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

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

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

revision 574 by ph10, Sat Nov 20 17:47:27 2010 UTC revision 751 by ph10, Fri Nov 18 11:13:30 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2010 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 393  static const char error_texts[] = Line 393  static const char error_texts[] =
393    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
# Line 408  static const char error_texts[] = Line 408  static const char error_texts[] =
408    "different names for subpatterns of the same number are not allowed\0"    "different names for subpatterns of the same number are not allowed\0"
409    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
410    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with PCRE_UCP support\0"
411    "\\c must be followed by an ASCII character\0"    "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413      /* 70 */
414      "internal error: unknown opcode in find_fixedlength()\0"
415    ;    ;
416    
417  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 545  static const unsigned char ebcdic_charta Line 548  static const unsigned char ebcdic_charta
548  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
549    
550  static BOOL  static BOOL
551    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
552      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
553    
554    
# Line 577  return s; Line 580  return s;
580    
581    
582  /*************************************************  /*************************************************
583    *            Check for counted repeat            *
584    *************************************************/
585    
586    /* This function is called when a '{' is encountered in a place where it might
587    start a quantifier. It looks ahead to see if it really is a quantifier or not.
588    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
589    where the ddds are digits.
590    
591    Arguments:
592      p         pointer to the first char after '{'
593    
594    Returns:    TRUE or FALSE
595    */
596    
597    static BOOL
598    is_counted_repeat(const uschar *p)
599    {
600    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
601    while ((digitab[*p] & ctype_digit) != 0) p++;
602    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
603    
604    if (*p++ != CHAR_COMMA) return FALSE;
605    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
606    
607    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
608    while ((digitab[*p] & ctype_digit) != 0) p++;
609    
610    return (*p == CHAR_RIGHT_CURLY_BRACKET);
611    }
612    
613    
614    
615    /*************************************************
616  *            Handle escapes                      *  *            Handle escapes                      *
617  *************************************************/  *************************************************/
618    
# Line 642  else Line 678  else
678    
679      case CHAR_l:      case CHAR_l:
680      case CHAR_L:      case CHAR_L:
681        *errorcodeptr = ERR37;
682        break;
683    
684      case CHAR_u:      case CHAR_u:
685        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
686          {
687          /* In JavaScript, \u must be followed by four hexadecimal numbers.
688          Otherwise it is a lowercase u letter. */
689          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
690               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
691            {
692            c = 0;
693            for (i = 0; i < 4; ++i)
694              {
695              register int cc = *(++ptr);
696    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
697              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
698              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
699    #else           /* EBCDIC coding */
700              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
701              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
702    #endif
703              }
704            }
705          }
706        else
707          *errorcodeptr = ERR37;
708        break;
709    
710      case CHAR_U:      case CHAR_U:
711      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
712        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
713      break;      break;
714    
715      /* \g must be followed by one of a number of specific things:      /* In a character class, \g is just a literal "g". Outside a character
716        class, \g must be followed by one of a number of specific things:
717    
718      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
719      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 664  else Line 730  else
730      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
731    
732      case CHAR_g:      case CHAR_g:
733        if (isclass) break;
734      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
735        {        {
736        c = -ESC_g;        c = -ESC_g;
# Line 792  else Line 859  else
859      treated as a data character. */      treated as a data character. */
860    
861      case CHAR_x:      case CHAR_x:
862        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
863          {
864          /* In JavaScript, \x must be followed by two hexadecimal numbers.
865          Otherwise it is a lowercase x letter. */
866          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
867            {
868            c = 0;
869            for (i = 0; i < 2; ++i)
870              {
871              register int cc = *(++ptr);
872    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
873              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
874              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
875    #else           /* EBCDIC coding */
876              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
877              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
878    #endif
879              }
880            }
881          break;
882          }
883    
884      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
885        {        {
886        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
# Line 857  else Line 946  else
946      if (c > 127)  /* Excludes all non-ASCII in either mode */      if (c > 127)  /* Excludes all non-ASCII in either mode */
947        {        {
948        *errorcodeptr = ERR68;        *errorcodeptr = ERR68;
949        break;        break;
950        }        }
951      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
952      c ^= 0x40;      c ^= 0x40;
953  #else             /* EBCDIC coding */  #else             /* EBCDIC coding */
# Line 885  else Line 974  else
974    }    }
975    
976  /* Perl supports \N{name} for character names, as well as plain \N for "not  /* Perl supports \N{name} for character names, as well as plain \N for "not
977  newline". PCRE does not support \N{name}. */  newline". PCRE does not support \N{name}. However, it does support
978    quantification such as \N{2,3}. */
979    
980  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
981         !is_counted_repeat(ptr+2))
982    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
983    
984  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
# Line 997  return -1; Line 1088  return -1;
1088    
1089    
1090  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if (*p++ != CHAR_COMMA) return FALSE;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == CHAR_RIGHT_CURLY_BRACKET);  
 }  
   
   
   
