/[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 265 by ph10, Wed Nov 14 11:35:48 2007 UTC revision 341 by ph10, Sat Apr 19 16:41:04 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 241  static const char error_texts[] = Line 241  static const char error_texts[] =
241    /* 10 */    /* 10 */
242    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
243    "internal error: unexpected repeat\0"    "internal error: unexpected repeat\0"
244    "unrecognized character after (?\0"    "unrecognized character after (? or (?-\0"
245    "POSIX named classes are supported only within a class\0"    "POSIX named classes are supported only within a class\0"
246    "missing )\0"    "missing )\0"
247    /* 15 */    /* 15 */
# 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";    "number is too big\0"
304      "subpattern name expected\0"
305      "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 496  ptr--; /* Set Line 499  ptr--; /* Set
499    
500  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
501    
502  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
503  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
504  Otherwise further processing may be required. */  Otherwise further processing may be required. */
505    
506  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
507  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */
508  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - '0']) != 0) c = i;
509    
510  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
511  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
512  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
513  #endif  #endif
514    
# Line 529  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 563  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        {        {
# Line 609  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 722  else Line 748  else
748      break;      break;
749    
750      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
751      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
752      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
753      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
754      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
755    
756      default:      default:
757      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 950  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 959  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 982  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 1506  for (;;) Line 1557  for (;;)
1557  can match the empty string or not. It is called from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
1558  below and from compile_branch() when checking for an unlimited repeat of a  below and from compile_branch() when checking for an unlimited repeat of a
1559  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1560  assertions. If we hit an unclosed bracket, we return "empty" - this means we've  backward and negative forward assertions when its final argument is TRUE. If we
1561  struck an inner bracket whose current branch will already have been scanned.  hit an unclosed bracket, we return "empty" - this means we've struck an inner
1562    bracket whose current branch will already have been scanned.
1563    
1564  Arguments:  Arguments:
1565    code        points to start of search    code        points to start of search
# Line 1529  for (code = first_significant_code(code Line 1581  for (code = first_significant_code(code
1581    
1582    c = *code;    c = *code;
1583    
1584      /* Skip over forward assertions; the other assertions are skipped by
1585      first_significant_code() with a TRUE final argument. */
1586    
1587      if (c == OP_ASSERT)
1588        {
1589        do code += GET(code, 1); while (*code == OP_ALT);
1590        c = *code;
1591        continue;
1592        }
1593    
1594    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1595    
1596    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1597      {      {
1598      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1599      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1724  return TRUE; Line 1786  return TRUE;
1786  *************************************************/  *************************************************/
1787    
1788  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1789  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
1790  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1791  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1792    
1793    Originally, this function only recognized a sequence of letters between the
1794    terminators, but it seems that Perl recognizes any sequence of characters,
1795    though of course unknown POSIX names are subsequently rejected. Perl gives an
1796    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1797    didn't consider this to be a POSIX class. Likewise for [:1234:].
1798    
1799    The problem in trying to be exactly like Perl is in the handling of escapes. We
1800    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1801    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1802    below handles the special case of \], but does not try to do any other escape
1803    processing. This makes it different from Perl for cases such as [:l\ower:]
1804    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1805    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1806    I think.
1807    
1808  Argument:  Arguments:
1809    ptr      pointer to the initial [    ptr      pointer to the initial [
1810    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1811    
1812  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1813  */  */
1814    
1815  static BOOL  static BOOL
1816  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1817  {  {
1818  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1819  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1820  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
1821    {    {
1822    *endptr = ptr;    if (*ptr == '\\' && ptr[1] == ']') ptr++; else
1823    return TRUE;      {
1824        if (*ptr == ']') return FALSE;
1825        if (*ptr == terminator && ptr[1] == ']')
1826          {
1827          *endptr = ptr;
1828          return TRUE;
1829          }
1830        }
1831    }    }
1832  return FALSE;  return FALSE;
