/[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 602 by ph10, Wed May 25 08:29:03 2011 UTC revision 762 by ph10, Tue Nov 22 13:36:51 2011 UTC
# 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 409  static const char error_texts[] = Line 409  static const char error_texts[] =
409    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
410    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with PCRE_UCP support\0"
411    "\\c must be followed by an ASCII character\0"    "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413      /* 70 */
414      "internal error: unknown opcode in find_fixedlength()\0"
415      "\\N is not supported in a class\0"
416      "too many forward references\0"
417    ;    ;
418    
419  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 545  static const unsigned char ebcdic_charta Line 550  static const unsigned char ebcdic_charta
550  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
551    
552  static BOOL  static BOOL
553    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
554      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
555    
556    
# Line 577  return s; Line 582  return s;
582    
583    
584  /*************************************************  /*************************************************
585    *            Check for counted repeat            *
586    *************************************************/
587    
588    /* This function is called when a '{' is encountered in a place where it might
589    start a quantifier. It looks ahead to see if it really is a quantifier or not.
590    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
591    where the ddds are digits.
592    
593    Arguments:
594      p         pointer to the first char after '{'
595    
596    Returns:    TRUE or FALSE
597    */
598    
599    static BOOL
600    is_counted_repeat(const uschar *p)
601    {
602    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
603    while ((digitab[*p] & ctype_digit) != 0) p++;
604    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
605    
606    if (*p++ != CHAR_COMMA) return FALSE;
607    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
608    
609    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
610    while ((digitab[*p] & ctype_digit) != 0) p++;
611    
612    return (*p == CHAR_RIGHT_CURLY_BRACKET);
613    }
614    
615    
616    
617    /*************************************************
618  *            Handle escapes                      *  *            Handle escapes                      *
619  *************************************************/  *************************************************/
620    
# Line 642  else Line 680  else
680    
681      case CHAR_l:      case CHAR_l:
682      case CHAR_L:      case CHAR_L:
683        *errorcodeptr = ERR37;
684        break;
685    
686      case CHAR_u:      case CHAR_u:
687        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
688          {
689          /* In JavaScript, \u must be followed by four hexadecimal numbers.
690          Otherwise it is a lowercase u letter. */
691          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
692               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
693            {
694            c = 0;
695            for (i = 0; i < 4; ++i)
696              {
697              register int cc = *(++ptr);
698    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
699              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
700              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
701    #else           /* EBCDIC coding */
702              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
703              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
704    #endif
705              }
706            }
707          }
708        else
709          *errorcodeptr = ERR37;
710        break;
711    
712      case CHAR_U:      case CHAR_U:
713      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
714        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
715      break;      break;
716    
717      /* \g must be followed by one of a number of specific things:      /* In a character class, \g is just a literal "g". Outside a character
718        class, \g must be followed by one of a number of specific things:
719    
720      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
721      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 664  else Line 732  else
732      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
733    
734      case CHAR_g:      case CHAR_g:
735        if (isclass) break;
736      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
737        {        {
738        c = -ESC_g;        c = -ESC_g;
# Line 792  else Line 861  else
861      treated as a data character. */      treated as a data character. */
862    
863      case CHAR_x:      case CHAR_x:
864        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
865          {
866          /* In JavaScript, \x must be followed by two hexadecimal numbers.
867          Otherwise it is a lowercase x letter. */
868          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
869            {
870            c = 0;
871            for (i = 0; i < 2; ++i)
872              {
873              register int cc = *(++ptr);
874    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
875              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
876              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
877    #else           /* EBCDIC coding */
878              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
879              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
880    #endif
881              }
882            }
883          break;
884          }
885    
886      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
887        {        {
888        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
# Line 885  else Line 976  else
976    }    }
977    
978  /* 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
979  newline". PCRE does not support \N{name}. */  newline". PCRE does not support \N{name}. However, it does support
980    quantification such as \N{2,3}. */
981    
982  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
983         !is_counted_repeat(ptr+2))
984    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
985    
986  /* 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 1090  return -1;
1090    
1091    
1092  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if (*p++ != CHAR_COMMA) return FALSE;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == CHAR_RIGHT_CURLY_BRACKET);  
 }  
   
   
   
 /*************************************************  
1093  *         Read repeat counts                     *  *         Read repeat counts                     *
1094  *************************************************/  *************************************************/
1095    
# Line 1403  does not. Line 1463  does not.
1463    
1464  Arguments:  Arguments:
1465    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1466    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1467    
1468  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1469  */  */
1470    
1471  static const uschar*  static const uschar*
1472  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1473  {  {
1474  for (;;)  for (;;)
1475    {    {
# Line 1468  and doing the check at the end; a flag s Line 1524  and doing the check at the end; a flag s
1524    
1525  Arguments:  Arguments:
1526    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1527    options  the compiling options    utf8     TRUE in UTF-8 mode
1528    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1529    cd       the "compile data" structure    cd       the "compile data" structure
1530    
1531  Returns:   the fixed length,  Returns:   the fixed length,
1532               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1533               or -2 if \C was encountered               or -2 if \C was encountered (in UTF-8 mode only)
1534               or -3 if an OP_RECURSE item was encountered and atend is FALSE               or -3 if an OP_RECURSE item was encountered and atend is FALSE
1535                 or -4 if an unknown opcode was encountered (internal error)
1536  */  */
1537    
1538  static int  static int
1539  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1540  {  {
1541  int length = -1;  int length = -1;
1542    
# Line 1496  for (;;) Line 1553  for (;;)
1553    register int op = *cc;    register int op = *cc;
1554    switch (op)    switch (op)
1555      {      {
1556        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1557        OP_BRA (normal non-capturing bracket) because the other variants of these
1558        opcodes are all concerned with unlimited repeated groups, which of course
1559        are not of fixed length. */
1560    
1561      case OP_CBRA:      case OP_CBRA:
1562      case OP_BRA:      case OP_BRA:
1563      case OP_ONCE:      case OP_ONCE:
1564        case OP_ONCE_NC:
1565      case OP_COND:      case OP_COND:
1566      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1567      if (d < 0) return d;      if (d < 0) return d;
1568      branchlength += d;      branchlength += d;
1569      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1570      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1571      break;      break;
1572    
1573      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested call.