 /*************************************************  
1091  *         Read repeat counts                     *  *         Read repeat counts                     *
1092  *************************************************/  *************************************************/
1093    
# Line 1105  top-level call starts at the beginning o Line 1163  top-level call starts at the beginning o
1163  start at a parenthesis. It scans along a pattern's text looking for capturing  start at a parenthesis. It scans along a pattern's text looking for capturing
1164  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1165  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1166  returns when it reaches a given numbered subpattern. We know that if (?P< is  returns when it reaches a given numbered subpattern. Recursion is used to keep
1167  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1168  first pass. Recursion is used to keep track of subpatterns that reset the  
1169  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1170    that if (?< or (?' or (?P< is encountered, the name will be correctly
1171    terminated because that is checked in the first pass. There is now one call to
1172    this function in the first pass, to check for a recursive back reference by
1173    name (so that we can make the whole group atomic). In this case, we need check
1174    only up to the current position in the pattern, and that is still OK because
1175    and previous occurrences will have been checked. To make this work, the test
1176    for "end of pattern" is a check against cd->end_pattern in the main loop,
1177    instead of looking for a binary zero. This means that the special first-pass
1178    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1179    processing items within the loop are OK, because afterwards the main loop will
1180    terminate.)
1181    
1182  Arguments:  Arguments:
1183    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1116  Arguments: Line 1185  Arguments:
1185    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1186    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1187    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1188    utf8         TRUE if we are in UTF-8 mode    utf8         TRUE if we are in UTF-8 mode
1189    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1190    
1191  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
# Line 1209  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1278  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1278    }    }
1279    
1280  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1281  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1282    first-pass call when this value is temporarily adjusted to stop at the current
1283    position. So DO NOT change this to a test for binary zero. */
1284    
1285  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1286    {    {
1287    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1288    
# Line 1285  for (; *ptr != 0; ptr++) Line 1356  for (; *ptr != 0; ptr++)
1356    
1357    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1358      {      {
1359      ptr++;      ptr++;
1360      while (*ptr != 0)      while (*ptr != 0)
1361        {        {
1362        if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }        if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1363        ptr++;        ptr++;
1364  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1365        if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;        if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1366  #endif  #endif
1367        }        }
# Line 1348  Arguments: Line 1419  Arguments:
1419    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1420    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1421    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1422    utf8         TRUE if we are in UTF-8 mode    utf8         TRUE if we are in UTF-8 mode
1423    
1424  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1425  */  */
# Line 1384  return rc; Line 1455  return rc;
1455    
1456  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1457  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1458  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1459  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1460  assertions, and also the \b assertion; for others it does not.  does not.
1461    
1462  Arguments:  Arguments:
1463    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1464    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1465    
1466  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1467  */  */
1468    
1469  static const uschar*  static const uschar*
1470  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1471  {  {
1472  for (;;)  for (;;)
1473    {    {
1474    switch ((int)*code)    switch ((int)*code)
1475      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1476      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1477      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1478      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1461  and doing the check at the end; a flag s Line 1522  and doing the check at the end; a flag s
1522    
1523  Arguments:  Arguments:
1524    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1525    options  the compiling options    utf8     TRUE in UTF-8 mode
1526    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1527    cd       the "compile data" structure    cd       the "compile data" structure
1528    
# Line 1469  Returns:   the fixed length, Line 1530  Returns:   the fixed length,
1530               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1531               or -2 if \C was encountered               or -2 if \C was encountered
1532               or -3 if an OP_RECURSE item was encountered and atend is FALSE               or -3 if an OP_RECURSE item was encountered and atend is FALSE
1533                 or -4 if an unknown opcode was encountered (internal error)
1534  */  */
1535    
1536  static int  static int
1537  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1538  {  {
1539  int length = -1;  int length = -1;
1540    
# Line 1489  for (;;) Line 1551  for (;;)
1551    register int op = *cc;    register int op = *cc;
1552    switch (op)    switch (op)
1553      {      {
1554        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1555        OP_BRA (normal non-capturing bracket) because the other variants of these
1556        opcodes are all concerned with unlimited repeated groups, which of course
1557        are not of fixed length. */
1558    
1559      case OP_CBRA:      case OP_CBRA:
1560      case OP_BRA:      case OP_BRA:
1561      case OP_ONCE:      case OP_ONCE:
1562        case OP_ONCE_NC:
1563      case OP_COND:      case OP_COND:
1564      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1565      if (d < 0) return d;      if (d < 0) return d;
1566      branchlength += d;      branchlength += d;
1567      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1568      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1569      break;      break;
1570    
1571      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested call.
1572      call. If it's ALT it is an alternation in a nested call. If it is      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1573      END it's the end of the outer call. All can be handled by the same code. */      an ALT. If it is END it's the end of the outer call. All can be handled by
1574        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1575        because they all imply an unlimited repeat. */
1576    
1577      case OP_ALT:      case OP_ALT:
1578      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1579      case OP_END:      case OP_END:
1580        case OP_ACCEPT:
1581        case OP_ASSERT_ACCEPT:
1582      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1583        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1584      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1525  for (;;) Line 1595  for (;;)
1595      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1596      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1597      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1598      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1599      if (d < 0) return d;      if (d < 0) return d;
1600      branchlength += d;      branchlength += d;
1601      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1542  for (;;) Line 1612  for (;;)
1612    
1613      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1614    
1615      case OP_REVERSE:      case OP_MARK:
1616        case OP_PRUNE_ARG:
1617        case OP_SKIP_ARG:
1618        case OP_THEN_ARG:
1619        cc += cc[1] + _pcre_OP_lengths[*cc];
1620        break;
1621    
1622        case OP_CALLOUT:
1623        case OP_CIRC:
1624        case OP_CIRCM:
1625        case OP_CLOSE:
1626        case OP_COMMIT:
1627      case OP_CREF:      case OP_CREF:
     case OP_NCREF:  
     case OP_RREF:  
     case OP_NRREF:  
1628      case OP_DEF:      case OP_DEF:
1629      case OP_OPT:      case OP_DOLL:
1630      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
     case OP_SET_SOM:  
1631      case OP_EOD:      case OP_EOD:
1632      case OP_EODN:      case OP_EODN:
1633      case OP_CIRC:      case OP_FAIL:
1634      case OP_DOLL:      case OP_NCREF:
1635        case OP_NRREF:
1636      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1637        case OP_PRUNE:
1638        case OP_REVERSE:
1639        case OP_RREF:
1640        case OP_SET_SOM:
1641        case OP_SKIP:
1642        case OP_SOD:
1643        case OP_SOM:
1644        case OP_THEN:
1645      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1646      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
1647      break;      break;
# Line 1565  for (;;) Line 1649  for (;;)
1649      /* Handle literal characters */      /* Handle literal characters */
1650    
1651      case OP_CHAR:      case OP_CHAR:
1652      case OP_CHARNC:      case OP_CHARI:
1653      case OP_NOT:      case OP_NOT:
1654        case OP_NOTI:
1655      branchlength++;      branchlength++;
1656      cc += 2;      cc += 2;
1657  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1658      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1659  #endif  #endif
1660      break;      break;
1661    
# Line 1579  for (;;) Line 1663  for (;;)
1663      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1664    
1665      case OP_EXACT:      case OP_EXACT:
1666        case OP_EXACTI:
1667        case OP_NOTEXACT:
1668        case OP_NOTEXACTI:
1669      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1670      cc += 4;      cc += 4;
1671  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1672      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1673  #endif  #endif
1674      break;      break;
1675    
# Line 1600  for (;;) Line 1686  for (;;)
1686      cc += 2;      cc += 2;
1687      /* Fall through */      /* Fall through */
1688    
1689        case OP_HSPACE:
1690        case OP_VSPACE:
1691        case OP_NOT_HSPACE:
1692        case OP_NOT_VSPACE:
1693      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1694      case OP_DIGIT:      case OP_DIGIT:
1695      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1631  for (;;) Line 1721  for (;;)
1721    
1722      switch (*cc)      switch (*cc)
1723        {        {
1724          case OP_CRPLUS:
1725          case OP_CRMINPLUS:
1726        case OP_CRSTAR:        case OP_CRSTAR:
1727        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1728        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1651  for (;;) Line 1743  for (;;)
1743    
1744      /* Anything else is variable length */      /* Anything else is variable length */
1745    
1746      default:      case OP_ANYNL:
1747        case OP_BRAMINZERO:
1748        case OP_BRAPOS:
1749        case OP_BRAPOSZERO:
1750        case OP_BRAZERO:
1751        case OP_CBRAPOS:
1752        case OP_EXTUNI:
1753        case OP_KETRMAX:
1754        case OP_KETRMIN:
1755        case OP_KETRPOS:
1756        case OP_MINPLUS:
1757        case OP_MINPLUSI:
1758        case OP_MINQUERY:
1759        case OP_MINQUERYI:
1760        case OP_MINSTAR:
1761        case OP_MINSTARI:
1762        case OP_MINUPTO:
1763        case OP_MINUPTOI:
1764        case OP_NOTMINPLUS:
1765        case OP_NOTMINPLUSI:
1766        case OP_NOTMINQUERY:
1767        case OP_NOTMINQUERYI:
1768        case OP_NOTMINSTAR:
1769        case OP_NOTMINSTARI:
1770        case OP_NOTMINUPTO:
1771        case OP_NOTMINUPTOI:
1772        case OP_NOTPLUS:
1773        case OP_NOTPLUSI:
1774        case OP_NOTPOSPLUS:
1775        case OP_NOTPOSPLUSI:
1776        case OP_NOTPOSQUERY:
1777        case OP_NOTPOSQUERYI:
1778        case OP_NOTPOSSTAR:
1779        case OP_NOTPOSSTARI:
1780        case OP_NOTPOSUPTO:
1781        case OP_NOTPOSUPTOI:
1782        case OP_NOTQUERY:
1783        case OP_NOTQUERYI:
1784        case OP_NOTSTAR:
1785        case OP_NOTSTARI:
1786        case OP_NOTUPTO:
1787        case OP_NOTUPTOI:
1788        case OP_PLUS:
1789        case OP_PLUSI:
1790        case OP_POSPLUS:
1791        case OP_POSPLUSI:
1792        case OP_POSQUERY:
1793        case OP_POSQUERYI:
1794        case OP_POSSTAR:
1795        case OP_POSSTARI:
1796        case OP_POSUPTO:
1797        case OP_POSUPTOI:
1798        case OP_QUERY:
1799        case OP_QUERYI:
1800        case OP_REF:
1801        case OP_REFI:
1802        case OP_SBRA:
1803        case OP_SBRAPOS:
1804        case OP_SCBRA:
1805        case OP_SCBRAPOS:
1806        case OP_SCOND:
1807        case OP_SKIPZERO:
1808        case OP_STAR:
1809        case OP_STARI:
1810        case OP_TYPEMINPLUS:
1811        case OP_TYPEMINQUERY:
1812        case OP_TYPEMINSTAR:
1813        case OP_TYPEMINUPTO:
1814        case OP_TYPEPLUS:
1815        case OP_TYPEPOSPLUS:
1816        case OP_TYPEPOSQUERY:
1817        case OP_TYPEPOSSTAR:
1818        case OP_TYPEPOSUPTO:
1819        case OP_TYPEQUERY:
1820        case OP_TYPESTAR:
1821        case OP_TYPEUPTO:
1822        case OP_UPTO:
1823        case OP_UPTOI:
1824      return -1;      return -1;
1825    
1826        /* Catch unrecognized opcodes so that when new ones are added they
1827        are not forgotten, as has happened in the past. */
1828    
1829        default:
1830        return -4;
1831      }      }
1832    }    }
1833  /* Control never gets here */  /* Control never gets here */
# Line 1685  _pcre_find_bracket(const uschar *code, B Line 1860  _pcre_find_bracket(const uschar *code, B
1860  for (;;)  for (;;)
1861    {    {
1862    register int c = *code;    register int c = *code;
1863    
1864    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1865    
1866    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1703  for (;;) Line 1879  for (;;)
1879    
1880    /* Handle capturing bracket */    /* Handle capturing bracket */
1881    
1882    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1883               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1884      {      {
1885      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1886      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1745  for (;;) Line 1922  for (;;)
1922        break;        break;
1923    
1924        case OP_THEN_ARG:        case OP_THEN_ARG:
1925        code += code[1+LINK_SIZE];        code += code[1];
1926        break;        break;
1927        }        }
1928    
# Line 1761  for (;;) Line 1938  for (;;)
1938      if (utf8) switch(c)      if (utf8) switch(c)
1939        {        {
1940        case OP_CHAR:        case OP_CHAR:
1941        case OP_CHARNC:        case OP_CHARI:
1942        case OP_EXACT:        case OP_EXACT:
1943          case OP_EXACTI:
1944        case OP_UPTO:        case OP_UPTO:
1945          case OP_UPTOI:
1946        case OP_MINUPTO:        case OP_MINUPTO:
1947          case OP_MINUPTOI:
1948        case OP_POSUPTO:        case OP_POSUPTO:
1949          case OP_POSUPTOI:
1950        case OP_STAR:        case OP_STAR:
1951          case OP_STARI:
1952        case OP_MINSTAR:        case OP_MINSTAR:
1953          case OP_MINSTARI:
1954        case OP_POSSTAR:        case OP_POSSTAR:
1955          case OP_POSSTARI:
1956        case OP_PLUS:        case OP_PLUS:
1957          case OP_PLUSI:
1958        case OP_MINPLUS:        case OP_MINPLUS:
1959          case OP_MINPLUSI:
1960        case OP_POSPLUS:        case OP_POSPLUS:
1961          case OP_POSPLUSI:
1962        case OP_QUERY:        case OP_QUERY:
1963          case OP_QUERYI:
1964        case OP_MINQUERY:        case OP_MINQUERY:
1965          case OP_MINQUERYI:
1966        case OP_POSQUERY:        case OP_POSQUERY:
1967          case OP_POSQUERYI:
1968        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1969        break;        break;
1970        }        }
# Line 1851  for (;;) Line 2041  for (;;)
2041        break;        break;
2042    
2043        case OP_THEN_ARG:        case OP_THEN_ARG:
2044        code += code[1+LINK_SIZE];        code += code[1];
2045        break;        break;
2046        }        }
2047    
# Line 1867  for (;;) Line 2057  for (;;)
2057      if (utf8) switch(c)      if (utf8) switch(c)
2058        {        {
2059        case OP_CHAR:        case OP_CHAR:
2060        case OP_CHARNC:        case OP_CHARI:
2061        case OP_EXACT:        case OP_EXACT:
2062          case OP_EXACTI:
2063        case OP_UPTO:        case OP_UPTO:
2064          case OP_UPTOI:
2065        case OP_MINUPTO:        case OP_MINUPTO:
2066          case OP_MINUPTOI:
2067        case OP_POSUPTO:        case OP_POSUPTO:
2068          case OP_POSUPTOI:
2069        case OP_STAR:        case OP_STAR:
2070          case OP_STARI:
2071        case OP_MINSTAR:        case OP_MINSTAR:
2072          case OP_MINSTARI:
2073        case OP_POSSTAR:        case OP_POSSTAR:
2074          case OP_POSSTARI:
2075        case OP_PLUS:        case OP_PLUS:
2076          case OP_PLUSI:
2077        case OP_MINPLUS:        case OP_MINPLUS:
2078          case OP_MINPLUSI:
2079        case OP_POSPLUS:        case OP_POSPLUS:
2080          case OP_POSPLUSI:
2081        case OP_QUERY:        case OP_QUERY:
2082          case OP_QUERYI:
2083        case OP_MINQUERY:        case OP_MINQUERY:
2084          case OP_MINQUERYI:
2085        case OP_POSQUERY:        case OP_POSQUERY:
2086          case OP_POSQUERYI:
2087        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
2088        break;        break;
2089        }        }
# Line 1919  could_be_empty_branch(const uschar *code Line 2122  could_be_empty_branch(const uschar *code
2122    compile_data *cd)    compile_data *cd)
2123  {  {
2124  register int c;  register int c;
2125  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2126       code < endcode;       code < endcode;
2127       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2128    {    {
2129    const uschar *ccode;    const uschar *ccode;
2130    
# Line 1937  for (code = first_significant_code(code Line 2140  for (code = first_significant_code(code
2140      continue;      continue;
2141      }      }
2142    
   /* Groups with zero repeats can of course be empty; skip them. */  
   
