634      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2635              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2636    
2637      case PT_SPACE:    /* Perl space */
2638      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2639              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2640              == negated;
2641    
2642      case PT_PXSPACE:  /* POSIX space */
2643      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2644              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2645              c == CHAR_FF || c == CHAR_CR)
2646              == negated;
2647    
2648      case PT_WORD:
2649      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2650              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2651              c == CHAR_UNDERSCORE) == negated;
2652      }
2653    return FALSE;
2654    }
2655  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2656    
2657    
# Line 2345  whether the next thing could possibly ma Line 2665  whether the next thing could possibly ma
2665  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2666    
2667  Arguments:  Arguments:
2668    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2669    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2670    ptr           next character in pattern    ptr           next character in pattern
2671    options       options bits    options       options bits
2672    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2357  Returns:        TRUE if possessifying is Line 2675  Returns:        TRUE if possessifying is
2675  */  */
2676    
2677  static BOOL  static BOOL
2678  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2679    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2680  {  {
2681  int next;  int c, next;
2682    int op_code = *previous++;
2683    
2684  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2685    
# Line 2371  if ((options & PCRE_EXTENDED) != 0) Line 2690  if ((options & PCRE_EXTENDED) != 0)
2690      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2691      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2692        {        {
2693        while (*(++ptr) != 0)        ptr++;
2694          while (*ptr != 0)
2695            {
2696          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2697            ptr++;
2698    #ifdef SUPPORT_UTF8
2699            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2700    #endif
2701            }
2702        }        }
2703      else break;      else break;
2704      }      }
# Line 2408  if ((options & PCRE_EXTENDED) != 0) Line 2734  if ((options & PCRE_EXTENDED) != 0)
2734      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2735      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2736        {        {
2737        while (*(++ptr) != 0)        ptr++;
2738          while (*ptr != 0)
2739            {
2740          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2741            ptr++;
2742    #ifdef SUPPORT_UTF8
2743            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2744    #endif
2745            }
2746        }        }
2747      else break;      else break;
2748      }      }
# Line 2421  if (*ptr == CHAR_ASTERISK || *ptr == CHA Line 2754  if (*ptr == CHAR_ASTERISK || *ptr == CHA
2754    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)
2755      return FALSE;      return FALSE;
2756    
2757  /* 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
2758  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. */  
2759    
2760  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2761    {    {
2762    case OP_CHAR:    case OP_CHAR:
2763  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2764    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2765  #else  #else
2766    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2767  #endif  #endif
2768    return item != next;    return c != next;
2769    
2770    /* 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
2771    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
2772    high-valued characters. */    high-valued characters. */
2773    
2774    case OP_CHARNC:    case OP_CHARI:
2775  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2776    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2777    #else
2778      c = *previous;
2779  #endif  #endif
2780    if (item == next) return FALSE;    if (c == next) return FALSE;
2781  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2782    if (utf8)    if (utf8)
2783      {      {
# Line 2458  if (next >= 0) switch(op_code) Line 2788  if (next >= 0) switch(op_code)
2788  #else  #else
2789      othercase = NOTACHAR;      othercase = NOTACHAR;
2790  #endif  #endif
2791      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2792      }      }
2793    else    else
2794  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2795    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2796    
2797    /* 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
2798      opcodes are not used for multi-byte characters, because they are coded using
2799      an XCLASS instead. */
2800    
2801    case OP_NOT:    case OP_NOT:
2802    if (item == next) return TRUE;    return (c = *previous) == next;
2803    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2804      case OP_NOTI:
2805      if ((c = *previous) == next) return TRUE;
2806  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2807    if (utf8)    if (utf8)
2808      {      {
# Line 2479  if (next >= 0) switch(op_code) Line 2813  if (next >= 0) switch(op_code)
2813  #else  #else
2814      othercase = NOTACHAR;      othercase = NOTACHAR;
2815  #endif  #endif
2816      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2817      }      }
2818    else    else
2819  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2820    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2821    
2822      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2823      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2824    
2825    case OP_DIGIT:    case OP_DIGIT:
2826    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2526  if (next >= 0) switch(op_code) Line 2863  if (next >= 0) switch(op_code)
2863      case 0x202f:      case 0x202f:
2864      case 0x205f:      case 0x205f:
2865      case 0x3000:      case 0x3000:
2866      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2867      default:      default:
2868      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2869      }      }
2870    
2871      case OP_ANYNL:
2872    case OP_VSPACE:    case OP_VSPACE:
2873    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2874    switch(next)    switch(next)
# Line 2542  if (next >= 0) switch(op_code) Line 2880  if (next >= 0) switch(op_code)
2880      case 0x85:      case 0x85:
2881      case 0x2028:      case 0x2028:
2882      case 0x2029:      case 0x2029:
2883      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2884      default:      default:
2885      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2886      }      }
2887    
2888    #ifdef SUPPORT_UCP
2889      case OP_PROP:
2890      return check_char_prop(next, previous[0], previous[1], FALSE);
2891    
2892      case OP_NOTPROP:
2893      return check_char_prop(next, previous[0], previous[1], TRUE);
2894    #endif
2895    
2896    default:    default:
2897    return FALSE;    return FALSE;
2898    }    }
2899    
2900    
2901  /* Handle the case when the next item is \d, \s, etc. */  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2902    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2903    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2904    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2905    replaced by OP_PROP codes when PCRE_UCP is set. */
2906    
2907  switch(op_code)  switch(op_code)
2908    {    {
2909    case OP_CHAR:    case OP_CHAR:
2910    case OP_CHARNC:    case OP_CHARI:
2911  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2912    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2913    #else
2914      c = *previous;
2915  #endif  #endif
2916    switch(-next)    switch(-next)
2917      {      {
2918      case ESC_d:      case ESC_d:
2919      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2920    
2921      case ESC_D:      case ESC_D:
2922      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2923    
2924      case ESC_s:      case ESC_s:
2925      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2926    
2927      case ESC_S:      case ESC_S:
2928      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2929    
2930      case ESC_w:      case ESC_w:
2931      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2932    
2933      case ESC_W:      case ESC_W:
2934      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2935    
2936      case ESC_h:      case ESC_h:
2937      case ESC_H:      case ESC_H:
2938      switch(item)      switch(c)
2939        {        {
2940        case 0x09:        case 0x09:
2941        case 0x20:        case 0x20:
# Line 2611  switch(op_code) Line 2963  switch(op_code)
2963    
2964      case ESC_v:      case ESC_v:
2965      case ESC_V:      case ESC_V:
2966      switch(item)      switch(c)
2967        {        {
2968        case 0x0a:        case 0x0a:
2969        case 0x0b:        case 0x0b:
# Line 2625  switch(op_code) Line 2977  switch(op_code)
2977        return -next == ESC_v;        return -next == ESC_v;
2978        }        }
2979    
2980        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2981        their substitutions and process them. The result will always be either
2982        -ESC_p or -ESC_P. Then fall through to process those values. */
2983    
2984    #ifdef SUPPORT_UCP
2985        case ESC_du:
2986        case ESC_DU:
2987        case ESC_wu:
2988        case ESC_WU:
2989        case ESC_su:
2990        case ESC_SU:
2991          {
2992          int temperrorcode = 0;
2993          ptr = substitutes[-next - ESC_DU];
2994          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2995          if (temperrorcode != 0) return FALSE;
2996          ptr++;    /* For compatibility */
2997          }
2998        /* Fall through */
2999    
3000        case ESC_p:
3001        case ESC_P:
3002          {
3003          int ptype, pdata, errorcodeptr;
3004          BOOL negated;
3005    
3006          ptr--;      /* Make ptr point at the p or P */
3007          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3008          if (ptype < 0) return FALSE;
3009          ptr++;      /* Point past the final curly ket */
3010    
3011          /* If the property item is optional, we have to give up. (When generated
3012          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3013          to the original \d etc. At this point, ptr will point to a zero byte. */
3014    
3015          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3016            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3017              return FALSE;
3018    
3019          /* Do the property check. */
3020    
3021          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3022          }
3023    #endif
3024    
3025      default:      default:
3026      return FALSE;      return FALSE;
3027      }      }
3028    
3029      /* In principle, support for Unicode properties should be integrated here as
3030      well. It means re-organizing the above code so as to get hold of the property
3031      values before switching on the op-code. However, I wonder how many patterns
3032      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3033      these op-codes are never generated.) */
3034    
3035    case OP_DIGIT:    case OP_DIGIT:
3036    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3037           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3038    
3039    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3040    return next == -ESC_d;    return next == -ESC_d;
3041    
3042    case OP_WHITESPACE:    case OP_WHITESPACE:
3043    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3044    
3045    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3046    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3047    
3048    case OP_HSPACE:    case OP_HSPACE:
3049    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3050             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3051    
3052    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3053    return next == -ESC_h;    return next == -ESC_h;
3054    
3055    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3056      case OP_ANYNL:
3057    case OP_VSPACE:    case OP_VSPACE:
3058    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3059    
3060    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3061    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3062    
3063    case OP_WORDCHAR:    case OP_WORDCHAR:
3064    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3065             next == -ESC_v || next == -ESC_R;
3066    
3067    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3068    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2688  Arguments: Line 3094  Arguments:
3094    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3095    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3096    bcptr          points to current branch chain    bcptr          points to current branch chain
3097      cond_depth     conditional nesting depth
3098    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3099    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3100                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2699  Returns:         TRUE on success Line 3106  Returns:         TRUE on success
3106  static BOOL  static BOOL
3107  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3108    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3109    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3110  {  {
3111  int repeat_type, op_type;  int repeat_type, op_type;
3112  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2708  int greedy_default, greedy_non_default; Line 3115  int greedy_default, greedy_non_default;
3115  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3116  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3117  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3118  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3119  int after_manual_callout = 0;  int after_manual_callout = 0;
3120  int length_prevgroup = 0;  int length_prevgroup = 0;
3121  register int c;  register int c;
# Line 2720  BOOL inescq = FALSE; Line 3127  BOOL inescq = FALSE;
3127  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3128  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3129  const uschar *tempptr;  const uschar *tempptr;
3130    const uschar *nestptr = NULL;
3131  uschar *previous = NULL;  uschar *previous = NULL;
3132  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3133  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3134  uschar classbits[32];  uschar classbits[32];
3135    
3136    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3137    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3138    dynamically as we process the pattern. */
3139    
3140  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3141  BOOL class_utf8;  BOOL class_utf8;
3142  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2733  uschar *class_utf8data_base; Line 3145  uschar *class_utf8data_base;
3145  uschar utf8_char[6];  uschar utf8_char[6];
3146  #else  #else
3147  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3148  #endif  #endif
3149    
3150  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 2784  for (;; ptr++) Line 3195  for (;; ptr++)
3195    int subfirstbyte;    int subfirstbyte;
3196    int terminator;    int terminator;
3197    int mclength;    int mclength;
3198      int tempbracount;
3199    uschar mcbuffer[8];    uschar mcbuffer[8];
3200    
3201    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3202    
3203    c = *ptr;    c = *ptr;
3204    
3205      /* If we are at the end of a nested substitution, revert to the outer level
3206      string. Nesting only happens one level deep. */
3207    
3208      if (c == 0 && nestptr != NULL)
3209        {
3210        ptr = nestptr;
3211        nestptr = NULL;
3212        c = *ptr;
3213        }
3214    
3215    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
3216    previous cycle of this loop. */    previous cycle of this loop. */
3217    
# Line 2820  for (;; ptr++) Line 3242  for (;; ptr++)
3242        goto FAILED;        goto FAILED;
3243        }        }
3244    
3245      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3246      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));
3247    
3248      /* 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 2895  for (;; ptr++) Line 3317  for (;; ptr++)
3317      previous_callout = NULL;      previous_callout = NULL;
3318      }      }
3319    
3320    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3321    
3322    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3323      {      {
3324      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3325      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3326        {        {
3327        while (*(++ptr) != 0)        ptr++;
3328          while (*ptr != 0)
3329          {          {
3330          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3331            ptr++;
3332    #ifdef SUPPORT_UTF8
3333            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3334    #endif
3335          }          }
3336        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3337    
# Line 2938  for (;; ptr++) Line 3365  for (;; ptr++)
3365          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3366          goto FAILED;          goto FAILED;
3367          }          }
3368        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3369        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3370        }        }
3371      return TRUE;      return TRUE;
# Line 2949  for (;; ptr++) Line 3376  for (;; ptr++)
3376      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3377    
3378      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3379        previous = NULL;
3380      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3381        {        {
3382        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3383          *code++ = OP_CIRCM;
3384        }        }
3385      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3386      break;      break;
3387    
3388      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3389      previous = NULL;      previous = NULL;
3390      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3391      break;      break;
3392    
3393      /* 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 3143  for (;; ptr++) Line 3571  for (;; ptr++)
3571            ptr++;            ptr++;
3572            }            }
3573    
3574          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3575          if (posix_class < 0)          if (posix_class < 0)
3576            {            {
3577            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 3157  for (;; ptr++) Line 3585  for (;; ptr++)
3585          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3586            posix_class = 0;            posix_class = 0;
3587    
3588          /* We build the bit map for the POSIX class in a chunk of local store          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3589          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3590          subtract bits that may be in the main map already. At the end we or the  
3591          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3592            if ((options & PCRE_UCP) != 0)
3593              {
3594              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3595              if (posix_substitutes[pc] != NULL)
3596                {
3597                nestptr = tempptr + 1;
3598                ptr = posix_substitutes[pc] - 1;
3599                continue;
3600                }
3601              }
3602    #endif
3603            /* In the non-UCP case, we build the bit map for the POSIX class in a
3604            chunk of local store because we may be adding and subtracting from it,
3605            and we don't want to subtract bits that may be in the main map already.
3606            At the end we or the result into the bit map that is being built. */
3607    
3608          posix_class *= 3;          posix_class *= 3;
3609    
# Line 3204  for (;; ptr++) Line 3647  for (;; ptr++)
3647    
3648        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3649        of the specials, which just set a flag. The sequence \b is a special        of the specials, which just set a flag. The sequence \b is a special
3650        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace. We
3651        Elsewhere it marks a word boundary. Other escapes have preset maps ready        assume that other escapes have more than one character in them, so set
3652        to 'or' into the one we are building. We assume they have more than one        class_charcount bigger than one. Unrecognized escapes fall through and
3653        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3654          PCRE_EXTRA is set. */
3655    
3656        if (c == CHAR_BACKSLASH)        if (c == CHAR_BACKSLASH)
3657          {          {
3658          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3659          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3660    
3661          if (-c == ESC_b) c = CHAR_BS;       /* \b is backspace in a class */          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
         else if (-c == ESC_X) c = CHAR_X;   /* \X is literal X in a class */  
         else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */  
3662          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3663            {            {
3664            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 3233  for (;; ptr++) Line 3675  for (;; ptr++)
3675            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3676            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3677    
3678            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3679              {              {
3680    #ifdef SUPPORT_UCP
3681                case ESC_du:     /* These are the values given for \d etc */
3682                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3683                case ESC_wu:     /* escape sequence with an appropriate \p */
3684                case ESC_WU:     /* or \P to test Unicode properties instead */
3685                case ESC_su:     /* of the default ASCII testing. */
3686                case ESC_SU:
3687                nestptr = ptr;
3688                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3689                class_charcount -= 2;                /* Undo! */
3690                continue;
3691    #endif
3692              case ESC_d:              case ESC_d:
3693              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3694              continue;              continue;
# Line 3255  for (;; ptr++) Line 3707  for (;; ptr++)
3707              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3708              continue;              continue;
3709    
3710                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3711                if it was previously set by something earlier in the character
3712                class. */
3713    
3714              case ESC_s:              case ESC_s:
3715              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3716              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3717                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3718              continue;              continue;
3719    
3720              case ESC_S:              case ESC_S:
# Line 3266  for (;; ptr++) Line 3723  for (;; ptr++)
3723              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3724              continue;              continue;
3725    
3726              default:    /* Not recognized; fall through */              case ESC_h:
             break;      /* Need "default" setting to stop compiler warning. */  
             }  
   