1574      call. If it's ALT it is an alternation in a nested call. If it is      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1575      END it's the end of the outer call. All can be handled by the same code. */      an ALT. If it is END it's the end of the outer call. All can be handled by
1576        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1577        because they all imply an unlimited repeat. */
1578    
1579      case OP_ALT:      case OP_ALT:
1580      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1581      case OP_END:      case OP_END:
1582        case OP_ACCEPT:
1583        case OP_ASSERT_ACCEPT:
1584      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1585        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1586      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1532  for (;;) Line 1597  for (;;)
1597      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1598      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1599      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1600      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1601      if (d < 0) return d;      if (d < 0) return d;
1602      branchlength += d;      branchlength += d;
1603      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1549  for (;;) Line 1614  for (;;)
1614    
1615      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1616    
1617      case OP_REVERSE:      case OP_MARK:
1618      case OP_CREF:      case OP_PRUNE_ARG:
1619      case OP_NCREF:      case OP_SKIP_ARG:
1620      case OP_RREF:      case OP_THEN_ARG:
1621      case OP_NRREF:      cc += cc[1] + _pcre_OP_lengths[*cc];
1622      case OP_DEF:      break;
1623    
1624      case OP_CALLOUT:      case OP_CALLOUT:
     case OP_SOD:  
     case OP_SOM:  
     case OP_SET_SOM:  
     case OP_EOD:  
     case OP_EODN:  
1625      case OP_CIRC:      case OP_CIRC:
1626      case OP_CIRCM:      case OP_CIRCM:
1627        case OP_CLOSE:
1628        case OP_COMMIT:
1629        case OP_CREF:
1630        case OP_DEF:
1631      case OP_DOLL:      case OP_DOLL:
1632      case OP_DOLLM:      case OP_DOLLM:
1633        case OP_EOD:
1634        case OP_EODN:
1635        case OP_FAIL:
1636        case OP_NCREF:
1637        case OP_NRREF:
1638      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1639        case OP_PRUNE:
1640        case OP_REVERSE:
1641        case OP_RREF:
1642        case OP_SET_SOM:
1643        case OP_SKIP:
1644        case OP_SOD:
1645        case OP_SOM:
1646        case OP_THEN:
1647      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1648      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
1649      break;      break;
# Line 1575  for (;;) Line 1653  for (;;)
1653      case OP_CHAR:      case OP_CHAR:
1654      case OP_CHARI:      case OP_CHARI:
1655      case OP_NOT:      case OP_NOT:
1656      case OP_NOTI:      case OP_NOTI:
1657      branchlength++;      branchlength++;
1658      cc += 2;      cc += 2;
1659  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1660      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1661  #endif  #endif
1662      break;      break;
1663    
# Line 1588  for (;;) Line 1665  for (;;)
1665      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1666    
1667      case OP_EXACT:      case OP_EXACT:
1668        case OP_EXACTI:
1669        case OP_NOTEXACT:
1670        case OP_NOTEXACTI:
1671      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1672      cc += 4;      cc += 4;
1673  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1674      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1675  #endif  #endif
1676      break;      break;
1677    
# Line 1609  for (;;) Line 1688  for (;;)
1688      cc += 2;      cc += 2;
1689      /* Fall through */      /* Fall through */
1690    
1691        case OP_HSPACE:
1692        case OP_VSPACE:
1693        case OP_NOT_HSPACE:
1694        case OP_NOT_VSPACE:
1695      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1696      case OP_DIGIT:      case OP_DIGIT:
1697      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1621  for (;;) Line 1704  for (;;)
1704      cc++;      cc++;
1705      break;      break;
1706    
1707      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1708        otherwise \C is coded as OP_ALLANY. */
1709    
1710      case OP_ANYBYTE:      case OP_ANYBYTE:
1711      return -2;      return -2;
# Line 1640  for (;;) Line 1724  for (;;)
1724    
1725      switch (*cc)      switch (*cc)
1726        {        {
1727          case OP_CRPLUS:
1728          case OP_CRMINPLUS:
1729        case OP_CRSTAR:        case OP_CRSTAR:
1730        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1731        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1660  for (;;) Line 1746  for (;;)
1746    
1747      /* Anything else is variable length */      /* Anything else is variable length */
1748    
1749      default:      case OP_ANYNL:
1750        case OP_BRAMINZERO:
1751        case OP_BRAPOS:
1752        case OP_BRAPOSZERO:
1753        case OP_BRAZERO:
1754        case OP_CBRAPOS:
1755        case OP_EXTUNI:
1756        case OP_KETRMAX:
1757        case OP_KETRMIN:
1758        case OP_KETRPOS:
1759        case OP_MINPLUS:
1760        case OP_MINPLUSI:
1761        case OP_MINQUERY:
1762        case OP_MINQUERYI:
1763        case OP_MINSTAR:
1764        case OP_MINSTARI:
1765        case OP_MINUPTO:
1766        case OP_MINUPTOI:
1767        case OP_NOTMINPLUS:
1768        case OP_NOTMINPLUSI:
1769        case OP_NOTMINQUERY:
1770        case OP_NOTMINQUERYI:
1771        case OP_NOTMINSTAR:
1772        case OP_NOTMINSTARI:
1773        case OP_NOTMINUPTO:
1774        case OP_NOTMINUPTOI:
1775        case OP_NOTPLUS:
1776        case OP_NOTPLUSI:
1777        case OP_NOTPOSPLUS:
1778        case OP_NOTPOSPLUSI:
1779        case OP_NOTPOSQUERY:
1780        case OP_NOTPOSQUERYI:
1781        case OP_NOTPOSSTAR:
1782        case OP_NOTPOSSTARI:
1783        case OP_NOTPOSUPTO:
1784        case OP_NOTPOSUPTOI:
1785        case OP_NOTQUERY:
1786        case OP_NOTQUERYI:
1787        case OP_NOTSTAR:
1788        case OP_NOTSTARI:
1789        case OP_NOTUPTO:
1790        case OP_NOTUPTOI:
1791        case OP_PLUS:
1792        case OP_PLUSI:
1793        case OP_POSPLUS:
1794        case OP_POSPLUSI:
1795        case OP_POSQUERY:
1796        case OP_POSQUERYI:
1797        case OP_POSSTAR:
1798        case OP_POSSTARI:
1799        case OP_POSUPTO:
1800        case OP_POSUPTOI:
1801        case OP_QUERY:
1802        case OP_QUERYI:
1803        case OP_REF:
1804        case OP_REFI:
1805        case OP_SBRA:
1806        case OP_SBRAPOS:
1807        case OP_SCBRA:
1808        case OP_SCBRAPOS:
1809        case OP_SCOND:
1810        case OP_SKIPZERO:
1811        case OP_STAR:
1812        case OP_STARI:
1813        case OP_TYPEMINPLUS:
1814        case OP_TYPEMINQUERY:
1815        case OP_TYPEMINSTAR:
1816        case OP_TYPEMINUPTO:
1817        case OP_TYPEPLUS:
1818        case OP_TYPEPOSPLUS:
1819        case OP_TYPEPOSQUERY:
1820        case OP_TYPEPOSSTAR:
1821        case OP_TYPEPOSUPTO:
1822        case OP_TYPEQUERY:
1823        case OP_TYPESTAR:
1824        case OP_TYPEUPTO:
1825        case OP_UPTO:
1826        case OP_UPTOI:
1827      return -1;      return -1;
1828    
1829        /* Catch unrecognized opcodes so that when new ones are added they
1830        are not forgotten, as has happened in the past. */
1831    
1832        default:
1833        return -4;
1834      }      }
1835    }    }
1836  /* Control never gets here */  /* Control never gets here */
# Line 1694  _pcre_find_bracket(const uschar *code, B Line 1863  _pcre_find_bracket(const uschar *code, B
1863  for (;;)  for (;;)
1864    {    {
1865    register int c = *code;    register int c = *code;
1866    
1867    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1868    
1869    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1712  for (;;) Line 1882  for (;;)
1882    
1883    /* Handle capturing bracket */    /* Handle capturing bracket */
1884    
1885    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1886               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1887      {      {
1888      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1889      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1754  for (;;) Line 1925  for (;;)
1925        break;        break;
1926    
1927        case OP_THEN_ARG:        case OP_THEN_ARG:
1928        code += code[1+LINK_SIZE];        code += code[1];
1929        break;        break;
1930        }        }
1931    
# Line 1873  for (;;) Line 2044  for (;;)
2044        break;        break;
2045    
2046        case OP_THEN_ARG:        case OP_THEN_ARG:
2047        code += code[1+LINK_SIZE];        code += code[1];
2048        break;        break;
2049        }        }
2050    
# Line 1954  could_be_empty_branch(const uschar *code Line 2125  could_be_empty_branch(const uschar *code
2125    compile_data *cd)    compile_data *cd)
2126  {  {
2127  register int c;  register int c;
2128  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2129       code < endcode;       code < endcode;
2130       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2131    {    {
2132    const uschar *ccode;    const uschar *ccode;
2133    
# Line 1972  for (code = first_significant_code(code Line 2143  for (code = first_significant_code(code
2143      continue;      continue;
2144      }      }
2145    
   /* 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;  
     }  
   
2146    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
2147    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
2148      forward reference subroutine call, we can't. To detect forward reference
2149      we have to scan up the list that is kept in the workspace. This function is
2150      called only when doing the real compile, not during the pre-compile that