1833  }  }
# Line 1792  return -1; Line 1873  return -1;
1873  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
1874  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
1875  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
1876  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
1877  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
1878  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
1879  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
1880  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
1881    OP_END.
1882    
1883  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
1884  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 1827  while ((ptr = (uschar *)find_recurse(ptr Line 1909  while ((ptr = (uschar *)find_recurse(ptr
1909    
1910    /* See if this recursion is on the forward reference list. If so, adjust the    /* See if this recursion is on the forward reference list. If so, adjust the
1911    reference. */    reference. */
1912    
1913    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
1914      {      {
1915      offset = GET(hc, 0);      offset = GET(hc, 0);
# Line 2081  if (next >= 0) switch(op_code) Line 2163  if (next >= 0) switch(op_code)
2163    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2164    
2165    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2166    if (item == next) return TRUE;    if (item == next) return TRUE;
2167    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2168  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 2344  uschar classbits[32]; Line 2425  uschar classbits[32];
2425  BOOL class_utf8;  BOOL class_utf8;
2426  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2427  uschar *class_utf8data;  uschar *class_utf8data;
2428    uschar *class_utf8data_base;
2429  uschar utf8_char[6];  uschar utf8_char[6];
2430  #else  #else
2431  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2383  req_caseopt = ((options & PCRE_CASELESS) Line 2465  req_caseopt = ((options & PCRE_CASELESS)
2465  for (;; ptr++)  for (;; ptr++)
2466    {    {
2467    BOOL negate_class;    BOOL negate_class;
2468    BOOL should_flip_negation;    BOOL should_flip_negation;
2469    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2470    BOOL is_quantifier;    BOOL is_quantifier;
2471    BOOL is_recurse;    BOOL is_recurse;
# Line 2403  for (;; ptr++) Line 2485  for (;; ptr++)
2485    /* Get next byte in the pattern */    /* Get next byte in the pattern */
2486    
2487    c = *ptr;    c = *ptr;
2488    
2489    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
2490    previous cycle of this loop. */    previous cycle of this loop. */
2491    
# Line 2598  for (;; ptr++) Line 2680  for (;; ptr++)
2680      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,
2681      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2682      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.
2683      */  
2684        In JavaScript compatibility mode, an isolated ']' causes an error. In
2685        default (Perl) mode, it is treated as a data character. */
2686    
2687        case ']':
2688        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2689          {
2690          *errorcodeptr = ERR64;
2691          goto FAILED;
2692          }
2693        goto NORMAL_CHAR;
2694    
2695      case '[':      case '[':
2696      previous = code;      previous = code;
# Line 2607  for (;; ptr++) Line 2699  for (;; ptr++)
2699      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. */
2700    
2701      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2702          check_posix_syntax(ptr, &tempptr, cd))          check_posix_syntax(ptr, &tempptr))
2703        {        {
2704        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;
2705        goto FAILED;        goto FAILED;
# Line 2631  for (;; ptr++) Line 2723  for (;; ptr++)
2723          negate_class = TRUE;          negate_class = TRUE;
2724        else break;        else break;
2725        }        }
2726    
2727        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2728        an initial ']' is taken as a data character -- the code below handles
2729        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2730        [^] must match any character, so generate OP_ALLANY. */
2731    
2732        if (c ==']' && (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2733          {
2734          *code++ = negate_class? OP_ALLANY : OP_FAIL;
2735          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2736          zerofirstbyte = firstbyte;
2737          break;
2738          }
2739    
2740      /* If a class contains a negative special such as \S, we need to flip the      /* If a class contains a negative special such as \S, we need to flip the
2741      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
2742      correctly (they are all included in the class). */      correctly (they are all included in the class). */
2743    
2744      should_flip_negation = FALSE;      should_flip_negation = FALSE;
# Line 2655  for (;; ptr++) Line 2760  for (;; ptr++)
2760  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2761      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2762      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2763        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2764  #endif  #endif
2765    
2766      /* 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 2670  for (;; ptr++) Line 2776  for (;; ptr++)
2776          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2777          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2778          }          }
2779    
2780          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2781          data and reset the pointer. This is so that very large classes that
2782          contain a zillion UTF-8 characters no longer overwrite the work space
2783          (which is on the stack). */
2784    
2785          if (lengthptr != NULL)
2786            {
2787            *lengthptr += class_utf8data - class_utf8data_base;
2788            class_utf8data = class_utf8data_base;
2789            }
2790    
2791  #endif  #endif
2792    
2793        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
# Line 2693  for (;; ptr++) Line 2811  for (;; ptr++)
2811    
2812        if (c == '[' &&        if (c == '[' &&
2813            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2814            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr))
2815          {          {
2816          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2817          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
# Line 2710  for (;; ptr++) Line 2828  for (;; ptr++)
2828          if (*ptr == '^')          if (*ptr == '^')
2829            {            {
2830            local_negate = TRUE;            local_negate = TRUE;
2831            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
2832            ptr++;            ptr++;
2833            }            }
2834    
# Line 2785  for (;; ptr++) Line 2903  for (;; ptr++)
2903          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2904          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
2905    
2906          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */
2907          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */
2908          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */
2909          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
# Line 2813  for (;; ptr++) Line 2931  for (;; ptr++)
2931              continue;              continue;
2932    
