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

Diff of /code/trunk/pcre_compile.c

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

revision 578 by ph10, Tue Nov 23 15:34:55 2010 UTC revision 744 by zherczeg, Sun Nov 13 16:31:38 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    ;    ;
414    
415  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 545  static const unsigned char ebcdic_charta Line 546  static const unsigned char ebcdic_charta
546  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
547    
548  static BOOL  static BOOL
549    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
550      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
551    
552    
# Line 577  return s; Line 578  return s;
578    
579    
580  /*************************************************  /*************************************************
581    *            Check for counted repeat            *
582    *************************************************/
583    
584    /* This function is called when a '{' is encountered in a place where it might
585    start a quantifier. It looks ahead to see if it really is a quantifier or not.
586    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
587    where the ddds are digits.
588    
589    Arguments:
590      p         pointer to the first char after '{'
591    
592    Returns:    TRUE or FALSE
593    */
594    
595    static BOOL
596    is_counted_repeat(const uschar *p)
597    {
598    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
599    while ((digitab[*p] & ctype_digit) != 0) p++;
600    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
601    
602    if (*p++ != CHAR_COMMA) return FALSE;
603    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
604    
605    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
606    while ((digitab[*p] & ctype_digit) != 0) p++;
607    
608    return (*p == CHAR_RIGHT_CURLY_BRACKET);
609    }
610    
611    
612    
613    /*************************************************
614  *            Handle escapes                      *  *            Handle escapes                      *
615  *************************************************/  *************************************************/
616    
# Line 642  else Line 676  else
676    
677      case CHAR_l:      case CHAR_l:
678      case CHAR_L:      case CHAR_L:
679        *errorcodeptr = ERR37;
680        break;
681    
682      case CHAR_u:      case CHAR_u:
683        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
684          {
685          /* In JavaScript, \u must be followed by four hexadecimal numbers.
686          Otherwise it is a lowercase u letter. */
687          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
688               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
689            {
690            int i;
691            c = 0;
692            for (i = 0; i < 4; ++i)
693              {
694              register int cc = *(++ptr);
695    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
696              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
697              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
698    #else           /* EBCDIC coding */
699              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
700              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
701    #endif
702              }
703            }
704          }
705        else
706          *errorcodeptr = ERR37;
707        break;
708    
709      case CHAR_U:      case CHAR_U:
710      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
711        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
712      break;      break;
713    
714      /* \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
715        class, \g must be followed by one of a number of specific things:
716    
717      (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
718      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 729  else
729      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
730    
731      case CHAR_g:      case CHAR_g:
732        if (isclass) break;
733      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
734        {        {
735        c = -ESC_g;        c = -ESC_g;
# Line 792  else Line 858  else
858      treated as a data character. */      treated as a data character. */
859    
860      case CHAR_x:      case CHAR_x:
861        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
862          {
863          /* In JavaScript, \x must be followed by two hexadecimal numbers.
864          Otherwise it is a lowercase x letter. */
865          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
866            {
867            int i;
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 1113  that if (?< or (?' or (?P< is encountere Line 1171  that if (?< or (?' or (?P< is encountere
1171  terminated because that is checked in the first pass. There is now one call to  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  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  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  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  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,  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  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  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  processing items within the loop are OK, because afterwards the main loop will
1180  terminate.)  terminate.)
1181    
1182  Arguments:  Arguments:
# Line 1127  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 1220  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. Stop if we get to cd->end_pattern. Note that this is important for the  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  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. */  position. So DO NOT change this to a test for binary zero. */
1284    
1285  for (; ptr < cd->end_pattern; ptr++)  for (; ptr < cd->end_pattern; ptr++)
# Line 1298  for (; ptr < cd->end_pattern; ptr++) Line 1356  for (; ptr < cd->end_pattern; 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 1361  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 1397  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 1474  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 1485  Returns:   the fixed length, Line 1533  Returns:   the fixed length,
1533  */  */
1534    
1535  static int  static int
1536  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1537  {  {
1538  int length = -1;  int length = -1;
1539    
# Line 1502  for (;;) Line 1550  for (;;)
1550    register int op = *cc;    register int op = *cc;
1551    switch (op)    switch (op)
1552      {      {
1553        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1554        OP_BRA (normal non-capturing bracket) because the other variants of these
1555        opcodes are all concerned with unlimited repeated groups, which of course
1556        are not of fixed length. They will cause a -1 response from the default
1557        case of this switch. */
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);
# Line 1515  for (;;) Line 1570  for (;;)
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
1572      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is
1573      END it's the end of the outer call. All can be handled by the same code. */      END it's the end of the outer call. All can be handled by the same code.