           /* In the pre-compile phase, just do the recognition. */  
   
           else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||  
                    c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;  
   
           /* We need to deal with \H, \h, \V, and \v in both phases because  
           they use extra memory. */  
   
           if (-c == ESC_h)  
             {  
3727              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3728              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3729              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 3303  for (;; ptr++) Line 3747  for (;; ptr++)
3747                }                }
3748  #endif  #endif
3749              continue;              continue;
             }  
3750    
3751            if (-c == ESC_H)              case ESC_H:
             {  
3752              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3753                {                {
3754                int x = 0xff;                int x = 0xff;
# Line 3348  for (;; ptr++) Line 3790  for (;; ptr++)
3790                }                }
3791  #endif  #endif
3792              continue;              continue;
             }  
3793    
3794            if (-c == ESC_v)              case ESC_v:
             {  
3795              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3796              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3797              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3367  for (;; ptr++) Line 3807  for (;; ptr++)
3807                }                }
3808  #endif  #endif
3809              continue;              continue;
             }  
3810    
3811            if (-c == ESC_V)              case ESC_V:
             {  
3812              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3813                {                {
3814                int x = 0xff;                int x = 0xff;
# Line 3400  for (;; ptr++) Line 3838  for (;; ptr++)
3838                }                }
3839  #endif  #endif
3840              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3841    
3842  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3843            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3844              {              case ESC_P:
3845              BOOL negated;                {
3846              int pdata;                BOOL negated;
3847              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3848              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3849              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3850              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3851                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3852              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3853              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3854              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3855              continue;                class_charcount -= 2;   /* Not a < 256 character */
3856              }                continue;
3857                  }
3858  #endif  #endif
3859            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3860            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3861            treated as literals. */              treated as literals. */
3862    
3863            if ((options & PCRE_EXTRA) != 0)              default:
3864              {              if ((options & PCRE_EXTRA) != 0)
3865              *errorcodeptr = ERR7;                {
3866              goto FAILED;                *errorcodeptr = ERR7;
3867                  goto FAILED;
3868                  }
3869                class_charcount -= 2;  /* Undo the default count from above */
3870                c = *ptr;              /* Get the final character and fall through */
3871                break;
3872              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3873            }            }
3874    
3875          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
# Line 3501  for (;; ptr++) Line 3939  for (;; ptr++)
3939            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3940            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3941    
3942            /* \b is backspace; \X is literal X; \R is literal R; any other            /* \b is backspace; any other special means the '-' was literal */
           special means the '-' was literal */  
3943    
3944            if (d < 0)            if (d < 0)
3945              {              {
3946              if (d == -ESC_b) d = CHAR_BS;              if (d == -ESC_b) d = CHAR_BS; else
             else if (d == -ESC_X) d = CHAR_X;  
             else if (d == -ESC_R) d = CHAR_R; else  
3947                {                {
3948                ptr = oldptr;                ptr = oldptr;
3949                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3674  for (;; ptr++) Line 4109  for (;; ptr++)
4109          }          }
4110        }        }
4111    
4112      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" far above.
4113        If we are at the end of an internal nested string, revert to the outer
4114        string. */
4115    
4116        while (((c = *(++ptr)) != 0 ||
4117               (nestptr != NULL &&
4118                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4119               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4120    
4121      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));      /* Check for missing terminating ']' */
4122    
4123      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4124        {        {
4125        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4126        goto FAILED;        goto FAILED;
4127        }        }
4128    
   
 /* This code has been disabled because it would mean that \s counts as  
 an explicit \r or \n reference, and that's not really what is wanted. Now  
 we set the flag only if there is a literal "\r" or "\n" in the class. */  
   
 #if 0  
     /* Remember whether \r or \n are in this class */  
   
     if (negate_class)  
       {  
       if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;  
       }  
     else  
       {  
       if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;  
       }  
 #endif  
   