2151      measures the size of the compiled pattern. */
2152    
2153    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2154      {      {
2155      BOOL empty_branch = FALSE;      const uschar *scode;
2156      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
2157    
2158        /* Test for forward reference */
2159    
2160        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2161          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2162    
2163        /* Not a forward reference, test for completed backward reference */
2164    
2165        empty_branch = FALSE;
2166        scode = cd->start_code + GET(code, 1);
2167      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2168    
2169        /* Completed backwards reference */
2170    
2171      do      do
2172        {        {
2173        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf8, cd))
# Line 2000  for (code = first_significant_code(code Line 2178  for (code = first_significant_code(code
2178        scode += GET(scode, 1);        scode += GET(scode, 1);
2179        }        }
2180      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2181    
2182      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2183      continue;      continue;
2184      }      }
2185    
2186      /* Groups with zero repeats can of course be empty; skip them. */
2187    
2188      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2189          c == OP_BRAPOSZERO)
2190        {
2191        code += _pcre_OP_lengths[c];
2192        do code += GET(code, 1); while (*code == OP_ALT);
2193        c = *code;
2194        continue;
2195        }
2196    
2197      /* A nested group that is already marked as "could be empty" can just be
2198      skipped. */
2199    
2200      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2201          c == OP_SCBRA || c == OP_SCBRAPOS)
2202        {
2203        do code += GET(code, 1); while (*code == OP_ALT);
2204        c = *code;
2205        continue;
2206        }
2207    
2208    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2209    
2210    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2211          c == OP_CBRA || c == OP_CBRAPOS ||
2212          c == OP_ONCE || c == OP_ONCE_NC ||
2213          c == OP_COND)
2214      {      {
2215      BOOL empty_branch;      BOOL empty_branch;
2216      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2135  for (code = first_significant_code(code Line 2339  for (code = first_significant_code(code
2339      case OP_KET:      case OP_KET:
2340      case OP_KETRMAX:      case OP_KETRMAX:
2341      case OP_KETRMIN:      case OP_KETRMIN:
2342        case OP_KETRPOS:
2343      case OP_ALT:      case OP_ALT:
2344      return TRUE;      return TRUE;
2345    
# Line 2177  for (code = first_significant_code(code Line 2382  for (code = first_significant_code(code
2382      break;      break;
2383    
2384      case OP_THEN_ARG:      case OP_THEN_ARG:
2385      code += code[1+LINK_SIZE];      code += code[1];
2386      break;      break;
2387    
2388      /* None of the remaining opcodes are required to match a character. */      /* None of the remaining opcodes are required to match a character. */
# Line 2200  return TRUE; Line 2405  return TRUE;
2405  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2406  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2407  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2408    This function is called only during the real compile, not during the
2409    pre-compile.
2410    
2411  Arguments:  Arguments:
2412    code        points to start of the recursion    code        points to start of the recursion
# Line 2250  where Perl recognizes it as the POSIX cl Line 2457  where Perl recognizes it as the POSIX cl
2457  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2458  I think.  I think.
2459    
2460    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2461    It seems that the appearance of a nested POSIX class supersedes an apparent
2462    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2463    a digit.
2464    
2465    In Perl, unescaped square brackets may also appear as part of class names. For
2466    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2467    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2468    seem right at all. PCRE does not allow closing square brackets in POSIX class
2469    names.
2470    
2471  Arguments:  Arguments:
2472    ptr      pointer to the initial [    ptr      pointer to the initial [
2473    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2264  int terminator;          /* Don't combin Line 2482  int terminator;          /* Don't combin
2482  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2483  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2484    {    {
2485    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2486        ptr++;
2487      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2488      else
2489      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2490      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2491        {        {
2492        *endptr = ptr;        *endptr = ptr;
2493        return TRUE;        return TRUE;
2494        }        }
2495        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2496             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2497              ptr[1] == CHAR_EQUALS_SIGN) &&
2498            check_posix_syntax(ptr, endptr))
2499          return FALSE;
2500      }      }
2501    }    }
2502  return FALSE;  return FALSE;
# Line 2682  if (next >= 0) switch(op_code) Line 2907  if (next >= 0) switch(op_code)
2907    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2908    
2909    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2910    opcodes are not used for multi-byte characters, because they are coded using    opcodes are not used for multi-byte characters, because they are coded using
2911    an XCLASS instead. */    an XCLASS instead. */
2912    
2913    case OP_NOT:    case OP_NOT:
2914    return (c = *previous) == next;    return (c = *previous) == next;
2915    
2916    case OP_NOTI:    case OP_NOTI:
2917    if ((c = *previous) == next) return TRUE;    if ((c = *previous) == next) return TRUE;
2918  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2919    if (utf8)    if (utf8)
# Line 2981  Arguments: Line 3206  Arguments:
3206    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3207    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3208    bcptr          points to current branch chain    bcptr          points to current branch chain
3209      cond_depth     conditional nesting depth
3210    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3211    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3212                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2992  Returns:         TRUE on success Line 3218  Returns:         TRUE on success
3218  static BOOL  static BOOL
3219  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3220    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3221    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3222  {  {
3223  int repeat_type, op_type;  int repeat_type, op_type;
3224  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 3001  int greedy_default, greedy_non_default; Line 3227  int greedy_default, greedy_non_default;
3227  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3228  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3229  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3230  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3231  int after_manual_callout = 0;  int after_manual_callout = 0;
3232  int length_prevgroup = 0;  int length_prevgroup = 0;
3233  register int c;  register int c;
# Line 3019  uschar *previous_callout = NULL; Line 3245  uschar *previous_callout = NULL;
3245  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3246  uschar classbits[32];  uschar classbits[32];
3247    
3248    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3249    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3250    dynamically as we process the pattern. */
3251    
3252  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3253  BOOL class_utf8;  BOOL class_utf8;
3254  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 3027  uschar *class_utf8data_base; Line 3257  uschar *class_utf8data_base;
3257  uschar utf8_char[6];  uschar utf8_char[6];
3258  #else  #else
3259  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3260  #endif  #endif
3261    
3262  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 3078  for (;; ptr++) Line 3307  for (;; ptr++)
3307    int subfirstbyte;    int subfirstbyte;
3308    int terminator;    int terminator;
3309    int mclength;    int mclength;
3310      int tempbracount;
3311    uschar mcbuffer[8];    uschar mcbuffer[8];
3312    
3313    /* Get next byte in the pattern */    /* Get next byte in the pattern */
# Line 3125  for (;; ptr++) Line 3355  for (;; ptr++)
3355        }        }
3356    
3357      *lengthptr += (int)(code - last_code);      *lengthptr += (int)(code - last_code);
3358      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, (int)(code - last_code),
3359          c));
3360    
3361      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
3362      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
# Line 3199  for (;; ptr++) Line 3430  for (;; ptr++)
3430      previous_callout = NULL;      previous_callout = NULL;
3431      }      }
3432    
3433    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3434    
3435    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3436      {      {
# Line 3541  for (;; ptr++) Line 3772  for (;; ptr++)
3772          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3773    
3774          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
3775            else if (-c == ESC_N)            /* \N is not supported in a class */
3776              {
