/[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 282 by ph10, Fri Dec 7 19:32:32 2007 UTC revision 359 by ph10, Wed Jul 9 16:20:19 2008 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2008 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 158  static const char verbnames[] = Line 158  static const char verbnames[] =
158    "SKIP\0"    "SKIP\0"
159    "THEN";    "THEN";
160    
161  static verbitem verbs[] = {  static const verbitem verbs[] = {
162    { 6, OP_ACCEPT },    { 6, OP_ACCEPT },
163    { 6, OP_COMMIT },    { 6, OP_COMMIT },
164    { 1, OP_FAIL },    { 1, OP_FAIL },
# Line 168  static verbitem verbs[] = { Line 168  static verbitem verbs[] = {
168    { 4, OP_THEN  }    { 4, OP_THEN  }
169  };  };
170    
171  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
172    
173    
174  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
# Line 295  static const char error_texts[] = Line 295  static const char error_texts[] =
295    /* 55 */    /* 55 */
296    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"
297    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
298    "\\g is not followed by a braced name or an optionally braced non-zero number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
299    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number\0"    "a numbered reference must not be zero\0"
300    "(*VERB) with an argument is not supported\0"    "(*VERB) with an argument is not supported\0"
301    /* 60 */    /* 60 */
302    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
303    "number is too big\0"    "number is too big\0"
304    "subpattern name expected\0"    "subpattern name expected\0"
305    "digit expected after (?+";    "digit expected after (?+\0"
306      "] is an invalid data character in JavaScript compatibility mode";
307    
308    
309  /* 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 531  else Line 532  else
532      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
533      break;      break;
534    
535      /* \g must be followed by a number, either plain or braced. If positive, it      /* \g must be followed by one of a number of specific things:
536      is an absolute backreference. If negative, it is a relative backreference.  
537      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a      (1) A number, either plain or braced. If positive, it is an absolute
538      reference to a named group. This is part of Perl's movement towards a      backreference. If negative, it is a relative backreference. This is a Perl
539      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
540      fudge it up by pretending it really was \k. */  
541        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
542        is part of Perl's movement towards a unified syntax for back references. As
543        this is synonymous with \k{name}, we fudge it up by pretending it really
544        was \k.
545    
546        (3) For Oniguruma compatibility we also support \g followed by a name or a
547        number either in angle brackets or in single quotes. However, these are
548        (possibly recursive) subroutine calls, _not_ backreferences. Just return
549        the -ESC_g code (cf \k). */
550    
551      case 'g':      case 'g':
552        if (ptr[1] == '<' || ptr[1] == '\'')
553          {
554          c = -ESC_g;
555          break;
556          }
557    
558        /* Handle the Perl-compatible cases */
559    
560      if (ptr[1] == '{')      if (ptr[1] == '{')
561        {        {
562        const uschar *p;        const uschar *p;
# Line 565  else Line 583  else
583      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
584        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - '0';
585    
586      if (c < 0)      if (c < 0)   /* Integer overflow */
587        {        {
588        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
589        break;        break;
590        }        }
591    
592      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != '}')
593        {        {
594        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
595        break;        break;
596        }        }
597    
598        if (c == 0)
599          {
600          *errorcodeptr = ERR58;
601          break;
602          }
603    
604      if (negated)      if (negated)
605        {        {
606        if (c > bracount)        if (c > bracount)
# Line 611  else Line 635  else
635        c -= '0';        c -= '0';
636        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
637          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
638        if (c < 0)        if (c < 0)    /* Integer overflow */
639          {          {
640          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
641          break;          break;
# Line 952  be terminated by '>' because that is che Line 976  be terminated by '>' because that is che
976    
977  Arguments:  Arguments:
978    ptr          current position in the pattern    ptr          current position in the pattern
979    count        current count of capturing parens so far encountered    cd           compile background data
