/[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 640 by ph10, Mon Jul 25 10:50:28 2011 UTC revision 751 by ph10, Fri Nov 18 11:13:30 2011 UTC
# 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"    "\\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 546  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, uschar **, const uschar **, int *, BOOL, BOOL, int, int *,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
552      int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
553    
554    
555    
# Line 676  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      /* In a character class, \g is just a literal "g". Outside a character      /* 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:      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
# Line 828  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 921  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}. However, it does support  newline". PCRE does not support \N{name}. However, it does support
978  quantification such as \N{2,3}. */  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 &&
# Line 1477  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
# Line 1500  for (;;) Line 1554  for (;;)
1554      /* We only need to continue for OP_CBRA (normal capturing bracket) and      /* 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      OP_BRA (normal non-capturing bracket) because the other variants of these
1556      opcodes are all concerned with unlimited repeated groups, which of course      opcodes are all concerned with unlimited repeated groups, which of course
1557      are not of fixed length. They will cause a -1 response from the default      are not of fixed length. */
     case of this switch. */  
1558    
1559      case OP_CBRA:      case OP_CBRA:
1560      case OP_BRA:      case OP_BRA:
1561      case OP_ONCE:      case OP_ONCE:
1562        case OP_ONCE_NC:
1563      case OP_COND:      case OP_COND:
1564      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, 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;
# Line 1514  for (;;) Line 1568  for (;;)
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      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,
1575      all imply an unlimited repeat. */      because they all imply an unlimited repeat. */
1576    
1577      case OP_ALT:      case OP_ALT:
1578      case OP_KET:      case OP_KET:
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 1556  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 1594  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
# Line 1614  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 1645  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 1665  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 1761  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 1880  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 1982  for (code = first_significant_code(code Line 2143  for (code = first_significant_code(code
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 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
2145    forward reference subroutine call, we can't. To detect forward reference    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    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    called only when doing the real compile, not during the pre-compile that
2148    measures the size of the compiled pattern. */    measures the size of the compiled pattern. */
2149    
2150    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2151      {      {
2152      const uschar *scode;      const uschar *scode;
2153      BOOL empty_branch;      BOOL empty_branch;
2154    
2155      /* Test for forward reference */      /* Test for forward reference */
2156    
2157      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2158        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2159    
2160      /* Not a forward reference, test for completed backward reference */      /* Not a forward reference, test for completed backward reference */
2161    
2162      empty_branch = FALSE;      empty_branch = FALSE;
2163      scode = cd->start_code + GET(code, 1);      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 */      /* 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 2014  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      }      }
# Line 2045  for (code = first_significant_code(code Line 2206  for (code = first_significant_code(code
2206    
2207    if (c == OP_BRA  || c == OP_BRAPOS ||    if (c == OP_BRA  || c == OP_BRAPOS ||
2208        c == OP_CBRA || c == OP_CBRAPOS ||        c == OP_CBRA || c == OP_CBRAPOS ||
2209        c == OP_ONCE || c == OP_COND)        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 2217  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 2240  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  This function is called only during the real compile, not during the
2406  pre-compile.  pre-compile.
2407    
2408  Arguments:  Arguments:
# Line 2295  I think. Line 2457  I think.
2457  A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.  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  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  external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2460  a digit. Also, unescaped square brackets may also appear as part of class  a digit.
2461  names. For example, [:a[:abc]b:] gives unknown class "[:abc]b:]"in Perl.  
