/[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 750 by ph10, Fri Nov 18 11:07:14 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    ;    ;
416    
417  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 545  static const unsigned char ebcdic_charta Line 548  static const unsigned char ebcdic_charta
548  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
549    
550  static BOOL  static BOOL
551    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
552      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
553    
554    
# Line 577  return s; Line 580  return s;
580    
581    
582  /*************************************************  /*************************************************
583    *            Check for counted repeat            *
584    *************************************************/
585    
586    /* This function is called when a '{' is encountered in a place where it might
587    start a quantifier. It looks ahead to see if it really is a quantifier or not.
588    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
589    where the ddds are digits.
590    
591    Arguments:
592      p         pointer to the first char after '{'
593    
594    Returns:    TRUE or FALSE
595    */
596    
597    static BOOL
598    is_counted_repeat(const uschar *p)
599    {
600    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
601    while ((digitab[*p] & ctype_digit) != 0) p++;
602    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
603    
604    if (*p++ != CHAR_COMMA) return FALSE;
605    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
606    
607    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
608    while ((digitab[*p] & ctype_digit) != 0) p++;
609    
610    return (*p == CHAR_RIGHT_CURLY_BRACKET);
611    }
612    
613    
614    
615    /*************************************************
616  *            Handle escapes                      *  *            Handle escapes                      *
617  *************************************************/  *************************************************/
618    
# Line 642  else Line 678  else
678    
679      case CHAR_l:      case CHAR_l:
680      case CHAR_L:      case CHAR_L:
681        *errorcodeptr = ERR37;
682        break;
683    
684      case CHAR_u:      case CHAR_u:
685        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
686          {
687          /* In JavaScript, \u must be followed by four hexadecimal numbers.
688          Otherwise it is a lowercase u letter. */
689          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
690               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
691            {
692            c = 0;
693            for (i = 0; i < 4; ++i)
694              {
695              register int cc = *(++ptr);
696    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
697              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
698              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
699    #else           /* EBCDIC coding */
700              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
701              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
702    #endif
703              }
704            }
705          }
706        else
707          *errorcodeptr = ERR37;
708        break;
709    
710      case CHAR_U:      case CHAR_U:
711      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
712        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
713      break;      break;
714    
715      /* \g must be followed by one of a number of specific things:      /* In a character class, \g is just a literal "g". Outside a character
716        class, \g must be followed by one of a number of specific things:
717    
718      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
719      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 664  else Line 730  else
730      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
731    
732      case CHAR_g:      case CHAR_g:
733        if (isclass) break;
734      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
735        {        {
736        c = -ESC_g;        c = -ESC_g;
# Line 792  else Line 859  else
859      treated as a data character. */      treated as a data character. */
860    
861      case CHAR_x:      case CHAR_x:
862        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
863          {
864          /* In JavaScript, \x must be followed by two hexadecimal numbers.
865          Otherwise it is a lowercase x letter. */
866          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
867            {
868            c = 0;
869            for (i = 0; i < 2; ++i)
870              {
871              register int cc = *(++ptr);
872    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
873              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
874              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
875    #else           /* EBCDIC coding */
876              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
877              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
878    #endif
879              }
880            }
881          break;
882          }
883    
884      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
885        {        {
886        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
# Line 885  else Line 974  else
974    }    }
975    
976  /* Perl supports \N{name} for character names, as well as plain \N for "not  /* Perl supports \N{name} for character names, as well as plain \N for "not
977  newline". PCRE does not support \N{name}. */  newline". PCRE does not support \N{name}. However, it does support
978    quantification such as \N{2,3}. */
979    
980  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
981         !is_counted_repeat(ptr+2))
982    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
983    
984  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
# Line 997  return -1; Line 1088  return -1;
1088    
1089    
1090  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if (*p++ != CHAR_COMMA) return FALSE;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == CHAR_RIGHT_CURLY_BRACKET);  
 }  
   
   
   
 /*************************************************  
1091  *         Read repeat counts                     *  *         Read repeat counts                     *
1092  *************************************************/  *************************************************/
1093    
# Line 1403  does not. Line 1461  does not.
1461    
1462  Arguments:  Arguments:
1463    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1464    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1465    
1466  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1467  */  */
1468    
1469  static const uschar*  static const uschar*
1470  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1471  {  {
1472  for (;;)  for (;;)
1473    {    {
# Line 1468  and doing the check at the end; a flag s Line 1522  and doing the check at the end; a flag s
1522    
1523  Arguments:  Arguments:
1524    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1525    options  the compiling options    utf8     TRUE in UTF-8 mode
1526    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1527    cd       the "compile data" structure    cd       the "compile data" structure
1528    
# Line 1476  Returns:   the fixed length, Line 1530  Returns:   the fixed length,
1530               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1531               or -2 if \C was encountered               or -2 if \C was encountered
1532               or -3 if an OP_RECURSE item was encountered and atend is FALSE               or -3 if an OP_RECURSE item was encountered and atend is FALSE
1533                 or -4 if an unknown opcode was encountered (internal error)
1534  */  */
1535    
1536  static int  static int
1537  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1538  {  {
1539  int length = -1;  int length = -1;
1540    
# Line 1496  for (;;) Line 1551  for (;;)
1551    register int op = *cc;    register int op = *cc;
1552    switch (op)    switch (op)
1553      {      {
1554        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1555        OP_BRA (normal non-capturing bracket) because the other variants of these
1556        opcodes are all concerned with unlimited repeated groups, which of course
1557        are not of fixed length. */
1558    
1559      case OP_CBRA:      case OP_CBRA:
1560      case OP_BRA:      case OP_BRA:
1561      case OP_ONCE:      case OP_ONCE:
1562        case OP_ONCE_NC:
1563      case OP_COND:      case OP_COND:
1564      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1565      if (d < 0) return d;      if (d < 0) return d;
1566      branchlength += d;      branchlength += d;
1567      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1568      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1569      break;      break;
1570    
1571      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested call.