   if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)  
     {  
     code += _pcre_OP_lengths[c];  
     do code += GET(code, 1); while (*code == OP_ALT);  
     c = *code;  
     continue;  
     }  
   
2143    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
2144    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
2145      forward reference subroutine call, we can't. To detect forward reference
2146      we have to scan up the list that is kept in the workspace. This function is
2147      called only when doing the real compile, not during the pre-compile that
2148      measures the size of the compiled pattern. */
2149    
2150    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2151      {      {
2152      BOOL empty_branch = FALSE;      const uschar *scode;
2153      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
2154    
2155        /* Test for forward reference */
2156    
2157        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2158          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2159    
2160        /* Not a forward reference, test for completed backward reference */
2161    
2162        empty_branch = FALSE;
2163        scode = cd->start_code + GET(code, 1);
2164      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2165    
2166        /* Completed backwards reference */
2167    
2168      do      do
2169        {        {
2170        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf8, cd))
# Line 1965  for (code = first_significant_code(code Line 2175  for (code = first_significant_code(code
2175        scode += GET(scode, 1);        scode += GET(scode, 1);
2176        }        }
2177      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2178    
2179      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2180      continue;      continue;
2181      }      }
2182    
2183      /* Groups with zero repeats can of course be empty; skip them. */
2184    
2185      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2186          c == OP_BRAPOSZERO)
2187        {
2188        code += _pcre_OP_lengths[c];
2189        do code += GET(code, 1); while (*code == OP_ALT);
2190        c = *code;
2191        continue;
2192        }
2193    
2194      /* A nested group that is already marked as "could be empty" can just be
2195      skipped. */
2196    
2197      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2198          c == OP_SCBRA || c == OP_SCBRAPOS)
2199        {
2200        do code += GET(code, 1); while (*code == OP_ALT);
2201        c = *code;
2202        continue;
2203        }
2204    
2205    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2206    
2207    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2208          c == OP_CBRA || c == OP_CBRAPOS ||
2209          c == OP_ONCE || c == OP_ONCE_NC ||
2210          c == OP_COND)
2211      {      {
2212      BOOL empty_branch;      BOOL empty_branch;
2213      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2058  for (code = first_significant_code(code Line 2294  for (code = first_significant_code(code
2294      case OP_ALLANY:      case OP_ALLANY:
2295      case OP_ANYBYTE:      case OP_ANYBYTE:
2296      case OP_CHAR:      case OP_CHAR:
2297      case OP_CHARNC:      case OP_CHARI:
2298      case OP_NOT:      case OP_NOT:
2299        case OP_NOTI:
2300      case OP_PLUS:      case OP_PLUS:
2301      case OP_MINPLUS:      case OP_MINPLUS:
2302      case OP_POSPLUS:      case OP_POSPLUS:
# Line 2099  for (code = first_significant_code(code Line 2336  for (code = first_significant_code(code
2336      case OP_KET:      case OP_KET:
2337      case OP_KETRMAX:      case OP_KETRMAX:
2338      case OP_KETRMIN:      case OP_KETRMIN:
2339        case OP_KETRPOS:
2340      case OP_ALT:      case OP_ALT:
2341      return TRUE;      return TRUE;
2342    
# Line 2107  for (code = first_significant_code(code Line 2345  for (code = first_significant_code(code
2345    
2346  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2347      case OP_STAR:      case OP_STAR:
2348        case OP_STARI:
2349      case OP_MINSTAR:      case OP_MINSTAR:
2350        case OP_MINSTARI:
2351      case OP_POSSTAR:      case OP_POSSTAR:
2352        case OP_POSSTARI:
2353      case OP_QUERY:      case OP_QUERY:
2354        case OP_QUERYI:
2355      case OP_MINQUERY:      case OP_MINQUERY:
2356        case OP_MINQUERYI:
2357      case OP_POSQUERY:      case OP_POSQUERY:
2358        case OP_POSQUERYI:
2359      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2360      break;      break;
2361    
2362      case OP_UPTO:      case OP_UPTO:
2363        case OP_UPTOI:
2364      case OP_MINUPTO:      case OP_MINUPTO:
2365        case OP_MINUPTOI:
2366      case OP_POSUPTO:      case OP_POSUPTO:
2367        case OP_POSUPTOI:
2368      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2369      break;      break;
2370  #endif  #endif
# Line 2132  for (code = first_significant_code(code Line 2379  for (code = first_significant_code(code
2379      break;      break;
2380    
2381      case OP_THEN_ARG:      case OP_THEN_ARG:
2382      code += code[1+LINK_SIZE];      code += code[1];
2383      break;      break;
2384    
2385      /* None of the remaining opcodes are required to match a character. */      /* None of the remaining opcodes are required to match a character. */
# Line 2155  return TRUE; Line 2402  return TRUE;
2402  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2403  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2404  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2405    This function is called only during the real compile, not during the
2406    pre-compile.
2407    
2408  Arguments:  Arguments:
2409    code        points to start of the recursion    code        points to start of the recursion
# Line 2205  where Perl recognizes it as the POSIX cl Line 2454  where Perl recognizes it as the POSIX cl
2454  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2455  I think.  I think.
2456    
2457    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2458    It seems that the appearance of a nested POSIX class supersedes an apparent
2459    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2460    a digit.
2461    
2462    In Perl, unescaped square brackets may also appear as part of class names. For
2463    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2464    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2465    seem right at all. PCRE does not allow closing square brackets in POSIX class
2466    names.
2467    
2468  Arguments:  Arguments:
2469    ptr      pointer to the initial [    ptr      pointer to the initial [
2470    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2219  int terminator;          /* Don't combin Line 2479  int terminator;          /* Don't combin
2479  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2480  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2481    {    {
2482    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2483        ptr++;
2484      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2485      else
2486      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2487      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2488        {        {
2489        *endptr = ptr;        *endptr = ptr;
2490        return TRUE;        return TRUE;
2491        }        }
2492        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2493             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2494              ptr[1] == CHAR_EQUALS_SIGN) &&
2495            check_posix_syntax(ptr, endptr))
2496          return FALSE;
2497      }      }
2498    }    }
2499  return FALSE;  return FALSE;
# Line 2532  if ((options & PCRE_EXTENDED) != 0) Line 2799  if ((options & PCRE_EXTENDED) != 0)
2799      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2800      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2801        {        {
2802        ptr++;        ptr++;
2803        while (*ptr != 0)        while (*ptr != 0)
2804          {          {
2805          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2806          ptr++;          ptr++;
2807  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2808          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2809  #endif  #endif
2810          }          }
# Line 2576  if ((options & PCRE_EXTENDED) != 0) Line 2843  if ((options & PCRE_EXTENDED) != 0)
2843      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2844      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2845        {        {
2846        ptr++;        ptr++;
2847        while (*ptr != 0)        while (*ptr != 0)
2848          {          {
2849          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2850          ptr++;          ptr++;
2851  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2852          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2853  #endif  #endif
2854          }          }
# Line 2609  if (next >= 0) switch(op_code) Line 2876  if (next >= 0) switch(op_code)
2876  #endif  #endif
2877    return c != next;    return c != next;
2878    
2879    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
2880    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
2881    high-valued characters. */    high-valued characters. */
2882    
2883    case OP_CHARNC:    case OP_CHARI:
2884  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2885    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
2886  #else  #else
# Line 2636  if (next >= 0) switch(op_code) Line 2903  if (next >= 0) switch(op_code)
2903  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2904    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2905    
2906    /* For OP_NOT, its data is always a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2907      opcodes are not used for multi-byte characters, because they are coded using
2908      an XCLASS instead. */
2909    
2910    case OP_NOT:    case OP_NOT:
2911      return (c = *previous) == next;
2912    
2913      case OP_NOTI:
2914    if ((c = *previous) == next) return TRUE;    if ((c = *previous) == next) return TRUE;
   if ((options & PCRE_CASELESS) == 0) return FALSE;  
2915  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2916    if (utf8)    if (utf8)
2917      {      {
# Line 2745  replaced by OP_PROP codes when PCRE_UCP Line 3016  replaced by OP_PROP codes when PCRE_UCP
3016  switch(op_code)  switch(op_code)
3017    {    {
3018    case OP_CHAR:    case OP_CHAR:
3019    case OP_CHARNC:    case OP_CHARI:
3020  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3021    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
3022  #else  #else
# Line 2932  Arguments: Line 3203  Arguments:
3203    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3204    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3205    bcptr          points to current branch chain    bcptr          points to current branch chain
3206      cond_depth     conditional nesting depth
3207    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3208    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3209                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2943  Returns:         TRUE on success Line 3215  Returns:         TRUE on success
3215  static BOOL  static BOOL
3216  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3217    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3218    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3219  {  {
3220  int repeat_type, op_type;  int repeat_type, op_type;
3221  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2952  int greedy_default, greedy_non_default; Line 3224  int greedy_default, greedy_non_default;
3224  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3225  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3226  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3227  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3228  int after_manual_callout = 0;  int after_manual_callout = 0;
3229  int length_prevgroup = 0;  int length_prevgroup = 0;
3230  register int c;  register int c;
# Line 2970  uschar *previous_callout = NULL; Line 3242  uschar *previous_callout = NULL;
3242  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3243  uschar classbits[32];  uschar classbits[32];
3244    
3245    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3246    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3247    dynamically as we process the pattern. */
3248    
3249  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3250  BOOL class_utf8;  BOOL class_utf8;
3251  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2978  uschar *class_utf8data_base; Line 3254  uschar *class_utf8data_base;
3254  uschar utf8_char[6];  uschar utf8_char[6];
3255  #else  #else
3256  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3257  #endif  #endif
3258    
3259  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 3029  for (;; ptr++) Line 3304  for (;; ptr++)
3304    int subfirstbyte;    int subfirstbyte;
3305    int terminator;    int terminator;
3306    int mclength;    int mclength;
3307      int tempbracount;
3308    uschar mcbuffer[8];    uschar mcbuffer[8];
3309    
3310    /* Get next byte in the pattern */    /* Get next byte in the pattern */
# Line 3076  for (;; ptr++) Line 3352  for (;; ptr++)
3352        }        }
3353    
3354      *lengthptr += (int)(code - last_code);      *lengthptr += (int)(code - last_code);
3355      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, (int)(code - last_code),
3356          c));
3357    
3358      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
3359      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
# Line 3150  for (;; ptr++) Line 3427  for (;; ptr++)
3427      previous_callout = NULL;      previous_callout = NULL;
3428      }      }
3429    
3430    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3431    
3432    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3433      {      {
3434      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3435      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3436        {        {
3437        ptr++;        ptr++;
3438        while (*ptr != 0)        while (*ptr != 0)
3439          {          {
3440          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3441          ptr++;          ptr++;
3442  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3443          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3444  #endif  #endif
3445          }          }
# Line 3209  for (;; ptr++) Line 3486  for (;; ptr++)
3486      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3487    
3488      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3489        previous = NULL;
3490      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3491        {        {
3492        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3493          *code++ = OP_CIRCM;
3494        }        }
3495      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3496      break;      break;
3497    
3498      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3499      previous = NULL;      previous = NULL;
3500      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3501      break;      break;
3502    
3503      /* There can never be a first char if '.' is first, whatever happens about      /* There can never be a first char if '.' is first, whatever happens about
# Line 3540  for (;; ptr++) Line 3818  for (;; ptr++)
3818              continue;              continue;
3819    
3820              /* Perl 5.004 onwards omits VT from \s, but we must preserve it              /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3821              if it was previously set by something earlier in the character              if it was previously set by something earlier in the character
3822              class. */              class. */
3823    
3824              case ESC_s:              case ESC_s:
3825              classbits[0] |= cbits[cbit_space];              classbits[0] |= cbits[cbit_space];
3826              classbits[1] |= cbits[cbit_space+1] & ~0x08;              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3827              for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3828              continue;              continue;
3829    
# Line 3965  for (;; ptr++) Line 4243  for (;; ptr++)
4243    
4244      In UTF-8 mode, we can optimize the negative case only if there were no      In UTF-8 mode, we can optimize the negative case only if there were no
4245      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4246      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4247      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4248    
4249      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
4250      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4251      that OP_NOT does not support multibyte characters. In the positive case, it      Note that OP_NOT[I] does not support multibyte characters. In the positive
4252      can cause firstbyte to be set. Otherwise, there can be no first char if      case, it can cause firstbyte to be set. Otherwise, there can be no first
4253      this item is first, whatever repeat count may follow. In the case of      char if this item is first, whatever repeat count may follow. In the case
4254      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4255    
4256  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4257      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3984  for (;; ptr++) Line 4262  for (;; ptr++)
4262        {        {
4263        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4264    
4265        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4266    
4267        if (negate_class)        if (negate_class)
4268          {          {
4269          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4270          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4271          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4272          *code++ = class_lastchar;          *code++ = class_lastchar;
4273          break;          break;
4274          }          }
# Line 4118  for (;; ptr++) Line 4396  for (;; ptr++)
4396      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4397      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4398    
4399      /* Save start of previous item, in case we have to move it up to make space      /* Save start of previous item, in case we have to move it up in order to
4400      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4401    
4402      tempcode = previous;      tempcode = previous;
4403    
# Line 4142  for (;; ptr++) Line 4420  for (;; ptr++)
4420        }        }
4421      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4422    
4423        /* If previous was a recursion call, wrap it in atomic brackets so that
4424        previous becomes the atomic group. All recursions were so wrapped in the
4425        past, but it no longer happens for non-repeated recursions. In fact, the
4426        repeated ones could be re-implemented independently so as not to need this,
4427        but for the moment we rely on the code for repeating groups. */
4428    
4429        if (*previous == OP_RECURSE)
4430          {
4431          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4432          *previous = OP_ONCE;
4433          PUT(previous, 1, 2 + 2*LINK_SIZE);
4434          previous[2 + 2*LINK_SIZE] = OP_KET;
4435          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4436          code += 2 + 2 * LINK_SIZE;
4437          length_prevgroup = 3 + 3*LINK_SIZE;
4438    
4439          /* When actually compiling, we need to check whether this was a forward
4440          reference, and if so, adjust the offset. */
4441    
4442          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4443            {
4444            int offset = GET(cd->hwm, -LINK_SIZE);
4445            if (offset == previous + 1 - cd->start_code)
4446              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4447            }
4448          }
4449    
4450        /* Now handle repetition for the different types of item. */
4451    
4452      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4453      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
4454      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
4455      the first thing in a branch because the x will have gone into firstbyte      the first thing in a branch because the x will have gone into firstbyte
4456      instead.  */      instead.  */
4457    
4458      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4459        {        {
4460          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4461    
4462        /* Deal with UTF-8 characters that take up more than one byte. It's        /* Deal with UTF-8 characters that take up more than one byte. It's
4463        easier to write this out separately than try to macrify it. Use c to        easier to write this out separately than try to macrify it. Use c to
4464        hold the length of the character in bytes, plus 0x80 to flag that it's a        hold the length of the character in bytes, plus 0x80 to flag that it's a
# Line 4194  for (;; ptr++) Line 4503  for (;; ptr++)
4503      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4504      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4505      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4506      repeat_type. We can also test for auto-possessification. OP_NOT is      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4507      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4508    
4509      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4510        {        {
4511        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4512        c = previous[1];        c = previous[1];
4513        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4514            repeat_max < 0 &&            repeat_max < 0 &&