   
4129      /* If class_charcount is 1, we saw precisely one character whose value is      /* If class_charcount is 1, we saw precisely one character whose value is
4130      less than 256. As long as there were no characters >= 128 and there was no      less than 256. As long as there were no characters >= 128 and there was no
4131      use of \p or \P, in other words, no use of any XCLASS features, we can      use of \p or \P, in other words, no use of any XCLASS features, we can
# Line 3710  we set the flag only if there is a liter Line 4133  we set the flag only if there is a liter
4133    
4134      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
4135      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4136      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4137      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4138    
4139      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
4140      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.
4141      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
4142      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
4143      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
4144      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4145    
4146  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4147      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3729  we set the flag only if there is a liter Line 4152  we set the flag only if there is a liter
4152        {        {
4153        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4154    
4155        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4156    
4157        if (negate_class)        if (negate_class)
4158          {          {
4159          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4160          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4161          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4162          *code++ = class_lastchar;          *code++ = class_lastchar;
4163          break;          break;
4164          }          }
# Line 3766  we set the flag only if there is a liter Line 4189  we set the flag only if there is a liter
4189    
4190      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4191      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4192      such as \S in the class, because in that case all characters > 255 are in      such as \S in the class, and PCRE_UCP is not set, because in that case all
4193      the class, so any that were explicitly given as well can be ignored. If      characters > 255 are in the class, so any that were explicitly given as
4194      (when there are explicit characters > 255 that must be listed) there are no      well can be ignored. If (when there are explicit characters > 255 that must
4195      characters < 256, we can omit the bitmap in the actual compiled code. */      be listed) there are no characters < 256, we can omit the bitmap in the
4196        actual compiled code. */
4197    
4198  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4199      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4200        {        {
4201        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4202        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3798  we set the flag only if there is a liter Line 4222  we set the flag only if there is a liter
4222        }        }
4223  #endif  #endif
4224    
4225      /* If there are no characters > 255, set the opcode to OP_CLASS or      /* If there are no characters > 255, or they are all to be included or
4226      OP_NCLASS, depending on whether the whole class was negated and whether      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
4227      there were negative specials such as \S in the class. Then copy the 32-byte      whole class was negated and whether there were negative specials such as \S
4228      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4229        negating it if necessary. */
4230    
4231      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4232      if (negate_class)      if (negate_class)
# Line 3861  we set the flag only if there is a liter Line 4286  we set the flag only if there is a liter
4286      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4287      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4288    
4289      /* 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
4290      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4291    
4292      tempcode = previous;      tempcode = previous;
4293    
# Line 3885  we set the flag only if there is a liter Line 4310  we set the flag only if there is a liter
4310        }        }
4311      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4312    
4313        /* If previous was a recursion call, wrap it in atomic brackets so that
4314        previous becomes the atomic group. All recursions were so wrapped in the
4315        past, but it no longer happens for non-repeated recursions. In fact, the
4316        repeated ones could be re-implemented independently so as not to need this,
4317        but for the moment we rely on the code for repeating groups. */
4318    
4319        if (*previous == OP_RECURSE)
4320          {
4321          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4322          *previous = OP_ONCE;
4323          PUT(previous, 1, 2 + 2*LINK_SIZE);
4324          previous[2 + 2*LINK_SIZE] = OP_KET;
4325          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4326          code += 2 + 2 * LINK_SIZE;
4327          length_prevgroup = 3 + 3*LINK_SIZE;
4328    
4329          /* When actually compiling, we need to check whether this was a forward
4330          reference, and if so, adjust the offset. */
4331    
4332          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4333            {
4334            int offset = GET(cd->hwm, -LINK_SIZE);
4335            if (offset == previous + 1 - cd->start_code)
4336              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4337            }
4338          }
4339    
4340        /* Now handle repetition for the different types of item. */
4341    
4342      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4343      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
4344      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
4345      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
4346      instead.  */      instead.  */
4347    
4348      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4349        {        {
4350          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4351    
4352        /* 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
4353        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
4354        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 3925  we set the flag only if there is a liter Line 4381  we set the flag only if there is a liter
4381    
4382        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4383            repeat_max < 0 &&            repeat_max < 0 &&
4384            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4385          {          {
4386          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4387          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3938  we set the flag only if there is a liter Line 4393  we set the flag only if there is a liter
4393      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4394      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-
4395      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4396      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
4397      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4398    
4399      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4400        {        {
4401        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4402        c = previous[1];        c = previous[1];
4403        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4404            repeat_max < 0 &&            repeat_max < 0 &&
4405            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4406          {          {
4407          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4408          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3971  we set the flag only if there is a liter Line 4426  we set the flag only if there is a liter
4426    
4427        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4428            repeat_max < 0 &&            repeat_max < 0 &&
4429            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4430          {          {
4431          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4432          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4140  we set the flag only if there is a liter Line 4595  we set the flag only if there is a liter
4595  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4596               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4597  #endif  #endif
4598               *previous == OP_REF)               *previous == OP_REF ||
4599                 *previous == OP_REFI)
4600        {        {
4601        if (repeat_max == 0)        if (repeat_max == 0)
4602          {          {
# Line 4174  we set the flag only if there is a liter Line 4630  we set the flag only if there is a liter
4630        }        }
4631    
4632      /* 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
4633      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4634        opcodes such as BRA and CBRA, as this is the place where they get converted
4635        into the more special varieties such as BRAPOS and SBRA. A test for >=
4636        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4637        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4638        repetition of assertions, but now it does, for Perl compatibility. */
4639    
4640      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4641        {        {
4642        register int i;        register int i;
4643        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4644        uschar *bralink = NULL;        uschar *bralink = NULL;
4645          uschar *brazeroptr = NULL;
4646    
4647        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4648          we just ignore the repeat. */
4649    
4650        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4651          {          goto END_REPEAT;
4652          *errorcodeptr = ERR55;  
4653          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4654          }        use of repetition is in cases when the assertion is optional. Therefore,
4655          if the minimum is greater than zero, just ignore the repeat. If the
4656          maximum is not not zero or one, set it to 1. */
4657    
4658        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4659        by scanning through from the start, and compute the offset back to it          {
4660        from the current code pointer. There may be an OP_OPT setting following          if (repeat_min > 0) goto END_REPEAT;
4661        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;  
4662          }          }
4663    
4664        /* 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 4223  we set the flag only if there is a liter Line 4679  we set the flag only if there is a liter
4679          **   goto END_REPEAT;          **   goto END_REPEAT;
4680          **   }          **   }
4681    
4682          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
4683          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
4684          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
4685          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4686            selectively.
4687    
4688          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
4689          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 4246  we set the flag only if there is a liter Line 4703  we set the flag only if there is a liter
4703              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4704              goto END_REPEAT;              goto END_REPEAT;
4705              }              }
4706              brazeroptr = previous;    /* Save for possessive optimizing */
4707            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4708            }            }
4709    
# Line 4270  we set the flag only if there is a liter Line 4728  we set the flag only if there is a liter
4728            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4729            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4730    
4731            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4732            bralink = previous;            bralink = previous;
4733            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4734            }            }
# Line 4379  we set the flag only if there is a liter Line 4837  we set the flag only if there is a liter
4837              {              {
4838              int offset;              int offset;
4839              *code++ = OP_BRA;              *code++ = OP_BRA;
4840              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4841              bralink = code;              bralink = code;
4842              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4843              }              }
# Line 4400  we set the flag only if there is a liter Line 4858  we set the flag only if there is a liter
4858          while (bralink != NULL)          while (bralink != NULL)
4859            {            {
4860            int oldlinkoffset;            int oldlinkoffset;
4861            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4862            uschar *bra = code - offset;            uschar *bra = code - offset;
4863            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4864            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4410  we set the flag only if there is a liter Line 4868  we set the flag only if there is a liter
4868            }            }
4869          }          }
4870    
4871        /* 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
4872        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4873        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
4874        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4875          deal with possessive ONCEs specially.
4876    
4877          Otherwise, if the quantifier was possessive, we convert the BRA code to