3777              *errorcodeptr = ERR71;
3778              goto FAILED;
3779              }
3780          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3781            {            {
3782            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 4168  for (;; ptr++) Line 4404  for (;; ptr++)
4404      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4405      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4406    
4407      /* Save start of previous item, in case we have to move it up to make space      /* Save start of previous item, in case we have to move it up in order to
4408      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4409    
4410      tempcode = previous;      tempcode = previous;
4411    
# Line 4192  for (;; ptr++) Line 4428  for (;; ptr++)
4428        }        }
4429      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4430    
4431        /* If previous was a recursion call, wrap it in atomic brackets so that
4432        previous becomes the atomic group. All recursions were so wrapped in the
4433        past, but it no longer happens for non-repeated recursions. In fact, the
4434        repeated ones could be re-implemented independently so as not to need this,
4435        but for the moment we rely on the code for repeating groups. */
4436    
4437        if (*previous == OP_RECURSE)
4438          {
4439          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4440          *previous = OP_ONCE;
4441          PUT(previous, 1, 2 + 2*LINK_SIZE);
4442          previous[2 + 2*LINK_SIZE] = OP_KET;
4443          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4444          code += 2 + 2 * LINK_SIZE;
4445          length_prevgroup = 3 + 3*LINK_SIZE;
4446    
4447          /* When actually compiling, we need to check whether this was a forward
4448          reference, and if so, adjust the offset. */
4449    
4450          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4451            {
4452            int offset = GET(cd->hwm, -LINK_SIZE);
4453            if (offset == previous + 1 - cd->start_code)
4454              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4455            }
4456          }
4457    
4458        /* Now handle repetition for the different types of item. */
4459    
4460      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4461      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
4462      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
# Line 4201  for (;; ptr++) Line 4466  for (;; ptr++)
4466      if (*previous == OP_CHAR || *previous == OP_CHARI)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4467        {        {
4468        op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;        op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4469    
4470        /* Deal with UTF-8 characters that take up more than one byte. It's        /* Deal with UTF-8 characters that take up more than one byte. It's
4471        easier to write this out separately than try to macrify it. Use c to        easier to write this out separately than try to macrify it. Use c to
4472        hold the length of the character in bytes, plus 0x80 to flag that it's a        hold the length of the character in bytes, plus 0x80 to flag that it's a
# Line 4246  for (;; ptr++) Line 4511  for (;; ptr++)
4511      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4512      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4513      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4514      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4515      are currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4516    
4517      else if (*previous == OP_NOT || *previous == OP_NOTI)      else if (*previous == OP_NOT || *previous == OP_NOTI)
# Line 4483  for (;; ptr++) Line 4748  for (;; ptr++)
4748        }        }
4749    
4750      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
4751      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4752        opcodes such as BRA and CBRA, as this is the place where they get converted
4753        into the more special varieties such as BRAPOS and SBRA. A test for >=
4754        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4755        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4756        repetition of assertions, but now it does, for Perl compatibility. */
4757    
4758      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4759        {        {
4760        register int i;        register int i;
       int ketoffset = 0;  
4761        int len = (int)(code - previous);        int len = (int)(code - previous);
4762        uschar *bralink = NULL;        uschar *bralink = NULL;
4763          uschar *brazeroptr = NULL;
4764    
4765        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4766          we just ignore the repeat. */
4767    
4768        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4769          {          goto END_REPEAT;
4770          *errorcodeptr = ERR55;  
4771          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4772          }        use of repetition is in cases when the assertion is optional. Therefore,
4773          if the minimum is greater than zero, just ignore the repeat. If the
4774          maximum is not not zero or one, set it to 1. */
4775    
4776        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4777        by scanning through from the start, and compute the offset back to it          {
4778        from the current code pointer. */          if (repeat_min > 0) goto END_REPEAT;
4779            if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = (int)(code - ket);  
4780          }          }
4781    
4782        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 4530  for (;; ptr++) Line 4797  for (;; ptr++)
4797          **   goto END_REPEAT;          **   goto END_REPEAT;
4798          **   }          **   }
4799    
4800          However, that fails when a group is referenced as a subroutine from          However, that fails when a group or a subgroup within it is referenced
4801          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it          as a subroutine from elsewhere in the pattern, so now we stick in
4802          so that it is skipped on execution. As we don't have a list of which          OP_SKIPZERO in front of it so that it is skipped on execution. As we
4803          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4804            selectively.
4805    
4806          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4807          and do no more at this point. However, we do need to adjust any          and do no more at this point. However, we do need to adjust any
# Line 4553  for (;; ptr++) Line 4821  for (;; ptr++)
4821              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4822              goto END_REPEAT;              goto END_REPEAT;
4823              }              }
4824              brazeroptr = previous;    /* Save for possessive optimizing */
4825            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4826            }            }
4827    
# Line 4615  for (;; ptr++) Line 4884  for (;; ptr++)
4884              *lengthptr += delta;              *lengthptr += delta;
4885              }              }
4886    
4887            /* This is compiling for real */            /* This is compiling for real. If there is a set first byte for
4888              the group, and we have not yet set a "required byte", set it. */
4889    
4890            else            else
4891              {              {
# Line 4627  for (;; ptr++) Line 4897  for (;; ptr++)
4897                memcpy(code, previous, len);                memcpy(code, previous, len);
4898                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
4899                  {                  {
4900                    if (cd->hwm >= cd->start_workspace + WORK_SIZE_CHECK)
4901                      {
4902                      *errorcodeptr = ERR72;
4903                      goto FAILED;
4904                      }
4905                  PUT(cd->hwm, 0, GET(hc, 0) + len);                  PUT(cd->hwm, 0, GET(hc, 0) + len);
4906                  cd->hwm += LINK_SIZE;                  cd->hwm += LINK_SIZE;
4907                  }                  }
# Line 4654  for (;; ptr++) Line 4929  for (;; ptr++)
4929          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4930          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4931          a 64-bit integer type when available, otherwise double. */          a 64-bit integer type when available, otherwise double. */
4932    
4933          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4934            {            {
4935            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
# Line 4694  for (;; ptr++) Line 4969  for (;; ptr++)
4969            memcpy(code, previous, len);            memcpy(code, previous, len);
4970            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
4971              {              {
4972                if (cd->hwm >= cd->start_workspace + WORK_SIZE_CHECK)
4973                  {
4974                  *errorcodeptr = ERR72;
4975                  goto FAILED;
4976                  }
4977              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
4978              cd->hwm += LINK_SIZE;              cd->hwm += LINK_SIZE;
4979              }              }
# Line 4717  for (;; ptr++) Line 4997  for (;; ptr++)
4997            }            }
4998          }          }
4999    
5000        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. For
5001        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
5002        don't know if there's been an options resetting after the ket. The        ONCE brackets can be converted into non-capturing brackets, as the
5003        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
5004          deal with possessive ONCEs specially.
