/[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 638 by ph10, Mon Jul 25 09:41:19 2011 UTC revision 773 by ph10, Wed Nov 30 18:10:27 2011 UTC
# Line 88  so this number is very generous. Line 88  so this number is very generous.
88  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
89  remembering forward references to groups so that they can be filled in at the  remembering forward references to groups so that they can be filled in at the
90  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
91  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
92    filled up by repetitions of forward references, for example patterns like
93    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
94    that the workspace is expanded using malloc() in this situation. The value
95    below is therefore a minimum, and we put a maximum on it for safety. The
96    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
97    kicks in at the same number of forward references in all cases. */
98    
99  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
100    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
101    
102  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
103  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
104    
105  #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)  #define WORK_SIZE_SAFETY_MARGIN (100)
106    
107    
108  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 409  static const char error_texts[] = Line 416  static const char error_texts[] =
416    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
417    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with PCRE_UCP support\0"
418    "\\c must be followed by an ASCII character\0"    "\\c must be followed by an ASCII character\0"
419    "\\k is not followed by a braced, angle-bracketed, or quoted name\0"    "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
420      /* 70 */
421      "internal error: unknown opcode in find_fixedlength()\0"
422      "\\N is not supported in a class\0"
423      "too many forward references\0"
424    ;    ;
425    
426  /* 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 546  static const unsigned char ebcdic_charta Line 557  static const unsigned char ebcdic_charta
557  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
558    
559  static BOOL  static BOOL
560    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int *,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
561      int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
562    
563    
564    
# Line 578  return s; Line 589  return s;
589    
590    
591  /*************************************************  /*************************************************
592    *           Expand the workspace                 *
593    *************************************************/
594    
595    /* This function is called during the second compiling phase, if the number of
596    forward references fills the existing workspace, which is originally a block on
597    the stack. A larger block is obtained from malloc() unless the ultimate limit
598    has been reached or the increase will be rather small.
599    
600    Argument: pointer to the compile data block
601    Returns:  0 if all went well, else an error number
602    */
603    
604    static int
605    expand_workspace(compile_data *cd)
606    {
607    uschar *newspace;
608    int newsize = cd->workspace_size * 2;
609    
610    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
611    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
612        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
613     return ERR72;
614    
615    newspace = (pcre_malloc)(newsize);
616    if (newspace == NULL) return ERR21;
617    
618    memcpy(newspace, cd->start_workspace, cd->workspace_size);
619    cd->hwm = (uschar *)newspace + (cd->hwm - cd->start_workspace);
620    if (cd->workspace_size > COMPILE_WORK_SIZE)
621      (pcre_free)((void *)cd->start_workspace);
622    cd->start_workspace = newspace;
623    cd->workspace_size = newsize;
624    return 0;
625    }
626    
627    
628    
629    /*************************************************
630    *            Check for counted repeat            *
631    *************************************************/
632    
633    /* This function is called when a '{' is encountered in a place where it might
634    start a quantifier. It looks ahead to see if it really is a quantifier or not.
635    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
636    where the ddds are digits.
637    
638    Arguments:
639      p         pointer to the first char after '{'
640    
641    Returns:    TRUE or FALSE
642    */
643    
644    static BOOL
645    is_counted_repeat(const uschar *p)
646    {
647    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
648    while ((digitab[*p] & ctype_digit) != 0) p++;
649    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
650    
651    if (*p++ != CHAR_COMMA) return FALSE;
652    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
653    
654    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
655    while ((digitab[*p] & ctype_digit) != 0) p++;
656    
657    return (*p == CHAR_RIGHT_CURLY_BRACKET);
658    }
659    
660    
661    
662    /*************************************************
663  *            Handle escapes                      *  *            Handle escapes                      *
664  *************************************************/  *************************************************/
665    
# Line 643  else Line 725  else
725    
726      case CHAR_l:      case CHAR_l:
727      case CHAR_L:      case CHAR_L:
728        *errorcodeptr = ERR37;
729        break;
730    
731      case CHAR_u:      case CHAR_u:
732        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
733          {
734          /* In JavaScript, \u must be followed by four hexadecimal numbers.
735          Otherwise it is a lowercase u letter. */
736          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
737               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
738            {
739            c = 0;
740            for (i = 0; i < 4; ++i)
741              {
742              register int cc = *(++ptr);
743    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
744              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
745              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
746    #else           /* EBCDIC coding */
747              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
748              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
749    #endif
750              }
751            }
752          }
753        else
754          *errorcodeptr = ERR37;
755        break;
756    
757      case CHAR_U:      case CHAR_U:
758      *errorcodeptr = ERR37;      /* In JavaScript, \U is an uppercase U letter. */
759        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
760      break;      break;
761    
762      /* \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
763        class, \g must be followed by one of a number of specific things:
764    
765      (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
766      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 665  else Line 777  else
777      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
778    
779      case CHAR_g:      case CHAR_g:
780        if (isclass) break;
781      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
782        {        {
783        c = -ESC_g;        c = -ESC_g;
# Line 793  else Line 906  else
906      treated as a data character. */      treated as a data character. */
907    
908      case CHAR_x:      case CHAR_x:
909        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
910          {
911          /* In JavaScript, \x must be followed by two hexadecimal numbers.
912          Otherwise it is a lowercase x letter. */
913          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
914            {
915            c = 0;
916            for (i = 0; i < 2; ++i)
917              {
918              register int cc = *(++ptr);
919    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
920              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
921              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
922    #else           /* EBCDIC coding */
923              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
924              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
925    #endif
926              }
927            }
928          break;
929          }
930    
931      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
932        {        {
933        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
# Line 886  else Line 1021  else
1021    }    }
1022    
1023  /* 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
1024  newline". PCRE does not support \N{name}. */  newline". PCRE does not support \N{name}. However, it does support
1025    quantification such as \N{2,3}. */
1026    
1027  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1028         !is_counted_repeat(ptr+2))
1029    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1030    
1031  /* 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 998  return -1; Line 1135  return -1;
1135    
1136    
1137  /*************************************************  /*************************************************
 *            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);  
 }  
   
   
   
 /*************************************************  
1138  *         Read repeat counts                     *  *         Read repeat counts                     *
1139  *************************************************/  *************************************************/
1140    
# Line 1471  Arguments: Line 1575  Arguments:
1575    
1576  Returns:   the fixed length,  Returns:   the fixed length,
1577               or -1 if there is no fixed length,               or -1 if there is no fixed length,
1578               or -2 if \C was encountered               or -2 if \C was encountered (in UTF-8 mode only)
1579               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
1580                 or -4 if an unknown opcode was encountered (internal error)
1581  */  */
1582    
1583  static int  static int
# Line 1496  for (;;) Line 1601  for (;;)
1601      /* We only need to continue for OP_CBRA (normal capturing bracket) and      /* We only need to continue for OP_CBRA (normal capturing bracket) and
1602      OP_BRA (normal non-capturing bracket) because the other variants of these      OP_BRA (normal non-capturing bracket) because the other variants of these
1603      opcodes are all concerned with unlimited repeated groups, which of course      opcodes are all concerned with unlimited repeated groups, which of course
1604      are not of fixed length. They will cause a -1 response from the default      are not of fixed length. */
     case of this switch. */  
1605    
1606      case OP_CBRA:      case OP_CBRA:
1607      case OP_BRA:      case OP_BRA:
1608      case OP_ONCE:      case OP_ONCE:
1609        case OP_ONCE_NC:
1610      case OP_COND:      case OP_COND:
1611      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1612      if (d < 0) return d;      if (d < 0) return d;
# Line 1510  for (;;) Line 1615  for (;;)
1615      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1616      break;      break;
1617    
1618      /* 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.