4878          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4879          at runtime to detect this kind of subpattern at both the start and at the
4880          end.) The use of special opcodes makes it possible to reduce greatly the
4881          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4882          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4883          the default action below, of wrapping everything inside atomic brackets,
4884          does not happen.
4885    
4886        Then, when we are doing the actual compile phase, check to see whether        Then, when we are doing the actual compile phase, check to see whether
4887        this group is a non-atomic one that could match an empty string. If so,        this group is one that could match an empty string. If so, convert the
4888        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so that runtime
4889        that runtime checking can be done. [This check is also applied to        checking can be done. [This check is also applied to ONCE groups at
4890        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4891    
4892        else        else
4893          {          {
4894          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4895          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4896          *ketcode = OP_KETRMAX + repeat_type;  
4897          if (lengthptr == NULL && *bracode != OP_ONCE)          if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
4898                possessive_quantifier) *bracode = OP_BRA;
4899    
4900            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
4901              *ketcode = OP_KETRMAX + repeat_type;
4902            else
4903            {            {
4904            uschar *scode = bracode;            if (possessive_quantifier)
           do  
4905              {              {
4906              if (could_be_empty_branch(scode, ketcode, utf8, cd))              *bracode += 1;                   /* Switch to xxxPOS opcodes */
4907                *ketcode = OP_KETRPOS;
4908                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4909                possessive_quantifier = FALSE;
4910                }
4911              else *ketcode = OP_KETRMAX + repeat_type;
4912    
4913              if (lengthptr == NULL)
4914                {
4915                uschar *scode = bracode;
4916                do
4917                {                {
4918                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4919                break;                  {
4920                    *bracode += OP_SBRA - OP_BRA;
4921                    break;
4922                    }
4923                  scode += GET(scode, 1);
4924                }                }
4925              scode += GET(scode, 1);              while (*scode == OP_ALT);
4926              }              }
           while (*scode == OP_ALT);  
4927            }            }
4928          }          }
4929        }        }
# Line 4459  we set the flag only if there is a liter Line 4944  we set the flag only if there is a liter
4944        }        }
4945    
4946      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4947      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4948      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4949      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4950      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
4951      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4952      tempcode, not at previous, which might be the first part of a string whose  
4953      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4954        just above, so possessive_quantifier is always FALSE for them at this
4955        stage.
4956    
4957        Note that the repeated item starts at tempcode, not at previous, which
4958        might be the first part of a string whose (former) last char we repeated.
4959    
4960      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
4961      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 4488  we set the flag only if there is a liter Line 4978  we set the flag only if there is a liter
4978  #endif  #endif
4979          }          }
4980    
4981        len = code - tempcode;        len = (int)(code - tempcode);
4982        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4983          {          {
4984          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4496  we set the flag only if there is a liter Line 4986  we set the flag only if there is a liter
4986          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4987          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4988    
4989          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4990          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4991          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4992          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4993    
4994          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4995          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4996          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4997          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4998    
4999            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
5000            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
5001            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
5002            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
5003    
5004            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
5005            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
5006            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
5007            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
5008    
5009          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
5010          pending recursive references are updated. */          pending recursive references are updated. */
5011    
# Line 4557  we set the flag only if there is a liter Line 5057  we set the flag only if there is a liter
5057        const uschar *arg = NULL;        const uschar *arg = NULL;
5058        previous = NULL;        previous = NULL;
5059        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5060        namelen = ptr - name;        namelen = (int)(ptr - name);
5061    
5062          /* It appears that Perl allows any characters whatsoever, other than
5063          a closing parenthesis, to appear in arguments, so we no longer insist on
5064          letters, digits, and underscores. */
5065    
5066        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5067          {          {
5068          arg = ++ptr;          arg = ++ptr;
5069          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5070            || *ptr == '_') ptr++;          arglen = (int)(ptr - arg);
         arglen = ptr - arg;  
5071          }          }
5072    
5073        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
# Line 4580  we set the flag only if there is a liter Line 5083  we set the flag only if there is a liter
5083          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5084              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5085            {            {
5086            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5087              ASSERT_ACCEPT if in an assertion. */
5088    
5089            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5090              {              {
5091              open_capitem *oc;              open_capitem *oc;
5092                if (arglen != 0)
5093                  {
5094                  *errorcodeptr = ERR59;
5095                  goto FAILED;
5096                  }
5097              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5098              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5099                {                {
5100                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5101                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5102                }                }
5103                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5104    
5105                /* Do not set firstbyte after *ACCEPT */
5106                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5107              }              }
5108    
5109            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
5110    
5111            if (arglen == 0)            else if (arglen == 0)
5112              {              {
5113              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
5114                {                {
5115                *errorcodeptr = ERR66;                *errorcodeptr = ERR66;
5116                goto FAILED;                goto FAILED;
5117                }                }
5118              *code++ = verbs[i].op;              *code = verbs[i].op;
5119                if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
5120              }              }
5121    
5122            else            else
# Line 4612  we set the flag only if there is a liter Line 5126  we set the flag only if there is a liter
5126                *errorcodeptr = ERR59;                *errorcodeptr = ERR59;
5127                goto FAILED;                goto FAILED;
5128                }                }
5129              *code++ = verbs[i].op_arg;              *code = verbs[i].op_arg;
5130                if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
5131              *code++ = arglen;              *code++ = arglen;
5132              memcpy(code, arg, arglen);              memcpy(code, arg, arglen);
5133              code += arglen;              code += arglen;
# Line 4745  we set the flag only if there is a liter Line 5260  we set the flag only if there is a liter
5260                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5261            ptr++;            ptr++;
5262            }            }
5263          namelen = ptr - name;          namelen = (int)(ptr - name);
5264    
5265          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
5266              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
# Line 4806  we set the flag only if there is a liter Line 5321  we set the flag only if there is a liter
5321          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5322    
5323          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5324                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5325            {            {
5326            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5327            code[1+LINK_SIZE]++;            code[1+LINK_SIZE]++;
# Line 4874  we set the flag only if there is a liter Line 5389  we set the flag only if there is a liter
5389          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5390          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5391          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5392            cd->assert_depth += 1;
5393          ptr++;          ptr++;
5394          break;          break;
5395    
# Line 4888  we set the flag only if there is a liter Line 5404  we set the flag only if there is a liter
5404            continue;            continue;
5405            }            }
5406          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5407            cd->assert_depth += 1;
5408          break;          break;
5409    
5410    
# Line 4897  we set the flag only if there is a liter Line 5414  we set the flag only if there is a liter
5414            {            {
5415            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5416            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5417              cd->assert_depth += 1;
5418            ptr += 2;            ptr += 2;
5419            break;            break;
5420    
5421            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5422            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5423              cd->assert_depth += 1;
5424            ptr += 2;            ptr += 2;
5425            break;            break;
5426    
# Line 4941  we set the flag only if there is a liter Line 5460  we set the flag only if there is a liter
5460              goto FAILED;              goto FAILED;
5461              }              }
5462            *code++ = n;            *code++ = n;
5463            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5464            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5465            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5466            }            }
5467          previous = NULL;          previous = NULL;
# Line 4975  we set the flag only if there is a liter Line 5494  we set the flag only if there is a liter
5494            name = ++ptr;            name = ++ptr;
5495    
5496            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5497            namelen = ptr - name;            namelen = (int)(ptr - name);
5498    
5499            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5500    
# Line 5105  we set the flag only if there is a liter Line 5624  we set the flag only if there is a liter
5624          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5625          name = ++ptr;          name = ++ptr;
5626          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5627          namelen = ptr - name;          namelen = (int)(ptr - name);
5628    
5629          /* 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
5630          reference number. */          a dummy reference number, because it was not used in the first pass.
5631            However, with the change of recursive back references to be atomic,
5632            we have to look for the number so that this state can be identified, as
5633            otherwise the incorrect length is computed. If it's not a backwards
5634            reference, the dummy number will do. */
5635    
5636          if (lengthptr != NULL)          if (lengthptr != NULL)
5637            {            {
5638              const uschar *temp;
5639    
5640            if (namelen == 0)            if (namelen == 0)
5641              {              {
5642              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5127  we set the flag only if there is a liter Line 5652  we set the flag only if there is a liter