980    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
981    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
982    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
# Line 961  Returns:       the number of the named s Line 985  Returns:       the number of the named s
985  */  */
986    
987  static int  static int
988  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  find_parens(const uschar *ptr, compile_data *cd, const uschar *name, int lorn,
989    BOOL xmode)    BOOL xmode)
990  {  {
991  const uschar *thisname;  const uschar *thisname;
992    int count = cd->bracount;
993    
994  for (; *ptr != 0; ptr++)  for (; *ptr != 0; ptr++)
995    {    {
# Line 984  for (; *ptr != 0; ptr++) Line 1009  for (; *ptr != 0; ptr++)
1009      continue;      continue;
1010      }      }
1011    
1012    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1013      are handled for real. If the first character is '^', skip it. Also, if the
1014      first few characters (either before or after ^) are \Q\E or \E we skip them
1015      too. This makes for compatibility with Perl. */
1016    
1017    if (*ptr == '[')    if (*ptr == '[')
1018      {      {
1019        BOOL negate_class = FALSE;
1020        for (;;)
1021          {
1022          int c = *(++ptr);
1023          if (c == '\\')
1024            {
1025            if (ptr[1] == 'E') ptr++;
1026              else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
1027                else break;
1028            }
1029          else if (!negate_class && c == '^')
1030            negate_class = TRUE;
1031          else break;
1032          }
1033    
1034        /* If the next character is ']', it is a data character that must be
1035        skipped, except in JavaScript compatibility mode. */
1036    
1037        if (ptr[1] == ']' && (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1038          ptr++;
1039    
1040      while (*(++ptr) != ']')      while (*(++ptr) != ']')
1041        {        {
1042        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
# Line 1252  for (;;) Line 1301  for (;;)
1301      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1302      case OP_WORDCHAR:      case OP_WORDCHAR:
1303      case OP_ANY:      case OP_ANY:
1304        case OP_ALLANY:
1305      branchlength++;      branchlength++;
1306      cc++;      cc++;
1307      break;      break;
# Line 1531  for (code = first_significant_code(code Line 1581  for (code = first_significant_code(code
1581    const uschar *ccode;    const uschar *ccode;
1582    
1583    c = *code;    c = *code;
1584    
1585    /* Skip over forward assertions; the other assertions are skipped by    /* Skip over forward assertions; the other assertions are skipped by
1586    first_significant_code() with a TRUE final argument. */    first_significant_code() with a TRUE final argument. */
1587    
1588    if (c == OP_ASSERT)    if (c == OP_ASSERT)
1589      {      {
1590      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1591      c = *code;      c = *code;
1592      continue;      continue;
1593      }      }
1594    
1595    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1596    
1597    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1598      {      {
1599      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1600      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1630  for (code = first_significant_code(code Line 1680  for (code = first_significant_code(code
1680      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1681      case OP_WORDCHAR:      case OP_WORDCHAR:
1682      case OP_ANY:      case OP_ANY:
1683        case OP_ALLANY:
1684      case OP_ANYBYTE:      case OP_ANYBYTE:
1685      case OP_CHAR:      case OP_CHAR:
1686      case OP_CHARNC:      case OP_CHARNC:
# Line 1737  return TRUE; Line 1788  return TRUE;
1788  *************************************************/  *************************************************/
1789    
1790  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1791  encountered in a character class. It checks whether this is followed by an  encountered in a character class. It checks whether this is followed by a
1792  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1793  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1794    
1795    Originally, this function only recognized a sequence of letters between the
1796    terminators, but it seems that Perl recognizes any sequence of characters,
1797    though of course unknown POSIX names are subsequently rejected. Perl gives an
1798    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1799    didn't consider this to be a POSIX class. Likewise for [:1234:].
1800    
1801    The problem in trying to be exactly like Perl is in the handling of escapes. We
1802    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1803    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1804    below handles the special case of \], but does not try to do any other escape
1805    processing. This makes it different from Perl for cases such as [:l\ower:]
1806    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1807    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1808    I think.