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 [
# Line 2312  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)    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2483      ptr++;      ptr++;
2484      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2485    else    else
2486      {      {
2487      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
# Line 2325  for (++ptr; *ptr != 0; ptr++) Line 2493  for (++ptr; *ptr != 0; ptr++)
2493           (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||           (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2494            ptr[1] == CHAR_EQUALS_SIGN) &&            ptr[1] == CHAR_EQUALS_SIGN) &&
2495          check_posix_syntax(ptr, endptr))          check_posix_syntax(ptr, endptr))
2496        return FALSE;        return FALSE;
2497      }      }
2498    }    }
2499  return FALSE;  return FALSE;
# Line 3035  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 3046  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 3074  uschar *save_hwm = NULL; Line 3243  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  /* 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  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3247  dynamically as we process the pattern. */  dynamically as we process the pattern. */
3248    
3249  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 3085  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 3136  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 3183  for (;; ptr++) Line 3352  for (;; ptr++)
3352        }        }
3353    
3354      *lengthptr += (int)(code - last_code);      *lengthptr += (int)(code - last_code);
3355      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, (int)(code - last_code),
3356          c));
3357    
3358      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
3359      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
# Line 4249  for (;; ptr++) Line 4419  for (;; ptr++)
4419        ptr++;        ptr++;
4420        }        }
4421      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4422    
4423      /* If previous was a recursion call, wrap it in atomic brackets so that      /* If previous was a recursion call, wrap it in atomic brackets so that
4424      previous becomes the atomic group. All recursions were so wrapped in the      previous becomes the atomic group. All recursions were so wrapped in the
4425      past, but it no longer happens for non-repeated recursions. In fact, the      past, but it no longer happens for non-repeated recursions. In fact, the
4426      repeated ones could be re-implemented independently so as not to need this,      repeated ones could be re-implemented independently so as not to need this,
# Line 4265  for (;; ptr++) Line 4435  for (;; ptr++)
4435        PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);        PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4436        code += 2 + 2 * LINK_SIZE;        code += 2 + 2 * LINK_SIZE;
4437        length_prevgroup = 3 + 3*LINK_SIZE;        length_prevgroup = 3 + 3*LINK_SIZE;
4438    
4439        /* When actually compiling, we need to check whether this was a forward        /* When actually compiling, we need to check whether this was a forward
4440        reference, and if so, adjust the offset. */        reference, and if so, adjust the offset. */
4441    
4442        if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)        if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4443          {          {
4444          int offset = GET(cd->hwm, -LINK_SIZE);          int offset = GET(cd->hwm, -LINK_SIZE);
4445          if (offset == previous + 1 - cd->start_code)          if (offset == previous + 1 - cd->start_code)
4446            PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);            PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4447          }          }
4448        }        }
4449    
4450      /* Now handle repetition for the different types of item. */      /* Now handle repetition for the different types of item. */
4451    
4452      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
# Line 4574  for (;; ptr++) Line 4744  for (;; ptr++)
4744      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
4745      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 >=
4746      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4747      ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow      ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4748      repetition of assertions, but now it does, for Perl compatibility. */      repetition of assertions, but now it does, for Perl compatibility. */
4749    
4750      else if (*previous >= OP_ASSERT && *previous <= OP_COND)      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
4751        {        {
4752        register int i;        register int i;
4753        int len = (int)(code - previous);        int len = (int)(code - previous);
4754        uschar *bralink = NULL;        uschar *bralink = NULL;
4755        uschar *brazeroptr = NULL;        uschar *brazeroptr = NULL;
4756    
4757        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4758        we just ignore the repeat. */        we just ignore the repeat. */
4759    
4760        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4761          goto END_REPEAT;          goto END_REPEAT;
4762    
4763        /* There is no sense in actually repeating assertions. The only potential        /* There is no sense in actually repeating assertions. The only potential
4764        use of repetition is in cases when the assertion is optional. Therefore,        use of repetition is in cases when the assertion is optional. Therefore,
4765        if the minimum is greater than zero, just ignore the repeat. If the        if the minimum is greater than zero, just ignore the repeat. If the
4766        maximum is not not zero or one, set it to 1. */        maximum is not not zero or one, set it to 1. */
4767    
4768        if (*previous < OP_ONCE)    /* Assertion */        if (*previous < OP_ONCE)    /* Assertion */
4769          {          {
4770          if (repeat_min > 0) goto END_REPEAT;          if (repeat_min > 0) goto END_REPEAT;
4771          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
4772          }          }
4773    
4774        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
4775        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 4813  for (;; ptr++) Line 4983  for (;; ptr++)
4983        ONCE brackets can be converted into non-capturing brackets, as the        ONCE brackets can be converted into non-capturing brackets, as the
4984        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
4985        deal with possessive ONCEs specially.        deal with possessive ONCEs specially.