2933              case ESC_D:              case ESC_D:
2934              should_flip_negation = TRUE;              should_flip_negation = TRUE;
2935              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
2936              continue;              continue;
2937    
# Line 2822  for (;; ptr++) Line 2940  for (;; ptr++)
2940              continue;              continue;
2941    
2942              case ESC_W:              case ESC_W:
2943              should_flip_negation = TRUE;              should_flip_negation = TRUE;
2944              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
2945              continue;              continue;
2946    
# Line 2832  for (;; ptr++) Line 2950  for (;; ptr++)
2950              continue;              continue;
2951    
2952              case ESC_S:              case ESC_S:
2953              should_flip_negation = TRUE;              should_flip_negation = TRUE;
2954              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
2955              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2956              continue;              continue;
2957    
             case ESC_E: /* Perl ignores an orphan \E */  
             continue;  
   
2958              default:    /* Not recognized; fall through */              default:    /* Not recognized; fall through */
2959              break;      /* Need "default" setting to stop compiler warning. */              break;      /* Need "default" setting to stop compiler warning. */
2960              }              }
# Line 3074  for (;; ptr++) Line 3189  for (;; ptr++)
3189            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3190            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3191    
3192            /* \b is backslash; \X is literal X; \R is literal R; any other            /* \b is backspace; \X is literal X; \R is literal R; any other
3193            special means the '-' was literal */            special means the '-' was literal */
3194    
3195            if (d < 0)            if (d < 0)
# Line 3338  we set the flag only if there is a liter Line 3453  we set the flag only if there is a liter
3453      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3454    
3455      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3456      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
3457      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
3458      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
3459      (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
3460      characters < 256, we can omit the bitmap in the actual compiled code. */      characters < 256, we can omit the bitmap in the actual compiled code. */
3461    
# Line 3371  we set the flag only if there is a liter Line 3486  we set the flag only if there is a liter
3486        }        }
3487  #endif  #endif
3488    
3489      /* 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
3490      OP_NCLASS, depending on whether the whole class was negated and whether      OP_NCLASS, depending on whether the whole class was negated and whether
3491      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
3492      map into the code vector, negating it if necessary. */      map into the code vector, negating it if necessary. */
3493    
3494      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3495      if (negate_class)      if (negate_class)
3496        {        {
# Line 3777  we set the flag only if there is a liter Line 3892  we set the flag only if there is a liter
3892    
3893        if (repeat_min == 0)        if (repeat_min == 0)
3894          {          {
3895          /* 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
3896          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
3897    
3898          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
3899          BRAZERO and do no more at this point. However, we do need to adjust          **   {
3900          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
3901          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
3902          the start of the whole regex. Temporarily terminate the pattern while          **   }
3903          doing this. */  
3904            However, that fails when a group is referenced as a subroutine from
3905            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
3906            so that it is skipped on execution. As we don't have a list of which
3907            groups are referenced, we cannot do this selectively.
3908    
3909            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
3910            and do no more at this point. However, we do need to adjust any
3911            OP_RECURSE calls inside the group that refer to the group itself or any
3912            internal or forward referenced group, because the offset is from the
3913            start of the whole regex. Temporarily terminate the pattern while doing
3914            this. */
3915    
3916          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
3917            {            {
3918            *code = OP_END;            *code = OP_END;
3919            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3920            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3921            code++;            code++;
3922              if (repeat_max == 0)
3923                {
3924                *previous++ = OP_SKIPZERO;
3925                goto END_REPEAT;
3926                }
3927            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
3928            }            }
3929    
# Line 3992  we set the flag only if there is a liter Line 4117  we set the flag only if there is a liter
4117            }            }
4118          }          }
4119        }        }
4120    
4121        /* If previous is OP_FAIL, it was generated by an empty class [] in
4122        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4123        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4124        error above. We can just ignore the repeat in JS case. */
4125    
4126        else if (*previous == OP_FAIL) goto END_REPEAT;
4127    
4128      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4129    
# Line 4019  we set the flag only if there is a liter Line 4151  we set the flag only if there is a liter
4151        int len;        int len;
4152        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4153            *tempcode == OP_NOTEXACT)            *tempcode == OP_NOTEXACT)
4154          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode] +
4155              ((*tempcode == OP_TYPEEXACT &&
4156                 (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);
4157        len = code - tempcode;        len = code - tempcode;
4158        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4159          {          {
# Line 4071  we set the flag only if there is a liter Line 4205  we set the flag only if there is a liter
4205      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4206      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4207      reset_bracount = FALSE;      reset_bracount = FALSE;
4208    
4209      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4210    
4211      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
# Line 4246  we set the flag only if there is a liter Line 4380  we set the flag only if there is a liter
4380              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4381              goto FAILED;              goto FAILED;
4382              }              }
4383            if (refsign == '-')            recno = (refsign == '-')?