1809    
1810  Argument:  Arguments:
1811    ptr      pointer to the initial [    ptr      pointer to the initial [
1812    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1813    
1814  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1815  */  */
1816    
1817  static BOOL  static BOOL
1818  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1819  {  {
1820  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1821  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1822  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
1823    {    {
1824    *endptr = ptr;    if (*ptr == '\\' && ptr[1] == ']') ptr++; else
1825    return TRUE;      {
1826        if (*ptr == ']') return FALSE;
1827        if (*ptr == terminator && ptr[1] == ']')
1828          {
1829          *endptr = ptr;
1830          return TRUE;
1831          }
1832        }
1833    }    }
1834  return FALSE;  return FALSE;
1835  }  }
# Line 1805  return -1; Line 1875  return -1;
1875  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
1876  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
1877  earlier groups that are outside the current group). However, when a group is  earlier groups that are outside the current group). However, when a group is
1878  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
1879  it, after it has been compiled. This means that any OP_RECURSE items within it  inserted before it, after it has been compiled. This means that any OP_RECURSE
1880  that refer to the group itself or any contained groups have to have their  items within it that refer to the group itself or any contained groups have to
1881  offsets adjusted. That one of the jobs of this function. Before it is called,  have their offsets adjusted. That one of the jobs of this function. Before it
1882  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
1883    OP_END.
1884    
1885  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
1886  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1944  get_othercase_range(unsigned int *cptr, Line 2015  get_othercase_range(unsigned int *cptr,
2015  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2016    
2017  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2018    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2019    
2020  if (c > d) return FALSE;  if (c > d) return FALSE;
2021    
# Line 1953  next = othercase + 1; Line 2024  next = othercase + 1;
2024    
2025  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2026    {    {
2027    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2028    next++;    next++;
2029    }    }
2030    
# Line 2081  if (next >= 0) switch(op_code) Line 2152  if (next >= 0) switch(op_code)
2152      unsigned int othercase;      unsigned int othercase;
2153      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2154  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2155      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2156  #else  #else
2157      othercase = NOTACHAR;      othercase = NOTACHAR;
2158  #endif  #endif
# Line 2094  if (next >= 0) switch(op_code) Line 2165  if (next >= 0) switch(op_code)
2165    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2166    
2167    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2168    if (item == next) return TRUE;    if (item == next) return TRUE;
2169    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2170  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 2103  if (next >= 0) switch(op_code) Line 2173  if (next >= 0) switch(op_code)
2173      unsigned int othercase;      unsigned int othercase;
2174      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2175  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2176      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2177  #else  #else
2178      othercase = NOTACHAR;      othercase = NOTACHAR;
2179  #endif  #endif
# Line 2357  uschar classbits[32]; Line 2427  uschar classbits[32];
2427  BOOL class_utf8;  BOOL class_utf8;
2428  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2429  uschar *class_utf8data;  uschar *class_utf8data;
2430    uschar *class_utf8data_base;
2431  uschar utf8_char[6];  uschar utf8_char[6];
2432  #else  #else
2433  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2396  req_caseopt = ((options & PCRE_CASELESS) Line 2467  req_caseopt = ((options & PCRE_CASELESS)
2467  for (;; ptr++)  for (;; ptr++)
2468    {    {
2469    BOOL negate_class;    BOOL negate_class;
2470    BOOL should_flip_negation;    BOOL should_flip_negation;
2471    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2472    BOOL is_quantifier;    BOOL is_quantifier;
2473    BOOL is_recurse;    BOOL is_recurse;
# Line 2596  for (;; ptr++) Line 2667  for (;; ptr++)
2667      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2668      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2669      previous = code;      previous = code;
2670      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2671      break;      break;
2672    
2673    
# Line 2611  for (;; ptr++) Line 2682  for (;; ptr++)
2682      opcode is compiled. It may optionally have a bit map for characters < 256,      opcode is compiled. It may optionally have a bit map for characters < 256,
2683      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2684      whether the bitmap is present, and whether this is a negated class or not.      whether the bitmap is present, and whether this is a negated class or not.