5005        Then, when we are doing the actual compile phase, check to see whether  
5006        this group is a non-atomic one that could match an empty string. If so,        Otherwise, when we are doing the actual compile phase, check to see
5007          whether this group is one that could match an empty string. If so,
5008        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
5009        that runtime checking can be done. [This check is also applied to        that runtime checking can be done. [This check is also applied to ONCE
5010        atomic groups at runtime, but in a different way.] */        groups at runtime, but in a different way.]
5011    
5012          Then, if the quantifier was possessive and the bracket is not a
5013          conditional, we convert the BRA code to the POS form, and the KET code to
5014          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
5015          subpattern at both the start and at the end.) The use of special opcodes
5016          makes it possible to reduce greatly the stack usage in pcre_exec(). If
5017          the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
5018    
5019          Then, if the minimum number of matches is 1 or 0, cancel the possessive
5020          flag so that the default action below, of wrapping everything inside
5021          atomic brackets, does not happen. When the minimum is greater than 1,
5022          there will be earlier copies of the group, and so we still have to wrap
5023          the whole thing. */
5024    
5025        else        else
5026          {          {
5027          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
5028          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
5029          *ketcode = OP_KETRMAX + repeat_type;  
5030          if (lengthptr == NULL && *bracode != OP_ONCE)          /* Convert possessive ONCE brackets to non-capturing */
5031    
5032            if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5033                possessive_quantifier) *bracode = OP_BRA;
5034    
5035            /* For non-possessive ONCE brackets, all we need to do is to
5036            set the KET. */
5037    
5038            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5039              *ketcode = OP_KETRMAX + repeat_type;
5040    
5041            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5042            converted to non-capturing above). */
5043    
5044            else
5045            {            {
5046            uschar *scode = bracode;            /* In the compile phase, check for empty string matching. */
5047            do  
5048              if (lengthptr == NULL)
5049              {              {
5050              if (could_be_empty_branch(scode, ketcode, utf8, cd))              uschar *scode = bracode;
5051                do
5052                {                {
5053                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
5054                break;                  {
5055                    *bracode += OP_SBRA - OP_BRA;
5056                    break;
5057                    }
5058                  scode += GET(scode, 1);
5059                }                }
5060              scode += GET(scode, 1);              while (*scode == OP_ALT);
5061              }              }
5062            while (*scode == OP_ALT);  
5063              /* Handle possessive quantifiers. */
5064    
5065              if (possessive_quantifier)
5066                {
5067                /* For COND brackets, we wrap the whole thing in a possessively
5068                repeated non-capturing bracket, because we have not invented POS
5069                versions of the COND opcodes. Because we are moving code along, we
5070                must ensure that any pending recursive references are updated. */
5071    
5072                if (*bracode == OP_COND || *bracode == OP_SCOND)
5073                  {
5074                  int nlen = (int)(code - bracode);
5075                  *code = OP_END;
5076                  adjust_recurse(bracode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5077                  memmove(bracode + 1+LINK_SIZE, bracode, nlen);
5078                  code += 1 + LINK_SIZE;
5079                  nlen += 1 + LINK_SIZE;
5080                  *bracode = OP_BRAPOS;
5081                  *code++ = OP_KETRPOS;
5082                  PUTINC(code, 0, nlen);
5083                  PUT(bracode, 1, nlen);
5084                  }
5085    
5086                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5087    
5088                else
5089                  {
5090                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5091                  *ketcode = OP_KETRPOS;
5092                  }
5093    
5094                /* If the minimum is zero, mark it as possessive, then unset the
5095                possessive flag when the minimum is 0 or 1. */
5096    
5097                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5098                if (repeat_min < 2) possessive_quantifier = FALSE;
5099                }
5100    
5101              /* Non-possessive quantifier */
5102    
5103              else *ketcode = OP_KETRMAX + repeat_type;
5104            }            }
5105          }          }
5106        }        }
# Line 4766  for (;; ptr++) Line 5121  for (;; ptr++)
5121        }        }
5122    
5123      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
5124      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
5125      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
5126      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
5127      The '+' notation is just syntactic sugar, taken from Sun's Java package,      notation is just syntactic sugar, taken from Sun's Java package, but the
5128      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
5129      tempcode, not at previous, which might be the first part of a string whose  
5130      (former) last char we repeated.      Some (but not all) possessively repeated subpatterns have already been
5131        completely handled in the code just above. For them, possessive_quantifier
5132        is always FALSE at this stage.
5133    
5134        Note that the repeated item starts at tempcode, not at previous, which
5135        might be the first part of a string whose (former) last char we repeated.