# Line 4396  for (;; ptr++) Line 4705  for (;; ptr++)
4705  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4706               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4707  #endif  #endif
4708               *previous == OP_REF)               *previous == OP_REF ||
4709                 *previous == OP_REFI)
4710        {        {
4711        if (repeat_max == 0)        if (repeat_max == 0)
4712          {          {
# Line 4430  for (;; ptr++) Line 4740  for (;; ptr++)
4740        }        }
4741    
4742      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
4743      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4744        opcodes such as BRA and CBRA, as this is the place where they get converted
4745        into the more special varieties such as BRAPOS and SBRA. A test for >=
4746        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4747        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4748        repetition of assertions, but now it does, for Perl compatibility. */
4749    
4750      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4751        {        {
4752        register int i;        register int i;
       int ketoffset = 0;  
4753        int len = (int)(code - previous);        int len = (int)(code - previous);
4754        uschar *bralink = NULL;        uschar *bralink = NULL;
4755          uschar *brazeroptr = NULL;
4756    
4757        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4758          we just ignore the repeat. */
4759    
4760        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4761          {          goto END_REPEAT;
4762          *errorcodeptr = ERR55;  
4763          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4764          }        use of repetition is in cases when the assertion is optional. Therefore,
4765          if the minimum is greater than zero, just ignore the repeat. If the
4766          maximum is not not zero or one, set it to 1. */
4767    
4768        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4769        by scanning through from the start, and compute the offset back to it          {
4770        from the current code pointer. There may be an OP_OPT setting following          if (repeat_min > 0) goto END_REPEAT;
4771        the final KET, so we can't find the end just by going back from the code          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
       pointer. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = (int)(code - ket);  
4772          }          }
4773    
4774        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 4479  for (;; ptr++) Line 4789  for (;; ptr++)
4789          **   goto END_REPEAT;          **   goto END_REPEAT;
4790          **   }          **   }
4791    
4792          However, that fails when a group is referenced as a subroutine from          However, that fails when a group or a subgroup within it is referenced
4793          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it          as a subroutine from elsewhere in the pattern, so now we stick in
4794          so that it is skipped on execution. As we don't have a list of which          OP_SKIPZERO in front of it so that it is skipped on execution. As we
4795          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4796            selectively.
4797    
4798          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4799          and do no more at this point. However, we do need to adjust any          and do no more at this point. However, we do need to adjust any
# Line 4502  for (;; ptr++) Line 4813  for (;; ptr++)
4813              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4814              goto END_REPEAT;              goto END_REPEAT;
4815              }              }
4816              brazeroptr = previous;    /* Save for possessive optimizing */
4817            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4818            }            }
4819    
# Line 4666  for (;; ptr++) Line 4978  for (;; ptr++)
4978            }            }
4979          }          }
4980    
4981        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. For
4982        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4983        don't know if there's been an options resetting after the ket. The        ONCE brackets can be converted into non-capturing brackets, as the
4984        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4985          deal with possessive ONCEs specially.
4986        Then, when we are doing the actual compile phase, check to see whether  
4987        this group is a non-atomic one that could match an empty string. If so,        Otherwise, when we are doing the actual compile phase, check to see
4988          whether this group is one that could match an empty string. If so,
4989        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
4990        that runtime checking can be done. [This check is also applied to        that runtime checking can be done. [This check is also applied to ONCE
4991        atomic groups at runtime, but in a different way.] */        groups at runtime, but in a different way.]
4992    
4993          Then, if the quantifier was possessive and the bracket is not a
4994          conditional, we convert the BRA code to the POS form, and the KET code to
4995          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
4996          subpattern at both the start and at the end.) The use of special opcodes
4997          makes it possible to reduce greatly the stack usage in pcre_exec(). If
4998          the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
4999    
5000          Then, if the minimum number of matches is 1 or 0, cancel the possessive
5001          flag so that the default action below, of wrapping everything inside
5002          atomic brackets, does not happen. When the minimum is greater than 1,
5003          there will be earlier copies of the group, and so we still have to wrap
5004          the whole thing. */
5005    
5006        else        else
5007          {          {
5008          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
5009          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
5010          *ketcode = OP_KETRMAX + repeat_type;  
5011          if (lengthptr == NULL && *bracode != OP_ONCE)          /* Convert possessive ONCE brackets to non-capturing */
5012    
5013            if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5014                possessive_quantifier) *bracode = OP_BRA;
5015    
5016            /* For non-possessive ONCE brackets, all we need to do is to
5017            set the KET. */
5018    
5019            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5020              *ketcode = OP_KETRMAX + repeat_type;
5021    
5022            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5023            converted to non-capturing above). */
5024    
5025            else
5026            {            {
5027            uschar *scode = bracode;            /* In the compile phase, check for empty string matching. */
5028            do  
5029              if (lengthptr == NULL)
5030              {              {
5031              if (could_be_empty_branch(scode, ketcode, utf8, cd))              uschar *scode = bracode;
5032                do
5033                {                {
5034                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
5035                break;                  {
5036                    *bracode += OP_SBRA - OP_BRA;
5037                    break;
5038                    }
5039                  scode += GET(scode, 1);
5040                }                }
5041              scode += GET(scode, 1);              while (*scode == OP_ALT);
5042              }              }
5043            while (*scode == OP_ALT);  
5044              /* Handle possessive quantifiers. */
5045    
5046              if (possessive_quantifier)
5047                {
5048                /* For COND brackets, we wrap the whole thing in a possessively
5049                repeated non-capturing bracket, because we have not invented POS
5050                versions of the COND opcodes. Because we are moving code along, we
5051                must ensure that any pending recursive references are updated. */
5052    
5053                if (*bracode == OP_COND || *bracode == OP_SCOND)
5054                  {
5055                  int nlen = (int)(code - bracode);
5056                  *code = OP_END;
5057                  adjust_recurse(bracode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5058                  memmove(bracode + 1+LINK_SIZE, bracode, nlen);
5059                  code += 1 + LINK_SIZE;
5060                  nlen += 1 + LINK_SIZE;
5061                  *bracode = OP_BRAPOS;
5062                  *code++ = OP_KETRPOS;
5063                  PUTINC(code, 0, nlen);
5064                  PUT(bracode, 1, nlen);
5065                  }
5066    
5067                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5068    
5069                else
5070                  {
5071                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5072                  *ketcode = OP_KETRPOS;
5073                  }
5074    
5075                /* If the minimum is zero, mark it as possessive, then unset the
5076                possessive flag when the minimum is 0 or 1. */
5077    
5078                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5079                if (repeat_min < 2) possessive_quantifier = FALSE;
5080                }
5081    
5082              /* Non-possessive quantifier */
5083    
5084              else *ketcode = OP_KETRMAX + repeat_type;
5085            }            }
5086          }          }
5087        }        }
# Line 4715  for (;; ptr++) Line 5102  for (;; ptr++)
5102        }        }
5103    
5104      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
5105      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
5106      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
5107      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
5108      The '+' notation is just syntactic sugar, taken from Sun's Java package,      notation is just syntactic sugar, taken from Sun's Java package, but the
5109      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
5110      tempcode, not at previous, which might be the first part of a string whose  
5111      (former) last char we repeated.      Some (but not all) possessively repeated subpatterns have already been
5112        completely handled in the code just above. For them, possessive_quantifier
5113        is always FALSE at this stage.
5114    
5115        Note that the repeated item starts at tempcode, not at previous, which
5116        might be the first part of a string whose (former) last char we repeated.
5117    
5118      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
5119      an 'upto' may follow. We skip over an 'exact' item, and then test the      an 'upto' may follow. We skip over an 'exact' item, and then test the
# Line 4752  for (;; ptr++) Line 5144  for (;; ptr++)
5144          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
5145          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
5146    
5147          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
5148          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
5149          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
5150          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
5151    
5152          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
5153          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
5154          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
5155          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
5156    
5157            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
5158            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
5159            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
5160            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
5161    
5162            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
5163            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
5164            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
5165            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
5166    
5167          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
5168          pending recursive references are updated. */          pending recursive references are updated. */
5169    
# Line 4815  for (;; ptr++) Line 5217  for (;; ptr++)
5217        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5218        namelen = (int)(ptr - name);        namelen = (int)(ptr - name);
5219    
5220          /* It appears that Perl allows any characters whatsoever, other than
5221          a closing parenthesis, to appear in arguments, so we no longer insist on
5222          letters, digits, and underscores. */
5223    
5224        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5225          {          {
5226          arg = ++ptr;          arg = ++ptr;
5227          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
           || *ptr == '_') ptr++;  
5228          arglen = (int)(ptr - arg);          arglen = (int)(ptr - arg);
5229          }          }
5230    
# Line 4836  for (;; ptr++) Line 5241  for (;; ptr++)
5241          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5242              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5243            {            {
5244            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5245              ASSERT_ACCEPT if in an assertion. */
5246    
5247            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5248              {              {
5249              open_capitem *oc;              open_capitem *oc;
5250                if (arglen != 0)
5251                  {
5252                  *errorcodeptr = ERR59;
5253                  goto FAILED;
5254                  }
5255              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5256              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5257                {                {
5258                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5259                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5260                }                }
5261                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5262    
5263                /* Do not set firstbyte after *ACCEPT */
5264                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5265              }              }
5266    
5267            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
5268    
5269            if (arglen == 0)            else if (arglen == 0)
5270              {              {
5271              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
5272                {                {
# Line 4859  for (;; ptr++) Line 5274  for (;; ptr++)
5274                goto FAILED;                goto FAILED;
5275                }                }
5276              *code = verbs[i].op;              *code = verbs[i].op;
5277              if (*code++ == OP_THEN)              if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
               {  
               PUT(code, 0, code - bcptr->current_branch - 1);  
               code += LINK_SIZE;  
               }  
5278              }              }
5279    
5280            else            else
# Line 4874  for (;; ptr++) Line 5285  for (;; ptr++)
5285                goto FAILED;                goto FAILED;
5286                }                }
5287              *code = verbs[i].op_arg;              *code = verbs[i].op_arg;
5288              if (*code++ == OP_THEN_ARG)              if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
               {  
               PUT(code, 0, code - bcptr->current_branch - 1);  
               code += LINK_SIZE;  
               }  
5289              *code++ = arglen;              *code++ = arglen;
5290              memcpy(code, arg, arglen);              memcpy(code, arg, arglen);
5291              code += arglen;              code += arglen;
# Line 5140  for (;; ptr++) Line 5547  for (;; ptr++)
5547          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5548          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5549          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5550            cd->assert_depth += 1;
5551          ptr++;          ptr++;
5552          break;          break;
5553    
# Line 5154  for (;; ptr++) Line 5562  for (;; ptr++)
5562            continue;            continue;
5563            }            }
5564          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5565            cd->assert_depth += 1;
5566          break;          break;
5567    
5568    
# Line 5163  for (;; ptr++) Line 5572  for (;; ptr++)
5572            {            {
5573            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5574            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5575              cd->assert_depth += 1;
5576            ptr += 2;            ptr += 2;
5577            break;            break;
5578    
5579            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5580            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5581              cd->assert_depth += 1;
5582            ptr += 2;            ptr += 2;
5583            break;            break;
5584    
# Line 5373  for (;; ptr++) Line 5784  for (;; ptr++)
5784          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5785          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
5786    
5787          /* In the pre-compile phase, do a syntax check and set a dummy          /* In the pre-compile phase, do a syntax check. We used to just set
5788          reference number. */          a dummy reference number, because it was not used in the first pass.
5789            However, with the change of recursive back references to be atomic,
5790            we have to look for the number so that this state can be identified, as
5791            otherwise the incorrect length is computed. If it's not a backwards
5792            reference, the dummy number will do. */
5793    
5794          if (lengthptr != NULL)          if (lengthptr != NULL)
5795            {            {
5796              const uschar *temp;
5797    
5798            if (namelen == 0)            if (namelen == 0)
5799              {              {
5800              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5393  for (;; ptr++) Line 5810  for (;; ptr++)
5810              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5811              goto FAILED;              goto FAILED;
5812              }              }
5813            recno = 0;  
5814              /* The name table does not exist in the first pass, so we cannot
5815              do a simple search as in the code below. Instead, we have to scan the
5816              pattern to find the number. It is important that we scan it only as
5817              far as we have got because the syntax of named subpatterns has not
5818              been checked for the rest of the pattern, and find_parens() assumes
5819              correct syntax. In any case, it's a waste of resources to scan
5820              further. We stop the scan at the current point by temporarily
5821              adjusting the value of cd->endpattern. */
5822    
5823              temp = cd->end_pattern;
5824              cd->end_pattern = ptr;
5825              recno = find_parens(cd, name, namelen,
5826                (options & PCRE_EXTENDED) != 0, utf8);
5827              cd->end_pattern = temp;
5828              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5829            }            }
5830    
5831          /* In the real compile, seek the name in the table. We check the name          /* In the real compile, seek the name in the table. We check the name
# Line 5537  for (;; ptr++) Line 5969  for (;; ptr++)
5969    
5970                /* 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
5971                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5972                of the group. */                of the group. Then remember the forward reference. */
5973    
5974                called = cd->start_code + recno;                called = cd->start_code + recno;
5975                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5976                }                }
5977    
5978              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
5979              this is a recursive call. We check to see if this is a left              this is a recursive call. We check to see if this is a left
5980              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
5981                must not, however, do this check if we are in a conditional
5982                subpattern because the condition might be testing for recursion in
5983                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
5984                Forever loops are also detected at runtime, so those that occur in
5985                conditional subpatterns will be picked up then. */
5986    
5987              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
5988                       could_be_empty(called, code, bcptr, utf8, cd))                       could_be_empty(called, code, bcptr, utf8, cd))
5989                {                {
5990                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
# Line 5555  for (;; ptr++) Line 5992  for (;; ptr++)
5992                }                }
5993              }              }
5994    
5995            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
           "once" brackets. Set up a "previous group" length so that a  
           subsequent quantifier will work. */  
   