2685      */  
2686        In JavaScript compatibility mode, an isolated ']' causes an error. In
2687        default (Perl) mode, it is treated as a data character. */
2688    
2689        case ']':
2690        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2691          {
2692          *errorcodeptr = ERR64;
2693          goto FAILED;
2694          }
2695        goto NORMAL_CHAR;
2696    
2697      case '[':      case '[':
2698      previous = code;      previous = code;
# Line 2620  for (;; ptr++) Line 2701  for (;; ptr++)
2701      they are encountered at the top level, so we'll do that too. */      they are encountered at the top level, so we'll do that too. */
2702    
2703      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2704          check_posix_syntax(ptr, &tempptr, cd))          check_posix_syntax(ptr, &tempptr))
2705        {        {
2706        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;
2707        goto FAILED;        goto FAILED;
# Line 2645  for (;; ptr++) Line 2726  for (;; ptr++)
2726        else break;        else break;
2727        }        }
2728    
2729      /* If a class contains a negative special such as \S, we need to flip the      /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2730      negation flag at the end, so that support for characters > 255 works      an initial ']' is taken as a data character -- the code below handles
2731        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2732        [^] must match any character, so generate OP_ALLANY. */
2733    
2734        if (c ==']' && (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2735          {
2736          *code++ = negate_class? OP_ALLANY : OP_FAIL;
2737          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2738          zerofirstbyte = firstbyte;
2739          break;
2740          }
2741    
2742        /* If a class contains a negative special such as \S, we need to flip the
2743        negation flag at the end, so that support for characters > 255 works
2744      correctly (they are all included in the class). */      correctly (they are all included in the class). */
2745    
2746      should_flip_negation = FALSE;      should_flip_negation = FALSE;
# Line 2668  for (;; ptr++) Line 2762  for (;; ptr++)
2762  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2763      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2764      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2765        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2766  #endif  #endif
2767    
2768      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2683  for (;; ptr++) Line 2778  for (;; ptr++)
2778          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2779          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2780          }          }
2781    
2782          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2783          data and reset the pointer. This is so that very large classes that
2784          contain a zillion UTF-8 characters no longer overwrite the work space
2785          (which is on the stack). */
2786    
2787          if (lengthptr != NULL)
2788            {
2789            *lengthptr += class_utf8data - class_utf8data_base;
2790            class_utf8data = class_utf8data_base;
2791            }
2792    
2793  #endif  #endif
2794    
2795        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
# Line 2706  for (;; ptr++) Line 2813  for (;; ptr++)
2813    
2814        if (c == '[' &&        if (c == '[' &&
2815            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2816            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr))
2817          {          {
2818          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2819          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
# Line 2723  for (;; ptr++) Line 2830  for (;; ptr++)
2830          if (*ptr == '^')          if (*ptr == '^')
2831            {            {
2832            local_negate = TRUE;            local_negate = TRUE;
2833            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
2834            ptr++;            ptr++;
2835            }            }
2836    
# Line 2826  for (;; ptr++) Line 2933  for (;; ptr++)
2933              continue;              continue;
2934    
2935              case ESC_D:              case ESC_D:
2936              should_flip_negation = TRUE;              should_flip_negation = TRUE;
2937              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
2938              continue;              continue;
2939    
# Line 2835  for (;; ptr++) Line 2942  for (;; ptr++)
2942              continue;              continue;
2943    
2944              case ESC_W:              case ESC_W:
2945              should_flip_negation = TRUE;              should_flip_negation = TRUE;
2946              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
2947              continue;              continue;
2948    
# Line 2845  for (;; ptr++) Line 2952  for (;; ptr++)
2952              continue;              continue;
2953    
2954              case ESC_S:              case ESC_S:
2955              should_flip_negation = TRUE;              should_flip_negation = TRUE;
2956              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
2957              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2958              continue;              continue;
# Line 3232  for (;; ptr++) Line 3339  for (;; ptr++)
3339          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3340            {            {
3341            unsigned int othercase;            unsigned int othercase;
3342            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3343              {              {
3344              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3345              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3348  we set the flag only if there is a liter Line 3455  we set the flag only if there is a liter