5136    
5137      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
5138      an 'upto' may follow. We skip over an 'exact' item, and then test the      an 'upto' may follow. We skip over an 'exact' item, and then test the
# Line 4876  for (;; ptr++) Line 5236  for (;; ptr++)
5236        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5237        namelen = (int)(ptr - name);        namelen = (int)(ptr - name);
5238    
5239          /* It appears that Perl allows any characters whatsoever, other than
5240          a closing parenthesis, to appear in arguments, so we no longer insist on
5241          letters, digits, and underscores. */
5242    
5243        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5244          {          {
5245          arg = ++ptr;          arg = ++ptr;
5246          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
           || *ptr == '_') ptr++;  
5247          arglen = (int)(ptr - arg);          arglen = (int)(ptr - arg);
5248          }          }
5249    
# Line 4897  for (;; ptr++) Line 5260  for (;; ptr++)
5260          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5261              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5262            {            {
5263            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5264              ASSERT_ACCEPT if in an assertion. */
5265    
5266            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5267              {              {
5268              open_capitem *oc;              open_capitem *oc;
5269                if (arglen != 0)
5270                  {
5271                  *errorcodeptr = ERR59;
5272                  goto FAILED;
5273                  }
5274              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5275              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5276                {                {
5277                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5278                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5279                }                }
5280                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5281    
5282                /* Do not set firstbyte after *ACCEPT */
5283                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5284              }              }
5285    
5286            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
5287    
5288            if (arglen == 0)            else if (arglen == 0)
5289              {              {
5290              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
5291                {                {
# Line 4920  for (;; ptr++) Line 5293  for (;; ptr++)
5293                goto FAILED;                goto FAILED;
5294                }                }
5295              *code = verbs[i].op;              *code = verbs[i].op;
5296              if (*code++ == OP_THEN)              if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
               {  
               PUT(code, 0, code - bcptr->current_branch - 1);  
               code += LINK_SIZE;  
               }  
5297              }              }
5298    
5299            else            else
# Line 4935  for (;; ptr++) Line 5304  for (;; ptr++)
5304                goto FAILED;                goto FAILED;
5305                }                }
5306              *code = verbs[i].op_arg;              *code = verbs[i].op_arg;
5307              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;  
               }  
5308              *code++ = arglen;              *code++ = arglen;
5309              memcpy(code, arg, arglen);              memcpy(code, arg, arglen);
5310              code += arglen;              code += arglen;
# Line 5201  for (;; ptr++) Line 5566  for (;; ptr++)
5566          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5567          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5568          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5569            cd->assert_depth += 1;
5570          ptr++;          ptr++;
5571          break;          break;
5572    
# Line 5215  for (;; ptr++) Line 5581  for (;; ptr++)
5581            continue;            continue;
5582            }            }
5583          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5584            cd->assert_depth += 1;
5585          break;          break;
5586    
5587    
# Line 5224  for (;; ptr++) Line 5591  for (;; ptr++)
5591            {            {
5592            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5593            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5594              cd->assert_depth += 1;
5595            ptr += 2;            ptr += 2;
5596            break;            break;
5597    
5598            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5599            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5600              cd->assert_depth += 1;
5601            ptr += 2;            ptr += 2;
5602            break;            break;
5603    
# Line 5250  for (;; ptr++) Line 5619  for (;; ptr++)
5619    
5620          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5621          case CHAR_C:                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
5622          previous_callout = code;  /* Save for later completion */          previous_callout = code;     /* Save for later completion */
5623          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1;    /* Skip one item before completing */
5624          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
5625            {            {
5626            int n = 0;            int n = 0;
# Line 5619  for (;; ptr++) Line 5988  for (;; ptr++)
5988    
5989                /* Fudge the value of "called" so that when it is inserted as an                /* Fudge the value of "called" so that when it is inserted as an
5990                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5991                of the group. */                of the group. Then remember the forward reference. */
5992    
5993                called = cd->start_code + recno;                called = cd->start_code + recno;
5994                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));                if (cd->hwm >= cd->start_workspace + WORK_SIZE_CHECK)
5995                    {
5996                    *errorcodeptr = ERR72;
5997                    goto FAILED;
5998                    }
5999                  PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
6000                }                }
6001    
6002              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
6003              this is a recursive call. We check to see if this is a left              this is a recursive call. We check to see if this is a left
6004              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
6005                must not, however, do this check if we are in a conditional
6006                subpattern because the condition might be testing for recursion in
6007                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
6008                Forever loops are also detected at runtime, so those that occur in
6009                conditional subpatterns will be picked up then. */
6010    
6011              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
6012                       could_be_empty(called, code, bcptr, utf8, cd))                       could_be_empty(called, code, bcptr, utf8, cd))
6013                {                {
6014                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
# Line 5637  for (;; ptr++) Line 6016  for (;; ptr++)
6016                }                }
6017              }              }
6018    
6019            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. It does not have a set first
6020            "once" brackets. Set up a "previous group" length so that a            byte (relevant if it is repeated, because it will then be wrapped
6021            subsequent quantifier will work. */            with ONCE brackets). */
   
           *code = OP_ONCE;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
6022    
6023            *code = OP_RECURSE;            *code = OP_RECURSE;
6024            PUT(code, 1, (int)(called - cd->start_code));            PUT(code, 1, (int)(called - cd->start_code));
6025            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
6026              groupsetfirstbyte = FALSE;
           *code = OP_KET;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
   
           length_prevgroup = 3 + 3*LINK_SIZE;  
6027            }            }
6028    
6029          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5714  for (;; ptr++) Line 6084  for (;; ptr++)
6084          is necessary to ensure we correctly detect the start of the pattern in          is necessary to ensure we correctly detect the start of the pattern in
6085          both phases.          both phases.
6086    
6087          If we are not at the pattern start, compile code to change the ims          If we are not at the pattern start, reset the greedy defaults and the
6088          options if this setting actually changes any of them, and reset the          case value for firstbyte and reqbyte. */
         greedy defaults and the case value for firstbyte and reqbyte. */  
6089    
6090          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
6091            {            {
# Line 5733  for (;; ptr++) Line 6102  for (;; ptr++)
6102              }              }
6103    
6104            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
6105            in subsequent branches. When not at the start of the pattern, this            in subsequent branches. */
           information is also necessary so that a resetting item can be  
           compiled at the end of a group (if we are in a group). */  
6106    
6107            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
6108            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5771  for (;; ptr++) Line 6138  for (;; ptr++)
6138        skipbytes = 2;        skipbytes = 2;
6139        }        }
6140    
6141      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
6142      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a      but this was changed for Perl compatibility, so all kinds can now be
6143      non-register variable in order to be able to pass its address because some      repeated. We copy code into a non-register variable (tempcode) in order to
6144      compilers complain otherwise. Pass in a new setting for the ims options if      be able to pass its address because some compilers complain otherwise. */
     they have changed. */  
6145    
6146      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                      /* For handling repetition */
6147      *code = bravalue;      *code = bravalue;
6148      tempcode = code;      tempcode = code;
6149      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
6150      length_prevgroup = 0;              /* Initialize for pre-compile phase */      tempbracount = cd->bracount;          /* Save value before bracket */
6151        length_prevgroup = 0;                 /* Initialize for pre-compile phase */
6152    
6153      if (!compile_regex(      if (!compile_regex(
6154           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
6155           options & PCRE_IMS,           /* The previous ims option state */           &tempcode,                       /* Where to put code (updated) */
6156           &tempcode,                    /* Where to put code (updated) */           &ptr,                            /* Input pointer (updated) */
6157           &ptr,                         /* Input pointer (updated) */           errorcodeptr,                    /* Where to put an error message */
          errorcodeptr,                 /* Where to put an error message */  
6158           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
6159            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
6160           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
6161           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
6162           &subfirstbyte,                /* For possible first char */           cond_depth +
6163           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
6164           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
6165           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
6166           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
6167             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
6168             (lengthptr == NULL)? NULL :      /* Actual compile phase */
6169               &length_prevgroup              /* Pre-compile phase */
6170           ))           ))
6171        goto FAILED;        goto FAILED;
6172    
6173        /* If this was an atomic group and there are no capturing groups within it,
6174        generate OP_ONCE_NC instead of OP_ONCE. */
6175    
6176        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
6177          *code = OP_ONCE_NC;
6178    
6179        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
6180          cd->assert_depth -= 1;
6181    
6182      /* At the end of compiling, code is still pointing to the start of the      /* At the end of compiling, code is still pointing to the start of the
6183      group, while tempcode has been updated to point past the end of the group      group, while tempcode has been updated to point past the end of the group.