1619      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
1620      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
1621      Note that we must not include the OP_KETRxxx opcodes here, because they      the same code. Note that we must not include the OP_KETRxxx opcodes here,
1622      all imply an unlimited repeat. */      because they all imply an unlimited repeat. */
1623    
1624      case OP_ALT:      case OP_ALT:
1625      case OP_KET:      case OP_KET:
1626      case OP_END:      case OP_END:
1627        case OP_ACCEPT:
1628        case OP_ASSERT_ACCEPT:
1629      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1630        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1631      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1552  for (;;) Line 1659  for (;;)
1659    
1660      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1661    
1662      case OP_REVERSE:      case OP_MARK:
1663      case OP_CREF:      case OP_PRUNE_ARG:
1664      case OP_NCREF:      case OP_SKIP_ARG:
1665      case OP_RREF:      case OP_THEN_ARG:
1666      case OP_NRREF:      cc += cc[1] + _pcre_OP_lengths[*cc];
1667      case OP_DEF:      break;
1668    
1669      case OP_CALLOUT:      case OP_CALLOUT:
     case OP_SOD:  
     case OP_SOM:  
     case OP_SET_SOM:  
     case OP_EOD:  
     case OP_EODN:  
1670      case OP_CIRC:      case OP_CIRC:
1671      case OP_CIRCM:      case OP_CIRCM:
1672        case OP_CLOSE:
1673        case OP_COMMIT:
1674        case OP_CREF:
1675        case OP_DEF:
1676      case OP_DOLL:      case OP_DOLL:
1677      case OP_DOLLM:      case OP_DOLLM:
1678        case OP_EOD:
1679        case OP_EODN:
1680        case OP_FAIL:
1681        case OP_NCREF:
1682        case OP_NRREF:
1683      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1684        case OP_PRUNE:
1685        case OP_REVERSE:
1686        case OP_RREF:
1687        case OP_SET_SOM:
1688        case OP_SKIP:
1689        case OP_SOD:
1690        case OP_SOM:
1691        case OP_THEN:
1692      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1693      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
1694      break;      break;
# Line 1590  for (;;) Line 1710  for (;;)
1710      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1711    
1712      case OP_EXACT:      case OP_EXACT:
1713        case OP_EXACTI:
1714        case OP_NOTEXACT:
1715        case OP_NOTEXACTI:
1716      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1717      cc += 4;      cc += 4;
1718  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1610  for (;;) Line 1733  for (;;)
1733      cc += 2;      cc += 2;
1734      /* Fall through */      /* Fall through */
1735    
1736        case OP_HSPACE:
1737        case OP_VSPACE:
1738        case OP_NOT_HSPACE:
1739        case OP_NOT_VSPACE:
1740      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1741      case OP_DIGIT:      case OP_DIGIT:
1742      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1622  for (;;) Line 1749  for (;;)
1749      cc++;      cc++;
1750      break;      break;
1751    
1752      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1753        otherwise \C is coded as OP_ALLANY. */
1754    
1755      case OP_ANYBYTE:      case OP_ANYBYTE:
1756      return -2;      return -2;
# Line 1641  for (;;) Line 1769  for (;;)
1769    
1770      switch (*cc)      switch (*cc)
1771        {        {
1772          case OP_CRPLUS:
1773          case OP_CRMINPLUS:
1774        case OP_CRSTAR:        case OP_CRSTAR:
1775        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1776        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1661  for (;;) Line 1791  for (;;)
1791    
1792      /* Anything else is variable length */      /* Anything else is variable length */
1793    
1794      default:      case OP_ANYNL:
1795        case OP_BRAMINZERO:
1796        case OP_BRAPOS:
1797        case OP_BRAPOSZERO:
1798        case OP_BRAZERO:
1799        case OP_CBRAPOS:
1800        case OP_EXTUNI:
1801        case OP_KETRMAX:
1802        case OP_KETRMIN:
1803        case OP_KETRPOS:
1804        case OP_MINPLUS:
1805        case OP_MINPLUSI:
1806        case OP_MINQUERY:
1807        case OP_MINQUERYI:
1808        case OP_MINSTAR:
1809        case OP_MINSTARI:
1810        case OP_MINUPTO:
1811        case OP_MINUPTOI:
1812        case OP_NOTMINPLUS:
1813        case OP_NOTMINPLUSI:
1814        case OP_NOTMINQUERY:
1815        case OP_NOTMINQUERYI:
1816        case OP_NOTMINSTAR:
1817        case OP_NOTMINSTARI:
1818        case OP_NOTMINUPTO:
1819        case OP_NOTMINUPTOI:
1820        case OP_NOTPLUS:
1821        case OP_NOTPLUSI:
1822        case OP_NOTPOSPLUS:
1823        case OP_NOTPOSPLUSI:
1824        case OP_NOTPOSQUERY:
1825        case OP_NOTPOSQUERYI:
1826        case OP_NOTPOSSTAR:
1827        case OP_NOTPOSSTARI:
1828        case OP_NOTPOSUPTO:
1829        case OP_NOTPOSUPTOI:
1830        case OP_NOTQUERY:
1831        case OP_NOTQUERYI:
1832        case OP_NOTSTAR:
1833        case OP_NOTSTARI:
1834        case OP_NOTUPTO:
1835        case OP_NOTUPTOI:
1836        case OP_PLUS:
1837        case OP_PLUSI:
1838        case OP_POSPLUS:
1839        case OP_POSPLUSI:
1840        case OP_POSQUERY:
1841        case OP_POSQUERYI:
1842        case OP_POSSTAR:
1843        case OP_POSSTARI:
1844        case OP_POSUPTO:
1845        case OP_POSUPTOI:
1846        case OP_QUERY:
1847        case OP_QUERYI:
1848        case OP_REF:
1849        case OP_REFI:
1850        case OP_SBRA:
1851        case OP_SBRAPOS:
1852        case OP_SCBRA:
1853        case OP_SCBRAPOS:
1854        case OP_SCOND:
1855        case OP_SKIPZERO:
1856        case OP_STAR:
1857        case OP_STARI:
1858        case OP_TYPEMINPLUS:
1859        case OP_TYPEMINQUERY:
1860        case OP_TYPEMINSTAR:
1861        case OP_TYPEMINUPTO:
1862        case OP_TYPEPLUS:
1863        case OP_TYPEPOSPLUS:
1864        case OP_TYPEPOSQUERY:
1865        case OP_TYPEPOSSTAR:
1866        case OP_TYPEPOSUPTO:
1867        case OP_TYPEQUERY:
1868        case OP_TYPESTAR:
1869        case OP_TYPEUPTO:
1870        case OP_UPTO:
1871        case OP_UPTOI:
1872      return -1;      return -1;
1873    
1874        /* Catch unrecognized opcodes so that when new ones are added they
1875        are not forgotten, as has happened in the past. */
1876    
1877        default:
1878        return -4;
1879      }      }
1880    }    }
1881  /* Control never gets here */  /* Control never gets here */
# Line 1757  for (;;) Line 1970  for (;;)
1970        break;        break;
1971    
1972        case OP_THEN_ARG:        case OP_THEN_ARG:
1973        code += code[1+LINK_SIZE];        code += code[1];
1974        break;        break;
1975        }        }
1976    
# Line 1876  for (;;) Line 2089  for (;;)
2089        break;        break;
2090    
2091        case OP_THEN_ARG:        case OP_THEN_ARG:
2092        code += code[1+LINK_SIZE];        code += code[1];
2093        break;        break;
2094        }        }
2095    
# Line 1978  for (code = first_significant_code(code Line 2191  for (code = first_significant_code(code
2191    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
2192    implies a backward reference subroutine call, we can scan it. If it's a    implies a backward reference subroutine call, we can scan it. If it's a
2193    forward reference subroutine call, we can't. To detect forward reference    forward reference subroutine call, we can't. To detect forward reference
2194    we have to scan up the list that is kept in the workspace. This function is    we have to scan up the list that is kept in the workspace. This function is
2195    called only when doing the real compile, not during the pre-compile that    called only when doing the real compile, not during the pre-compile that
2196    measures the size of the compiled pattern. */    measures the size of the compiled pattern. */
2197    
2198    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2199      {      {
2200      const uschar *scode;      const uschar *scode;
2201      BOOL empty_branch;      BOOL empty_branch;
2202    
2203      /* Test for forward reference */      /* Test for forward reference */
2204    
2205      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2206        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2207    
2208      /* Not a forward reference, test for completed backward reference */      /* Not a forward reference, test for completed backward reference */
2209    
2210      empty_branch = FALSE;      empty_branch = FALSE;
2211      scode = cd->start_code + GET(code, 1);      scode = cd->start_code + GET(code, 1);
2212      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2213    
2214      /* Completed backwards reference */      /* Completed backwards reference */
2215    
2216      do      do
2217        {        {
2218        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf8, cd))
# Line 2010  for (code = first_significant_code(code Line 2223  for (code = first_significant_code(code
2223        scode += GET(scode, 1);        scode += GET(scode, 1);
2224        }        }
2225      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2226    
2227      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2228      continue;      continue;
2229      }      }
# Line 2041  for (code = first_significant_code(code Line 2254  for (code = first_significant_code(code
2254    
2255    if (c == OP_BRA  || c == OP_BRAPOS ||    if (c == OP_BRA  || c == OP_BRAPOS ||
2256        c == OP_CBRA || c == OP_CBRAPOS ||        c == OP_CBRA || c == OP_CBRAPOS ||
2257        c == OP_ONCE || c == OP_COND)        c == OP_ONCE || c == OP_ONCE_NC ||
2258          c == OP_COND)
2259      {      {
2260      BOOL empty_branch;      BOOL empty_branch;
2261      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2213  for (code = first_significant_code(code Line 2427  for (code = first_significant_code(code
2427      break;      break;
2428    
2429      case OP_THEN_ARG:      case OP_THEN_ARG:
2430      code += code[1+LINK_SIZE];      code += code[1];
2431      break;      break;
2432    
2433      /* None of the remaining opcodes are required to match a character. */      /* None of the remaining opcodes are required to match a character. */
# Line 2236  return TRUE; Line 2450  return TRUE;
2450  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
2451  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,
2452  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.
2453  This function is called only during the real compile, not during the  This function is called only during the real compile, not during the
2454  pre-compile.  pre-compile.
2455    
2456  Arguments:  Arguments:
# Line 2288  where Perl recognizes it as the POSIX cl Line 2502  where Perl recognizes it as the POSIX cl
2502  "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,
2503  I think.  I think.
2504    
2505    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2506    It seems that the appearance of a nested POSIX class supersedes an apparent
2507    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2508    a digit.
2509    
2510    In Perl, unescaped square brackets may also appear as part of class names. For
2511    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2512    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2513    seem right at all. PCRE does not allow closing square brackets in POSIX class
2514    names.