           *code = OP_ONCE;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
5996    
5997            *code = OP_RECURSE;            *code = OP_RECURSE;
5998            PUT(code, 1, (int)(called - cd->start_code));            PUT(code, 1, (int)(called - cd->start_code));
5999            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;  
6000            }            }
6001    
6002          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5632  for (;; ptr++) Line 6057  for (;; ptr++)
6057          is necessary to ensure we correctly detect the start of the pattern in          is necessary to ensure we correctly detect the start of the pattern in
6058          both phases.          both phases.
6059    
6060          If we are not at the pattern start, compile code to change the ims          If we are not at the pattern start, reset the greedy defaults and the
6061          options if this setting actually changes any of them, and reset the          case value for firstbyte and reqbyte. */
         greedy defaults and the case value for firstbyte and reqbyte. */  
6062    
6063          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
6064            {            {
# Line 5645  for (;; ptr++) Line 6069  for (;; ptr++)
6069              }              }
6070            else            else
6071              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
6072              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
6073              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
6074              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
6075              }              }
6076    
6077            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
6078            in subsequent branches. When not at the start of the pattern, this            in subsequent branches. */
           information is also necessary so that a resetting item can be  
           compiled at the end of a group (if we are in a group). */  
6079    
6080            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
6081            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5694  for (;; ptr++) Line 6111  for (;; ptr++)
6111        skipbytes = 2;        skipbytes = 2;
6112        }        }
6113    
6114      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
6115      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a      but this was changed for Perl compatibility, so all kinds can now be
6116      non-register variable in order to be able to pass its address because some      repeated. We copy code into a non-register variable (tempcode) in order to
6117      compilers complain otherwise. Pass in a new setting for the ims options if      be able to pass its address because some compilers complain otherwise. */
     they have changed. */  
6118    
6119      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                      /* For handling repetition */
6120      *code = bravalue;      *code = bravalue;
6121      tempcode = code;      tempcode = code;
6122      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
6123      length_prevgroup = 0;              /* Initialize for pre-compile phase */      tempbracount = cd->bracount;          /* Save value before bracket */
6124        length_prevgroup = 0;                 /* Initialize for pre-compile phase */
6125    
6126      if (!compile_regex(      if (!compile_regex(
6127           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
6128           options & PCRE_IMS,           /* The previous ims option state */           &tempcode,                       /* Where to put code (updated) */
6129           &tempcode,                    /* Where to put code (updated) */           &ptr,                            /* Input pointer (updated) */
6130           &ptr,                         /* Input pointer (updated) */           errorcodeptr,                    /* Where to put an error message */
          errorcodeptr,                 /* Where to put an error message */  
6131           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
6132            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
6133           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
6134           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
6135           &subfirstbyte,                /* For possible first char */           cond_depth +
6136           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
6137           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
6138           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
6139           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
6140             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
6141             (lengthptr == NULL)? NULL :      /* Actual compile phase */
6142               &length_prevgroup              /* Pre-compile phase */
6143           ))           ))
6144        goto FAILED;        goto FAILED;
6145    
6146        /* If this was an atomic group and there are no capturing groups within it,
6147        generate OP_ONCE_NC instead of OP_ONCE. */
6148    
6149        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
6150          *code = OP_ONCE_NC;
6151    
6152        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
6153          cd->assert_depth -= 1;
6154    
6155      /* At the end of compiling, code is still pointing to the start of the      /* At the end of compiling, code is still pointing to the start of the
6156      group, while tempcode has been updated to point past the end of the group      group, while tempcode has been updated to point past the end of the group.
6157      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
6158    
6159      /* If this is a conditional bracket, check that there are no more than      If this is a conditional bracket, check that there are no more than
6160      two branches in the group, or just one if it's a DEFINE group. We do this      two branches in the group, or just one if it's a DEFINE group. We do this
6161      in the real compile phase, not in the pre-pass, where the whole group may      in the real compile phase, not in the pre-pass, where the whole group may
6162      not be available. */      not be available. */
# Line 5795  for (;; ptr++) Line 6221  for (;; ptr++)
6221          goto FAILED;          goto FAILED;
6222          }          }
6223        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
6224        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
6225        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
6226        *code++ = OP_KET;        *code++ = OP_KET;
6227        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5963  for (;; ptr++) Line 6389  for (;; ptr++)
6389          }          }
6390    
6391        /* \k<name> or \k'name' is a back reference by name (Perl syntax).        /* \k<name> or \k'name' is a back reference by name (Perl syntax).
6392        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6393    
6394        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6395          {          {
6396            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6397              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6398              {
6399              *errorcodeptr = ERR69;
6400              break;
6401              }
6402          is_recurse = FALSE;          is_recurse = FALSE;
6403          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6404            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
# Line 5987  for (;; ptr++) Line 6418  for (;; ptr++)
6418          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6419          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6420          previous = code;          previous = code;
6421          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6422          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6423          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6424          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
# Line 6095  for (;; ptr++) Line 6526  for (;; ptr++)
6526    
6527      ONE_CHAR:      ONE_CHAR:
6528      previous = code;      previous = code;
6529      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6530      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6531    
6532      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 6124  for (;; ptr++) Line 6555  for (;; ptr++)
6555        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6556        }        }
6557    
6558      /* firstbyte was previously set; we can set reqbyte only the length is      /* firstbyte was previously set; we can set reqbyte only if the length is
6559      1 or the matching is caseful. */      1 or the matching is caseful. */
6560    
6561      else      else
# Line 6159  return FALSE; Line 6590  return FALSE;
6590  /* On entry, ptr is pointing past the bracket character, but on return it  /* On entry, ptr is pointing past the bracket character, but on return it
6591  points to the closing bracket, or vertical bar, or end of string. The code  points to the closing bracket, or vertical bar, or end of string. The code
6592  variable is pointing at the byte into which the BRA operator has been stored.  variable is pointing at the byte into which the BRA operator has been stored.
 If the ims options are changed at the start (for a (?ims: group) or during any  
 branch, we need to insert an OP_OPT item at the start of every following branch  
 to ensure they get set correctly at run time, and also pass the new options  
 into every subsequent branch compile.  
   