4986    
4987        Otherwise, if the quantifier was possessive, we convert the BRA code to        Otherwise, when we are doing the actual compile phase, check to see
4988        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,
4989        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
4990        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
4991        stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,        groups at runtime, but in a different way.]
4992        convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that  
4993        the default action below, of wrapping everything inside atomic brackets,        Then, if the quantifier was possessive and the bracket is not a
4994        does not happen.        conditional, we convert the BRA code to the POS form, and the KET code to
4995          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
4996        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
4997        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
4998        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.
4999        checking can be done. [This check is also applied to ONCE groups at  
5000        runtime, but in a different way.] */        Then, if the minimum number of matches is 1 or 0, cancel the possessive
5001          flag so that the default action below, of wrapping everything inside
5002          atomic brackets, does not happen. When the minimum is greater than 1,
5003          there will be earlier copies of the group, and so we still have to wrap
5004          the whole thing. */
5005    
5006        else        else
5007          {          {
5008          uschar *ketcode = code - 1 - LINK_SIZE;          uschar *ketcode = code - 1 - LINK_SIZE;
5009          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
5010    
5011            /* Convert possessive ONCE brackets to non-capturing */
5012    
5013          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;          if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5014          if (*bracode == OP_ONCE)              possessive_quantifier) *bracode = OP_BRA;
5015    
5016            /* For non-possessive ONCE brackets, all we need to do is to
5017            set the KET. */
5018    
5019            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5020            *ketcode = OP_KETRMAX + repeat_type;            *ketcode = OP_KETRMAX + repeat_type;
5021    
5022            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5023            converted to non-capturing above). */
5024    
5025          else          else
5026            {            {
5027            if (possessive_quantifier)            /* In the compile phase, check for empty string matching. */
5028              {  
             *bracode += 1;                   /* Switch to xxxPOS opcodes */  
             *ketcode = OP_KETRPOS;  
             if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;  
             possessive_quantifier = FALSE;  
             }  
           else *ketcode = OP_KETRMAX + repeat_type;  
   
5029            if (lengthptr == NULL)            if (lengthptr == NULL)
5030              {              {
5031              uschar *scode = bracode;              uschar *scode = bracode;
# Line 4862  for (;; ptr++) Line 5040  for (;; ptr++)
5040                }                }
5041              while (*scode == OP_ALT);              while (*scode == OP_ALT);
5042              }              }
5043    
5044              /* Handle possessive quantifiers. */
5045    
5046              if (possessive_quantifier)
5047                {
5048                /* For COND brackets, we wrap the whole thing in a possessively
5049                repeated non-capturing bracket, because we have not invented POS
5050                versions of the COND opcodes. Because we are moving code along, we
5051                must ensure that any pending recursive references are updated. */
5052    
5053                if (*bracode == OP_COND || *bracode == OP_SCOND)
5054                  {
5055                  int nlen = (int)(code - bracode);
5056                  *code = OP_END;
5057                  adjust_recurse(bracode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5058                  memmove(bracode + 1+LINK_SIZE, bracode, nlen);
5059                  code += 1 + LINK_SIZE;
5060                  nlen += 1 + LINK_SIZE;
5061                  *bracode = OP_BRAPOS;
5062                  *code++ = OP_KETRPOS;
5063                  PUTINC(code, 0, nlen);
5064                  PUT(bracode, 1, nlen);
5065                  }
5066    
5067                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5068    
5069                else
5070                  {
5071                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5072                  *ketcode = OP_KETRPOS;
5073                  }
5074    
5075                /* If the minimum is zero, mark it as possessive, then unset the
5076                possessive flag when the minimum is 0 or 1. */
5077    
5078                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5079                if (repeat_min < 2) possessive_quantifier = FALSE;
5080                }
5081    
5082              /* Non-possessive quantifier */
5083    
5084              else *ketcode = OP_KETRMAX + repeat_type;
5085            }            }
5086          }          }
5087        }        }
# Line 4886  for (;; ptr++) Line 5106  for (;; ptr++)
5106      there are special alternative opcodes for this case. For anything else, we      there are special alternative opcodes for this case. For anything else, we
5107      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
5108      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
5109      special opcodes can optimize it.      special opcodes can optimize it.