4384                cd->bracount - recno + 1 : recno +cd->bracount;
4385              if (recno <= 0 || recno > cd->final_bracount)
4386              {              {
4387              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
4388              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
4389              }              }
           else recno += cd->bracount;  
4390            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4391            break;            break;
4392            }            }
# Line 4280  we set the flag only if there is a liter Line 4411  we set the flag only if there is a liter
4411    
4412          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4413    
4414          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(ptr, cd, name, namelen,
4415                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4416            {            {
4417            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
# Line 4327  we set the flag only if there is a liter Line 4458  we set the flag only if there is a liter
4458            skipbytes = 1;            skipbytes = 1;
4459            }            }
4460    
4461          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4462            in the second pass here, so final_bracount is set. */
4463    
4464          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4465            {            {
4466            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4467            }            }
# Line 4523  we set the flag only if there is a liter Line 4655  we set the flag only if there is a liter
4655    
4656          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4657          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4658          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4659            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4660            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4661    
4662          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4663          name = ++ptr;          name = ++ptr;
# Line 4535  we set the flag only if there is a liter Line 4669  we set the flag only if there is a liter
4669    
4670          if (lengthptr != NULL)          if (lengthptr != NULL)
4671            {            {
4672              if (namelen == 0)
4673                {
4674                *errorcodeptr = ERR62;
4675                goto FAILED;
4676                }
4677            if (*ptr != terminator)            if (*ptr != terminator)
4678              {              {
4679              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4548  we set the flag only if there is a liter Line 4687  we set the flag only if there is a liter
4687            recno = 0;            recno = 0;
4688            }            }
4689    
4690          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
4691            first, and then check that we have reached the end of the name in the
4692            table. That way, if the name that is longer than any in the table,
4693            the comparison will fail without reading beyond the table entry. */
4694    
4695          else          else
4696            {            {
4697            slot = cd->name_table;            slot = cd->name_table;
4698            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
4699              {              {
4700              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
4701                    slot[2+namelen] == 0)
4702                  break;
4703              slot += cd->name_entry_size;              slot += cd->name_entry_size;
4704              }              }
4705    
# Line 4564  we set the flag only if there is a liter Line 4708  we set the flag only if there is a liter
4708              recno = GET2(slot, 0);              recno = GET2(slot, 0);
4709              }              }
4710            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
4711                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(ptr, cd, name, namelen,
4712                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
4713              {              {
4714              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4591  we set the flag only if there is a liter Line 4735  we set the flag only if there is a liter
4735          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case '5': case '6': case '7': case '8': case '9':   /* subroutine */
4736            {            {
4737            const uschar *called;            const uschar *called;
4738              terminator = ')';
4739    
4740              /* Come here from the \g<...> and \g'...' code (Oniguruma
4741              compatibility). However, the syntax has been checked to ensure that
4742              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
4743              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
4744              ever be taken. */
4745    
4746              HANDLE_NUMERICAL_RECURSION:
4747    
4748            if ((refsign = *ptr) == '+') ptr++;            if ((refsign = *ptr) == '+')
4749                {
4750                ptr++;
4751                if ((digitab[*ptr] & ctype_digit) == 0)
4752                  {
4753                  *errorcodeptr = ERR63;
4754                  goto FAILED;
4755                  }
4756                }
4757            else if (refsign == '-')            else if (refsign == '-')
4758              {              {
4759              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
# Line 4604  we set the flag only if there is a liter Line 4765  we set the flag only if there is a liter
4765            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
4766              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - '0';
4767    
4768            if (*ptr != ')')            if (*ptr != terminator)
4769              {              {
4770              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
4771              goto FAILED;              goto FAILED;
# Line 4657  we set the flag only if there is a liter Line 4818  we set the flag only if there is a liter
4818    
4819              if (called == NULL)              if (called == NULL)
4820                {                {
4821                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(ptr, cd, NULL, recno,
4822                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
4823                  {                  {
4824                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
4825                  goto FAILED;                  goto FAILED;
# Line 5000  we set the flag only if there is a liter Line 5161  we set the flag only if there is a liter
5161      back references and those types that consume a character may be repeated.      back references and those types that consume a character may be repeated.