2515    
2516  Arguments:  Arguments:
2517    ptr      pointer to the initial [    ptr      pointer to the initial [
2518    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2302  int terminator;          /* Don't combin Line 2527  int terminator;          /* Don't combin
2527  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2528  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2529    {    {
2530    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2531        ptr++;
2532      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2533      else
2534      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2535      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2536        {        {
2537        *endptr = ptr;        *endptr = ptr;
2538        return TRUE;        return TRUE;
2539        }        }
2540        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2541             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2542              ptr[1] == CHAR_EQUALS_SIGN) &&
2543            check_posix_syntax(ptr, endptr))
2544          return FALSE;
2545      }      }
2546    }    }
2547  return FALSE;  return FALSE;
# Line 3019  Arguments: Line 3251  Arguments:
3251    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3252    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3253    bcptr          points to current branch chain    bcptr          points to current branch chain
3254      cond_depth     conditional nesting depth
3255    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3256    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3257                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 3030  Returns:         TRUE on success Line 3263  Returns:         TRUE on success
3263  static BOOL  static BOOL
3264  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3265    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3266    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3267  {  {
3268  int repeat_type, op_type;  int repeat_type, op_type;
3269  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 3058  uschar *save_hwm = NULL; Line 3291  uschar *save_hwm = NULL;
3291  uschar classbits[32];  uschar classbits[32];
3292    
3293  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3294  must not do this for other options (e.g. PCRE_EXTENDED) because they may change  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3295  dynamically as we process the pattern. */  dynamically as we process the pattern. */
3296    
3297  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 3069  uschar *class_utf8data_base; Line 3302  uschar *class_utf8data_base;
3302  uschar utf8_char[6];  uschar utf8_char[6];
3303  #else  #else
3304  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3305  #endif  #endif
3306    
3307  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 3120  for (;; ptr++) Line 3352  for (;; ptr++)
3352    int subfirstbyte;    int subfirstbyte;
3353    int terminator;    int terminator;
3354    int mclength;    int mclength;
3355      int tempbracount;
3356    uschar mcbuffer[8];    uschar mcbuffer[8];
3357    
3358    /* Get next byte in the pattern */    /* Get next byte in the pattern */
# Line 3144  for (;; ptr++) Line 3377  for (;; ptr++)
3377  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
3378      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3379  #endif  #endif
3380      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */      if (code > cd->start_workspace + cd->workspace_size -
3381            WORK_SIZE_SAFETY_MARGIN)                       /* Check for overrun */
3382        {        {
3383        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3384        goto FAILED;        goto FAILED;
# Line 3167  for (;; ptr++) Line 3401  for (;; ptr++)
3401        }        }
3402    
3403      *lengthptr += (int)(code - last_code);      *lengthptr += (int)(code - last_code);
3404      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, (int)(code - last_code),
3405          c));
3406    
3407      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
3408      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
# Line 3193  for (;; ptr++) Line 3428  for (;; ptr++)
3428    /* In the real compile phase, just check the workspace used by the forward    /* In the real compile phase, just check the workspace used by the forward
3429    reference list. */    reference list. */
3430    
3431    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)    else if (cd->hwm > cd->start_workspace + cd->workspace_size -
3432               WORK_SIZE_SAFETY_MARGIN)
3433      {      {
3434      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3435      goto FAILED;      goto FAILED;
# Line 3583  for (;; ptr++) Line 3819  for (;; ptr++)
3819          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3820    
3821          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
3822            else if (-c == ESC_N)            /* \N is not supported in a class */
3823              {
3824              *errorcodeptr = ERR71;
3825              goto FAILED;
3826              }
3827          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3828            {            {
3829            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
# Line 4233  for (;; ptr++) Line 4474  for (;; ptr++)
4474        ptr++;        ptr++;
4475        }        }
4476      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4477    
4478      /* If previous was a recursion call, wrap it in atomic brackets so that      /* If previous was a recursion call, wrap it in atomic brackets so that
4479      previous becomes the atomic group. All recursions were so wrapped in the      previous becomes the atomic group. All recursions were so wrapped in the
4480      past, but it no longer happens for non-repeated recursions. In fact, the      past, but it no longer happens for non-repeated recursions. In fact, the
4481      repeated ones could be re-implemented independently so as not to need this,      repeated ones could be re-implemented independently so as not to need this,
# Line 4249  for (;; ptr++) Line 4490  for (;; ptr++)
4490        PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);        PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4491        code += 2 + 2 * LINK_SIZE;        code += 2 + 2 * LINK_SIZE;
4492        length_prevgroup = 3 + 3*LINK_SIZE;        length_prevgroup = 3 + 3*LINK_SIZE;
4493    
4494        /* When actually compiling, we need to check whether this was a forward        /* When actually compiling, we need to check whether this was a forward
4495        reference, and if so, adjust the offset. */        reference, and if so, adjust the offset. */
4496    
4497        if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)        if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4498          {          {
4499          int offset = GET(cd->hwm, -LINK_SIZE);          int offset = GET(cd->hwm, -LINK_SIZE);
4500          if (offset == previous + 1 - cd->start_code)          if (offset == previous + 1 - cd->start_code)
4501            PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);            PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4502          }          }
4503        }        }
4504    
4505      /* Now handle repetition for the different types of item. */      /* Now handle repetition for the different types of item. */
4506    
4507      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
# Line 4558  for (;; ptr++) Line 4799  for (;; ptr++)
4799      opcodes such as BRA and CBRA, as this is the place where they get converted      opcodes such as BRA and CBRA, as this is the place where they get converted
4800      into the more special varieties such as BRAPOS and SBRA. A test for >=      into the more special varieties such as BRAPOS and SBRA. A test for >=
4801      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4802      ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow      ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4803      repetition of assertions, but now it does, for Perl compatibility. */      repetition of assertions, but now it does, for Perl compatibility. */
4804    
4805      else if (*previous >= OP_ASSERT && *previous <= OP_COND)      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
4806        {        {
4807        register int i;        register int i;
4808        int len = (int)(code - previous);        int len = (int)(code - previous);
4809        uschar *bralink = NULL;        uschar *bralink = NULL;
4810        uschar *brazeroptr = NULL;        uschar *brazeroptr = NULL;
4811    
4812        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4813        we just ignore the repeat. */        we just ignore the repeat. */
4814    
4815        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4816          goto END_REPEAT;          goto END_REPEAT;
4817    
4818        /* There is no sense in actually repeating assertions. The only potential        /* There is no sense in actually repeating assertions. The only potential
4819        use of repetition is in cases when the assertion is optional. Therefore,        use of repetition is in cases when the assertion is optional. Therefore,
4820        if the minimum is greater than zero, just ignore the repeat. If the        if the minimum is greater than zero, just ignore the repeat. If the
4821        maximum is not not zero or one, set it to 1. */        maximum is not not zero or one, set it to 1. */
4822    
4823        if (*previous < OP_ONCE)    /* Assertion */        if (*previous < OP_ONCE)    /* Assertion */
4824          {          {
4825          if (repeat_min > 0) goto END_REPEAT;          if (repeat_min > 0) goto END_REPEAT;
4826          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
4827          }          }
4828    
4829        /* 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
4830        OP_BRAZERO in front of it, and because the group appears once in the        OP_BRAZERO in front of it, and because the group appears once in the
# Line 4690  for (;; ptr++) Line 4931  for (;; ptr++)
4931              *lengthptr += delta;              *lengthptr += delta;
4932              }              }