1572      call. If it's ALT it is an alternation in a nested call. If it is      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1573      END it's the end of the outer call. All can be handled by the same code. */      an ALT. If it is END it's the end of the outer call. All can be handled by
1574        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1575        because they all imply an unlimited repeat. */
1576    
1577      case OP_ALT:      case OP_ALT:
1578      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1579      case OP_END:      case OP_END:
1580        case OP_ACCEPT:
1581        case OP_ASSERT_ACCEPT:
1582      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1583        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1584      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1532  for (;;) Line 1595  for (;;)
1595      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1596      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1597      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1598      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1599      if (d < 0) return d;      if (d < 0) return d;
1600      branchlength += d;      branchlength += d;
1601      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1549  for (;;) Line 1612  for (;;)
1612    
1613      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1614    
1615      case OP_REVERSE:      case OP_MARK:
1616      case OP_CREF:      case OP_PRUNE_ARG:
1617      case OP_NCREF:      case OP_SKIP_ARG:
1618      case OP_RREF:      case OP_THEN_ARG:
1619      case OP_NRREF:      cc += cc[1] + _pcre_OP_lengths[*cc];
1620      case OP_DEF:      break;
1621    
1622      case OP_CALLOUT:      case OP_CALLOUT:
     case OP_SOD:  
     case OP_SOM:  
     case OP_SET_SOM:  
     case OP_EOD:  
     case OP_EODN:  
1623      case OP_CIRC:      case OP_CIRC:
1624      case OP_CIRCM:      case OP_CIRCM:
1625        case OP_CLOSE:
1626        case OP_COMMIT:
1627        case OP_CREF:
1628        case OP_DEF:
1629      case OP_DOLL:      case OP_DOLL:
1630      case OP_DOLLM:      case OP_DOLLM:
1631        case OP_EOD:
1632        case OP_EODN:
1633        case OP_FAIL:
1634        case OP_NCREF:
1635        case OP_NRREF:
1636      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1637        case OP_PRUNE:
1638        case OP_REVERSE:
1639        case OP_RREF:
1640        case OP_SET_SOM:
1641        case OP_SKIP:
1642        case OP_SOD:
1643        case OP_SOM:
1644        case OP_THEN:
1645      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1646      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
1647      break;      break;
# Line 1575  for (;;) Line 1651  for (;;)
1651      case OP_CHAR:      case OP_CHAR:
1652      case OP_CHARI:      case OP_CHARI:
1653      case OP_NOT:      case OP_NOT:
1654      case OP_NOTI:      case OP_NOTI:
1655      branchlength++;      branchlength++;
1656      cc += 2;      cc += 2;
1657  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1658      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1659  #endif  #endif
1660      break;      break;
1661    
# Line 1588  for (;;) Line 1663  for (;;)
1663      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1664    
1665      case OP_EXACT:      case OP_EXACT:
1666        case OP_EXACTI:
1667        case OP_NOTEXACT:
1668        case OP_NOTEXACTI:
1669      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1670      cc += 4;      cc += 4;
1671  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1672      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1673  #endif  #endif
1674      break;      break;
1675    
# Line 1609  for (;;) Line 1686  for (;;)
1686      cc += 2;      cc += 2;
1687      /* Fall through */      /* Fall through */
1688    
1689        case OP_HSPACE:
1690        case OP_VSPACE:
1691        case OP_NOT_HSPACE:
1692        case OP_NOT_VSPACE:
1693      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1694      case OP_DIGIT:      case OP_DIGIT:
1695      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1640  for (;;) Line 1721  for (;;)
1721    
1722      switch (*cc)      switch (*cc)
1723        {        {
1724          case OP_CRPLUS:
1725          case OP_CRMINPLUS:
1726        case OP_CRSTAR:        case OP_CRSTAR:
1727        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1728        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1660  for (;;) Line 1743  for (;;)
1743    
1744      /* Anything else is variable length */      /* Anything else is variable length */
1745    
1746      default:      case OP_ANYNL:
1747        case OP_BRAMINZERO:
1748        case OP_BRAPOS:
1749        case OP_BRAPOSZERO:
1750        case OP_BRAZERO:
1751        case OP_CBRAPOS:
1752        case OP_EXTUNI:
1753        case OP_KETRMAX:
1754        case OP_KETRMIN:
1755        case OP_KETRPOS:
1756        case OP_MINPLUS:
1757        case OP_MINPLUSI:
1758        case OP_MINQUERY:
1759        case OP_MINQUERYI:
1760        case OP_MINSTAR:
1761        case OP_MINSTARI:
1762        case OP_MINUPTO:
1763        case OP_MINUPTOI:
1764        case OP_NOTMINPLUS:
1765        case OP_NOTMINPLUSI:
1766        case OP_NOTMINQUERY:
1767        case OP_NOTMINQUERYI:
1768        case OP_NOTMINSTAR:
1769        case OP_NOTMINSTARI:
1770        case OP_NOTMINUPTO:
1771        case OP_NOTMINUPTOI:
1772        case OP_NOTPLUS:
1773        case OP_NOTPLUSI:
1774        case OP_NOTPOSPLUS:
1775        case OP_NOTPOSPLUSI:
1776        case OP_NOTPOSQUERY:
1777        case OP_NOTPOSQUERYI:
1778        case OP_NOTPOSSTAR:
1779        case OP_NOTPOSSTARI:
1780        case OP_NOTPOSUPTO:
1781        case OP_NOTPOSUPTOI:
1782        case OP_NOTQUERY:
1783        case OP_NOTQUERYI:
1784        case OP_NOTSTAR:
1785        case OP_NOTSTARI:
1786        case OP_NOTUPTO:
1787        case OP_NOTUPTOI:
1788        case OP_PLUS:
1789        case OP_PLUSI:
1790        case OP_POSPLUS:
1791        case OP_POSPLUSI:
1792        case OP_POSQUERY:
1793        case OP_POSQUERYI:
1794        case OP_POSSTAR:
1795        case OP_POSSTARI:
1796        case OP_POSUPTO:
1797        case OP_POSUPTOI:
1798        case OP_QUERY:
1799        case OP_QUERYI:
1800        case OP_REF:
1801        case OP_REFI:
1802        case OP_SBRA:
1803        case OP_SBRAPOS:
1804        case OP_SCBRA:
1805        case OP_SCBRAPOS:
1806        case OP_SCOND:
1807        case OP_SKIPZERO:
1808        case OP_STAR:
1809        case OP_STARI:
1810        case OP_TYPEMINPLUS:
1811        case OP_TYPEMINQUERY:
1812        case OP_TYPEMINSTAR:
1813        case OP_TYPEMINUPTO:
1814        case OP_TYPEPLUS:
1815        case OP_TYPEPOSPLUS:
1816        case OP_TYPEPOSQUERY:
1817        case OP_TYPEPOSSTAR:
1818        case OP_TYPEPOSUPTO:
1819        case OP_TYPEQUERY:
1820        case OP_TYPESTAR:
1821        case OP_TYPEUPTO:
1822        case OP_UPTO:
1823        case OP_UPTOI:
1824      return -1;      return -1;
1825    
1826        /* Catch unrecognized opcodes so that when new ones are added they
1827        are not forgotten, as has happened in the past. */
1828    
1829        default:
1830        return -4;
1831      }      }
1832    }    }
1833  /* Control never gets here */  /* Control never gets here */
# Line 1694  _pcre_find_bracket(const uschar *code, B Line 1860  _pcre_find_bracket(const uschar *code, B
1860  for (;;)  for (;;)
1861    {    {
1862    register int c = *code;    register int c = *code;
1863    
1864    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1865    
1866    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1712  for (;;) Line 1879  for (;;)
1879    
1880    /* Handle capturing bracket */    /* Handle capturing bracket */
1881    
1882    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1883               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1884      {      {
1885      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1886      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1754  for (;;) Line 1922  for (;;)
1922        break;        break;
1923    
1924        case OP_THEN_ARG:        case OP_THEN_ARG:
1925        code += code[1+LINK_SIZE];        code += code[1];
1926        break;        break;
1927        }        }
1928    
# Line 1873  for (;;) Line 2041  for (;;)
2041        break;        break;
2042    
2043        case OP_THEN_ARG:        case OP_THEN_ARG:
2044        code += code[1+LINK_SIZE];        code += code[1];
2045        break;        break;
2046        }        }
2047    
# Line 1954  could_be_empty_branch(const uschar *code Line 2122  could_be_empty_branch(const uschar *code
2122    compile_data *cd)    compile_data *cd)
2123  {  {
2124  register int c;  register int c;
2125  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2126       code < endcode;       code < endcode;
2127       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2128    {    {
2129    const uschar *ccode;    const uschar *ccode;
2130    
# Line 1972  for (code = first_significant_code(code Line 2140  for (code = first_significant_code(code
2140      continue;      continue;
2141      }      }
2142    
   /* Groups with zero repeats can of course be empty; skip them. */  
   
   if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)  
     {  
     code += _pcre_OP_lengths[c];  
     do code += GET(code, 1); while (*code == OP_ALT);  
     c = *code;  
     continue;  
     }  
   
2143    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
2144    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
2145      forward reference subroutine call, we can't. To detect forward reference
2146      we have to scan up the list that is kept in the workspace. This function is
2147      called only when doing the real compile, not during the pre-compile that
2148      measures the size of the compiled pattern. */
2149    
2150    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2151      {      {
2152      BOOL empty_branch = FALSE;      const uschar *scode;
2153      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
2154    
2155        /* Test for forward reference */
2156    
2157        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2158          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2159    
2160        /* Not a forward reference, test for completed backward reference */
2161    
2162        empty_branch = FALSE;
2163        scode = cd->start_code + GET(code, 1);
2164      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2165    
2166        /* Completed backwards reference */
2167    
2168      do      do
2169        {        {
2170        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf8, cd))
# Line 2000  for (code = first_significant_code(code Line 2175  for (code = first_significant_code(code
2175        scode += GET(scode, 1);        scode += GET(scode, 1);
2176        }        }
2177      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2178    
2179      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2180      continue;      continue;
2181      }      }
2182    
2183      /* Groups with zero repeats can of course be empty; skip them. */
2184    
2185      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2186          c == OP_BRAPOSZERO)
2187        {
2188        code += _pcre_OP_lengths[c];
2189        do code += GET(code, 1); while (*code == OP_ALT);
2190        c = *code;
2191        continue;
2192        }
2193    
2194      /* A nested group that is already marked as "could be empty" can just be
2195      skipped. */
2196    
2197      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2198          c == OP_SCBRA || c == OP_SCBRAPOS)
2199        {
2200        do code += GET(code, 1); while (*code == OP_ALT);
2201        c = *code;
2202        continue;
2203        }
2204    
2205    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2206    
2207    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2208          c == OP_CBRA || c == OP_CBRAPOS ||
2209          c == OP_ONCE || c == OP_ONCE_NC ||
2210          c == OP_COND)
2211      {      {
2212      BOOL empty_branch;      BOOL empty_branch;
2213      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2135  for (code = first_significant_code(code Line 2336  for (code = first_significant_code(code
2336      case OP_KET:      case OP_KET:
2337      case OP_KETRMAX:      case OP_KETRMAX:
2338      case OP_KETRMIN:      case OP_KETRMIN:
2339        case OP_KETRPOS:
2340      case OP_ALT:      case OP_ALT:
2341      return TRUE;      return TRUE;
2342    
# Line 2177  for (code = first_significant_code(code Line 2379  for (code = first_significant_code(code
2379      break;      break;
2380    
2381      case OP_THEN_ARG:      case OP_THEN_ARG:
2382      code += code[1+LINK_SIZE];      code += code[1];
2383      break;      break;
2384    
2385      /* None of the remaining opcodes are required to match a character. */      /* None of the remaining opcodes are required to match a character. */
# Line 2200  return TRUE; Line 2402  return TRUE;
2402  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2403  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2404  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2405    This function is called only during the real compile, not during the
2406    pre-compile.
2407    
2408  Arguments:  Arguments:
2409    code        points to start of the recursion    code        points to start of the recursion
# Line 2250  where Perl recognizes it as the POSIX cl Line 2454  where Perl recognizes it as the POSIX cl
2454  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2455  I think.  I think.
2456    
2457    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2458    It seems that the appearance of a nested POSIX class supersedes an apparent
2459    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2460    a digit.