1574        Note that we must not include the OP_KETRxxx opcodes here, because they
1575        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      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1581        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1538  for (;;) Line 1593  for (;;)
1593      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1594      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1595      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1596      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1597      if (d < 0) return d;      if (d < 0) return d;
1598      branchlength += d;      branchlength += d;
1599      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1561  for (;;) Line 1616  for (;;)
1616      case OP_RREF:      case OP_RREF:
1617      case OP_NRREF:      case OP_NRREF:
1618      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1619      case OP_CALLOUT:      case OP_CALLOUT:
1620      case OP_SOD:      case OP_SOD:
1621      case OP_SOM:      case OP_SOM:
# Line 1569  for (;;) Line 1623  for (;;)
1623      case OP_EOD:      case OP_EOD:
1624      case OP_EODN:      case OP_EODN:
1625      case OP_CIRC:      case OP_CIRC:
1626        case OP_CIRCM:
1627      case OP_DOLL:      case OP_DOLL:
1628        case OP_DOLLM:
1629      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1630      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1631      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1578  for (;;) Line 1634  for (;;)
1634      /* Handle literal characters */      /* Handle literal characters */
1635    
1636      case OP_CHAR:      case OP_CHAR:
1637      case OP_CHARNC:      case OP_CHARI:
1638      case OP_NOT:      case OP_NOT:
1639        case OP_NOTI:
1640      branchlength++;      branchlength++;
1641      cc += 2;      cc += 2;
1642  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1643      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];  
1644  #endif  #endif
1645      break;      break;
1646    
# Line 1595  for (;;) Line 1651  for (;;)
1651      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1652      cc += 4;      cc += 4;
1653  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1654      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];  
1655  #endif  #endif
1656      break;      break;
1657    
# Line 1698  _pcre_find_bracket(const uschar *code, B Line 1753  _pcre_find_bracket(const uschar *code, B
1753  for (;;)  for (;;)
1754    {    {
1755    register int c = *code;    register int c = *code;
1756    
1757    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1758    
1759    /* 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 1716  for (;;) Line 1772  for (;;)
1772    
1773    /* Handle capturing bracket */    /* Handle capturing bracket */
1774    
1775    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1776               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1777      {      {
1778      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1779      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1758  for (;;) Line 1815  for (;;)
1815        break;        break;
1816    
1817        case OP_THEN_ARG:        case OP_THEN_ARG:
1818        code += code[1+LINK_SIZE];        code += code[1];
1819        break;        break;
1820        }        }
1821    
# Line 1774  for (;;) Line 1831  for (;;)
1831      if (utf8) switch(c)      if (utf8) switch(c)
1832        {        {
1833        case OP_CHAR:        case OP_CHAR:
1834        case OP_CHARNC:        case OP_CHARI:
1835        case OP_EXACT:        case OP_EXACT:
1836          case OP_EXACTI:
1837        case OP_UPTO:        case OP_UPTO:
1838          case OP_UPTOI:
1839        case OP_MINUPTO:        case OP_MINUPTO:
1840          case OP_MINUPTOI:
1841        case OP_POSUPTO:        case OP_POSUPTO:
1842          case OP_POSUPTOI:
1843        case OP_STAR:        case OP_STAR:
1844          case OP_STARI:
1845        case OP_MINSTAR:        case OP_MINSTAR:
1846          case OP_MINSTARI:
1847        case OP_POSSTAR:        case OP_POSSTAR:
1848          case OP_POSSTARI:
1849        case OP_PLUS:        case OP_PLUS:
1850          case OP_PLUSI:
1851        case OP_MINPLUS:        case OP_MINPLUS:
1852          case OP_MINPLUSI:
1853        case OP_POSPLUS:        case OP_POSPLUS:
1854          case OP_POSPLUSI:
1855        case OP_QUERY:        case OP_QUERY:
1856          case OP_QUERYI:
1857        case OP_MINQUERY:        case OP_MINQUERY:
1858          case OP_MINQUERYI:
1859        case OP_POSQUERY:        case OP_POSQUERY:
1860          case OP_POSQUERYI:
1861        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1862        break;        break;
1863        }        }
# Line 1864  for (;;) Line 1934  for (;;)
1934        break;        break;
1935    
1936        case OP_THEN_ARG:        case OP_THEN_ARG:
1937        code += code[1+LINK_SIZE];        code += code[1];
1938        break;        break;
1939        }        }
1940    
# Line 1880  for (;;) Line 1950  for (;;)
1950      if (utf8) switch(c)      if (utf8) switch(c)
1951        {        {
1952        case OP_CHAR:        case OP_CHAR:
1953        case OP_CHARNC:        case OP_CHARI:
1954        case OP_EXACT:        case OP_EXACT:
1955          case OP_EXACTI:
1956        case OP_UPTO:        case OP_UPTO:
1957          case OP_UPTOI:
1958        case OP_MINUPTO:        case OP_MINUPTO:
1959          case OP_MINUPTOI:
1960        case OP_POSUPTO:        case OP_POSUPTO:
1961          case OP_POSUPTOI:
1962        case OP_STAR:        case OP_STAR:
1963          case OP_STARI:
1964        case OP_MINSTAR:        case OP_MINSTAR:
1965          case OP_MINSTARI:
1966        case OP_POSSTAR:        case OP_POSSTAR:
1967          case OP_POSSTARI:
1968        case OP_PLUS:        case OP_PLUS:
1969          case OP_PLUSI:
1970        case OP_MINPLUS:        case OP_MINPLUS:
1971          case OP_MINPLUSI:
1972        case OP_POSPLUS:        case OP_POSPLUS:
1973          case OP_POSPLUSI:
1974        case OP_QUERY:        case OP_QUERY:
1975          case OP_QUERYI:
1976        case OP_MINQUERY:        case OP_MINQUERY:
1977          case OP_MINQUERYI:
1978        case OP_POSQUERY:        case OP_POSQUERY:
1979          case OP_POSQUERYI:
1980        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1981        break;        break;
1982        }        }
# Line 1932  could_be_empty_branch(const uschar *code Line 2015  could_be_empty_branch(const uschar *code
2015    compile_data *cd)    compile_data *cd)
2016  {  {
2017  register int c;  register int c;
2018  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2019       code < endcode;       code < endcode;
2020       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2021    {    {
2022    const uschar *ccode;    const uschar *ccode;
2023    
# Line 1950  for (code = first_significant_code(code Line 2033  for (code = first_significant_code(code
2033      continue;      continue;
2034      }      }
2035    
   /* 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;  
     }  
   
2036    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
2037    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
2038      forward reference subroutine call, we can't. To detect forward reference
2039      we have to scan up the list that is kept in the workspace. This function is
2040      called only when doing the real compile, not during the pre-compile that
2041      measures the size of the compiled pattern. */
2042    
2043    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2044      {      {
2045      BOOL empty_branch = FALSE;      const uschar *scode;
2046      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
2047    
2048        /* Test for forward reference */
2049    
2050        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2051          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2052    
2053        /* Not a forward reference, test for completed backward reference */
2054    
2055        empty_branch = FALSE;
2056        scode = cd->start_code + GET(code, 1);
2057      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2058    
2059        /* Completed backwards reference */
2060    
2061      do      do
2062        {        {
2063        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf8, cd))
# Line 1978  for (code = first_significant_code(code Line 2068  for (code = first_significant_code(code
2068        scode += GET(scode, 1);        scode += GET(scode, 1);
2069        }        }
2070      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2071    
2072      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2073      continue;      continue;
2074      }      }
2075    
2076      /* Groups with zero repeats can of course be empty; skip them. */
2077    
2078      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2079          c == OP_BRAPOSZERO)
2080        {
2081        code += _pcre_OP_lengths[c];
2082        do code += GET(code, 1); while (*code == OP_ALT);
2083        c = *code;
2084        continue;
2085        }
2086    
2087      /* A nested group that is already marked as "could be empty" can just be
2088      skipped. */
2089    
2090      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2091          c == OP_SCBRA || c == OP_SCBRAPOS)
2092        {
2093        do code += GET(code, 1); while (*code == OP_ALT);
2094        c = *code;
2095        continue;
2096        }
2097    
2098    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2099    
2100    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2101          c == OP_CBRA || c == OP_CBRAPOS ||
2102          c == OP_ONCE || c == OP_ONCE_NC ||
2103          c == OP_COND)
2104      {      {
2105      BOOL empty_branch;      BOOL empty_branch;
2106      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2071  for (code = first_significant_code(code Line 2187  for (code = first_significant_code(code
2187      case OP_ALLANY:      case OP_ALLANY:
2188      case OP_ANYBYTE:      case OP_ANYBYTE:
2189      case OP_CHAR:      case OP_CHAR:
2190      case OP_CHARNC:      case OP_CHARI:
2191      case OP_NOT:      case OP_NOT:
2192        case OP_NOTI:
2193      case OP_PLUS:      case OP_PLUS:
2194      case OP_MINPLUS:      case OP_MINPLUS:
2195      case OP_POSPLUS:      case OP_POSPLUS:
# Line 2112  for (code = first_significant_code(code Line 2229  for (code = first_significant_code(code
2229      case OP_KET:      case OP_KET:
2230      case OP_KETRMAX:      case OP_KETRMAX:
2231      case OP_KETRMIN:      case OP_KETRMIN:
2232        case OP_KETRPOS:
2233      case OP_ALT:      case OP_ALT:
2234      return TRUE;      return TRUE;
2235    
# Line 2120  for (code = first_significant_code(code Line 2238  for (code = first_significant_code(code
2238    
2239  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2240      case OP_STAR:      case OP_STAR:
2241        case OP_STARI:
2242      case OP_MINSTAR:      case OP_MINSTAR:
2243        case OP_MINSTARI:
2244      case OP_POSSTAR:      case OP_POSSTAR:
2245        case OP_POSSTARI:
2246      case OP_QUERY:      case OP_QUERY:
2247        case OP_QUERYI:
2248      case OP_MINQUERY:      case OP_MINQUERY:
2249        case OP_MINQUERYI:
2250      case OP_POSQUERY:      case OP_POSQUERY:
2251        case OP_POSQUERYI:
2252      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2253      break;      break;
2254    
2255      case OP_UPTO:      case OP_UPTO:
2256        case OP_UPTOI:
2257      case OP_MINUPTO:      case OP_MINUPTO:
2258        case OP_MINUPTOI:
2259      case OP_POSUPTO:      case OP_POSUPTO:
2260        case OP_POSUPTOI:
2261      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2262      break;      break;
2263  #endif  #endif
# Line 2145  for (code = first_significant_code(code Line 2272  for (code = first_significant_code(code
2272      break;      break;
2273    
2274      case OP_THEN_ARG:      case OP_THEN_ARG:
2275      code += code[1+LINK_SIZE];      code += code[1];
2276      break;      break;
2277    
2278      /* None of the remaining opcodes are required to match a character. */      /* None of the remaining opcodes are required to match a character. */
# Line 2168  return TRUE; Line 2295  return TRUE;
2295  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
2296  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,
2297  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.
2298    This function is called only during the real compile, not during the
2299    pre-compile.
2300    
2301  Arguments:  Arguments:
2302    code        points to start of the recursion    code        points to start of the recursion
# Line 2218  where Perl recognizes it as the POSIX cl Line 2347  where Perl recognizes it as the POSIX cl
2347  "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,
2348  I think.  I think.
2349    
2350    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2351    It seems that the appearance of a nested POSIX class supersedes an apparent
2352    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2353    a digit.
2354    
2355    In Perl, unescaped square brackets may also appear as part of class names. For
2356    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2357    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2358    seem right at all. PCRE does not allow closing square brackets in POSIX class
2359    names.
2360    
2361  Arguments:  Arguments:
2362    ptr      pointer to the initial [    ptr      pointer to the initial [
2363    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2232  int terminator;          /* Don't combin Line 2372  int terminator;          /* Don't combin
2372  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2373  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2374    {    {
2375    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2376        ptr++;
2377      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2378      else
2379      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2380      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2381        {        {
2382        *endptr = ptr;        *endptr = ptr;
2383        return TRUE;        return TRUE;
2384        }        }
2385        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2386             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2387              ptr[1] == CHAR_EQUALS_SIGN) &&
2388            check_posix_syntax(ptr, endptr))
2389          return FALSE;
2390      }      }
2391    }    }
2392  return FALSE;  return FALSE;
# Line 2545  if ((options & PCRE_EXTENDED) != 0) Line 2692  if ((options & PCRE_EXTENDED) != 0)
2692      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2693      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2694        {        {
2695        ptr++;        ptr++;
2696        while (*ptr != 0)        while (*ptr != 0)
2697          {          {
2698          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2699          ptr++;          ptr++;
2700  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2701          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2702  #endif  #endif
2703          }          }
# Line 2589  if ((options & PCRE_EXTENDED) != 0) Line 2736  if ((options & PCRE_EXTENDED) != 0)
2736      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2737      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2738        {        {
2739        ptr++;        ptr++;
2740        while (*ptr != 0)        while (*ptr != 0)
2741          {          {
2742          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2743          ptr++;          ptr++;
2744  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2745          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2746  #endif  #endif
2747          }          }
# Line 2622  if (next >= 0) switch(op_code) Line 2769  if (next >= 0) switch(op_code)
2769  #endif  #endif
2770    return c != next;    return c != next;
2771    
2772    /* 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
2773    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
2774    high-valued characters. */    high-valued characters. */
2775    
2776    case OP_CHARNC:    case OP_CHARI:
2777  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2778    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
2779  #else  #else
# Line 2649  if (next >= 0) switch(op_code) Line 2796  if (next >= 0) switch(op_code)
2796  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2797    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2798    
2799    /* 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
2800      opcodes are not used for multi-byte characters, because they are coded using
2801      an XCLASS instead. */
2802    
2803    case OP_NOT:    case OP_NOT:
2804      return (c = *previous) == next;
2805    
2806      case OP_NOTI:
2807    if ((c = *previous) == next) return TRUE;    if ((c = *previous) == next) return TRUE;
   if ((options & PCRE_CASELESS) == 0) return FALSE;  
2808  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2809    if (utf8)    if (utf8)
2810      {      {
# Line 2758  replaced by OP_PROP codes when PCRE_UCP Line 2909  replaced by OP_PROP codes when PCRE_UCP
2909  switch(op_code)  switch(op_code)
2910    {    {
2911    case OP_CHAR:    case OP_CHAR:
2912    case OP_CHARNC:    case OP_CHARI:
2913  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2914    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
2915  #else  #else
# Line 2945  Arguments: Line 3096  Arguments:
3096    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3097    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3098    bcptr          points to current branch chain    bcptr          points to current branch chain
3099      cond_depth     conditional nesting depth
3100    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3101    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3102                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2956  Returns:         TRUE on success Line 3108  Returns:         TRUE on success
3108  static BOOL  static BOOL
3109  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3110    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3111    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3112  {  {
3113  int repeat_type, op_type;  int repeat_type, op_type;
3114  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2965  int greedy_default, greedy_non_default; Line 3117  int greedy_default, greedy_non_default;
3117  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3118  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3119  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3120  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3121  int after_manual_callout = 0;  int after_manual_callout = 0;
3122  int length_prevgroup = 0;  int length_prevgroup = 0;
3123  register int c;  register int c;
# Line 2983  uschar *previous_callout = NULL; Line 3135  uschar *previous_callout = NULL;
3135  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3136  uschar classbits[32];  uschar classbits[32];
3137    
3138    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3139    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3140    dynamically as we process the pattern. */
3141    
3142  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3143  BOOL class_utf8;  BOOL class_utf8;
3144  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2991  uschar *class_utf8data_base; Line 3147  uschar *class_utf8data_base;
3147  uschar utf8_char[6];  uschar utf8_char[6];
3148  #else  #else
3149  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3150  #endif  #endif
3151    
3152  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 3042  for (;; ptr++) Line 3197  for (;; ptr++)
3197    int subfirstbyte;    int subfirstbyte;
3198    int terminator;    int terminator;
3199    int mclength;    int mclength;
3200      int tempbracount;
3201    uschar mcbuffer[8];    uschar mcbuffer[8];
3202    
3203    /* Get next byte in the pattern */    /* Get next byte in the pattern */