6593  This function is used during the pre-compile phase when we are trying to find  This function is used during the pre-compile phase when we are trying to find
6594  out the amount of memory needed, as well as during the real compile phase. The  out the amount of memory needed, as well as during the real compile phase. The
6595  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6596    
6597  Arguments:  Arguments:
6598    options        option bits, including any changes for this subpattern    options        option bits, including any changes for this subpattern
   oldims         previous settings of ims option bits  
6599    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6600    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6601    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
6602    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6603    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6604    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6605      cond_depth     depth of nesting for conditional subpatterns
6606    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6607    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6608    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 6188  Returns:         TRUE on success Line 6614  Returns:         TRUE on success
6614  */  */
6615    
6616  static BOOL  static BOOL
6617  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6618    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6619    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6620    int *lengthptr)    compile_data *cd, int *lengthptr)
6621  {  {
6622  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6623  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 6205  int branchfirstbyte, branchreqbyte; Line 6631  int branchfirstbyte, branchreqbyte;
6631  int length;  int length;
6632  int orig_bracount;  int orig_bracount;
6633  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6634  branch_chain bc;  branch_chain bc;
6635    
6636  bc.outer = bcptr;  bc.outer = bcptr;
# Line 6229  pre-compile phase to find out whether an Line 6654  pre-compile phase to find out whether an
6654    
6655  /* 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
6656  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
6657  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6658    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6659    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6660    
6661  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6662    {    {
# Line 6255  for (;;) Line 6682  for (;;)
6682    
6683    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6684    
   /* Handle a change of ims options at the start of the branch */  
   