2461    
2462    In Perl, unescaped square brackets may also appear as part of class names. For
2463    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2464    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2465    seem right at all. PCRE does not allow closing square brackets in POSIX class
2466    names.
2467    
2468  Arguments:  Arguments:
2469    ptr      pointer to the initial [    ptr      pointer to the initial [
2470    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2264  int terminator;          /* Don't combin Line 2479  int terminator;          /* Don't combin
2479  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2480  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2481    {    {
2482    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2483        ptr++;
2484      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2485      else
2486      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2487      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2488        {        {
2489        *endptr = ptr;        *endptr = ptr;
2490        return TRUE;        return TRUE;
2491        }        }
2492        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2493             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2494              ptr[1] == CHAR_EQUALS_SIGN) &&
2495            check_posix_syntax(ptr, endptr))
2496          return FALSE;
2497      }      }
2498    }    }
2499  return FALSE;  return FALSE;
# Line 2682  if (next >= 0) switch(op_code) Line 2904  if (next >= 0) switch(op_code)
2904    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2905    
2906    /* 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
2907    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
2908    an XCLASS instead. */    an XCLASS instead. */
2909    
2910    case OP_NOT:    case OP_NOT:
2911    return (c = *previous) == next;    return (c = *previous) == next;
2912    
2913    case OP_NOTI:    case OP_NOTI:
2914    if ((c = *previous) == next) return TRUE;    if ((c = *previous) == next) return TRUE;
2915  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2916    if (utf8)    if (utf8)
# Line 2981  Arguments: Line 3203  Arguments:
3203    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3204    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3205    bcptr          points to current branch chain    bcptr          points to current branch chain
3206      cond_depth     conditional nesting depth
3207    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3208    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3209                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2992  Returns:         TRUE on success Line 3215  Returns:         TRUE on success
3215  static BOOL  static BOOL
3216  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3217    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3218    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3219  {  {
3220  int repeat_type, op_type;  int repeat_type, op_type;
3221  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 3001  int greedy_default, greedy_non_default; Line 3224  int greedy_default, greedy_non_default;
3224  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3225  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3226  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3227  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3228  int after_manual_callout = 0;  int after_manual_callout = 0;
3229  int length_prevgroup = 0;  int length_prevgroup = 0;
3230  register int c;  register int c;
# Line 3019  uschar *previous_callout = NULL; Line 3242  uschar *previous_callout = NULL;
3242  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3243  uschar classbits[32];  uschar classbits[32];
3244    
3245    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3246    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3247    dynamically as we process the pattern. */
3248    
3249  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3250  BOOL class_utf8;  BOOL class_utf8;
3251  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 3027  uschar *class_utf8data_base; Line 3254  uschar *class_utf8data_base;
3254  uschar utf8_char[6];  uschar utf8_char[6];
3255  #else  #else
3256  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3257  #endif  #endif
3258    
3259  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 3078  for (;; ptr++) Line 3304  for (;; ptr++)
3304    int subfirstbyte;    int subfirstbyte;
3305    int terminator;    int terminator;
3306    int mclength;    int mclength;
3307      int tempbracount;
3308    uschar mcbuffer[8];    uschar mcbuffer[8];
3309    
3310    /* Get next byte in the pattern */    /* Get next byte in the pattern */
# Line 3199  for (;; ptr++) Line 3426  for (;; ptr++)
3426      previous_callout = NULL;      previous_callout = NULL;
3427      }      }
3428    
3429    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3430    
3431    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3432      {      {
# Line 4168  for (;; ptr++) Line 4395  for (;; ptr++)
4395      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4396      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4397    
4398      /* Save start of previous item, in case we have to move it up to make space      /* Save start of previous item, in case we have to move it up in order to
4399      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4400    
4401      tempcode = previous;      tempcode = previous;
4402    
# Line 4192  for (;; ptr++) Line 4419  for (;; ptr++)
4419        }        }
4420      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4421    
4422        /* If previous was a recursion call, wrap it in atomic brackets so that
4423        previous becomes the atomic group. All recursions were so wrapped in the
4424        past, but it no longer happens for non-repeated recursions. In fact, the
4425        repeated ones could be re-implemented independently so as not to need this,
4426        but for the moment we rely on the code for repeating groups. */
4427    
4428        if (*previous == OP_RECURSE)
4429          {
4430          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4431          *previous = OP_ONCE;
4432          PUT(previous, 1, 2 + 2*LINK_SIZE);
4433          previous[2 + 2*LINK_SIZE] = OP_KET;
4434          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4435          code += 2 + 2 * LINK_SIZE;
4436          length_prevgroup = 3 + 3*LINK_SIZE;
4437    
4438          /* When actually compiling, we need to check whether this was a forward
4439          reference, and if so, adjust the offset. */
4440    
4441          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4442            {
4443            int offset = GET(cd->hwm, -LINK_SIZE);
4444            if (offset == previous + 1 - cd->start_code)
4445              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4446            }
4447          }
4448    
4449        /* Now handle repetition for the different types of item. */
4450    
4451      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4452      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
4453      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
# Line 4201  for (;; ptr++) Line 4457  for (;; ptr++)
4457      if (*previous == OP_CHAR || *previous == OP_CHARI)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4458        {        {
4459        op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;        op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4460    
4461        /* Deal with UTF-8 characters that take up more than one byte. It's        /* Deal with UTF-8 characters that take up more than one byte. It's
4462        easier to write this out separately than try to macrify it. Use c to        easier to write this out separately than try to macrify it. Use c to
4463        hold the length of the character in bytes, plus 0x80 to flag that it's a        hold the length of the character in bytes, plus 0x80 to flag that it's a
# Line 4246  for (;; ptr++) Line 4502  for (;; ptr++)
4502      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4503      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4504      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4505      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4506      are currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4507    
4508      else if (*previous == OP_NOT || *previous == OP_NOTI)      else if (*previous == OP_NOT || *previous == OP_NOTI)
# Line 4483  for (;; ptr++) Line 4739  for (;; ptr++)
4739        }        }
4740    
4741      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
4742      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4743        opcodes such as BRA and CBRA, as this is the place where they get converted
4744        into the more special varieties such as BRAPOS and SBRA. A test for >=
4745        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4746        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4747        repetition of assertions, but now it does, for Perl compatibility. */
4748    
4749      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4750        {        {
4751        register int i;        register int i;
       int ketoffset = 0;  
4752        int len = (int)(code - previous);        int len = (int)(code - previous);
4753        uschar *bralink = NULL;        uschar *bralink = NULL;
4754          uschar *brazeroptr = NULL;
4755    
4756        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4757          we just ignore the repeat. */
4758    
4759        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4760          {          goto END_REPEAT;
4761          *errorcodeptr = ERR55;  
4762          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4763          }        use of repetition is in cases when the assertion is optional. Therefore,
4764          if the minimum is greater than zero, just ignore the repeat. If the
4765          maximum is not not zero or one, set it to 1. */
4766    
4767        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4768        by scanning through from the start, and compute the offset back to it          {
4769        from the current code pointer. */          if (repeat_min > 0) goto END_REPEAT;
4770            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);  
4771          }          }
4772    
4773        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 4530  for (;; ptr++) Line 4788  for (;; ptr++)
4788          **   goto END_REPEAT;          **   goto END_REPEAT;
4789          **   }          **   }
4790    
4791          However, that fails when a group is referenced as a subroutine from          However, that fails when a group or a subgroup within it is referenced
4792          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it          as a subroutine from elsewhere in the pattern, so now we stick in
4793          so that it is skipped on execution. As we don't have a list of which          OP_SKIPZERO in front of it so that it is skipped on execution. As we
4794          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4795            selectively.