# Line 3163  for (;; ptr++) Line 3319  for (;; ptr++)
3319      previous_callout = NULL;      previous_callout = NULL;
3320      }      }
3321    
3322    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3323    
3324    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3325      {      {
3326      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3327      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3328        {        {
3329        ptr++;        ptr++;
3330        while (*ptr != 0)        while (*ptr != 0)
3331          {          {
3332          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3333          ptr++;          ptr++;
3334  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3335          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3336  #endif  #endif
3337          }          }
# Line 3222  for (;; ptr++) Line 3378  for (;; ptr++)
3378      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3379    
3380      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3381        previous = NULL;
3382      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3383        {        {
3384        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3385          *code++ = OP_CIRCM;
3386        }        }
3387      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3388      break;      break;
3389    
3390      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3391      previous = NULL;      previous = NULL;
3392      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3393      break;      break;
3394    
3395      /* 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 3553  for (;; ptr++) Line 3710  for (;; ptr++)
3710              continue;              continue;
3711    
3712              /* 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
3713              if it was previously set by something earlier in the character              if it was previously set by something earlier in the character
3714              class. */              class. */
3715    
3716              case ESC_s:              case ESC_s:
3717              classbits[0] |= cbits[cbit_space];              classbits[0] |= cbits[cbit_space];
3718              classbits[1] |= cbits[cbit_space+1] & ~0x08;              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3719              for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3720              continue;              continue;
3721    
# Line 3978  for (;; ptr++) Line 4135  for (;; ptr++)
4135    
4136      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
4137      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4138      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4139      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4140    
4141      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
4142      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.
4143      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
4144      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
4145      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
4146      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4147    
4148  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4149      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3997  for (;; ptr++) Line 4154  for (;; ptr++)
4154        {        {
4155        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4156    
4157        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4158    
4159        if (negate_class)        if (negate_class)
4160          {          {
4161          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4162          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4163          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4164          *code++ = class_lastchar;          *code++ = class_lastchar;
4165          break;          break;
4166          }          }
# Line 4131  for (;; ptr++) Line 4288  for (;; ptr++)
4288      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4289      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4290    
4291      /* 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
4292      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4293    
4294      tempcode = previous;      tempcode = previous;
4295    
# Line 4155  for (;; ptr++) Line 4312  for (;; ptr++)
4312        }        }
4313      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4314    
4315        /* If previous was a recursion call, wrap it in atomic brackets so that
4316        previous becomes the atomic group. All recursions were so wrapped in the
4317        past, but it no longer happens for non-repeated recursions. In fact, the
4318        repeated ones could be re-implemented independently so as not to need this,
4319        but for the moment we rely on the code for repeating groups. */
4320    
4321        if (*previous == OP_RECURSE)
4322          {
4323          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4324          *previous = OP_ONCE;
4325          PUT(previous, 1, 2 + 2*LINK_SIZE);
4326          previous[2 + 2*LINK_SIZE] = OP_KET;
4327          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4328          code += 2 + 2 * LINK_SIZE;
4329          length_prevgroup = 3 + 3*LINK_SIZE;
4330    
4331          /* When actually compiling, we need to check whether this was a forward
4332          reference, and if so, adjust the offset. */
4333    
4334          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4335            {
4336            int offset = GET(cd->hwm, -LINK_SIZE);
4337            if (offset == previous + 1 - cd->start_code)
4338              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4339            }
4340          }
4341    
4342        /* Now handle repetition for the different types of item. */
4343    
4344      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4345      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
4346      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
4347      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
4348      instead.  */      instead.  */
4349    
4350      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4351        {        {
4352          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4353    
4354        /* 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
4355        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
4356        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 4207  for (;; ptr++) Line 4395  for (;; ptr++)
4395      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4396      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-
4397      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4398      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
4399      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4400    
4401      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4402        {        {
4403        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4404        c = previous[1];        c = previous[1];
4405        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4406            repeat_max < 0 &&            repeat_max < 0 &&
# Line 4409  for (;; ptr++) Line 4597  for (;; ptr++)
4597  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4598               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4599  #endif  #endif
4600               *previous == OP_REF)               *previous == OP_REF ||
4601                 *previous == OP_REFI)
4602        {        {
4603        if (repeat_max == 0)        if (repeat_max == 0)
4604          {          {
# Line 4443  for (;; ptr++) Line 4632  for (;; ptr++)
4632        }        }
4633    
4634      /* 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
4635      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4636        opcodes such as BRA and CBRA, as this is the place where they get converted
4637        into the more special varieties such as BRAPOS and SBRA. A test for >=
4638        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4639        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4640        repetition of assertions, but now it does, for Perl compatibility. */
4641    
4642      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4643        {        {
4644        register int i;        register int i;
       int ketoffset = 0;  
4645        int len = (int)(code - previous);        int len = (int)(code - previous);
4646        uschar *bralink = NULL;        uschar *bralink = NULL;
4647          uschar *brazeroptr = NULL;
4648    
4649        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4650          we just ignore the repeat. */
4651    
4652        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4653          {          goto END_REPEAT;
4654          *errorcodeptr = ERR55;  
4655          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4656          }        use of repetition is in cases when the assertion is optional. Therefore,
4657          if the minimum is greater than zero, just ignore the repeat. If the
4658          maximum is not not zero or one, set it to 1. */
4659    
4660        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4661        by scanning through from the start, and compute the offset back to it          {
4662        from the current code pointer. There may be an OP_OPT setting following          if (repeat_min > 0) goto END_REPEAT;
4663        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);  
4664          }          }
4665    
4666        /* 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 4492  for (;; ptr++) Line 4681  for (;; ptr++)
4681          **   goto END_REPEAT;          **   goto END_REPEAT;
4682          **   }          **   }
4683    
4684          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
4685          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
4686          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
4687          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4688            selectively.