3455      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3456    
3457      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3458      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
3459      such as \S in the class, because in that case all characters > 255 are in      such as \S in the class, because in that case all characters > 255 are in
3460      the class, so any that were explicitly given as well can be ignored. If      the class, so any that were explicitly given as well can be ignored. If
3461      (when there are explicit characters > 255 that must be listed) there are no      (when there are explicit characters > 255 that must be listed) there are no
3462      characters < 256, we can omit the bitmap in the actual compiled code. */      characters < 256, we can omit the bitmap in the actual compiled code. */
3463    
# Line 3381  we set the flag only if there is a liter Line 3488  we set the flag only if there is a liter
3488        }        }
3489  #endif  #endif
3490    
3491      /* If there are no characters > 255, set the opcode to OP_CLASS or      /* If there are no characters > 255, set the opcode to OP_CLASS or
3492      OP_NCLASS, depending on whether the whole class was negated and whether      OP_NCLASS, depending on whether the whole class was negated and whether
3493      there were negative specials such as \S in the class. Then copy the 32-byte      there were negative specials such as \S in the class. Then copy the 32-byte
3494      map into the code vector, negating it if necessary. */      map into the code vector, negating it if necessary. */
3495    
3496      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3497      if (negate_class)      if (negate_class)
3498        {        {
# Line 3787  we set the flag only if there is a liter Line 3894  we set the flag only if there is a liter
3894    
3895        if (repeat_min == 0)        if (repeat_min == 0)
3896          {          {
3897          /* If the maximum is also zero, we just omit the group from the output          /* If the maximum is also zero, we used to just omit the group from the
3898          altogether. */          output altogether, like this:
3899    
3900          if (repeat_max == 0)          ** if (repeat_max == 0)
3901            {          **   {
3902            code = previous;          **   code = previous;
3903            goto END_REPEAT;          **   goto END_REPEAT;
3904            }          **   }
3905    
3906          /* If the maximum is 1 or unlimited, we just have to stick in the          However, that fails when a group is referenced as a subroutine from
3907          BRAZERO and do no more at this point. However, we do need to adjust          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
3908          any OP_RECURSE calls inside the group that refer to the group itself or          so that it is skipped on execution. As we don't have a list of which
3909          any internal or forward referenced group, because the offset is from          groups are referenced, we cannot do this selectively.
3910          the start of the whole regex. Temporarily terminate the pattern while  
3911          doing this. */          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
3912            and do no more at this point. However, we do need to adjust any
3913            OP_RECURSE calls inside the group that refer to the group itself or any
3914            internal or forward referenced group, because the offset is from the
3915            start of the whole regex. Temporarily terminate the pattern while doing
3916            this. */
3917    
3918          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
3919            {            {
3920            *code = OP_END;            *code = OP_END;
3921            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3922            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3923            code++;            code++;
3924              if (repeat_max == 0)
3925                {
3926                *previous++ = OP_SKIPZERO;
3927                goto END_REPEAT;
3928                }
3929            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
3930            }            }
3931    
# Line 4003  we set the flag only if there is a liter Line 4120  we set the flag only if there is a liter
4120          }          }
4121        }        }
4122    
4123        /* If previous is OP_FAIL, it was generated by an empty class [] in
4124        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4125        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4126        error above. We can just ignore the repeat in JS case. */
4127    
4128        else if (*previous == OP_FAIL) goto END_REPEAT;
4129    
4130      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4131    
4132      else      else
# Line 4029  we set the flag only if there is a liter Line 4153  we set the flag only if there is a liter
4153        int len;        int len;
4154        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4155            *tempcode == OP_NOTEXACT)            *tempcode == OP_NOTEXACT)
4156          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode] +
4157              ((*tempcode == OP_TYPEEXACT &&
4158                 (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);
4159        len = code - tempcode;        len = code - tempcode;
4160        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4161          {          {
# Line 4256  we set the flag only if there is a liter Line 4382  we set the flag only if there is a liter
4382              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4383              goto FAILED;              goto FAILED;
4384              }              }
4385            recno = (refsign == '-')?            recno = (refsign == '-')?