4796    
4797          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4798          and do no more at this point. However, we do need to adjust any          and do no more at this point. However, we do need to adjust any
# Line 4553  for (;; ptr++) Line 4812  for (;; ptr++)
4812              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4813              goto END_REPEAT;              goto END_REPEAT;
4814              }              }
4815              brazeroptr = previous;    /* Save for possessive optimizing */
4816            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4817            }            }
4818    
# Line 4717  for (;; ptr++) Line 4977  for (;; ptr++)
4977            }            }
4978          }          }
4979    
4980        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. For
4981        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4982        don't know if there's been an options resetting after the ket. The        ONCE brackets can be converted into non-capturing brackets, as the
4983        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4984          deal with possessive ONCEs specially.
4985        Then, when we are doing the actual compile phase, check to see whether  
4986        this group is a non-atomic one that could match an empty string. If so,        Otherwise, when we are doing the actual compile phase, check to see
4987          whether this group is one that could match an empty string. If so,
4988        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
4989        that runtime checking can be done. [This check is also applied to        that runtime checking can be done. [This check is also applied to ONCE
4990        atomic groups at runtime, but in a different way.] */        groups at runtime, but in a different way.]
4991    
4992          Then, if the quantifier was possessive and the bracket is not a
4993          conditional, we convert the BRA code to the POS form, and the KET code to
4994          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
4995          subpattern at both the start and at the end.) The use of special opcodes
4996          makes it possible to reduce greatly the stack usage in pcre_exec(). If
4997          the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
4998    
4999          Then, if the minimum number of matches is 1 or 0, cancel the possessive
5000          flag so that the default action below, of wrapping everything inside
5001          atomic brackets, does not happen. When the minimum is greater than 1,
5002          there will be earlier copies of the group, and so we still have to wrap
5003          the whole thing. */
5004    
5005        else        else
5006          {          {
5007          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
5008          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
5009          *ketcode = OP_KETRMAX + repeat_type;  
5010          if (lengthptr == NULL && *bracode != OP_ONCE)          /* Convert possessive ONCE brackets to non-capturing */
5011    
5012            if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5013                possessive_quantifier) *bracode = OP_BRA;
5014    
5015            /* For non-possessive ONCE brackets, all we need to do is to
5016            set the KET. */
5017    
5018            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5019              *ketcode = OP_KETRMAX + repeat_type;
5020    
5021            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5022            converted to non-capturing above). */
5023    
5024            else
5025            {            {
5026            uschar *scode = bracode;            /* In the compile phase, check for empty string matching. */
5027            do  
5028              if (lengthptr == NULL)
5029              {              {
5030              if (could_be_empty_branch(scode, ketcode, utf8, cd))              uschar *scode = bracode;
5031                do
5032                {                {
5033                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
5034                break;                  {
5035                    *bracode += OP_SBRA - OP_BRA;
5036                    break;
5037                    }
5038                  scode += GET(scode, 1);
5039                }                }
5040              scode += GET(scode, 1);              while (*scode == OP_ALT);
5041              }              }
5042            while (*scode == OP_ALT);  
5043              /* Handle possessive quantifiers. */
5044    
5045              if (possessive_quantifier)
5046                {
5047                /* For COND brackets, we wrap the whole thing in a possessively
5048                repeated non-capturing bracket, because we have not invented POS
5049                versions of the COND opcodes. Because we are moving code along, we
5050                must ensure that any pending recursive references are updated. */
5051    
5052                if (*bracode == OP_COND || *bracode == OP_SCOND)
5053                  {
5054                  int nlen = (int)(code - bracode);
5055                  *code = OP_END;
5056                  adjust_recurse(bracode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5057                  memmove(bracode + 1+LINK_SIZE, bracode, nlen);
5058                  code += 1 + LINK_SIZE;
5059                  nlen += 1 + LINK_SIZE;
5060                  *bracode = OP_BRAPOS;
5061                  *code++ = OP_KETRPOS;
5062                  PUTINC(code, 0, nlen);
5063                  PUT(bracode, 1, nlen);
5064                  }
5065    
5066                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5067    
5068                else
5069                  {
5070                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5071                  *ketcode = OP_KETRPOS;
5072                  }
5073    
5074                /* If the minimum is zero, mark it as possessive, then unset the
5075                possessive flag when the minimum is 0 or 1. */
5076    
5077                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5078                if (repeat_min < 2) possessive_quantifier = FALSE;
5079                }
5080    
5081              /* Non-possessive quantifier */
5082    
5083              else *ketcode = OP_KETRMAX + repeat_type;
5084            }            }
5085          }          }
5086        }        }
# Line 4766  for (;; ptr++) Line 5101  for (;; ptr++)
5101        }        }
5102    
5103      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
5104      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
5105      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
5106      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
5107      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
5108      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
5109      tempcode, not at previous, which might be the first part of a string whose  
5110      (former) last char we repeated.      Some (but not all) possessively repeated subpatterns have already been
5111        completely handled in the code just above. For them, possessive_quantifier
5112        is always FALSE at this stage.