4689    
4690          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
4691          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 4515  for (;; ptr++) Line 4705  for (;; ptr++)
4705              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4706              goto END_REPEAT;              goto END_REPEAT;
4707              }              }
4708              brazeroptr = previous;    /* Save for possessive optimizing */
4709            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4710            }            }
4711    
# Line 4679  for (;; ptr++) Line 4870  for (;; ptr++)
4870            }            }
4871          }          }
4872    
4873        /* 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
4874        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4875        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
4876        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4877          deal with possessive ONCEs specially.
4878    
4879          Otherwise, if the quantifier was possessive, we convert the BRA code to
4880          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4881          at runtime to detect this kind of subpattern at both the start and at the
4882          end.) The use of special opcodes makes it possible to reduce greatly the
4883          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4884          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4885          the default action below, of wrapping everything inside atomic brackets,
4886          does not happen.
4887    
4888        Then, when we are doing the actual compile phase, check to see whether        Then, when we are doing the actual compile phase, check to see whether
4889        this group is a non-atomic one that could match an empty string. If so,        this group is one that could match an empty string. If so, convert the
4890        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so that runtime
4891        that runtime checking can be done. [This check is also applied to        checking can be done. [This check is also applied to ONCE groups at
4892        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4893    
4894        else        else
4895          {          {
4896          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4897          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4898          *ketcode = OP_KETRMAX + repeat_type;  
4899          if (lengthptr == NULL && *bracode != OP_ONCE)          if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
4900                possessive_quantifier) *bracode = OP_BRA;
4901    
4902            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
4903              *ketcode = OP_KETRMAX + repeat_type;
4904            else
4905            {            {
4906            uschar *scode = bracode;            if (possessive_quantifier)
           do  
4907              {              {
4908              if (could_be_empty_branch(scode, ketcode, utf8, cd))              *bracode += 1;                   /* Switch to xxxPOS opcodes */
4909                *ketcode = OP_KETRPOS;
4910                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4911                possessive_quantifier = FALSE;
4912                }
4913              else *ketcode = OP_KETRMAX + repeat_type;
4914    
4915              if (lengthptr == NULL)
4916                {
4917                uschar *scode = bracode;
4918                do
4919                {                {
4920                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4921                break;                  {
4922                    *bracode += OP_SBRA - OP_BRA;
4923                    break;
4924                    }
4925                  scode += GET(scode, 1);
4926                }                }
4927              scode += GET(scode, 1);              while (*scode == OP_ALT);
4928              }              }
           while (*scode == OP_ALT);  
4929            }            }
4930          }          }
4931        }        }
# Line 4728  for (;; ptr++) Line 4946  for (;; ptr++)
4946        }        }
4947    
4948      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4949      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4950      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4951      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4952      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
4953      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4954      tempcode, not at previous, which might be the first part of a string whose  
4955      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4956        just above, so possessive_quantifier is always FALSE for them at this
4957        stage.
4958    
4959        Note that the repeated item starts at tempcode, not at previous, which
4960        might be the first part of a string whose (former) last char we repeated.
4961    
4962      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
4963      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 4765  for (;; ptr++) Line 4988  for (;; ptr++)
4988          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4989          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4990    
4991          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4992          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4993          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4994          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4995    
4996          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4997          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4998          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4999          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
5000    
5001            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
5002            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
5003            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
5004            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
5005    
5006            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
5007            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
5008            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
5009            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
5010    
5011          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
5012          pending recursive references are updated. */          pending recursive references are updated. */
5013    
# Line 4828  for (;; ptr++) Line 5061  for (;; ptr++)
5061        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5062        namelen = (int)(ptr - name);        namelen = (int)(ptr - name);
5063    
5064          /* It appears that Perl allows any characters whatsoever, other than
5065          a closing parenthesis, to appear in arguments, so we no longer insist on
5066          letters, digits, and underscores. */
5067    
5068        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5069          {          {
5070          arg = ++ptr;          arg = ++ptr;
5071          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
           || *ptr == '_') ptr++;  
5072          arglen = (int)(ptr - arg);          arglen = (int)(ptr - arg);
5073          }          }
5074    
# Line 4849  for (;; ptr++) Line 5085  for (;; ptr++)
5085          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5086              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5087            {            {
5088            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5089              ASSERT_ACCEPT if in an assertion. */
5090    
5091            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5092              {              {
5093              open_capitem *oc;              open_capitem *oc;
5094                if (arglen != 0)
5095                  {
5096                  *errorcodeptr = ERR59;
5097                  goto FAILED;
5098                  }
5099              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5100              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5101                {                {
5102                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5103                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5104                }                }
5105                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5106    
5107                /* Do not set firstbyte after *ACCEPT */
5108                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5109              }              }
5110    
5111            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
5112    
5113            if (arglen == 0)            else if (arglen == 0)
5114              {              {
5115              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
5116                {                {
# Line 4872  for (;; ptr++) Line 5118  for (;; ptr++)
5118                goto FAILED;                goto FAILED;
5119                }                }
5120              *code = verbs[i].op;              *code = verbs[i].op;
5121              if (*code++ == OP_THEN)              if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
               {  
               PUT(code, 0, code - bcptr->current_branch - 1);  
               code += LINK_SIZE;  
               }  
5122              }              }
5123    
5124            else            else
# Line 4887  for (;; ptr++) Line 5129  for (;; ptr++)
5129                goto FAILED;                goto FAILED;
5130                }                }
5131              *code = verbs[i].op_arg;              *code = verbs[i].op_arg;
5132              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;  
               }  
5133              *code++ = arglen;              *code++ = arglen;
5134              memcpy(code, arg, arglen);              memcpy(code, arg, arglen);
5135              code += arglen;              code += arglen;