5110    
5111      Possessively repeated subpatterns have already been handled in the code      Some (but not all) possessively repeated subpatterns have already been
5112      just above, so possessive_quantifier is always FALSE for them at this      completely handled in the code just above. For them, possessive_quantifier
5113      stage.      is always FALSE at this stage.
5114    
5115      Note that the repeated item starts at tempcode, not at previous, which      Note that the repeated item starts at tempcode, not at previous, which
5116      might be the first part of a string whose (former) last char we repeated.      might be the first part of a string whose (former) last char we repeated.
5117    
# Line 4997  for (;; ptr++) Line 5217  for (;; ptr++)
5217        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5218        namelen = (int)(ptr - name);        namelen = (int)(ptr - name);
5219    
5220          /* It appears that Perl allows any characters whatsoever, other than
5221          a closing parenthesis, to appear in arguments, so we no longer insist on
5222          letters, digits, and underscores. */
5223    
5224        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5225          {          {
5226          arg = ++ptr;          arg = ++ptr;
5227          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
           || *ptr == '_') ptr++;  
5228          arglen = (int)(ptr - arg);          arglen = (int)(ptr - arg);
5229          }          }
5230    
# Line 5018  for (;; ptr++) Line 5241  for (;; ptr++)
5241          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5242              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5243            {            {
5244            /* Check for open captures before ACCEPT and convert it to            /* Check for open captures before ACCEPT and convert it to
5245            ASSERT_ACCEPT if in an assertion. */            ASSERT_ACCEPT if in an assertion. */
5246    
5247            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
# Line 5028  for (;; ptr++) Line 5251  for (;; ptr++)
5251                {                {
5252                *errorcodeptr = ERR59;                *errorcodeptr = ERR59;
5253                goto FAILED;                goto FAILED;
5254                }                }
5255              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5256              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5257                {                {
# Line 5036  for (;; ptr++) Line 5259  for (;; ptr++)
5259                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5260                }                }
5261              *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;              *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5262    
5263                /* Do not set firstbyte after *ACCEPT */
5264                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5265              }              }
5266    
5267            /* Handle other cases with/without an argument */            /* Handle other cases with/without an argument */
# Line 5048  for (;; ptr++) Line 5274  for (;; ptr++)
5274                goto FAILED;                goto FAILED;
5275                }                }
5276              *code = verbs[i].op;              *code = verbs[i].op;
5277              if (*code++ == OP_THEN)              if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
               {  
               PUT(code, 0, code - bcptr->current_branch - 1);  
               code += LINK_SIZE;  
               }  
5278              }              }
5279    
5280            else            else
# Line 5063  for (;; ptr++) Line 5285  for (;; ptr++)
5285                goto FAILED;                goto FAILED;
5286                }                }
5287              *code = verbs[i].op_arg;              *code = verbs[i].op_arg;
5288              if (*code++ == OP_THEN_ARG)              if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
               {  
               PUT(code, 0, code - bcptr->current_branch - 1);  
               code += LINK_SIZE;  
               }  
5289              *code++ = arglen;              *code++ = arglen;
5290              memcpy(code, arg, arglen);              memcpy(code, arg, arglen);
5291              code += arglen;              code += arglen;
# Line 5329  for (;; ptr++) Line 5547  for (;; ptr++)
5547          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5548          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5549          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5550          cd->assert_depth += 1;          cd->assert_depth += 1;
5551          ptr++;          ptr++;
5552          break;          break;
5553    
# Line 5344  for (;; ptr++) Line 5562  for (;; ptr++)
5562            continue;            continue;
5563            }            }
5564          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5565          cd->assert_depth += 1;          cd->assert_depth += 1;
5566          break;          break;
5567    
5568    
# Line 5354  for (;; ptr++) Line 5572  for (;; ptr++)
5572            {            {
5573            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5574            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5575            cd->assert_depth += 1;            cd->assert_depth += 1;
5576            ptr += 2;            ptr += 2;
5577            break;            break;
5578    