6184      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
6185    
6186      /* If this is a conditional bracket, check that there are no more than      If this is a conditional bracket, check that there are no more than
6187      two branches in the group, or just one if it's a DEFINE group. We do this      two branches in the group, or just one if it's a DEFINE group. We do this
6188      in the real compile phase, not in the pre-pass, where the whole group may      in the real compile phase, not in the pre-pass, where the whole group may
6189      not be available. */      not be available. */
# Line 5872  for (;; ptr++) Line 6248  for (;; ptr++)
6248          goto FAILED;          goto FAILED;
6249          }          }
6250        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
6251        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
6252        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
6253        *code++ = OP_KET;        *code++ = OP_KET;
6254        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 6040  for (;; ptr++) Line 6416  for (;; ptr++)
6416          }          }
6417    
6418        /* \k<name> or \k'name' is a back reference by name (Perl syntax).        /* \k<name> or \k'name' is a back reference by name (Perl syntax).
6419        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6420    
6421        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6422          {          {
6423            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6424              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6425              {
6426              *errorcodeptr = ERR69;
6427              break;
6428              }
6429          is_recurse = FALSE;          is_recurse = FALSE;
6430          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6431            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
# Line 6124  for (;; ptr++) Line 6505  for (;; ptr++)
6505            }            }
6506          else          else
6507  #endif  #endif
6508            {          /* In non-UTF-8 mode, we turn \C into OP_ALLANY instead of OP_ANYBYTE
6509            so that it works in DFA mode and in lookbehinds. */
6510    
6511              {
6512            previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;            previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6513            *code++ = -c;            *code++ = (!utf8 && c == -ESC_C)? OP_ALLANY : -c;
6514            }            }
6515          }          }
6516        continue;        continue;
# Line 6201  for (;; ptr++) Line 6585  for (;; ptr++)
6585        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6586        }        }
6587    
6588      /* firstbyte was previously set; we can set reqbyte only the length is      /* firstbyte was previously set; we can set reqbyte only if the length is
6589      1 or the matching is caseful. */      1 or the matching is caseful. */
6590    
6591      else      else
# Line 6242  value of lengthptr distinguishes the two Line 6626  value of lengthptr distinguishes the two
6626    
6627  Arguments:  Arguments:
6628    options        option bits, including any changes for this subpattern    options        option bits, including any changes for this subpattern
   oldims         previous settings of ims option bits  
6629    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6630    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6631    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
6632    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6633    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6634    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6635      cond_depth     depth of nesting for conditional subpatterns
6636    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6637    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6638    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 6260  Returns:         TRUE on success Line 6644  Returns:         TRUE on success
6644  */  */
6645    
6646  static BOOL  static BOOL
6647  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6648    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6649    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6650    int *lengthptr)    compile_data *cd, int *lengthptr)
6651  {  {
6652  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6653  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 6277  int branchfirstbyte, branchreqbyte; Line 6661  int branchfirstbyte, branchreqbyte;
6661  int length;  int length;
6662  int orig_bracount;  int orig_bracount;
6663  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6664  branch_chain bc;  branch_chain bc;
6665    
6666  bc.outer = bcptr;  bc.outer = bcptr;
# Line 6301  pre-compile phase to find out whether an Line 6684  pre-compile phase to find out whether an
6684    
6685  /* If this is a capturing subpattern, add to the chain of open capturing items  /* If this is a capturing subpattern, add to the chain of open capturing items
6686  so that we can detect them if (*ACCEPT) is encountered. This is also used to  so that we can detect them if (*ACCEPT) is encountered. This is also used to
6687  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6688    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6689    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6690    
6691  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6692    {    {
# Line 6341  for (;;) Line 6726  for (;;)
6726    into the length. */    into the length. */
6727    
6728    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6729          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6730            (lengthptr == NULL)? NULL : &length))
6731      {      {
6732      *ptrptr = ptr;      *ptrptr = ptr;
6733      return FALSE;      return FALSE;
6734      }      }
6735    
   /* If the external options have changed during this branch, it means that we  
   are at the top level, and a leading option setting has been encountered. We  
   need to re-set the original option values to take account of this so that,  
   during the pre-compile phase, we know to allow for a re-set at the start of  
   subsequent branches. */  
   
   if (old_external_options != cd->external_options)  
     oldims = cd->external_options & PCRE_IMS;  
   
6736    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6737    has fewer than the rest. */    has fewer than the rest. */
6738    
# Line 6416  for (;;) Line 6793  for (;;)
6793        {        {
6794        int fixed_length;        int fixed_length;
6795        *code = OP_END;        *code = OP_END;
6796        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6797            FALSE, cd);
6798        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6799        if (fixed_length == -3)        if (fixed_length == -3)
6800          {          {
# Line 6424  for (;;) Line 6802  for (;;)
6802          }          }
6803        else if (fixed_length < 0)        else if (fixed_length < 0)
6804          {          {
6805          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 :
6806                            (fixed_length == -4)? ERR70: ERR25;
6807          *ptrptr = ptr;          *ptrptr = ptr;
6808          return FALSE;          return FALSE;
6809          }          }
# Line 6437  for (;;) Line 6816  for (;;)
6816    of offsets, with the field in the BRA item now becoming an offset to the    of offsets, with the field in the BRA item now becoming an offset to the
6817    first alternative. If there are no alternatives, it points to the end of the    first alternative. If there are no alternatives, it points to the end of the
6818    group. The length in the terminating ket is always the length of the whole    group. The length in the terminating ket is always the length of the whole
6819    bracketed item. If any of the ims options were changed inside the group,    bracketed item. Return leaving the pointer at the terminating char. */
   compile a resetting op-code following, except at the very end of the pattern.  
   Return leaving the pointer at the terminating char. */  
6820    
6821    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6822      {      {
# Line 6564  of the more common cases more precisely. Line 6941  of the more common cases more precisely.
6941    
6942  Arguments:  Arguments:
6943    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6944    bracket_map    a bitmap of which brackets we are inside while testing; this    bracket_map    a bitmap of which brackets we are inside while testing; this
6945                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6946                    the less precise approach                    the less precise approach
# Line 6574  Returns:     TRUE or FALSE Line 6950  Returns:     TRUE or FALSE
6950  */  */
6951    
6952  static BOOL  static BOOL
6953  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6954    unsigned int backref_map)    unsigned int backref_map)
6955  {  {
6956  do {  do {
6957     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6958       options, PCRE_MULTILINE, FALSE);       FALSE);
6959     register int op = *scode;     register int op = *scode;
6960    
6961     /* Non-capturing brackets */     /* Non-capturing brackets */
6962    
6963     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6964           op == OP_SBRA || op == OP_SBRAPOS)
6965       {       {
6966       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6967       }       }
6968    
6969     /* Capturing brackets */     /* Capturing brackets */
6970    
6971     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6972                op == OP_SCBRA || op == OP_SCBRAPOS)
6973       {       {
6974       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6975       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6976       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6977       }       }
6978    
6979     /* Other brackets */     /* Other brackets */
6980    
6981     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC ||
6982                op == OP_COND)
6983       {       {
6984       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6985       }       }
6986    
6987     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
# Line 6653  is_startline(const uschar *code, unsigne Line 7032  is_startline(const uschar *code, unsigne
7032  {  {
7033  do {  do {
7034     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
7035       NULL, 0, FALSE);       FALSE);
7036     register int op = *scode;     register int op = *scode;
7037    
7038     /* If we are at the start of a conditional assertion group, *both* the     /* If we are at the start of a conditional assertion group, *both* the
# Line 6680  do { Line 7059  do {
7059         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
7060         break;         break;
7061         }         }
7062       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
7063       op = *scode;       op = *scode;
7064       }       }
7065    
7066     /* Non-capturing brackets */     /* Non-capturing brackets */
7067    
7068     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
7069           op == OP_SBRA || op == OP_SBRAPOS)
7070       {       {
7071       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
7072       }       }
7073    
7074     /* Capturing brackets */     /* Capturing brackets */
7075    
7076     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
7077                op == OP_SCBRA || op == OP_SCBRAPOS)
7078       {       {
7079       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
7080       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6702  do { Line 7083  do {
7083    
7084     /* Other brackets */     /* Other brackets */
7085    
7086     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC)