5652              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5653              goto FAILED;              goto FAILED;
5654              }              }
5655            recno = 0;  
5656              /* The name table does not exist in the first pass, so we cannot
5657              do a simple search as in the code below. Instead, we have to scan the
5658              pattern to find the number. It is important that we scan it only as
5659              far as we have got because the syntax of named subpatterns has not
5660              been checked for the rest of the pattern, and find_parens() assumes
5661              correct syntax. In any case, it's a waste of resources to scan
5662              further. We stop the scan at the current point by temporarily
5663              adjusting the value of cd->endpattern. */
5664    
5665              temp = cd->end_pattern;
5666              cd->end_pattern = ptr;
5667              recno = find_parens(cd, name, namelen,
5668                (options & PCRE_EXTENDED) != 0, utf8);
5669              cd->end_pattern = temp;
5670              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5671            }            }
5672    
5673          /* 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 5152  we set the flag only if there is a liter Line 5692  we set the flag only if there is a liter
5692              }              }
5693            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5694                      find_parens(cd, name, namelen,                      find_parens(cd, name, namelen,
5695                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5696              {              {
5697              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5698              goto FAILED;              goto FAILED;
# Line 5263  we set the flag only if there is a liter Line 5803  we set the flag only if there is a liter
5803              if (called == NULL)              if (called == NULL)
5804                {                {
5805                if (find_parens(cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5806                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5807                  {                  {
5808                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5809                  goto FAILED;                  goto FAILED;
# Line 5271  we set the flag only if there is a liter Line 5811  we set the flag only if there is a liter
5811    
5812                /* 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
5813                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5814                of the group. */                of the group. Then remember the forward reference. */
5815    
5816                called = cd->start_code + recno;                called = cd->start_code + recno;
5817                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5818                }                }
5819    
5820              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
5821              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
5822              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
5823                must not, however, do this check if we are in a conditional
5824                subpattern because the condition might be testing for recursion in
5825                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
5826                Forever loops are also detected at runtime, so those that occur in
5827                conditional subpatterns will be picked up then. */
5828    
5829              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
5830                       could_be_empty(called, code, bcptr, utf8, cd))                       could_be_empty(called, code, bcptr, utf8, cd))
5831                {                {
5832                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
# Line 5289  we set the flag only if there is a liter Line 5834  we set the flag only if there is a liter
5834                }                }
5835              }              }
5836    
5837            /* 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;  
5838    
5839            *code = OP_RECURSE;            *code = OP_RECURSE;
5840            PUT(code, 1, called - cd->start_code);            PUT(code, 1, (int)(called - cd->start_code));
5841            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;  
5842            }            }
5843    
5844          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5366  we set the flag only if there is a liter Line 5899  we set the flag only if there is a liter
5899          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
5900          both phases.          both phases.
5901    
5902          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
5903          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. */  
5904    
5905          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5906            {            {
# Line 5379  we set the flag only if there is a liter Line 5911  we set the flag only if there is a liter
5911              }              }
5912            else            else
5913              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
5914              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5915              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5916              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5917              }              }
5918    
5919            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
5920            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). */  
5921    
5922            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
5923            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5428  we set the flag only if there is a liter Line 5953  we set the flag only if there is a liter
5953        skipbytes = 2;        skipbytes = 2;
5954        }        }
5955    
5956      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
5957      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
5958      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
5959      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. */  
5960    
5961      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                      /* For handling repetition */
5962      *code = bravalue;      *code = bravalue;
5963      tempcode = code;      tempcode = code;
5964      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
5965      length_prevgroup = 0;              /* Initialize for pre-compile phase */      tempbracount = cd->bracount;          /* Save value before bracket */
5966        length_prevgroup = 0;                 /* Initialize for pre-compile phase */
5967    
5968      if (!compile_regex(      if (!compile_regex(
5969           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
5970           options & PCRE_IMS,           /* The previous ims option state */           &tempcode,                       /* Where to put code (updated) */
5971           &tempcode,                    /* Where to put code (updated) */           &ptr,                            /* Input pointer (updated) */
5972           &ptr,                         /* Input pointer (updated) */           errorcodeptr,                    /* Where to put an error message */
          errorcodeptr,                 /* Where to put an error message */  
5973           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
5974            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
5975           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
5976           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
5977           &subfirstbyte,                /* For possible first char */           cond_depth +
5978           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
5979           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
5980           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
5981           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
5982             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
5983             (lengthptr == NULL)? NULL :      /* Actual compile phase */
5984               &length_prevgroup              /* Pre-compile phase */
5985           ))           ))
5986        goto FAILED;        goto FAILED;
5987    
5988        /* If this was an atomic group and there are no capturing groups within it,
5989        generate OP_ONCE_NC instead of OP_ONCE. */
5990    
5991        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
5992          *code = OP_ONCE_NC;
5993    
5994        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5995          cd->assert_depth -= 1;
5996    
5997      /* 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
5998      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.
5999      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
6000    
6001      /* 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
6002      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
6003      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
6004      not be available. */      not be available. */
# Line 5529  we set the flag only if there is a liter Line 6063  we set the flag only if there is a liter
6063          goto FAILED;          goto FAILED;
6064          }          }
6065        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
6066        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
6067        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
6068        *code++ = OP_KET;        *code++ = OP_KET;
6069        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5602  we set the flag only if there is a liter Line 6136  we set the flag only if there is a liter
6136    
6137      /* ===================================================================*/      /* ===================================================================*/
6138      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
6139      are arranged to be the negation of the corresponding OP_values. For the      are arranged to be the negation of the corresponding OP_values in the
6140      back references, the values are ESC_REF plus the reference number. Only      default case when PCRE_UCP is not set. For the back references, the values
6141      back references and those types that consume a character may be repeated.      are ESC_REF plus the reference number. Only back references and those types
6142      We can test for values between ESC_b and ESC_Z for the latter; this may      that consume a character may be repeated. We can test for values between
6143      have to change if any new ones are ever created. */      ESC_b and ESC_Z for the latter; this may have to change if any new ones are
6144        ever created. */
6145    
6146      case CHAR_BACKSLASH:      case CHAR_BACKSLASH:
6147      tempptr = ptr;      tempptr = ptr;
# Line 5696  we set the flag only if there is a liter Line 6231  we set the flag only if there is a liter
6231          }          }
6232    
6233        /* \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).
6234        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6235    
6236        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6237          {          {
6238            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6239              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6240              {
6241              *errorcodeptr = ERR69;
6242              break;
6243              }
6244          is_recurse = FALSE;          is_recurse = FALSE;
6245          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6246            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
# Line 5720  we set the flag only if there is a liter Line 6260  we set the flag only if there is a liter
6260          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6261          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6262          previous = code;          previous = code;
6263          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6264          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6265          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6266          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
# Line 5766  we set the flag only if there is a liter Line 6306  we set the flag only if there is a liter
6306  #endif  #endif
6307    
6308        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
6309        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
6310          situation when PCRE_UCP is not set. When it *is* set, we substitute
6311          Unicode property tests. */
6312    
6313        else        else
6314          {          {
6315          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
6316          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
6317              {
6318              nestptr = ptr + 1;                   /* Where to resume */
6319              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
6320              }
6321            else
6322    #endif
6323              {
6324              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6325              *code++ = -c;
6326              }
6327          }          }
6328        continue;        continue;
6329        }        }
# Line 5816  we set the flag only if there is a liter Line 6368  we set the flag only if there is a liter
6368    
6369      ONE_CHAR:      ONE_CHAR:
6370      previous = code;      previous = code;
6371      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6372      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6373    
6374      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 5845  we set the flag only if there is a liter Line 6397  we set the flag only if there is a liter
6397        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6398        }        }
6399    
6400      /* 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
6401      1 or the matching is caseful. */      1 or the matching is caseful. */
6402    
6403      else      else
# Line 5880  return FALSE; Line 6432  return FALSE;
6432  /* 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
6433  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
6434  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.  
   