5579            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5580            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5581            cd->assert_depth += 1;            cd->assert_depth += 1;
5582            ptr += 2;            ptr += 2;
5583            break;            break;
5584    
# Line 5604  for (;; ptr++) Line 5822  for (;; ptr++)
5822    
5823            temp = cd->end_pattern;            temp = cd->end_pattern;
5824            cd->end_pattern = ptr;            cd->end_pattern = ptr;
5825            recno = find_parens(cd, name, namelen,            recno = find_parens(cd, name, namelen,
5826              (options & PCRE_EXTENDED) != 0, utf8);              (options & PCRE_EXTENDED) != 0, utf8);
5827            cd->end_pattern = temp;            cd->end_pattern = temp;
5828            if (recno < 0) recno = 0;    /* Forward ref; set dummy number */            if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
# Line 5759  for (;; ptr++) Line 5977  for (;; ptr++)
5977    
5978              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
5979              this is a recursive call. We check to see if this is a left              this is a recursive call. We check to see if this is a left
5980              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
5981                must not, however, do this check if we are in a conditional
5982                subpattern because the condition might be testing for recursion in
5983                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
5984                Forever loops are also detected at runtime, so those that occur in
5985                conditional subpatterns will be picked up then. */
5986    
5987              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
5988                       could_be_empty(called, code, bcptr, utf8, cd))                       could_be_empty(called, code, bcptr, utf8, cd))
5989                {                {
5990                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
# Line 5770  for (;; ptr++) Line 5993  for (;; ptr++)
5993              }              }
5994    
5995            /* Insert the recursion/subroutine item. */            /* Insert the recursion/subroutine item. */
5996    
5997            *code = OP_RECURSE;            *code = OP_RECURSE;
5998            PUT(code, 1, (int)(called - cd->start_code));            PUT(code, 1, (int)(called - cd->start_code));
5999            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
# Line 5893  for (;; ptr++) Line 6116  for (;; ptr++)
6116      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
6117      be able to pass its address because some compilers complain otherwise. */      be able to pass its address because some compilers complain otherwise. */
6118    
6119      previous = code;                   /* For handling repetition */      previous = code;                      /* For handling repetition */
6120      *code = bravalue;      *code = bravalue;
6121      tempcode = code;      tempcode = code;
6122      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
6123      length_prevgroup = 0;              /* Initialize for pre-compile phase */      tempbracount = cd->bracount;          /* Save value before bracket */
6124        length_prevgroup = 0;                 /* Initialize for pre-compile phase */
6125    
6126      if (!compile_regex(      if (!compile_regex(
6127           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
6128           &tempcode,                    /* Where to put code (updated) */           &tempcode,                       /* Where to put code (updated) */
6129           &ptr,                         /* Input pointer (updated) */           &ptr,                            /* Input pointer (updated) */
6130           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                    /* Where to put an error message */
6131           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
6132            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
6133           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
6134           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
6135           &subfirstbyte,                /* For possible first char */           cond_depth +
6136           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
6137           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
6138           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
6139           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
6140             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
6141             (lengthptr == NULL)? NULL :      /* Actual compile phase */
6142               &length_prevgroup              /* Pre-compile phase */
6143           ))           ))
6144        goto FAILED;        goto FAILED;
6145    
6146        /* If this was an atomic group and there are no capturing groups within it,
6147        generate OP_ONCE_NC instead of OP_ONCE. */
6148    
6149        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
6150          *code = OP_ONCE_NC;
6151    
6152      if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)      if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
6153        cd->assert_depth -= 1;        cd->assert_depth -= 1;
6154    
6155      /* At the end of compiling, code is still pointing to the start of the      /* At the end of compiling, code is still pointing to the start of the
6156      group, while tempcode has been updated to point past the end of the group      group, while tempcode has been updated to point past the end of the group.