4386              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno +cd->bracount;
4387            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
4388              {              {
# Line 4287  we set the flag only if there is a liter Line 4413  we set the flag only if there is a liter
4413    
4414          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4415    
4416          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(ptr, cd, name, namelen,
4417                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4418            {            {
4419            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
# Line 4335  we set the flag only if there is a liter Line 4461  we set the flag only if there is a liter
4461            }            }
4462    
4463          /* Check for the "name" actually being a subpattern number. We are          /* Check for the "name" actually being a subpattern number. We are
4464          in the second pass here, so final_bracount is set. */          in the second pass here, so final_bracount is set. */
4465    
4466          else if (recno > 0 && recno <= cd->final_bracount)          else if (recno > 0 && recno <= cd->final_bracount)
4467            {            {
# Line 4533  we set the flag only if there is a liter Line 4659  we set the flag only if there is a liter
4659          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4660          through from the Perl recursion syntax (?&name). We also come here from          through from the Perl recursion syntax (?&name). We also come here from
4661          the Perl \k<name> or \k'name' back reference syntax and the \k{name}          the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4662          .NET syntax. */          .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4663    
4664          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4665          name = ++ptr;          name = ++ptr;
# Line 4549  we set the flag only if there is a liter Line 4675  we set the flag only if there is a liter
4675              {              {
4676              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
4677              goto FAILED;              goto FAILED;
4678              }              }
4679            if (*ptr != terminator)            if (*ptr != terminator)
4680              {              {
4681              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4563  we set the flag only if there is a liter Line 4689  we set the flag only if there is a liter
4689            recno = 0;            recno = 0;
4690            }            }
4691    
4692          /* In the real compile, seek the name in the table. We check the name          /* In the real compile, seek the name in the table. We check the name
4693          first, and then check that we have reached the end of the name in the          first, and then check that we have reached the end of the name in the
4694          table. That way, if the name that is longer than any in the table,          table. That way, if the name that is longer than any in the table,
4695          the comparison will fail without reading beyond the table entry. */          the comparison will fail without reading beyond the table entry. */
4696    
# Line 4574  we set the flag only if there is a liter Line 4700  we set the flag only if there is a liter
4700            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
4701              {              {
4702              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
4703                  slot[2+namelen] == 0)                  slot[2+namelen] == 0)
4704                break;                break;
4705              slot += cd->name_entry_size;              slot += cd->name_entry_size;
4706              }              }
# Line 4584  we set the flag only if there is a liter Line 4710  we set the flag only if there is a liter
4710              recno = GET2(slot, 0);              recno = GET2(slot, 0);
4711              }              }
4712            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
4713                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(ptr, cd, name, namelen,
4714                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
4715              {              {
4716              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4611  we set the flag only if there is a liter Line 4737  we set the flag only if there is a liter
4737          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case '5': case '6': case '7': case '8': case '9':   /* subroutine */
4738            {            {
4739            const uschar *called;            const uschar *called;
4740              terminator = ')';
4741    
4742            if ((refsign = *ptr) == '+')            /* Come here from the \g<...> and \g'...' code (Oniguruma
4743              compatibility). However, the syntax has been checked to ensure that
4744              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
4745              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
4746              ever be taken. */
4747    
4748              HANDLE_NUMERICAL_RECURSION:
4749    
4750              if ((refsign = *ptr) == '+')
4751              {              {
4752              ptr++;              ptr++;
4753              if ((digitab[*ptr] & ctype_digit) == 0)              if ((digitab[*ptr] & ctype_digit) == 0)
4754                {                {
4755                *errorcodeptr = ERR63;                *errorcodeptr = ERR63;
4756                goto FAILED;                goto FAILED;
4757                }                }
4758              }              }
4759            else if (refsign == '-')            else if (refsign == '-')
4760              {              {
4761              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
# Line 4632  we set the flag only if there is a liter Line 4767  we set the flag only if there is a liter
4767            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