6435  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
6436  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
6437  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6438    
6439  Arguments:  Arguments:
6440    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  
6441    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6442    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6443    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
6444    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6445    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6446    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6447      cond_depth     depth of nesting for conditional subpatterns
6448    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6449    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6450    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 5909  Returns:         TRUE on success Line 6456  Returns:         TRUE on success
6456  */  */
6457    
6458  static BOOL  static BOOL
6459  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6460    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6461    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6462    int *lengthptr)    compile_data *cd, int *lengthptr)
6463  {  {
6464  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6465  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 5926  int branchfirstbyte, branchreqbyte; Line 6473  int branchfirstbyte, branchreqbyte;
6473  int length;  int length;
6474  int orig_bracount;  int orig_bracount;
6475  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6476  branch_chain bc;  branch_chain bc;
6477    
6478  bc.outer = bcptr;  bc.outer = bcptr;
# Line 5950  pre-compile phase to find out whether an Line 6496  pre-compile phase to find out whether an
6496    
6497  /* 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
6498  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
6499  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6500    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6501    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6502    
6503  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6504    {    {
# Line 5976  for (;;) Line 6524  for (;;)
6524    
6525    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6526    
   /* 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;  
     }  
   
6527    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6528    
6529    if (lookbehind)    if (lookbehind)
# Line 5999  for (;;) Line 6538  for (;;)
6538    into the length. */    into the length. */
6539    
6540    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6541          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6542            (lengthptr == NULL)? NULL : &length))
6543      {      {
6544      *ptrptr = ptr;      *ptrptr = ptr;
6545      return FALSE;      return FALSE;
6546      }      }
6547    
   /* 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;  
   