# Line 5153  for (;; ptr++) Line 5391  for (;; ptr++)
5391          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5392          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5393          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5394            cd->assert_depth += 1;
5395          ptr++;          ptr++;
5396          break;          break;
5397    
# Line 5167  for (;; ptr++) Line 5406  for (;; ptr++)
5406            continue;            continue;
5407            }            }
5408          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5409            cd->assert_depth += 1;
5410          break;          break;
5411    
5412    
# Line 5176  for (;; ptr++) Line 5416  for (;; ptr++)
5416            {            {
5417            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5418            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5419              cd->assert_depth += 1;
5420            ptr += 2;            ptr += 2;
5421            break;            break;
5422    
5423            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5424            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5425              cd->assert_depth += 1;
5426            ptr += 2;            ptr += 2;
5427            break;            break;
5428    
# Line 5395  for (;; ptr++) Line 5637  for (;; ptr++)
5637    
5638          if (lengthptr != NULL)          if (lengthptr != NULL)
5639            {            {
5640            const uschar *temp;            const uschar *temp;
5641    
5642            if (namelen == 0)            if (namelen == 0)
5643              {              {
5644              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5412  for (;; ptr++) Line 5654  for (;; ptr++)
5654              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5655              goto FAILED;              goto FAILED;
5656              }              }
5657    
5658            /* The name table does not exist in the first pass, so we cannot            /* The name table does not exist in the first pass, so we cannot
5659            do a simple search as in the code below. Instead, we have to scan the            do a simple search as in the code below. Instead, we have to scan the
5660            pattern to find the number. It is important that we scan it only as            pattern to find the number. It is important that we scan it only as
5661            far as we have got because the syntax of named subpatterns has not            far as we have got because the syntax of named subpatterns has not
5662            been checked for the rest of the pattern, and find_parens() assumes            been checked for the rest of the pattern, and find_parens() assumes
5663            correct syntax. In any case, it's a waste of resources to scan            correct syntax. In any case, it's a waste of resources to scan
5664            further. We stop the scan at the current point by temporarily            further. We stop the scan at the current point by temporarily
5665            adjusting the value of cd->endpattern. */            adjusting the value of cd->endpattern. */
5666    
5667            temp = cd->end_pattern;            temp = cd->end_pattern;
5668            cd->end_pattern = ptr;            cd->end_pattern = ptr;
5669            recno = find_parens(cd, name, namelen,            recno = find_parens(cd, name, namelen,
5670              (options & PCRE_EXTENDED) != 0, utf8);              (options & PCRE_EXTENDED) != 0, utf8);
5671            cd->end_pattern = temp;            cd->end_pattern = temp;
5672            if (recno < 0) recno = 0;    /* Forward ref; set dummy number */            if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5673            }            }
5674    
# Line 5571  for (;; ptr++) Line 5813  for (;; ptr++)
5813    
5814                /* 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
5815                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5816                of the group. */                of the group. Then remember the forward reference. */
5817    
5818                called = cd->start_code + recno;                called = cd->start_code + recno;
5819                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5820                }                }
5821    
5822              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
5823              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
5824              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
5825                must not, however, do this check if we are in a conditional
5826                subpattern because the condition might be testing for recursion in
5827                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
5828                Forever loops are also detected at runtime, so those that occur in
5829                conditional subpatterns will be picked up then. */
5830    
5831              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
5832                       could_be_empty(called, code, bcptr, utf8, cd))                       could_be_empty(called, code, bcptr, utf8, cd))
5833                {                {
5834                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
# Line 5589  for (;; ptr++) Line 5836  for (;; ptr++)
5836                }                }
5837              }              }
5838    
5839            /* 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;  
5840    
5841            *code = OP_RECURSE;            *code = OP_RECURSE;
5842            PUT(code, 1, (int)(called - cd->start_code));            PUT(code, 1, (int)(called - cd->start_code));
5843            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;  
5844            }            }
5845    
5846          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5666  for (;; ptr++) Line 5901  for (;; ptr++)
5901          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
5902          both phases.          both phases.
5903    
5904          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
5905          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. */  
5906    
5907          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5908            {            {
# Line 5679  for (;; ptr++) Line 5913  for (;; ptr++)
5913              }              }
5914            else            else
5915              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
5916              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5917              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5918              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5919              }              }
5920    
5921            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
5922            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). */  
5923    
5924            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
5925            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5728  for (;; ptr++) Line 5955  for (;; ptr++)
5955        skipbytes = 2;        skipbytes = 2;
5956        }        }
5957    
5958      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
5959      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
5960      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
5961      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. */  
5962    
5963      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                      /* For handling repetition */
5964      *code = bravalue;      *code = bravalue;
5965      tempcode = code;      tempcode = code;
5966      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
5967      length_prevgroup = 0;              /* Initialize for pre-compile phase */      tempbracount = cd->bracount;          /* Save value before bracket */
5968        length_prevgroup = 0;                 /* Initialize for pre-compile phase */
5969    
5970      if (!compile_regex(      if (!compile_regex(
5971           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
5972           options & PCRE_IMS,           /* The previous ims option state */           &tempcode,                       /* Where to put code (updated) */
5973           &tempcode,                    /* Where to put code (updated) */           &ptr,                            /* Input pointer (updated) */
5974           &ptr,                         /* Input pointer (updated) */           errorcodeptr,                    /* Where to put an error message */
          errorcodeptr,                 /* Where to put an error message */  
5975           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
5976            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
5977           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
5978           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
5979           &subfirstbyte,                /* For possible first char */           cond_depth +
5980           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
5981           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
5982           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
5983           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
5984             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
5985             (lengthptr == NULL)? NULL :      /* Actual compile phase */
5986               &length_prevgroup              /* Pre-compile phase */
5987           ))           ))
5988        goto FAILED;        goto FAILED;
5989    
5990        /* If this was an atomic group and there are no capturing groups within it,
5991        generate OP_ONCE_NC instead of OP_ONCE. */
5992    
5993        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
5994          *code = OP_ONCE_NC;
5995    
5996        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5997          cd->assert_depth -= 1;
5998    
5999      /* 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
6000      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.