6157      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
6158    
6159      /* If this is a conditional bracket, check that there are no more than      If this is a conditional bracket, check that there are no more than
6160      two branches in the group, or just one if it's a DEFINE group. We do this      two branches in the group, or just one if it's a DEFINE group. We do this
6161      in the real compile phase, not in the pre-pass, where the whole group may      in the real compile phase, not in the pre-pass, where the whole group may
6162      not be available. */      not be available. */
# Line 6166  for (;; ptr++) Line 6397  for (;; ptr++)
6397            ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))            ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6398            {            {
6399            *errorcodeptr = ERR69;            *errorcodeptr = ERR69;
6400            break;            break;
6401            }            }
6402          is_recurse = FALSE;          is_recurse = FALSE;
6403          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6404            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
6405            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
6406          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
6407          }          }
6408    
6409        /* Back references are handled specially; must disable firstbyte if        /* Back references are handled specially; must disable firstbyte if
6410        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 6324  for (;; ptr++) Line 6555  for (;; ptr++)
6555        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6556        }        }
6557    
6558      /* firstbyte was previously set; we can set reqbyte only the length is      /* firstbyte was previously set; we can set reqbyte only if the length is
6559      1 or the matching is caseful. */      1 or the matching is caseful. */
6560    
6561      else      else
# Line 6371  Arguments: Line 6602  Arguments:
6602    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6603    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6604    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6605      cond_depth     depth of nesting for conditional subpatterns
6606    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6607    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6608    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 6384  Returns:         TRUE on success Line 6616  Returns:         TRUE on success
6616  static BOOL  static BOOL
6617  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6618    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6619    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6620    int *lengthptr)    compile_data *cd, int *lengthptr)
6621  {  {
6622  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6623  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 6422  pre-compile phase to find out whether an Line 6654  pre-compile phase to find out whether an
6654    
6655  /* If this is a capturing subpattern, add to the chain of open capturing items  /* If this is a capturing subpattern, add to the chain of open capturing items
6656  so that we can detect them if (*ACCEPT) is encountered. This is also used to  so that we can detect them if (*ACCEPT) is encountered. This is also used to
6657  detect groups that contain recursive back references to themselves. Note that  detect groups that contain recursive back references to themselves. Note that
6658  only OP_CBRA need be tested here; changing this opcode to one of its variants,  only OP_CBRA need be tested here; changing this opcode to one of its variants,
6659  e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */  e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6660    
6661  if (*code == OP_CBRA)  if (*code == OP_CBRA)
# Line 6464  for (;;) Line 6696  for (;;)
6696    into the length. */    into the length. */
6697    
6698    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6699          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6700            (lengthptr == NULL)? NULL : &length))
6701      {      {
6702      *ptrptr = ptr;      *ptrptr = ptr;
6703      return FALSE;      return FALSE;
# Line 6539  for (;;) Line 6772  for (;;)
6772          }          }
6773        else if (fixed_length < 0)        else if (fixed_length < 0)
6774          {          {
6775          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 :
6776                            (fixed_length == -4)? ERR70: ERR25;