4933    
4934            /* This is compiling for real */            /* This is compiling for real. If there is a set first byte for
4935              the group, and we have not yet set a "required byte", set it. Make
4936              sure there is enough workspace for copying forward references before
4937              doing the copy. */
4938    
4939            else            else
4940              {              {
4941              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;              if (groupsetfirstbyte && reqbyte < 0) reqbyte = firstbyte;
4942    
4943              for (i = 1; i < repeat_min; i++)              for (i = 1; i < repeat_min; i++)
4944                {                {
4945                uschar *hc;                uschar *hc;
4946                uschar *this_hwm = cd->hwm;                uschar *this_hwm = cd->hwm;
4947                memcpy(code, previous, len);                memcpy(code, previous, len);
4948    
4949                  while (cd->hwm > cd->start_workspace + cd->workspace_size -
4950                         WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
4951                    {
4952                    int save_offset = save_hwm - cd->start_workspace;
4953                    int this_offset = this_hwm - cd->start_workspace;
4954                    *errorcodeptr = expand_workspace(cd);
4955                    if (*errorcodeptr != 0) goto FAILED;
4956                    save_hwm = (uschar *)cd->start_workspace + save_offset;
4957                    this_hwm = (uschar *)cd->start_workspace + this_offset;
4958                    }
4959    
4960                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
4961                  {                  {
4962                  PUT(cd->hwm, 0, GET(hc, 0) + len);                  PUT(cd->hwm, 0, GET(hc, 0) + len);
# Line 4729  for (;; ptr++) Line 4986  for (;; ptr++)
4986          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4987          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4988          a 64-bit integer type when available, otherwise double. */          a 64-bit integer type when available, otherwise double. */
4989    
4990          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4991            {            {
4992            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
# Line 4767  for (;; ptr++) Line 5024  for (;; ptr++)
5024              }              }
5025    
5026            memcpy(code, previous, len);            memcpy(code, previous, len);
5027    
5028              /* Ensure there is enough workspace for forward references before
5029              copying them. */
5030    
5031              while (cd->hwm > cd->start_workspace + cd->workspace_size -
5032                     WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5033                {
5034                int save_offset = save_hwm - cd->start_workspace;
5035                int this_offset = this_hwm - cd->start_workspace;
5036                *errorcodeptr = expand_workspace(cd);
5037                if (*errorcodeptr != 0) goto FAILED;
5038                save_hwm = (uschar *)cd->start_workspace + save_offset;
5039                this_hwm = (uschar *)cd->start_workspace + this_offset;
5040                }
5041    
5042            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5043              {              {
5044              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
# Line 4797  for (;; ptr++) Line 5069  for (;; ptr++)
5069        ONCE brackets can be converted into non-capturing brackets, as the        ONCE brackets can be converted into non-capturing brackets, as the
5070        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
5071        deal with possessive ONCEs specially.        deal with possessive ONCEs specially.
5072    
5073        Otherwise, if the quantifier was possessive, we convert the BRA code to        Otherwise, when we are doing the actual compile phase, check to see
5074        the POS form, and the KET code to KETRPOS. (It turns out to be convenient        whether this group is one that could match an empty string. If so,
5075        at runtime to detect this kind of subpattern at both the start and at the        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
5076        end.) The use of special opcodes makes it possible to reduce greatly the        that runtime checking can be done. [This check is also applied to ONCE
5077        stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,        groups at runtime, but in a different way.]
5078        convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that  
5079        the default action below, of wrapping everything inside atomic brackets,        Then, if the quantifier was possessive and the bracket is not a
5080        does not happen.        conditional, we convert the BRA code to the POS form, and the KET code to
5081          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
5082        Then, when we are doing the actual compile phase, check to see whether        subpattern at both the start and at the end.) The use of special opcodes
5083        this group is one that could match an empty string. If so, convert the        makes it possible to reduce greatly the stack usage in pcre_exec(). If
5084        initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so that runtime        the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
5085        checking can be done. [This check is also applied to ONCE groups at  
5086        runtime, but in a different way.] */        Then, if the minimum number of matches is 1 or 0, cancel the possessive
5087          flag so that the default action below, of wrapping everything inside
5088          atomic brackets, does not happen. When the minimum is greater than 1,
5089          there will be earlier copies of the group, and so we still have to wrap
5090          the whole thing. */
5091    
5092        else        else
5093          {          {
5094          uschar *ketcode = code - 1 - LINK_SIZE;          uschar *ketcode = code - 1 - LINK_SIZE;
5095          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
5096    
5097            /* Convert possessive ONCE brackets to non-capturing */
5098    
5099          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;          if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5100          if (*bracode == OP_ONCE)              possessive_quantifier) *bracode = OP_BRA;
5101    
5102            /* For non-possessive ONCE brackets, all we need to do is to
5103            set the KET. */
5104    
5105            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5106            *ketcode = OP_KETRMAX + repeat_type;            *ketcode = OP_KETRMAX + repeat_type;
5107    
5108            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5109            converted to non-capturing above). */
5110    
5111          else          else
5112            {            {
5113            if (possessive_quantifier)            /* In the compile phase, check for empty string matching. */
5114              {  
             *bracode += 1;                   /* Switch to xxxPOS opcodes */  
             *ketcode = OP_KETRPOS;  
             if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;  
             possessive_quantifier = FALSE;  
             }  
           else *ketcode = OP_KETRMAX + repeat_type;  
   
5115            if (lengthptr == NULL)            if (lengthptr == NULL)
5116              {              {
5117              uschar *scode = bracode;              uschar *scode = bracode;
# Line 4846  for (;; ptr++) Line 5126  for (;; ptr++)
5126                }                }
5127              while (*scode == OP_ALT);              while (*scode == OP_ALT);
5128              }              }
5129    
5130              /* Handle possessive quantifiers. */
5131    
5132              if (possessive_quantifier)
5133                {
5134                /* For COND brackets, we wrap the whole thing in a possessively
5135                repeated non-capturing bracket, because we have not invented POS
5136                versions of the COND opcodes. Because we are moving code along, we
5137                must ensure that any pending recursive references are updated. */
5138    
5139                if (*bracode == OP_COND || *bracode == OP_SCOND)
5140                  {
5141                  int nlen = (int)(code - bracode);
5142                  *code = OP_END;
5143                  adjust_recurse(bracode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5144                  memmove(bracode + 1+LINK_SIZE, bracode, nlen);
5145                  code += 1 + LINK_SIZE;
5146                  nlen += 1 + LINK_SIZE;
5147                  *bracode = OP_BRAPOS;
5148                  *code++ = OP_KETRPOS;
5149                  PUTINC(code, 0, nlen);
5150                  PUT(bracode, 1, nlen);
5151                  }
5152    
5153                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5154    
5155                else
5156                  {
5157                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5158                  *ketcode = OP_KETRPOS;
5159                  }
5160    
5161                /* If the minimum is zero, mark it as possessive, then unset the
5162                possessive flag when the minimum is 0 or 1. */
5163    
5164                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5165                if (repeat_min < 2) possessive_quantifier = FALSE;
5166                }
5167    
5168              /* Non-possessive quantifier */
5169    
5170              else *ketcode = OP_KETRMAX + repeat_type;
5171            }            }
5172          }          }
5173        }        }
# Line 4870  for (;; ptr++) Line 5192  for (;; ptr++)
5192      there are special alternative opcodes for this case. For anything else, we      there are special alternative opcodes for this case. For anything else, we
5193      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
5194      notation is just syntactic sugar, taken from Sun's Java package, but the      notation is just syntactic sugar, taken from Sun's Java package, but the
5195      special opcodes can optimize it.      special opcodes can optimize it.