6001      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
6002    
6003      /* 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
6004      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
6005      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
6006      not be available. */      not be available. */
# Line 5829  for (;; ptr++) Line 6065  for (;; ptr++)
6065          goto FAILED;          goto FAILED;
6066          }          }
6067        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
6068        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
6069        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
6070        *code++ = OP_KET;        *code++ = OP_KET;
6071        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5997  for (;; ptr++) Line 6233  for (;; ptr++)
6233          }          }
6234    
6235        /* \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).
6236        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6237    
6238        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6239          {          {
6240            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6241              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6242              {
6243              *errorcodeptr = ERR69;
6244              break;
6245              }
6246          is_recurse = FALSE;          is_recurse = FALSE;
6247          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6248            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
# Line 6021  for (;; ptr++) Line 6262  for (;; ptr++)
6262          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6263          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6264          previous = code;          previous = code;
6265          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6266          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6267          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6268          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
# Line 6129  for (;; ptr++) Line 6370  for (;; ptr++)
6370    
6371      ONE_CHAR:      ONE_CHAR:
6372      previous = code;      previous = code;
6373      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6374      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6375    
6376      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 6158  for (;; ptr++) Line 6399  for (;; ptr++)
6399        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6400        }        }
6401    
6402      /* 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
6403      1 or the matching is caseful. */      1 or the matching is caseful. */
6404    
6405      else      else
# Line 6193  return FALSE; Line 6434  return FALSE;
6434  /* 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
6435  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
6436  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.  
   
6437  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
6438  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
6439  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6440    
6441  Arguments:  Arguments:
6442    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  
6443    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6444    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6445    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
6446    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6447    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6448    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6449      cond_depth     depth of nesting for conditional subpatterns
6450    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6451    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6452    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 6222  Returns:         TRUE on success Line 6458  Returns:         TRUE on success
6458  */  */
6459    
6460  static BOOL  static BOOL
6461  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6462    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6463    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6464    int *lengthptr)    compile_data *cd, int *lengthptr)
6465  {  {
6466  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6467  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 6239  int branchfirstbyte, branchreqbyte; Line 6475  int branchfirstbyte, branchreqbyte;
6475  int length;  int length;
6476  int orig_bracount;  int orig_bracount;
6477  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6478  branch_chain bc;  branch_chain bc;
6479    
6480  bc.outer = bcptr;  bc.outer = bcptr;
# Line 6263  pre-compile phase to find out whether an Line 6498  pre-compile phase to find out whether an
6498    
6499  /* 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
6500  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
6501  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6502    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6503    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6504    
6505  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6506    {    {
# Line 6289  for (;;) Line 6526  for (;;)
6526    
6527    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6528    
   /* 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;  
     }  
   
6529    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6530    
6531    if (lookbehind)    if (lookbehind)
# Line 6312  for (;;) Line 6540  for (;;)
6540    into the length. */    into the length. */
6541    
6542    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6543          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6544            (lengthptr == NULL)? NULL : &length))
6545      {      {
6546      *ptrptr = ptr;      *ptrptr = ptr;
6547      return FALSE;      return FALSE;
6548      }      }
6549    
   /* 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;  
   
6550    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6551    has fewer than the rest. */    has fewer than the rest. */
6552    
# Line 6387  for (;;) Line 6607  for (;;)
6607        {        {
6608        int fixed_length;        int fixed_length;
6609        *code = OP_END;        *code = OP_END;
6610        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6611            FALSE, cd);
6612        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6613        if (fixed_length == -3)        if (fixed_length == -3)
6614          {          {
# Line 6408  for (;;) Line 6629  for (;;)
6629    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
6630    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
6631    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
6632    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. */  
6633    
6634    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6635      {      {
# Line 6454  for (;;) Line 6673  for (;;)
6673        cd->open_caps = cd->open_caps->next;        cd->open_caps = cd->open_caps->next;
6674        }        }
6675    
     /* Reset options if needed. */  
   
     if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)  
       {  
       *code++ = OP_OPT;  
       *code++ = oldims;  
       length += 2;  
       }  
   
6676      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
6677    
6678      cd->bracount = max_bracount;      cd->bracount = max_bracount;
# Line 6522  for (;;) Line 6732  for (;;)
6732  /* 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
6733  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
6734  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
6735  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
6736  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6737    
6738  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.
6739  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 6544  of the more common cases more precisely. Line 6754  of the more common cases more precisely.
6754    
6755  Arguments:  Arguments:
6756    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6757    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
6758                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6759                    the less precise approach                    the less precise approach
# Line 6554  Returns:     TRUE or FALSE Line 6763  Returns:     TRUE or FALSE
6763  */  */
6764    
6765  static BOOL  static BOOL
6766  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6767    unsigned int backref_map)    unsigned int backref_map)
6768  {  {
6769  do {  do {
6770     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6771       options, PCRE_MULTILINE, FALSE);       FALSE);
6772     register int op = *scode;     register int op = *scode;
6773    
6774     /* Non-capturing brackets */     /* Non-capturing brackets */
6775    
6776     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6777           op == OP_SBRA || op == OP_SBRAPOS)
6778       {       {
6779       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6780       }       }
6781    
6782     /* Capturing brackets */     /* Capturing brackets */
6783    
6784     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6785                op == OP_SCBRA || op == OP_SCBRAPOS)
6786       {       {
6787       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6788       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6789       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6790       }       }
6791    
6792     /* Other brackets */     /* Other brackets */
6793    
6794     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC ||
6795                op == OP_COND)
6796       {       {
6797       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6798       }       }
6799    
6800     /* .* 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 6597  do { Line 6809  do {
6809    
6810     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6811    
6812     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;  
6813     code += GET(code, 1);     code += GET(code, 1);
6814     }     }
6815  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6635  is_startline(const uschar *code, unsigne Line 6845  is_startline(const uschar *code, unsigne
6845  {  {
6846  do {  do {
6847     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6848       NULL, 0, FALSE);       FALSE);
6849     register int op = *scode;     register int op = *scode;
6850    
6851     /* 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 6662  do { Line 6872  do {
6872         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6873         break;         break;
6874         }         }
6875       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
6876       op = *scode;       op = *scode;
6877       }       }
6878    
6879     /* Non-capturing brackets */     /* Non-capturing brackets */
6880    
6881     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6882           op == OP_SBRA || op == OP_SBRAPOS)
6883       {       {
6884       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6885       }       }
6886    
6887     /* Capturing brackets */     /* Capturing brackets */
6888    
6889     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6890                op == OP_SCBRA || op == OP_SCBRAPOS)
6891       {       {
6892       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6893       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6684  do { Line 6896  do {
6896    
6897     /* Other brackets */     /* Other brackets */
6898    
6899     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC)
6900       {       {
6901       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6902       }       }
# Line 6699  do { Line 6911  do {
6911    
6912     /* Check for explicit circumflex */     /* Check for explicit circumflex */
6913    
6914     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
6915    
6916     /* Move on to the next alternative */     /* Move on to the next alternative */