7087       {       {
7088       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
7089       }       }
# Line 6743  we return that char, otherwise -1. Line 7124  we return that char, otherwise -1.
7124    
7125  Arguments:  Arguments:
7126    code       points to start of expression (the bracket)    code       points to start of expression (the bracket)
   options    pointer to the options (used to check casing changes)  
7127    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
7128    
7129  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
7130  */  */
7131    
7132  static int  static int
7133  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
7134  {  {
7135  register int c = -1;  register int c = -1;
7136  do {  do {
7137     int d;     int d;
7138     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
7139       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
7140       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
7141     register int op = *scode;     register int op = *scode;
7142    
7143     switch(op)     switch(op)
# Line 6765  do { Line 7146  do {
7146       return -1;       return -1;
7147    
7148       case OP_BRA:       case OP_BRA:
7149         case OP_BRAPOS:
7150       case OP_CBRA:       case OP_CBRA:
7151         case OP_SCBRA:
7152         case OP_CBRAPOS:
7153         case OP_SCBRAPOS:
7154       case OP_ASSERT:       case OP_ASSERT:
7155       case OP_ONCE:       case OP_ONCE:
7156         case OP_ONCE_NC:
7157       case OP_COND:       case OP_COND:
7158       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
7159         return -1;         return -1;
7160       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
7161       break;       break;
7162    
7163       case OP_EXACT:       /* Fall through */       case OP_EXACT:
7164       scode += 2;       scode += 2;
7165         /* Fall through */
7166    
7167       case OP_CHAR:       case OP_CHAR:
      case OP_CHARI:  
7168       case OP_PLUS:       case OP_PLUS:
7169       case OP_MINPLUS:       case OP_MINPLUS:
7170       case OP_POSPLUS:       case OP_POSPLUS:
7171       if (!inassert) return -1;       if (!inassert) return -1;
7172       if (c < 0)       if (c < 0) c = scode[1];
7173         {         else if (c != scode[1]) return -1;
7174         c = scode[1];       break;
7175         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
7176         }       case OP_EXACTI:
7177       else if (c != scode[1]) return -1;       scode += 2;
7178         /* Fall through */
7179    
7180         case OP_CHARI:
7181         case OP_PLUSI:
7182         case OP_MINPLUSI:
7183         case OP_POSPLUSI:
7184         if (!inassert) return -1;
7185         if (c < 0) c = scode[1] | REQ_CASELESS;
7186           else if (c != scode[1]) return -1;
7187       break;       break;
7188       }       }
7189    
# Line 6939  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7334  while (ptr[skipatstart] == CHAR_LEFT_PAR
7334    
7335  utf8 = (options & PCRE_UTF8) != 0;  utf8 = (options & PCRE_UTF8) != 0;
7336    
7337  /* Can't support UTF8 unless PCRE has been compiled to include the code. The  /* Can't support UTF8 unless PCRE has been compiled to include the code. The
7338  return of an error code from _pcre_valid_utf8() is a new feature, introduced in  return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7339  release 8.13. The only use we make of it here is to adjust the offset value to  release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7340  the end of the string for a short string error, for compatibility with previous  not used here. */
 versions. */  
7341    
7342  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7343  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7344       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1, &errorcode)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7345    {    {
7346    errorcode = ERR44;    errorcode = ERR44;
7347    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6972  if ((options & PCRE_UCP) != 0) Line 7366  if ((options & PCRE_UCP) != 0)
7366    
7367  /* Check validity of \R options. */  /* Check validity of \R options. */
7368    
7369  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
7370         (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
7371    {    {
7372    case 0:    errorcode = ERR56;
7373    case PCRE_BSR_ANYCRLF:    goto PCRE_EARLY_ERROR_RETURN;
   case PCRE_BSR_UNICODE:  
   break;  
   default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;  
7374    }    }
7375    
7376  /* 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 7063  outside can help speed up starting point Line 7455  outside can help speed up starting point
7455  ptr += skipatstart;  ptr += skipatstart;
7456  code = cworkspace;  code = cworkspace;
7457  *code = OP_BRA;  *code = OP_BRA;
7458  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7459    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7460  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7461    
7462  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 7119  field; this time it's used for rememberi Line 7510  field; this time it's used for rememberi
7510  */  */
7511    
7512  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7513    cd->assert_depth = 0;
7514  cd->bracount = 0;  cd->bracount = 0;
7515  cd->names_found = 0;  cd->names_found = 0;
7516  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 7137  of the function here. */ Line 7529  of the function here. */
7529  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7530  code = (uschar *)codestart;  code = (uschar *)codestart;
7531  *code = OP_BRA;  *code = OP_BRA;
7532  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0, 0,
7533    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7534  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7535  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7536  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7537    
7538  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */  if (cd->had_accept) reqbyte = REQ_NONE;   /* Must disable after (*ACCEPT) */
7539    
7540  /* 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. */
7541    
# Line 7204  if (cd->check_lookbehind) Line 7596  if (cd->check_lookbehind)
7596        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7597        int end_op = *be;        int end_op = *be;
7598        *be = OP_END;        *be = OP_END;
7599        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7600            cd);
7601        *be = end_op;        *be = end_op;
7602        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7603        if (fixed_length < 0)        if (fixed_length < 0)
7604          {          {
7605          errorcode = (fixed_length == -2)? ERR36 : ERR25;          errorcode = (fixed_length == -2)? ERR36 :
7606                        (fixed_length == -4)? ERR70 : ERR25;
7607          break;          break;
7608          }          }
7609        PUT(cc, 1, fixed_length);        PUT(cc, 1, fixed_length);
# Line 7243  start with ^. and also when all branches Line 7637  start with ^. and also when all branches
7637    
7638  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7639    {    {
7640    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))  
7641      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7642    else    else
7643      {      {
7644      if (firstbyte < 0)      if (firstbyte < 0)
7645        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7646      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7647        {        {
7648        int ch = firstbyte & 255;        int ch = firstbyte & 255;

Legend:
Removed from v.602  
changed lines
  Added in v.762

  ViewVC Help
Powered by ViewVC 1.1.5