5196    
5197      Possessively repeated subpatterns have already been handled in the code      Some (but not all) possessively repeated subpatterns have already been
5198      just above, so possessive_quantifier is always FALSE for them at this      completely handled in the code just above. For them, possessive_quantifier
5199      stage.      is always FALSE at this stage.
5200    
5201      Note that the repeated item starts at tempcode, not at previous, which      Note that the repeated item starts at tempcode, not at previous, which
5202      might be the first part of a string whose (former) last char we repeated.      might be the first part of a string whose (former) last char we repeated.
5203    
# Line 4981  for (;; ptr++) Line 5303  for (;; ptr++)
5303        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5304        namelen = (int)(ptr - name);        namelen = (int)(ptr - name);
5305    
5306          /* It appears that Perl allows any characters whatsoever, other than
5307          a closing parenthesis, to appear in arguments, so we no longer insist on
5308          letters, digits, and underscores. */
5309    
5310        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5311          {          {
5312          arg = ++ptr;          arg = ++ptr;
5313          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
           || *ptr == '_') ptr++;  
5314          arglen = (int)(ptr - arg);          arglen = (int)(ptr - arg);
5315          }          }
5316    
# Line 5002  for (;; ptr++) Line 5327  for (;; ptr++)
5327          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5328              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5329            {            {
5330            /* Check for open captures before ACCEPT and convert it to            /* Check for open captures before ACCEPT and convert it to
5331            ASSERT_ACCEPT if in an assertion. */            ASSERT_ACCEPT if in an assertion. */
5332    
5333            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
# Line 5012  for (;; ptr++) Line 5337  for (;; ptr++)
5337                {                {
5338                *errorcodeptr = ERR59;                *errorcodeptr = ERR59;
5339                goto FAILED;                goto FAILED;
5340                }                }
5341              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5342              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5343                {                {
# Line 5020  for (;; ptr++) Line 5345  for (;; ptr++)
5345                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5346                }                }
5347              *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;              *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5348    
5349                /* Do not set firstbyte after *ACCEPT */
5350                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5351              }              }
5352    
5353            /* Handle other cases with/without an argument */            /* Handle other cases with/without an argument */
# Line 5032  for (;; ptr++) Line 5360  for (;; ptr++)
5360                goto FAILED;                goto FAILED;
5361                }                }
5362              *code = verbs[i].op;              *code = verbs[i].op;
5363              if (*code++ == OP_THEN)              if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
               {  
               PUT(code, 0, code - bcptr->current_branch - 1);  
               code += LINK_SIZE;  
               }  
5364              }              }
5365    
5366            else            else
# Line 5047  for (;; ptr++) Line 5371  for (;; ptr++)
5371                goto FAILED;                goto FAILED;
5372                }                }
5373              *code = verbs[i].op_arg;              *code = verbs[i].op_arg;
5374              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;  
               }  
5375              *code++ = arglen;              *code++ = arglen;
5376              memcpy(code, arg, arglen);              memcpy(code, arg, arglen);
5377              code += arglen;              code += arglen;
# Line 5313  for (;; ptr++) Line 5633  for (;; ptr++)
5633          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5634          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5635          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5636          cd->assert_depth += 1;          cd->assert_depth += 1;
5637          ptr++;          ptr++;
5638          break;          break;
5639    
# Line 5328  for (;; ptr++) Line 5648  for (;; ptr++)
5648            continue;            continue;
5649            }            }
5650          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5651          cd->assert_depth += 1;          cd->assert_depth += 1;
5652          break;          break;
5653    
5654    
# Line 5338  for (;; ptr++) Line 5658  for (;; ptr++)
5658            {            {
5659            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5660            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5661            cd->assert_depth += 1;            cd->assert_depth += 1;
5662            ptr += 2;            ptr += 2;
5663            break;            break;
5664    
5665            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5666            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5667            cd->assert_depth += 1;            cd->assert_depth += 1;
5668            ptr += 2;            ptr += 2;
5669            break;            break;
5670    
# Line 5366  for (;; ptr++) Line 5686  for (;; ptr++)
5686    
5687          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5688          case CHAR_C:                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
5689          previous_callout = code;  /* Save for later completion */          previous_callout = code;     /* Save for later completion */
5690          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1;    /* Skip one item before completing */
5691          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
5692            {            {
5693            int n = 0;            int n = 0;
# Line 5588  for (;; ptr++) Line 5908  for (;; ptr++)
5908    
5909            temp = cd->end_pattern;            temp = cd->end_pattern;
5910            cd->end_pattern = ptr;            cd->end_pattern = ptr;
5911            recno = find_parens(cd, name, namelen,            recno = find_parens(cd, name, namelen,
5912              (options & PCRE_EXTENDED) != 0, utf8);              (options & PCRE_EXTENDED) != 0, utf8);
5913            cd->end_pattern = temp;            cd->end_pattern = temp;
5914            if (recno < 0) recno = 0;    /* Forward ref; set dummy number */            if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
# Line 5736  for (;; ptr++) Line 6056  for (;; ptr++)
6056                /* 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
6057                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
6058                of the group. Then remember the forward reference. */                of the group. Then remember the forward reference. */
6059    
6060                called = cd->start_code + recno;                called = cd->start_code + recno;
6061                  if (cd->hwm >= cd->start_workspace + cd->workspace_size -
6062                      WORK_SIZE_SAFETY_MARGIN)
6063                    {
6064                    *errorcodeptr = expand_workspace(cd);
6065                    if (*errorcodeptr != 0) goto FAILED;
6066                    }
6067                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
6068                }                }
6069    
6070              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
6071              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
6072              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
6073                must not, however, do this check if we are in a conditional
6074                subpattern because the condition might be testing for recursion in
6075                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
6076                Forever loops are also detected at runtime, so those that occur in
6077                conditional subpatterns will be picked up then. */
6078    
6079              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
6080                       could_be_empty(called, code, bcptr, utf8, cd))                       could_be_empty(called, code, bcptr, utf8, cd))