5113    
5114        Note that the repeated item starts at tempcode, not at previous, which
5115        might be the first part of a string whose (former) last char we repeated.
5116    
5117      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
5118      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 5216  for (;; ptr++)
5216        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5217        namelen = (int)(ptr - name);        namelen = (int)(ptr - name);
5218    
5219          /* It appears that Perl allows any characters whatsoever, other than
5220          a closing parenthesis, to appear in arguments, so we no longer insist on
5221          letters, digits, and underscores. */
5222    
5223        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5224          {          {
5225          arg = ++ptr;          arg = ++ptr;
5226          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
           || *ptr == '_') ptr++;  
5227          arglen = (int)(ptr - arg);          arglen = (int)(ptr - arg);
5228          }          }
5229    
# Line 4897  for (;; ptr++) Line 5240  for (;; ptr++)
5240          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5241              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5242            {            {
5243            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5244              ASSERT_ACCEPT if in an assertion. */
5245    
5246            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5247              {              {
5248              open_capitem *oc;              open_capitem *oc;
5249                if (arglen != 0)
5250                  {
5251                  *errorcodeptr = ERR59;
5252                  goto FAILED;
5253                  }
5254              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5255              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5256                {                {
5257                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5258                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5259                }                }
5260                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5261    
5262                /* Do not set firstbyte after *ACCEPT */
5263                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5264              }              }
5265    
5266            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
5267    
5268            if (arglen == 0)            else if (arglen == 0)
5269              {              {
5270              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
5271                {                {
# Line 4920  for (;; ptr++) Line 5273  for (;; ptr++)
5273                goto FAILED;                goto FAILED;
5274                }                }
5275              *code = verbs[i].op;              *code = verbs[i].op;
5276              if (*code++ == OP_THEN)              if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
               {  
               PUT(code, 0, code - bcptr->current_branch - 1);  
               code += LINK_SIZE;  
               }  
5277              }              }
5278    
5279            else            else
# Line 4935  for (;; ptr++) Line 5284  for (;; ptr++)
5284                goto FAILED;                goto FAILED;
5285                }                }
5286              *code = verbs[i].op_arg;              *code = verbs[i].op_arg;
5287              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;  
               }  
5288              *code++ = arglen;              *code++ = arglen;
5289              memcpy(code, arg, arglen);              memcpy(code, arg, arglen);
5290              code += arglen;              code += arglen;
# Line 5201  for (;; ptr++) Line 5546  for (;; ptr++)
5546          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5547          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5548          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5549            cd->assert_depth += 1;
5550          ptr++;          ptr++;
5551          break;          break;
5552    
# Line 5215  for (;; ptr++) Line 5561  for (;; ptr++)
5561            continue;            continue;
5562            }            }
5563          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5564            cd->assert_depth += 1;
5565          break;          break;
5566    
5567    
# Line 5224  for (;; ptr++) Line 5571  for (;; ptr++)
5571            {            {
5572            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5573            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5574              cd->assert_depth += 1;
5575            ptr += 2;            ptr += 2;
5576            break;            break;
5577    
5578            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5579            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5580              cd->assert_depth += 1;
5581            ptr += 2;            ptr += 2;
5582            break;            break;
5583    
# Line 5619  for (;; ptr++) Line 5968  for (;; ptr++)
5968    
5969                /* 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
5970                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5971                of the group. */                of the group. Then remember the forward reference. */
5972    
5973                called = cd->start_code + recno;                called = cd->start_code + recno;
5974                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5975                }                }
5976    
5977              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
5978              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
5979              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
5980                must not, however, do this check if we are in a conditional
5981                subpattern because the condition might be testing for recursion in
5982                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
5983                Forever loops are also detected at runtime, so those that occur in
5984                conditional subpatterns will be picked up then. */
5985    
5986              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
5987                       could_be_empty(called, code, bcptr, utf8, cd))                       could_be_empty(called, code, bcptr, utf8, cd))
5988                {                {
5989                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
# Line 5637  for (;; ptr++) Line 5991  for (;; ptr++)
5991                }                }
5992              }              }
5993    
5994            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
           "once" brackets. Set up a "previous group" length so that a  
           subsequent quantifier will work. */  
   
           *code = OP_ONCE;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
5995    
5996            *code = OP_RECURSE;            *code = OP_RECURSE;
5997            PUT(code, 1, (int)(called - cd->start_code));            PUT(code, 1, (int)(called - cd->start_code));
5998            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
   
           *code = OP_KET;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
   
           length_prevgroup = 3 + 3*LINK_SIZE;  
5999            }            }
6000    
6001          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5714  for (;; ptr++) Line 6056  for (;; ptr++)
6056          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
6057          both phases.          both phases.