6548    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6549    has fewer than the rest. */    has fewer than the rest. */
6550    
# Line 6074  for (;;) Line 6605  for (;;)
6605        {        {
6606        int fixed_length;        int fixed_length;
6607        *code = OP_END;        *code = OP_END;
6608        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6609            FALSE, cd);
6610        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6611        if (fixed_length == -3)        if (fixed_length == -3)
6612          {          {
# Line 6095  for (;;) Line 6627  for (;;)
6627    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
6628    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
6629    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
6630    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. */  
6631    
6632    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6633      {      {
6634      if (lengthptr == NULL)      if (lengthptr == NULL)
6635        {        {
6636        int branch_length = code - last_branch;        int branch_length = (int)(code - last_branch);
6637        do        do
6638          {          {
6639          int prev_length = GET(last_branch, 1);          int prev_length = GET(last_branch, 1);
# Line 6117  for (;;) Line 6647  for (;;)
6647      /* Fill in the ket */      /* Fill in the ket */
6648    
6649      *code = OP_KET;      *code = OP_KET;
6650      PUT(code, 1, code - start_bracket);      PUT(code, 1, (int)(code - start_bracket));
6651      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6652    
6653      /* 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 6132  for (;;) Line 6662  for (;;)
6662            code - start_bracket);            code - start_bracket);
6663          *start_bracket = OP_ONCE;          *start_bracket = OP_ONCE;
6664          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
6665          PUT(start_bracket, 1, code - start_bracket);          PUT(start_bracket, 1, (int)(code - start_bracket));
6666          *code = OP_KET;          *code = OP_KET;
6667          PUT(code, 1, code - start_bracket);          PUT(code, 1, (int)(code - start_bracket));
6668          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
6669          length += 2 + 2*LINK_SIZE;          length += 2 + 2*LINK_SIZE;
6670          }          }
6671        cd->open_caps = cd->open_caps->next;        cd->open_caps = cd->open_caps->next;
6672        }        }
6673    
     /* Reset options if needed. */  
   
     if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)  
       {  
       *code++ = OP_OPT;  
       *code++ = oldims;  
       length += 2;  
       }  
   