   if ((options & PCRE_IMS) != oldims)  
     {  
     *code++ = OP_OPT;  
     *code++ = options & PCRE_IMS;  
     length += 2;  
     }  
   
6685    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6686    
6687    if (lookbehind)    if (lookbehind)
# Line 6278  for (;;) Line 6696  for (;;)
6696    into the length. */    into the length. */
6697    
6698    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6699          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6700            (lengthptr == NULL)? NULL : &length))
6701      {      {
6702      *ptrptr = ptr;      *ptrptr = ptr;
6703      return FALSE;      return FALSE;
6704      }      }
6705    
   /* 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;  
   
6706    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6707    has fewer than the rest. */    has fewer than the rest. */
6708    
# Line 6353  for (;;) Line 6763  for (;;)
6763        {        {
6764        int fixed_length;        int fixed_length;
6765        *code = OP_END;        *code = OP_END;
6766        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6767            FALSE, cd);
6768        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6769        if (fixed_length == -3)        if (fixed_length == -3)
6770          {          {
# Line 6361  for (;;) Line 6772  for (;;)
6772          }          }
6773        else if (fixed_length < 0)        else if (fixed_length < 0)
6774          {          {
6775          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 :
6776                            (fixed_length == -4)? ERR70: ERR25;
6777          *ptrptr = ptr;          *ptrptr = ptr;
6778          return FALSE;          return FALSE;
6779          }          }
# Line 6374  for (;;) Line 6786  for (;;)
6786    of offsets, with the field in the BRA item now becoming an offset to the    of offsets, with the field in the BRA item now becoming an offset to the
6787    first alternative. If there are no alternatives, it points to the end of the    first alternative. If there are no alternatives, it points to the end of the
6788    group. The length in the terminating ket is always the length of the whole    group. The length in the terminating ket is always the length of the whole
6789    bracketed item. If any of the ims options were changed inside the group,    bracketed item. Return leaving the pointer at the terminating char. */
   compile a resetting op-code following, except at the very end of the pattern.  
   Return leaving the pointer at the terminating char. */  
6790    
6791    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6792      {      {
# Line 6420  for (;;) Line 6830  for (;;)
6830        cd->open_caps = cd->open_caps->next;        cd->open_caps = cd->open_caps->next;
6831        }        }
6832    
     /* Reset options if needed. */  
   
     if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)  
       {  
       *code++ = OP_OPT;  
       *code++ = oldims;  
       length += 2;  
       }  
   