6081                {                {
6082                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
# Line 5753  for (;; ptr++) Line 6084  for (;; ptr++)
6084                }                }
6085              }              }
6086    
6087            /* Insert the recursion/subroutine item. */            /* Insert the recursion/subroutine item. It does not have a set first
6088              byte (relevant if it is repeated, because it will then be wrapped
6089              with ONCE brackets). */
6090    
6091            *code = OP_RECURSE;            *code = OP_RECURSE;
6092            PUT(code, 1, (int)(called - cd->start_code));            PUT(code, 1, (int)(called - cd->start_code));
6093            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
6094              groupsetfirstbyte = FALSE;
6095            }            }
6096    
6097          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5877  for (;; ptr++) Line 6211  for (;; ptr++)
6211      repeated. We copy code into a non-register variable (tempcode) in order to      repeated. We copy code into a non-register variable (tempcode) in order to
6212      be able to pass its address because some compilers complain otherwise. */      be able to pass its address because some compilers complain otherwise. */
6213    
6214      previous = code;                   /* For handling repetition */      previous = code;                      /* For handling repetition */
6215      *code = bravalue;      *code = bravalue;
6216      tempcode = code;      tempcode = code;
6217      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
6218      length_prevgroup = 0;              /* Initialize for pre-compile phase */      tempbracount = cd->bracount;          /* Save value before bracket */
6219        length_prevgroup = 0;                 /* Initialize for pre-compile phase */
6220    
6221      if (!compile_regex(      if (!compile_regex(
6222           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
6223           &tempcode,                    /* Where to put code (updated) */           &tempcode,                       /* Where to put code (updated) */
6224           &ptr,                         /* Input pointer (updated) */           &ptr,                            /* Input pointer (updated) */
6225           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                    /* Where to put an error message */
6226           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
6227            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
6228           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
6229           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
6230           &subfirstbyte,                /* For possible first char */           cond_depth +
6231           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
6232           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
6233           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
6234           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
6235             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
6236             (lengthptr == NULL)? NULL :      /* Actual compile phase */
6237               &length_prevgroup              /* Pre-compile phase */
6238           ))           ))
6239        goto FAILED;        goto FAILED;
6240    
6241        /* If this was an atomic group and there are no capturing groups within it,
6242        generate OP_ONCE_NC instead of OP_ONCE. */
6243    
6244        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
6245          *code = OP_ONCE_NC;
6246    
6247      if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)      if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
6248        cd->assert_depth -= 1;        cd->assert_depth -= 1;
6249    
6250      /* 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
6251      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.
6252      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
6253    
6254      /* 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
6255      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
6256      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
6257      not be available. */      not be available. */
# Line 6150  for (;; ptr++) Line 6492  for (;; ptr++)
6492            ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))            ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6493            {            {
6494            *errorcodeptr = ERR69;            *errorcodeptr = ERR69;
6495            break;            break;
6496            }            }
6497          is_recurse = FALSE;          is_recurse = FALSE;
6498          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6499            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
6500            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
6501          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
6502          }          }
6503    
6504        /* Back references are handled specially; must disable firstbyte if        /* Back references are handled specially; must disable firstbyte if
6505        not set to cope with cases like (?=(\w+))\1: which would otherwise set        not set to cope with cases like (?=(\w+))\1: which would otherwise set
# Line 6231  for (;; ptr++) Line 6573  for (;; ptr++)
6573            }            }
6574          else          else
6575  #endif  #endif
6576            {          /* In non-UTF-8 mode, we turn \C into OP_ALLANY instead of OP_ANYBYTE
6577            so that it works in DFA mode and in lookbehinds. */
6578    
6579              {
6580            previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;            previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6581            *code++ = -c;            *code++ = (!utf8 && c == -ESC_C)? OP_ALLANY : -c;
6582            }            }
6583          }          }
6584        continue;        continue;
# Line 6308  for (;; ptr++) Line 6653  for (;; ptr++)
6653        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6654        }        }
6655    
6656      /* 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
6657      1 or the matching is caseful. */      1 or the matching is caseful. */
6658    
6659      else      else
# Line 6355  Arguments: Line 6700  Arguments:
6700    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6701    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6702    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6703      cond_depth     depth of nesting for conditional subpatterns
6704    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6705    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6706    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 6368  Returns:         TRUE on success Line 6714  Returns:         TRUE on success
6714  static BOOL  static BOOL
6715  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6716    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6717    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6718    int *lengthptr)    compile_data *cd, int *lengthptr)
6719  {  {
6720  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6721  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 6406  pre-compile phase to find out whether an Line 6752  pre-compile phase to find out whether an
6752    
6753  /* 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
6754  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
6755  detect groups that contain recursive back references to themselves. Note that  detect groups that contain recursive back references to themselves. Note that
6756  only OP_CBRA need be tested here; changing this opcode to one of its variants,  only OP_CBRA need be tested here; changing this opcode to one of its variants,
6757  e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */  e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6758    
6759  if (*code == OP_CBRA)  if (*code == OP_CBRA)
# Line 6448  for (;;) Line 6794  for (;;)
6794    into the length. */    into the length. */
6795    
6796    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6797          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6798            (lengthptr == NULL)? NULL : &length))
6799      {      {
6800      *ptrptr = ptr;      *ptrptr = ptr;
6801      return FALSE;      return FALSE;
# Line 6523  for (;;) Line 6870  for (;;)
6870          }          }
6871        else if (fixed_length < 0)        else if (fixed_length < 0)
6872          {          {
6873          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 :