6058    
6059          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
6060          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. */  
6061    
6062          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
6063            {            {
# Line 5733  for (;; ptr++) Line 6074  for (;; ptr++)
6074              }              }
6075    
6076            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
6077            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). */  
6078    
6079            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
6080            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5771  for (;; ptr++) Line 6110  for (;; ptr++)
6110        skipbytes = 2;        skipbytes = 2;
6111        }        }
6112    
6113      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
6114      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
6115      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
6116      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. */  
6117    
6118      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                      /* For handling repetition */
6119      *code = bravalue;      *code = bravalue;
6120      tempcode = code;      tempcode = code;
6121      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
6122      length_prevgroup = 0;              /* Initialize for pre-compile phase */      tempbracount = cd->bracount;          /* Save value before bracket */
6123        length_prevgroup = 0;                 /* Initialize for pre-compile phase */
6124    
6125      if (!compile_regex(      if (!compile_regex(
6126           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
6127           options & PCRE_IMS,           /* The previous ims option state */           &tempcode,                       /* Where to put code (updated) */
6128           &tempcode,                    /* Where to put code (updated) */           &ptr,                            /* Input pointer (updated) */
6129           &ptr,                         /* Input pointer (updated) */           errorcodeptr,                    /* Where to put an error message */
          errorcodeptr,                 /* Where to put an error message */  
6130           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
6131            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
6132           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
6133           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
6134           &subfirstbyte,                /* For possible first char */           cond_depth +
6135           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
6136           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
6137           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
6138           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
6139             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
6140             (lengthptr == NULL)? NULL :      /* Actual compile phase */
6141               &length_prevgroup              /* Pre-compile phase */
6142           ))           ))
6143        goto FAILED;        goto FAILED;
6144    
6145        /* If this was an atomic group and there are no capturing groups within it,
6146        generate OP_ONCE_NC instead of OP_ONCE. */
6147    
6148        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
6149          *code = OP_ONCE_NC;
6150    
6151        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
6152          cd->assert_depth -= 1;
6153    
6154      /* 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
6155      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.
6156      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
6157    
6158      /* 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
6159      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
6160      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
6161      not be available. */      not be available. */
# Line 5872  for (;; ptr++) Line 6220  for (;; ptr++)
6220          goto FAILED;          goto FAILED;
6221          }          }
6222        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
6223        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
6224        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
6225        *code++ = OP_KET;        *code++ = OP_KET;
6226        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 6040  for (;; ptr++) Line 6388  for (;; ptr++)
6388          }          }
6389    
6390        /* \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).
6391        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6392    
6393        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6394          {          {
6395            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6396              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6397              {
6398              *errorcodeptr = ERR69;
6399              break;
6400              }
6401          is_recurse = FALSE;          is_recurse = FALSE;
6402          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6403            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
# Line 6201  for (;; ptr++) Line 6554  for (;; ptr++)
6554        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6555        }        }
6556    
6557      /* 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
6558      1 or the matching is caseful. */      1 or the matching is caseful. */
6559    
6560      else      else
# Line 6242  value of lengthptr distinguishes the two Line 6595  value of lengthptr distinguishes the two
6595    
6596  Arguments:  Arguments:
6597    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  
6598    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6599    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6600    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
6601    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6602    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6603    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6604      cond_depth     depth of nesting for conditional subpatterns
6605    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6606    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6607    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 6613  Returns:         TRUE on success
6613  */  */
6614    
6615  static BOOL  static BOOL
6616  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6617    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6618    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6619    int *lengthptr)    compile_data *cd, int *lengthptr)
6620  {  {
6621  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6622  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 6277  int branchfirstbyte, branchreqbyte; Line 6630  int branchfirstbyte, branchreqbyte;
6630  int length;  int length;
6631  int orig_bracount;  int orig_bracount;
6632  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6633  branch_chain bc;  branch_chain bc;
6634    
6635  bc.outer = bcptr;  bc.outer = bcptr;
# Line 6301  pre-compile phase to find out whether an Line 6653  pre-compile phase to find out whether an
6653    
6654  /* 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
6655  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
6656  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6657    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6658    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6659    
6660  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6661    {    {
# Line 6341  for (;;) Line 6695  for (;;)
6695    into the length. */    into the length. */
6696    
6697    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6698          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6699            (lengthptr == NULL)? NULL : &length))
6700      {      {
6701      *ptrptr = ptr;      *ptrptr = ptr;
6702      return FALSE;      return FALSE;
6703      }      }
6704    
   /* 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;  
   
6705    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6706    has fewer than the rest. */    has fewer than the rest. */
6707    
# Line 6416  for (;;) Line 6762  for (;;)
6762        {        {
6763        int fixed_length;        int fixed_length;
6764        *code = OP_END;        *code = OP_END;
6765        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6766            FALSE, cd);
6767        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6768        if (fixed_length == -3)        if (fixed_length == -3)
6769          {          {
# Line 6424  for (;;) Line 6771  for (;;)
6771          }          }
6772        else if (fixed_length < 0)        else if (fixed_length < 0)
6773          {          {
6774          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 :
6775                            (fixed_length == -4)? ERR70: ERR25;
6776          *ptrptr = ptr;          *ptrptr = ptr;
6777          return FALSE;          return FALSE;
6778          }          }
# Line 6437  for (;;) Line 6785  for (;;)
6785    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
6786    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
6787    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
6788    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. */  
6789    
6790    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6791      {      {
# Line 6564  of the more common cases more precisely. Line 6910  of the more common cases more precisely.
6910    
6911  Arguments:  Arguments:
6912    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6913    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
6914                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6915                    the less precise approach                    the less precise approach
# Line 6574  Returns:     TRUE or FALSE Line 6919  Returns:     TRUE or FALSE
6919  */  */
6920    
6921  static BOOL  static BOOL
6922  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6923    unsigned int backref_map)    unsigned int backref_map)
6924  {  {
6925  do {  do {
6926     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6927       options, PCRE_MULTILINE, FALSE);       FALSE);
6928     register int op = *scode;     register int op = *scode;
6929    
6930     /* Non-capturing brackets */     /* Non-capturing brackets */
6931    
6932     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6933           op == OP_SBRA || op == OP_SBRAPOS)
6934       {       {
6935       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6936       }       }
6937    
6938     /* Capturing brackets */     /* Capturing brackets */
6939    
6940     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6941                op == OP_SCBRA || op == OP_SCBRAPOS)
6942       {       {
6943       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6944       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6945       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6946       }       }
6947    
6948     /* Other brackets */     /* Other brackets */
6949    
6950     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC ||
6951                op == OP_COND)
6952       {       {
6953       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6954       }       }
6955    
6956     /* .* 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 7001  is_startline(const uschar *code, unsigne
7001  {  {
7002  do {  do {
7003     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
7004       NULL, 0, FALSE);       FALSE);
7005     register int op = *scode;     register int op = *scode;
7006    
7007     /* 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 7028  do {
7028         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
7029         break;         break;
7030         }         }
7031       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
7032       op = *scode;       op = *scode;
7033       }       }
7034    
7035     /* Non-capturing brackets */     /* Non-capturing brackets */
7036    
7037     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
7038           op == OP_SBRA || op == OP_SBRAPOS)
7039       {       {
7040       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
7041       }       }
7042    
7043     /* Capturing brackets */     /* Capturing brackets */
7044    
7045     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
7046                op == OP_SCBRA || op == OP_SCBRAPOS)
7047       {       {
7048       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
7049       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6702  do { Line 7052  do {
7052    
7053     /* Other brackets */     /* Other brackets */
7054    
7055     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC)