5162      We can test for values between ESC_b and ESC_Z for the latter; this may      We can test for values between ESC_b and ESC_Z for the latter; this may
5163      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5164    
5165      case '\\':      case '\\':
5166      tempptr = ptr;      tempptr = ptr;
5167      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
# Line 5027  we set the flag only if there is a liter Line 5188  we set the flag only if there is a liter
5188    
5189        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5190        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5191    
5192          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5193          is a subroutine call by number (Oniguruma syntax). In fact, the value
5194          -ESC_g is returned only for these cases. So we don't need to check for <
5195          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5196          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5197          that is a synonym for a named back reference). */
5198    
5199          if (-c == ESC_g)
5200            {
5201            const uschar *p;
5202            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5203            terminator = (*(++ptr) == '<')? '>' : '\'';
5204    
5205            /* These two statements stop the compiler for warning about possibly
5206            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5207            fact, because we actually check for a number below, the paths that
5208            would actually be in error are never taken. */
5209    
5210            skipbytes = 0;
5211            reset_bracount = FALSE;
5212    
5213            /* Test for a name */
5214    
5215            if (ptr[1] != '+' && ptr[1] != '-')
5216              {
5217              BOOL isnumber = TRUE;
5218              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5219                {
5220                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5221                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5222                }
5223              if (*p != terminator)
5224                {
5225                *errorcodeptr = ERR57;
5226                break;
5227                }
5228              if (isnumber)
5229                {
5230                ptr++;
5231                goto HANDLE_NUMERICAL_RECURSION;
5232                }
5233              is_recurse = TRUE;
5234              goto NAMED_REF_OR_RECURSE;
5235              }
5236    
5237            /* Test a signed number in angle brackets or quotes. */
5238    
5239            p = ptr + 2;
5240            while ((digitab[*p] & ctype_digit) != 0) p++;
5241            if (*p != terminator)
5242              {
5243              *errorcodeptr = ERR57;
5244              break;
5245              }
5246            ptr++;
5247            goto HANDLE_NUMERICAL_RECURSION;
5248            }
5249    
5250        /* \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).
5251        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
# Line 5758  to fill in forward references to subpatt Line 5977  to fill in forward references to subpatt
5977    
5978  uschar cworkspace[COMPILE_WORK_SIZE];  uschar cworkspace[COMPILE_WORK_SIZE];
5979    
   
5980  /* Set this early so that early errors get offset 0. */  /* Set this early so that early errors get offset 0. */
5981    
5982  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern;
# Line 5919  to compile parts of the pattern into; th Line 6137  to compile parts of the pattern into; th
6137  no longer needed, so hopefully this workspace will never overflow, though there  no longer needed, so hopefully this workspace will never overflow, though there
6138  is a test for its doing so. */  is a test for its doing so. */
6139    
6140  cd->bracount = 0;  cd->bracount = cd->final_bracount = 0;
6141  cd->names_found = 0;  cd->names_found = 0;
6142  cd->name_entry_size = 0;  cd->name_entry_size = 0;
6143  cd->name_table = NULL;  cd->name_table = NULL;
# Line 5996  field. Reset the bracket count and the n Line 6214  field. Reset the bracket count and the n
6214  field; this time it's used for remembering forward references to subpatterns.  field; this time it's used for remembering forward references to subpatterns.
6215  */  */
6216    
6217    cd->final_bracount = cd->bracount;  /* Save for checking forward references */
6218  cd->bracount = 0;  cd->bracount = 0;
6219  cd->names_found = 0;  cd->names_found = 0;
6220  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 6046  while (errorcode == 0 && cd->hwm > cwork Line 6265  while (errorcode == 0 && cd->hwm > cwork
6265    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6266      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6267    }    }
6268    
6269  /* Give an error if there's back reference to a non-existent capturing  /* Give an error if there's back reference to a non-existent capturing
6270  subpattern. */  subpattern. */
6271    

Legend:
Removed from v.265  
changed lines
  Added in v.341

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12