4768              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - '0';
4769    
4770            if (*ptr != ')')            if (*ptr != terminator)
4771              {              {
4772              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
4773              goto FAILED;              goto FAILED;
# Line 4685  we set the flag only if there is a liter Line 4820  we set the flag only if there is a liter
4820    
4821              if (called == NULL)              if (called == NULL)
4822                {                {
4823                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(ptr, cd, NULL, recno,
4824                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
4825                  {                  {
4826                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
4827                  goto FAILED;                  goto FAILED;
# Line 5056  we set the flag only if there is a liter Line 5191  we set the flag only if there is a liter
5191        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5192        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5193    
5194          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5195          is a subroutine call by number (Oniguruma syntax). In fact, the value
5196          -ESC_g is returned only for these cases. So we don't need to check for <
5197          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5198          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5199          that is a synonym for a named back reference). */
5200    
5201          if (-c == ESC_g)
5202            {
5203            const uschar *p;
5204            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5205            terminator = (*(++ptr) == '<')? '>' : '\'';
5206    
5207            /* These two statements stop the compiler for warning about possibly
5208            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5209            fact, because we actually check for a number below, the paths that
5210            would actually be in error are never taken. */
5211    
5212            skipbytes = 0;
5213            reset_bracount = FALSE;
5214    
5215            /* Test for a name */
5216    
5217            if (ptr[1] != '+' && ptr[1] != '-')
5218              {
5219              BOOL isnumber = TRUE;
5220              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5221                {
5222                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5223                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5224                }
5225              if (*p != terminator)
5226                {
5227                *errorcodeptr = ERR57;
5228                break;
5229                }
5230              if (isnumber)
5231                {
5232                ptr++;
5233                goto HANDLE_NUMERICAL_RECURSION;
5234                }
5235              is_recurse = TRUE;
5236              goto NAMED_REF_OR_RECURSE;
5237              }
5238    
5239            /* Test a signed number in angle brackets or quotes. */
5240    
5241            p = ptr + 2;
5242            while ((digitab[*p] & ctype_digit) != 0) p++;
5243            if (*p != terminator)
5244              {
5245              *errorcodeptr = ERR57;
5246              break;
5247              }
5248            ptr++;
5249            goto HANDLE_NUMERICAL_RECURSION;
5250            }
5251    
5252        /* \k<name> or \k'name' is a back reference by name (Perl syntax).        /* \k<name> or \k'name' is a back reference by name (Perl syntax).
5253        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5254    
# Line 5562  do { Line 5755  do {
5755       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;
5756       }       }
5757    
5758     /* .* is not anchored unless DOTALL is set and it isn't in brackets that     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
5759     are or may be referenced. */     it isn't in brackets that are or may be referenced. */
5760    
5761     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
5762               op == OP_TYPEPOSSTAR) &&               op == OP_TYPEPOSSTAR))
             (*options & PCRE_DOTALL) != 0)  
5763       {       {
5764       if (scode[1] != OP_ANY || (bracket_map & backref_map) != 0) return FALSE;       if (scode[1] != OP_ALLANY || (bracket_map & backref_map) != 0)
5765           return FALSE;
5766       }       }
5767    
5768     /* Check for explicit anchoring */     /* Check for explicit anchoring */
# Line 5751  Returns:        pointer to compiled data Line 5944  Returns:        pointer to compiled data
5944                  with errorptr and erroroffset set                  with errorptr and erroroffset set
5945  */  */
5946    
5947  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
5948  pcre_compile(const char *pattern, int options, const char **errorptr,  pcre_compile(const char *pattern, int options, const char **errorptr,
5949    int *erroroffset, const unsigned char *tables)    int *erroroffset, const unsigned char *tables)
5950  {  {
# Line 5759  return pcre_compile2(pattern, options, N Line 5952  return pcre_compile2(pattern, options, N
5952  }  }
5953    
5954    
5955  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
5956  pcre_compile2(const char *pattern, int options, int *errorcodeptr,  pcre_compile2(const char *pattern, int options, int *errorcodeptr,
5957    const char **errorptr, int *erroroffset, const unsigned char *tables)    const char **errorptr, int *erroroffset, const unsigned char *tables)
5958  {  {
# Line 5786  to fill in forward references to subpatt Line 5979  to fill in forward references to subpatt
5979    
5980  uschar cworkspace[COMPILE_WORK_SIZE];  uschar cworkspace[COMPILE_WORK_SIZE];
5981    
   
5982  /* Set this early so that early errors get offset 0. */  /* Set this early so that early errors get offset 0. */
5983    
5984  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern;

Legend:
Removed from v.282  
changed lines
  Added in v.359

  ViewVC Help
Powered by ViewVC 1.1.5