7056       {       {
7057       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
7058       }       }
# Line 6743  we return that char, otherwise -1. Line 7093  we return that char, otherwise -1.
7093    
7094  Arguments:  Arguments:
7095    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)  
7096    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
7097    
7098  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
7099  */  */
7100    
7101  static int  static int
7102  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
7103  {  {
7104  register int c = -1;  register int c = -1;
7105  do {  do {
7106     int d;     int d;
7107     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
7108       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
7109       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
7110     register int op = *scode;     register int op = *scode;
7111    
7112     switch(op)     switch(op)
# Line 6765  do { Line 7115  do {
7115       return -1;       return -1;
7116    
7117       case OP_BRA:       case OP_BRA:
7118         case OP_BRAPOS:
7119       case OP_CBRA:       case OP_CBRA:
7120         case OP_SCBRA:
7121         case OP_CBRAPOS:
7122         case OP_SCBRAPOS:
7123       case OP_ASSERT:       case OP_ASSERT:
7124       case OP_ONCE:       case OP_ONCE:
7125         case OP_ONCE_NC:
7126       case OP_COND:       case OP_COND:
7127       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
7128         return -1;         return -1;
7129       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
7130       break;       break;
7131    
7132       case OP_EXACT:       /* Fall through */       case OP_EXACT:
7133       scode += 2;       scode += 2;
7134         /* Fall through */
7135    
7136       case OP_CHAR:       case OP_CHAR:
      case OP_CHARI:  
7137       case OP_PLUS:       case OP_PLUS:
7138       case OP_MINPLUS:       case OP_MINPLUS:
7139       case OP_POSPLUS:       case OP_POSPLUS:
7140       if (!inassert) return -1;       if (!inassert) return -1;
7141       if (c < 0)       if (c < 0) c = scode[1];
7142         {         else if (c != scode[1]) return -1;
7143         c = scode[1];       break;
7144         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
7145         }       case OP_EXACTI:
7146       else if (c != scode[1]) return -1;       scode += 2;
7147         /* Fall through */
7148    
7149         case OP_CHARI:
7150         case OP_PLUSI:
7151         case OP_MINPLUSI:
7152         case OP_POSPLUSI:
7153         if (!inassert) return -1;
7154         if (c < 0) c = scode[1] | REQ_CASELESS;
7155           else if (c != scode[1]) return -1;
7156       break;       break;
7157       }       }
7158    
# Line 6939  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7303  while (ptr[skipatstart] == CHAR_LEFT_PAR
7303    
7304  utf8 = (options & PCRE_UTF8) != 0;  utf8 = (options & PCRE_UTF8) != 0;
7305    
7306  /* 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
7307  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
7308  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
7309  the end of the string for a short string error, for compatibility with previous  not used here. */
 versions. */  
7310    
7311  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7312  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7313       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1, &errorcode)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7314    {    {
7315    errorcode = ERR44;    errorcode = ERR44;
7316    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6972  if ((options & PCRE_UCP) != 0) Line 7335  if ((options & PCRE_UCP) != 0)
7335    
7336  /* Check validity of \R options. */  /* Check validity of \R options. */
7337    
7338  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
7339         (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
7340    {    {
7341    case 0:    errorcode = ERR56;
7342    case PCRE_BSR_ANYCRLF:    goto PCRE_EARLY_ERROR_RETURN;
   case PCRE_BSR_UNICODE:  
   break;  
   default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;  
7343    }    }
7344    
7345  /* 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 7424  outside can help speed up starting point
7424  ptr += skipatstart;  ptr += skipatstart;
7425  code = cworkspace;  code = cworkspace;
7426  *code = OP_BRA;  *code = OP_BRA;
7427  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7428    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7429  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7430    
7431  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 7479  field; this time it's used for rememberi
7479  */  */
7480    
7481  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7482    cd->assert_depth = 0;
7483  cd->bracount = 0;  cd->bracount = 0;
7484  cd->names_found = 0;  cd->names_found = 0;
7485  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 7498  of the function here. */
7498  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7499  code = (uschar *)codestart;  code = (uschar *)codestart;
7500  *code = OP_BRA;  *code = OP_BRA;
7501  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0, 0,
7502    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7503  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7504  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7505  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7506    
7507  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */  if (cd->had_accept) reqbyte = REQ_NONE;   /* Must disable after (*ACCEPT) */
7508    
7509  /* 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. */
7510    
# Line 7204  if (cd->check_lookbehind) Line 7565  if (cd->check_lookbehind)
7565        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7566        int end_op = *be;        int end_op = *be;
7567        *be = OP_END;        *be = OP_END;
7568        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7569            cd);
7570        *be = end_op;        *be = end_op;
7571        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7572        if (fixed_length < 0)        if (fixed_length < 0)
7573          {          {
7574          errorcode = (fixed_length == -2)? ERR36 : ERR25;          errorcode = (fixed_length == -2)? ERR36 :
7575                        (fixed_length == -4)? ERR70 : ERR25;
7576          break;          break;
7577          }          }
7578        PUT(cc, 1, fixed_length);        PUT(cc, 1, fixed_length);
# Line 7243  start with ^. and also when all branches Line 7606  start with ^. and also when all branches
7606    
7607  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7608    {    {
7609    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))  
7610      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7611    else    else
7612      {      {
7613      if (firstbyte < 0)      if (firstbyte < 0)
7614        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7615      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7616        {        {
7617        int ch = firstbyte & 255;        int ch = firstbyte & 255;

Legend:
Removed from v.602  
changed lines
  Added in v.750

  ViewVC Help
Powered by ViewVC 1.1.5