6917    
# Line 6725  we return that char, otherwise -1. Line 6937  we return that char, otherwise -1.
6937    
6938  Arguments:  Arguments:
6939    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)  
6940    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
6941    
6942  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
6943  */  */
6944    
6945  static int  static int
6946  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
6947  {  {
6948  register int c = -1;  register int c = -1;
6949  do {  do {
6950     int d;     int d;
6951     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
6952       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
6953       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
6954     register int op = *scode;     register int op = *scode;
6955    
6956     switch(op)     switch(op)
# Line 6747  do { Line 6959  do {
6959       return -1;       return -1;
6960    
6961       case OP_BRA:       case OP_BRA:
6962         case OP_BRAPOS:
6963       case OP_CBRA:       case OP_CBRA:
6964         case OP_SCBRA:
6965         case OP_CBRAPOS:
6966         case OP_SCBRAPOS:
6967       case OP_ASSERT:       case OP_ASSERT:
6968       case OP_ONCE:       case OP_ONCE:
6969         case OP_ONCE_NC:
6970       case OP_COND:       case OP_COND:
6971       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
6972         return -1;         return -1;
6973       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
6974       break;       break;
6975    
6976       case OP_EXACT:       /* Fall through */       case OP_EXACT:
6977       scode += 2;       scode += 2;
6978         /* Fall through */
6979    
6980       case OP_CHAR:       case OP_CHAR:
      case OP_CHARNC:  
6981       case OP_PLUS:       case OP_PLUS:
6982       case OP_MINPLUS:       case OP_MINPLUS:
6983       case OP_POSPLUS:       case OP_POSPLUS:
6984       if (!inassert) return -1;       if (!inassert) return -1;
6985       if (c < 0)       if (c < 0) c = scode[1];
6986         {         else if (c != scode[1]) return -1;
6987         c = scode[1];       break;
6988         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
6989         }       case OP_EXACTI:
6990       else if (c != scode[1]) return -1;       scode += 2;
6991         /* Fall through */
6992    
6993         case OP_CHARI:
6994         case OP_PLUSI:
6995         case OP_MINPLUSI:
6996         case OP_POSPLUSI:
6997         if (!inassert) return -1;
6998         if (c < 0) c = scode[1] | REQ_CASELESS;
6999           else if (c != scode[1]) return -1;
7000       break;       break;
7001       }       }
7002    
# Line 6921  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7147  while (ptr[skipatstart] == CHAR_LEFT_PAR
7147    
7148  utf8 = (options & PCRE_UTF8) != 0;  utf8 = (options & PCRE_UTF8) != 0;
7149    
7150  /* 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
7151    return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7152    release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7153    not used here. */
7154    
7155  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7156  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7157       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7158    {    {
7159    errorcode = ERR44;    errorcode = ERR44;
7160    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6950  if ((options & PCRE_UCP) != 0) Line 7179  if ((options & PCRE_UCP) != 0)
7179    
7180  /* Check validity of \R options. */  /* Check validity of \R options. */
7181    
7182  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
7183         (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
7184    {    {
7185    case 0:    errorcode = ERR56;
7186    case PCRE_BSR_ANYCRLF:    goto PCRE_EARLY_ERROR_RETURN;
   case PCRE_BSR_UNICODE:  
   break;  
   default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;  
7187    }    }
7188    
7189  /* 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 7041  outside can help speed up starting point Line 7268  outside can help speed up starting point
7268  ptr += skipatstart;  ptr += skipatstart;
7269  code = cworkspace;  code = cworkspace;
7270  *code = OP_BRA;  *code = OP_BRA;
7271  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7272    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7273  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7274    
7275  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 7097  field; this time it's used for rememberi Line 7323  field; this time it's used for rememberi
7323  */  */
7324    
7325  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7326    cd->assert_depth = 0;
7327  cd->bracount = 0;  cd->bracount = 0;
7328  cd->names_found = 0;  cd->names_found = 0;
7329  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 7115  of the function here. */ Line 7342  of the function here. */
7342  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7343  code = (uschar *)codestart;  code = (uschar *)codestart;
7344  *code = OP_BRA;  *code = OP_BRA;
7345  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0, 0,
7346    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7347  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7348  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7349  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7350    
7351  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */  if (cd->had_accept) reqbyte = REQ_NONE;   /* Must disable after (*ACCEPT) */
7352    
7353  /* 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. */
7354    
# Line 7182  if (cd->check_lookbehind) Line 7409  if (cd->check_lookbehind)
7409        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7410        int end_op = *be;        int end_op = *be;
7411        *be = OP_END;        *be = OP_END;
7412        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7413            cd);
7414        *be = end_op;        *be = end_op;
7415        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7416        if (fixed_length < 0)        if (fixed_length < 0)
# Line 7221  start with ^. and also when all branches Line 7449  start with ^. and also when all branches
7449    
7450  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7451    {    {
7452    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))  
7453      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7454    else    else
7455      {      {
7456      if (firstbyte < 0)      if (firstbyte < 0)
7457        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7458      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7459        {        {
7460        int ch = firstbyte & 255;        int ch = firstbyte & 255;

Legend:
Removed from v.578  
changed lines
  Added in v.744

  ViewVC Help
Powered by ViewVC 1.1.5