6777          *ptrptr = ptr;          *ptrptr = ptr;
6778          return FALSE;          return FALSE;
6779          }          }
# Line 6714  do { Line 6948  do {
6948    
6949     /* Other brackets */     /* Other brackets */
6950    
6951     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC ||
6952                op == OP_COND)
6953       {       {
6954       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6955       }       }
# Line 6818  do { Line 7053  do {
7053    
7054     /* Other brackets */     /* Other brackets */
7055    
7056     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC)
7057       {       {
7058       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
7059       }       }
# Line 6888  do { Line 7123  do {
7123       case OP_SCBRAPOS:       case OP_SCBRAPOS:
7124       case OP_ASSERT:       case OP_ASSERT:
7125       case OP_ONCE:       case OP_ONCE:
7126         case OP_ONCE_NC:
7127       case OP_COND:       case OP_COND:
7128       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
7129         return -1;         return -1;
# Line 6897  do { Line 7133  do {
7133       case OP_EXACT:       case OP_EXACT:
7134       scode += 2;       scode += 2;
7135       /* Fall through */       /* Fall through */
7136    
7137       case OP_CHAR:       case OP_CHAR:
7138       case OP_PLUS:       case OP_PLUS:
7139       case OP_MINPLUS:       case OP_MINPLUS:
# Line 6910  do { Line 7146  do {
7146       case OP_EXACTI:       case OP_EXACTI:
7147       scode += 2;       scode += 2;
7148       /* Fall through */       /* Fall through */
7149    
7150       case OP_CHARI:       case OP_CHARI:
7151       case OP_PLUSI:       case OP_PLUSI:
7152       case OP_MINPLUSI:       case OP_MINPLUSI:
# Line 7070  utf8 = (options & PCRE_UTF8) != 0; Line 7306  utf8 = (options & PCRE_UTF8) != 0;
7306    
7307  /* 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
7308  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
7309  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
7310  not used here. */  not used here. */
7311    
7312  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 7103  if ((options & PCRE_UCP) != 0) Line 7339  if ((options & PCRE_UCP) != 0)
7339  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
7340       (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))       (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
7341    {    {
7342    errorcode = ERR56;    errorcode = ERR56;
7343    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
7344    }    }
7345    
# Line 7189  outside can help speed up starting point Line 7425  outside can help speed up starting point
7425  ptr += skipatstart;  ptr += skipatstart;
7426  code = cworkspace;  code = cworkspace;
7427  *code = OP_BRA;  *code = OP_BRA;
7428  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7429    FALSE, 0, &firstbyte, &reqbyte, NULL, cd, &length);    FALSE, 0, 0, &firstbyte, &reqbyte, NULL, cd, &length);
7430  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7431    
7432  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 7263  of the function here. */ Line 7499  of the function here. */
7499  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7500  code = (uschar *)codestart;  code = (uschar *)codestart;
7501  *code = OP_BRA;  *code = OP_BRA;
7502  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0, 0,
7503    &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7504  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7505  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7506  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7507    
7508  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */  if (cd->had_accept) reqbyte = REQ_NONE;   /* Must disable after (*ACCEPT) */
7509    
7510  /* If not reached end of pattern on success, there's an excess bracket. */  /* If not reached end of pattern on success, there's an excess bracket. */
7511    
# Line 7336  if (cd->check_lookbehind) Line 7572  if (cd->check_lookbehind)
7572        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7573        if (fixed_length < 0)        if (fixed_length < 0)
7574          {          {
7575          errorcode = (fixed_length == -2)? ERR36 : ERR25;          errorcode = (fixed_length == -2)? ERR36 :
7576                        (fixed_length == -4)? ERR70 : ERR25;
7577          break;          break;
7578          }          }
7579        PUT(cc, 1, fixed_length);        PUT(cc, 1, fixed_length);

Legend:
Removed from v.640  
changed lines
  Added in v.751

  ViewVC Help
Powered by ViewVC 1.1.5