6674      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
6675    
6676      cd->bracount = max_bracount;      cd->bracount = max_bracount;
# Line 6189  for (;;) Line 6710  for (;;)
6710    else    else
6711      {      {
6712      *code = OP_ALT;      *code = OP_ALT;
6713      PUT(code, 1, code - last_branch);      PUT(code, 1, (int)(code - last_branch));
6714      bc.current_branch = last_branch = code;      bc.current_branch = last_branch = code;
6715      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6716      }      }
# Line 6209  for (;;) Line 6730  for (;;)
6730  /* 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
6731  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
6732  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
6733  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
6734  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6735    
6736  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.
6737  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 6231  of the more common cases more precisely. Line 6752  of the more common cases more precisely.
6752    
6753  Arguments:  Arguments:
6754    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6755    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
6756                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6757                    the less precise approach                    the less precise approach
# Line 6241  Returns:     TRUE or FALSE Line 6761  Returns:     TRUE or FALSE
6761  */  */
6762    
6763  static BOOL  static BOOL
6764  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6765    unsigned int backref_map)    unsigned int backref_map)
6766  {  {
6767  do {  do {
6768     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6769       options, PCRE_MULTILINE, FALSE);       FALSE);
6770     register int op = *scode;     register int op = *scode;
6771    
6772     /* Non-capturing brackets */     /* Non-capturing brackets */
6773    
6774     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6775           op == OP_SBRA || op == OP_SBRAPOS)
6776       {       {
6777       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6778       }       }
6779    
6780     /* Capturing brackets */     /* Capturing brackets */
6781    
6782     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6783                op == OP_SCBRA || op == OP_SCBRAPOS)
6784       {       {
6785       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6786       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6787       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6788       }       }
6789    
6790     /* Other brackets */     /* Other brackets */
6791    
6792     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC ||
6793                op == OP_COND)
6794       {       {
6795       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6796       }       }
6797    
6798     /* .* 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 6284  do { Line 6807  do {
6807    
6808     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6809    
6810     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;  
6811     code += GET(code, 1);     code += GET(code, 1);
6812     }     }
6813  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6322  is_startline(const uschar *code, unsigne Line 6843  is_startline(const uschar *code, unsigne
6843  {  {
6844  do {  do {
6845     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6846       NULL, 0, FALSE);       FALSE);
6847     register int op = *scode;     register int op = *scode;
6848    
6849     /* 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 6349  do { Line 6870  do {
6870         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6871         break;         break;
6872         }         }
6873       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
6874       op = *scode;       op = *scode;
6875       }       }
6876    
6877     /* Non-capturing brackets */     /* Non-capturing brackets */
6878    
6879     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6880           op == OP_SBRA || op == OP_SBRAPOS)
6881       {       {
6882       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6883       }       }
6884    
6885     /* Capturing brackets */     /* Capturing brackets */
6886    
6887     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6888                op == OP_SCBRA || op == OP_SCBRAPOS)
6889       {       {
6890       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6891       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6371  do { Line 6894  do {
6894    
6895     /* Other brackets */     /* Other brackets */
6896    
6897     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC)
6898       {       {
6899       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6900       }       }
# Line 6386  do { Line 6909  do {
6909    
6910     /* Check for explicit circumflex */     /* Check for explicit circumflex */
6911    
6912     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
6913    
6914     /* Move on to the next alternative */     /* Move on to the next alternative */
6915    
# Line 6412  we return that char, otherwise -1. Line 6935  we return that char, otherwise -1.
6935    
6936  Arguments:  Arguments:
6937    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)  
6938    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
6939    
6940  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
6941  */  */
6942    
6943  static int  static int
6944  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
6945  {  {
6946  register int c = -1;  register int c = -1;
6947  do {  do {
6948     int d;     int d;
6949     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
6950       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
6951       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
6952     register int op = *scode;     register int op = *scode;
6953    
6954     switch(op)     switch(op)
# Line 6434  do { Line 6957  do {
6957       return -1;       return -1;
6958    
6959       case OP_BRA:       case OP_BRA:
6960         case OP_BRAPOS:
6961       case OP_CBRA:       case OP_CBRA:
6962         case OP_SCBRA:
6963         case OP_CBRAPOS:
6964         case OP_SCBRAPOS:
6965       case OP_ASSERT:       case OP_ASSERT:
6966       case OP_ONCE:       case OP_ONCE:
6967         case OP_ONCE_NC:
6968       case OP_COND:       case OP_COND:
6969       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
6970         return -1;         return -1;
6971       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
6972       break;       break;
6973    
6974       case OP_EXACT:       /* Fall through */       case OP_EXACT:
6975       scode += 2;       scode += 2;
6976         /* Fall through */
6977    
6978       case OP_CHAR:       case OP_CHAR:
      case OP_CHARNC:  
6979       case OP_PLUS:       case OP_PLUS:
6980       case OP_MINPLUS:       case OP_MINPLUS:
6981       case OP_POSPLUS:       case OP_POSPLUS:
6982       if (!inassert) return -1;       if (!inassert) return -1;
6983       if (c < 0)       if (c < 0) c = scode[1];
6984         {         else if (c != scode[1]) return -1;
6985         c = scode[1];       break;
6986         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
6987         }       case OP_EXACTI:
6988       else if (c != scode[1]) return -1;       scode += 2;
6989         /* Fall through */
6990    
6991         case OP_CHARI:
6992         case OP_PLUSI:
6993         case OP_MINPLUSI:
6994         case OP_POSPLUSI:
6995         if (!inassert) return -1;
6996         if (c < 0) c = scode[1] | REQ_CASELESS;
6997           else if (c != scode[1]) return -1;
6998       break;       break;
6999       }       }
7000    
# Line 6508  int length = 1;  /* For final END opcode Line 7045  int length = 1;  /* For final END opcode
7045  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
7046  int errorcode = 0;  int errorcode = 0;
7047  int skipatstart = 0;  int skipatstart = 0;
7048  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8;
7049  size_t size;  size_t size;
7050  uschar *code;  uschar *code;
7051  const uschar *codestart;  const uschar *codestart;
# Line 6578  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7115  while (ptr[skipatstart] == CHAR_LEFT_PAR
7115    
7116    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
7117      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
7118      else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
7119        { skipatstart += 6; options |= PCRE_UCP; continue; }
7120      else if (strncmp((char *)(ptr+skipatstart+2), STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
7121        { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
7122    
7123    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
7124      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6602  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7143  while (ptr[skipatstart] == CHAR_LEFT_PAR
7143    else break;    else break;
7144    }    }
7145    
7146  /* Can't support UTF8 unless PCRE has been compiled to include the code. */  utf8 = (options & PCRE_UTF8) != 0;
7147    
7148    /* Can't support UTF8 unless PCRE has been compiled to include the code. The
7149    return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7150    release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7151    not used here. */
7152    
7153  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7154  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7155       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7156    {    {
7157    errorcode = ERR44;    errorcode = ERR44;
7158    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6619  if (utf8) Line 7165  if (utf8)
7165    }    }
7166  #endif  #endif
7167    
7168    /* Can't support UCP unless PCRE has been compiled to include the code. */
7169    
7170    #ifndef SUPPORT_UCP
7171    if ((options & PCRE_UCP) != 0)
7172      {
7173      errorcode = ERR67;
7174      goto PCRE_EARLY_ERROR_RETURN;
7175      }
7176    #endif
7177    
7178  /* Check validity of \R options. */  /* Check validity of \R options. */
7179    
7180  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
7181         (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
7182    {    {
7183    case 0:    errorcode = ERR56;
7184    case PCRE_BSR_ANYCRLF:    goto PCRE_EARLY_ERROR_RETURN;
   case PCRE_BSR_UNICODE:  
   break;  
   default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;  
7185    }    }
7186    
7187  /* 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 6712  outside can help speed up starting point Line 7266  outside can help speed up starting point
7266  ptr += skipatstart;  ptr += skipatstart;
7267  code = cworkspace;  code = cworkspace;
7268  *code = OP_BRA;  *code = OP_BRA;
7269  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7270    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7271  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7272    
7273  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 6747  regex compiled on a system with 4-byte p Line 7300  regex compiled on a system with 4-byte p
7300  pointers. */  pointers. */
7301    
7302  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
7303  re->size = size;  re->size = (int)size;
7304  re->options = cd->external_options;  re->options = cd->external_options;
7305  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7306  re->dummy1 = 0;  re->dummy1 = 0;
# Line 6768  field; this time it's used for rememberi Line 7321  field; this time it's used for rememberi
7321  */  */
7322    
7323  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7324    cd->assert_depth = 0;
7325  cd->bracount = 0;  cd->bracount = 0;
7326  cd->names_found = 0;  cd->names_found = 0;
7327  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 6786  of the function here. */ Line 7340  of the function here. */
7340  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7341  code = (uschar *)codestart;  code = (uschar *)codestart;
7342  *code = OP_BRA;  *code = OP_BRA;
7343  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0, 0,
7344    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7345  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7346  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7347  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7348    
7349  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */  if (cd->had_accept) reqbyte = REQ_NONE;   /* Must disable after (*ACCEPT) */
7350    
7351  /* 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. */
7352    
# Line 6818  while (errorcode == 0 && cd->hwm > cwork Line 7372  while (errorcode == 0 && cd->hwm > cwork
7372    recno = GET(codestart, offset);    recno = GET(codestart, offset);
7373    groupptr = _pcre_find_bracket(codestart, utf8, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
7374    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
7375      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, (int)(groupptr - codestart));
7376    }    }
7377    
7378  /* 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 6853  if (cd->check_lookbehind) Line 7407  if (cd->check_lookbehind)
7407        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7408        int end_op = *be;        int end_op = *be;
7409        *be = OP_END;        *be = OP_END;
7410        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7411            cd);
7412        *be = end_op;        *be = end_op;
7413        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7414        if (fixed_length < 0)        if (fixed_length < 0)
# Line 6873  if (errorcode != 0) Line 7428  if (errorcode != 0)
7428    {    {
7429    (pcre_free)(re);    (pcre_free)(re);
7430    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
7431    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = (int)(ptr - (const uschar *)pattern);
7432    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
7433    *errorptr = find_error_text(errorcode);    *errorptr = find_error_text(errorcode);
7434    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
# Line 6892  start with ^. and also when all branches Line 7447  start with ^. and also when all branches
7447    
7448  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7449    {    {
7450    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))  
7451      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7452    else    else
7453      {      {
7454      if (firstbyte < 0)      if (firstbyte < 0)
7455        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7456      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7457        {        {
7458        int ch = firstbyte & 255;        int ch = firstbyte & 255;

Legend:
Removed from v.512  
changed lines
  Added in v.745

  ViewVC Help
Powered by ViewVC 1.1.5