6874                            (fixed_length == -4)? ERR70: ERR25;
6875          *ptrptr = ptr;          *ptrptr = ptr;
6876          return FALSE;          return FALSE;
6877          }          }
# Line 6698  do { Line 7046  do {
7046    
7047     /* Other brackets */     /* Other brackets */
7048    
7049     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC ||
7050                op == OP_COND)
7051       {       {
7052       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
7053       }       }
# Line 6802  do { Line 7151  do {
7151    
7152     /* Other brackets */     /* Other brackets */
7153    
7154     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC)
7155       {       {
7156       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
7157       }       }
# Line 6872  do { Line 7221  do {
7221       case OP_SCBRAPOS:       case OP_SCBRAPOS:
7222       case OP_ASSERT:       case OP_ASSERT:
7223       case OP_ONCE:       case OP_ONCE:
7224         case OP_ONCE_NC:
7225       case OP_COND:       case OP_COND:
7226       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
7227         return -1;         return -1;
# Line 6881  do { Line 7231  do {
7231       case OP_EXACT:       case OP_EXACT:
7232       scode += 2;       scode += 2;
7233       /* Fall through */       /* Fall through */
7234    
7235       case OP_CHAR:       case OP_CHAR:
7236       case OP_PLUS:       case OP_PLUS:
7237       case OP_MINPLUS:       case OP_MINPLUS:
# Line 6894  do { Line 7244  do {
7244       case OP_EXACTI:       case OP_EXACTI:
7245       scode += 2;       scode += 2;
7246       /* Fall through */       /* Fall through */
7247    
7248       case OP_CHARI:       case OP_CHARI:
7249       case OP_PLUSI:       case OP_PLUSI:
7250       case OP_MINPLUSI:       case OP_MINPLUSI:
# Line 6964  compile_data *cd = &compile_block; Line 7314  compile_data *cd = &compile_block;
7314  computing the amount of memory that is needed. Compiled items are thrown away  computing the amount of memory that is needed. Compiled items are thrown away
7315  as soon as possible, so that a fairly large buffer should be sufficient for  as soon as possible, so that a fairly large buffer should be sufficient for
7316  this purpose. The same space is used in the second phase for remembering where  this purpose. The same space is used in the second phase for remembering where
7317  to fill in forward references to subpatterns. */  to fill in forward references to subpatterns. That may overflow, in which case
7318    new memory is obtained from malloc(). */
7319    
7320  uschar cworkspace[COMPILE_WORK_SIZE];  uschar cworkspace[COMPILE_WORK_SIZE];
7321    
# Line 7054  utf8 = (options & PCRE_UTF8) != 0; Line 7405  utf8 = (options & PCRE_UTF8) != 0;
7405    
7406  /* 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
7407  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
7408  release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is  release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7409  not used here. */  not used here. */
7410    
7411  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 7087  if ((options & PCRE_UCP) != 0) Line 7438  if ((options & PCRE_UCP) != 0)
7438  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
7439       (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))       (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
7440    {    {
7441    errorcode = ERR56;    errorcode = ERR56;
7442    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
7443    }    }
7444    
# Line 7154  cd->bracount = cd->final_bracount = 0; Line 7505  cd->bracount = cd->final_bracount = 0;
7505  cd->names_found = 0;  cd->names_found = 0;
7506  cd->name_entry_size = 0;  cd->name_entry_size = 0;
7507  cd->name_table = NULL;  cd->name_table = NULL;
 cd->start_workspace = cworkspace;  
7508  cd->start_code = cworkspace;  cd->start_code = cworkspace;
7509  cd->hwm = cworkspace;  cd->hwm = cworkspace;
7510    cd->start_workspace = cworkspace;
7511    cd->workspace_size = COMPILE_WORK_SIZE;
7512  cd->start_pattern = (const uschar *)pattern;  cd->start_pattern = (const uschar *)pattern;
7513  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));
7514  cd->req_varyopt = 0;  cd->req_varyopt = 0;
# Line 7173  outside can help speed up starting point Line 7525  outside can help speed up starting point
7525  ptr += skipatstart;  ptr += skipatstart;
7526  code = cworkspace;  code = cworkspace;
7527  *code = OP_BRA;  *code = OP_BRA;
7528  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7529    FALSE, 0, &firstbyte, &reqbyte, NULL, cd, &length);    FALSE, 0, 0, &firstbyte, &reqbyte, NULL, cd, &length);
7530  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7531    
7532  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 7234  cd->names_found = 0; Line 7586  cd->names_found = 0;
7586  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
7587  codestart = cd->name_table + re->name_entry_size * re->name_count;  codestart = cd->name_table + re->name_entry_size * re->name_count;
7588  cd->start_code = codestart;  cd->start_code = codestart;
7589  cd->hwm = cworkspace;  cd->hwm = (uschar *)(cd->start_workspace);
7590  cd->req_varyopt = 0;  cd->req_varyopt = 0;
7591  cd->had_accept = FALSE;  cd->had_accept = FALSE;
7592  cd->check_lookbehind = FALSE;  cd->check_lookbehind = FALSE;
# Line 7247  of the function here. */ Line 7599  of the function here. */
7599  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7600  code = (uschar *)codestart;  code = (uschar *)codestart;
7601  *code = OP_BRA;  *code = OP_BRA;
7602  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0, 0,
7603    &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7604  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7605  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7606  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7607    
7608  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */  if (cd->had_accept) reqbyte = REQ_NONE;   /* Must disable after (*ACCEPT) */
7609    
7610  /* 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. */
7611    
# Line 7268  if debugging, leave the test till after Line 7620  if debugging, leave the test till after
7620  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
7621  #endif  #endif
7622    
7623  /* Fill in any forward references that are required. */  /* Fill in any forward references that are required. There may be repeated
7624    references; optimize for them, as searching a large regex takes time. */
7625    
7626  while (errorcode == 0 && cd->hwm > cworkspace)  if (cd->hwm > cd->start_workspace)
7627    {    {
7628    int offset, recno;    int prev_recno = -1;
7629    const uschar *groupptr;    const uschar *groupptr = NULL;
7630    cd->hwm -= LINK_SIZE;    while (errorcode == 0 && cd->hwm > cd->start_workspace)
7631    offset = GET(cd->hwm, 0);      {
7632    recno = GET(codestart, offset);      int offset, recno;
7633    groupptr = _pcre_find_bracket(codestart, utf8, recno);      cd->hwm -= LINK_SIZE;
7634    if (groupptr == NULL) errorcode = ERR53;      offset = GET(cd->hwm, 0);
7635      else PUT(((uschar *)codestart), offset, (int)(groupptr - codestart));      recno = GET(codestart, offset);
7636    }      if (recno != prev_recno)
7637          {
7638          groupptr = _pcre_find_bracket(codestart, utf8, recno);
7639          prev_recno = recno;
7640          }
7641        if (groupptr == NULL) errorcode = ERR53;
7642          else PUT(((uschar *)codestart), offset, (int)(groupptr - codestart));
7643        }
7644      }
7645    
7646    /* If the workspace had to be expanded, free the new memory. */
7647    
7648    if (cd->workspace_size > COMPILE_WORK_SIZE)
7649      (pcre_free)((void *)cd->start_workspace);
7650    
7651  /* Give an error if there's back reference to a non-existent capturing  /* Give an error if there's back reference to a non-existent capturing
7652  subpattern. */  subpattern. */
# Line 7320  if (cd->check_lookbehind) Line 7686  if (cd->check_lookbehind)
7686        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7687        if (fixed_length < 0)        if (fixed_length < 0)
7688          {          {
7689          errorcode = (fixed_length == -2)? ERR36 : ERR25;          errorcode = (fixed_length == -2)? ERR36 :
7690                        (fixed_length == -4)? ERR70 : ERR25;
7691          break;          break;
7692          }          }
7693        PUT(cc, 1, fixed_length);        PUT(cc, 1, fixed_length);

Legend:
Removed from v.638  
changed lines
  Added in v.773

  ViewVC Help
Powered by ViewVC 1.1.5