6833      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
6834    
6835      cd->bracount = max_bracount;      cd->bracount = max_bracount;
# Line 6488  for (;;) Line 6889  for (;;)
6889  /* Try to find out if this is an anchored regular expression. Consider each  /* Try to find out if this is an anchored regular expression. Consider each
6890  alternative branch. If they all start with OP_SOD or OP_CIRC, or with a bracket  alternative branch. If they all start with OP_SOD or OP_CIRC, or with a bracket
6891  all of whose alternatives start with OP_SOD or OP_CIRC (recurse ad lib), then  all of whose alternatives start with OP_SOD or OP_CIRC (recurse ad lib), then
6892  it's anchored. However, if this is a multiline pattern, then only OP_SOD  it's anchored. However, if this is a multiline pattern, then only OP_SOD will
6893  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6894    
6895  We can also consider a regex to be anchored if OP_SOM starts all its branches.  We can also consider a regex to be anchored if OP_SOM starts all its branches.
6896  This is the code for \G, which means "match at start of match position, taking  This is the code for \G, which means "match at start of match position, taking
# Line 6510  of the more common cases more precisely. Line 6911  of the more common cases more precisely.
6911    
6912  Arguments:  Arguments:
6913    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6914    bracket_map    a bitmap of which brackets we are inside while testing; this    bracket_map    a bitmap of which brackets we are inside while testing; this
6915                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6916                    the less precise approach                    the less precise approach
# Line 6520  Returns:     TRUE or FALSE Line 6920  Returns:     TRUE or FALSE
6920  */  */
6921    
6922  static BOOL  static BOOL
6923  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6924    unsigned int backref_map)    unsigned int backref_map)
6925  {  {
6926  do {  do {
6927     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6928       options, PCRE_MULTILINE, FALSE);       FALSE);
6929     register int op = *scode;     register int op = *scode;
6930    
6931     /* Non-capturing brackets */     /* Non-capturing brackets */
6932    
6933     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6934           op == OP_SBRA || op == OP_SBRAPOS)
6935       {       {
6936       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6937       }       }
6938    
6939     /* Capturing brackets */     /* Capturing brackets */
6940    
6941     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6942                op == OP_SCBRA || op == OP_SCBRAPOS)
6943       {       {
6944       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6945       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6946       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6947       }       }
6948    
6949     /* Other brackets */     /* Other brackets */
6950    
6951     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC ||
6952                op == OP_COND)
6953       {       {
6954       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6955       }       }
6956    
6957     /* .* 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 6563  do { Line 6966  do {
6966    
6967     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6968    
6969     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;  
6970     code += GET(code, 1);     code += GET(code, 1);
6971     }     }
6972  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6601  is_startline(const uschar *code, unsigne Line 7002  is_startline(const uschar *code, unsigne
7002  {  {
7003  do {  do {
7004     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
7005       NULL, 0, FALSE);       FALSE);
7006     register int op = *scode;     register int op = *scode;
7007    
7008     /* 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 6628  do { Line 7029  do {
7029         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
7030         break;         break;
7031         }         }
7032       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
7033       op = *scode;       op = *scode;
7034       }       }
7035    
7036     /* Non-capturing brackets */     /* Non-capturing brackets */
7037    
7038     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
7039           op == OP_SBRA || op == OP_SBRAPOS)
7040       {       {
7041       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
7042       }       }
7043    
7044     /* Capturing brackets */     /* Capturing brackets */
7045    
7046     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
7047                op == OP_SCBRA || op == OP_SCBRAPOS)
7048       {       {
7049       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
7050       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6650  do { Line 7053  do {
7053    
7054     /* Other brackets */     /* Other brackets */
7055    
7056     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC)
7057       {       {
7058       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
7059       }       }
# Line 6665  do { Line 7068  do {
7068    
7069     /* Check for explicit circumflex */     /* Check for explicit circumflex */
7070    
7071     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
7072    
7073     /* Move on to the next alternative */     /* Move on to the next alternative */
7074    
# Line 6691  we return that char, otherwise -1. Line 7094  we return that char, otherwise -1.
7094    
7095  Arguments:  Arguments:
7096    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)  
7097    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
7098    
7099  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
7100  */  */
7101    
7102  static int  static int
7103  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
7104  {  {
7105  register int c = -1;  register int c = -1;
7106  do {  do {
7107     int d;     int d;
7108     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
7109       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
7110       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
7111     register int op = *scode;     register int op = *scode;
7112    
7113     switch(op)     switch(op)
# Line 6713  do { Line 7116  do {
7116       return -1;       return -1;
7117    
7118       case OP_BRA:       case OP_BRA:
7119         case OP_BRAPOS:
7120       case OP_CBRA:       case OP_CBRA:
7121         case OP_SCBRA:
7122         case OP_CBRAPOS:
7123         case OP_SCBRAPOS:
7124       case OP_ASSERT:       case OP_ASSERT:
7125       case OP_ONCE:       case OP_ONCE:
7126         case OP_ONCE_NC:
7127       case OP_COND:       case OP_COND:
7128       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
7129         return -1;         return -1;
7130       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
7131       break;       break;
7132    
7133       case OP_EXACT:       /* Fall through */       case OP_EXACT:
7134       scode += 2;       scode += 2;
7135         /* Fall through */
7136    
7137       case OP_CHAR:       case OP_CHAR:
      case OP_CHARNC:  
7138       case OP_PLUS:       case OP_PLUS:
7139       case OP_MINPLUS:       case OP_MINPLUS:
7140       case OP_POSPLUS:       case OP_POSPLUS:
7141       if (!inassert) return -1;       if (!inassert) return -1;
7142       if (c < 0)       if (c < 0) c = scode[1];
7143         {         else if (c != scode[1]) return -1;
7144         c = scode[1];       break;
7145         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
7146         }       case OP_EXACTI:
7147       else if (c != scode[1]) return -1;       scode += 2;
7148         /* Fall through */
7149    
7150         case OP_CHARI:
7151         case OP_PLUSI:
7152         case OP_MINPLUSI:
7153         case OP_POSPLUSI:
7154         if (!inassert) return -1;
7155         if (c < 0) c = scode[1] | REQ_CASELESS;
7156           else if (c != scode[1]) return -1;
7157       break;       break;
7158       }       }
7159    
# Line 6859  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7276  while (ptr[skipatstart] == CHAR_LEFT_PAR
7276      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
7277    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
7278      { skipatstart += 6; options |= PCRE_UCP; continue; }      { skipatstart += 6; options |= PCRE_UCP; continue; }
7279      else if (strncmp((char *)(ptr+skipatstart+2), STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
7280        { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
7281    
7282    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
7283      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6885  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7304  while (ptr[skipatstart] == CHAR_LEFT_PAR
7304    
7305  utf8 = (options & PCRE_UTF8) != 0;  utf8 = (options & PCRE_UTF8) != 0;
7306    
7307  /* Can't support UTF8 unless PCRE has been compiled to include the code. */  /* Can't support UTF8 unless PCRE has been compiled to include the code. The
7308    return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7309    release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7310    not used here. */
7311    
7312  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7313  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7314       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7315    {    {
7316    errorcode = ERR44;    errorcode = ERR44;
7317    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6914  if ((options & PCRE_UCP) != 0) Line 7336  if ((options & PCRE_UCP) != 0)
7336    
7337  /* Check validity of \R options. */  /* Check validity of \R options. */
7338    
7339  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
7340         (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
7341    {    {
7342    case 0:    errorcode = ERR56;
7343    case PCRE_BSR_ANYCRLF:    goto PCRE_EARLY_ERROR_RETURN;
   case PCRE_BSR_UNICODE:  
   break;  
   default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;  
7344    }    }
7345    
7346  /* 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 7005  outside can help speed up starting point Line 7425  outside can help speed up starting point
7425  ptr += skipatstart;  ptr += skipatstart;
7426  code = cworkspace;  code = cworkspace;
7427  *code = OP_BRA;  *code = OP_BRA;
7428  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7429    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7430  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7431    
7432  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 7061  field; this time it's used for rememberi Line 7480  field; this time it's used for rememberi
7480  */  */
7481    
7482  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7483    cd->assert_depth = 0;
7484  cd->bracount = 0;  cd->bracount = 0;
7485  cd->names_found = 0;  cd->names_found = 0;
7486  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 7079  of the function here. */ Line 7499  of the function here. */
7499  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7500  code = (uschar *)codestart;  code = (uschar *)codestart;
7501  *code = OP_BRA;  *code = OP_BRA;
7502  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0, 0,
7503    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7504  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7505  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7506  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7507    
7508  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */  if (cd->had_accept) reqbyte = REQ_NONE;   /* Must disable after (*ACCEPT) */
7509    
7510  /* 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. */
7511    
# Line 7146  if (cd->check_lookbehind) Line 7566  if (cd->check_lookbehind)
7566        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7567        int end_op = *be;        int end_op = *be;
7568        *be = OP_END;        *be = OP_END;
7569        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7570            cd);
7571        *be = end_op;        *be = end_op;
7572        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7573        if (fixed_length < 0)        if (fixed_length < 0)
7574          {          {
7575          errorcode = (fixed_length == -2)? ERR36 : ERR25;          errorcode = (fixed_length == -2)? ERR36 :
7576                        (fixed_length == -4)? ERR70 : ERR25;
7577          break;          break;
7578          }          }
7579        PUT(cc, 1, fixed_length);        PUT(cc, 1, fixed_length);
# Line 7185  start with ^. and also when all branches Line 7607  start with ^. and also when all branches
7607    
7608  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7609    {    {
7610    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))  
7611      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7612    else    else
7613      {      {
7614      if (firstbyte < 0)      if (firstbyte < 0)
7615        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7616      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7617        {        {
7618        int ch = firstbyte & 255;        int ch = firstbyte & 255;

Legend:
Removed from v.574  
changed lines
  Added in v.751

  ViewVC Help
Powered by ViewVC 1.1.5