/[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 295 by ph10, Mon Dec 31 17:00:24 2007 UTC revision 380 by ph10, Tue Mar 3 12:32:47 2009 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 454  static const char * Line 455  static const char *
455  find_error_text(int n)  find_error_text(int n)
456  {  {
457  const char *s = error_texts;  const char *s = error_texts;
458  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--) while (*s++ != 0) {};
459  return s;  return s;
460  }  }
461    
# 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 977  for (; *ptr != 0; ptr++) Line 1002  for (; *ptr != 0; ptr++)
1002      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) return -1;
1003      if (*ptr == 'Q') for (;;)      if (*ptr == 'Q') for (;;)
1004        {        {
1005        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != '\\') {};
1006        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1007        if (*(++ptr) == 'E') break;        if (*(++ptr) == 'E') break;
1008        }        }
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 996  for (; *ptr != 0; ptr++) Line 1045  for (; *ptr != 0; ptr++)
1045          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) return -1;
1046          if (*ptr == 'Q') for (;;)          if (*ptr == 'Q') for (;;)
1047            {            {
1048            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != '\\') {};
1049            if (*ptr == 0) return -1;            if (*ptr == 0) return -1;
1050            if (*(++ptr) == 'E') break;            if (*(++ptr) == 'E') break;
1051            }            }
# Line 1010  for (; *ptr != 0; ptr++) Line 1059  for (; *ptr != 0; ptr++)
1059    
1060    if (xmode && *ptr == '#')    if (xmode && *ptr == '#')
1061      {      {
1062      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != '\n') {};
1063      if (*ptr == 0) return -1;      if (*ptr == 0) return -1;
1064      continue;      continue;
1065      }      }
# 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 1400  for (;;) Line 1450  for (;;)
1450        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1451        break;        break;
1452        }        }
1453    #else
1454        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1455  #endif  #endif
1456      }      }
1457    }    }
# Line 1493  for (;;) Line 1545  for (;;)
1545        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1546        break;        break;
1547        }        }
1548    #else
1549        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1550  #endif  #endif
1551      }      }
1552    }    }
# Line 1544  for (code = first_significant_code(code Line 1598  for (code = first_significant_code(code
1598    
1599    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1600    
1601    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1602      {      {
1603      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1604      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 1684  for (code = first_significant_code(code
1684      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1685      case OP_WORDCHAR:      case OP_WORDCHAR:
1686      case OP_ANY:      case OP_ANY:
1687        case OP_ALLANY:
1688      case OP_ANYBYTE:      case OP_ANYBYTE:
1689      case OP_CHAR:      case OP_CHAR:
1690      case OP_CHARNC:      case OP_CHARNC:
# Line 1738  return TRUE; Line 1793  return TRUE;
1793    
1794  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1795  encountered in a character class. It checks whether this is followed by a  encountered in a character class. It checks whether this is followed by a
1796  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1797  reach an unescaped ']' without the special preceding character, return FALSE.  reach an unescaped ']' without the special preceding character, return FALSE.
1798    
1799  Originally, this function only recognized a sequence of letters between the  Originally, this function only recognized a sequence of letters between the
1800  terminators, but it seems that Perl recognizes any sequence of characters,  terminators, but it seems that Perl recognizes any sequence of characters,
1801  though of course unknown POSIX names are subsequently rejected. Perl gives an  though of course unknown POSIX names are subsequently rejected. Perl gives an
1802  "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE  "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1803  didn't consider this to be a POSIX class. Likewise for [:1234:].  didn't consider this to be a POSIX class. Likewise for [:1234:].
1804    
1805  The problem in trying to be exactly like Perl is in the handling of escapes. We  The problem in trying to be exactly like Perl is in the handling of escapes. We
1806  have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1807  class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1808  below handles the special case of \], but does not try to do any other escape  below handles the special case of \], but does not try to do any other escape
1809  processing. This makes it different from Perl for cases such as [:l\ower:]  processing. This makes it different from Perl for cases such as [:l\ower:]
1810  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1811  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1812  I think.  I think.
1813    
1814  Arguments:  Arguments:
# Line 1771  terminator = *(++ptr);   /* compiler war Line 1826  terminator = *(++ptr);   /* compiler war
1826  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
1827    {    {
1828    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == '\\' && ptr[1] == ']') ptr++; else
1829      {      {
1830      if (*ptr == ']') return FALSE;      if (*ptr == ']') return FALSE;
1831      if (*ptr == terminator && ptr[1] == ']')      if (*ptr == terminator && ptr[1] == ']')
1832        {        {
1833        *endptr = ptr;        *endptr = ptr;
1834        return TRUE;        return TRUE;
1835        }        }
1836      }      }
1837    }    }
1838  return FALSE;  return FALSE;
1839  }  }
1840    
# Line 1824  return -1; Line 1879  return -1;
1879  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
1880  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
1881  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
1882  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
1883  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
1884  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
1885  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
1886  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
1887    OP_END.
1888    
1889  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
1890  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 1963  get_othercase_range(unsigned int *cptr, Line 2019  get_othercase_range(unsigned int *cptr,
2019  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2020    
2021  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2022    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2023    
2024  if (c > d) return FALSE;  if (c > d) return FALSE;
2025    
# Line 1972  next = othercase + 1; Line 2028  next = othercase + 1;
2028    
2029  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2030    {    {
2031    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2032    next++;    next++;
2033    }    }
2034    
# Line 2082  if (next >= 0) switch(op_code) Line 2138  if (next >= 0) switch(op_code)
2138    case OP_CHAR:    case OP_CHAR:
2139  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2140    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2141    #else
2142      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2143  #endif  #endif
2144    return item != next;    return item != next;
2145    
# Line 2100  if (next >= 0) switch(op_code) Line 2158  if (next >= 0) switch(op_code)
2158      unsigned int othercase;      unsigned int othercase;
2159      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2160  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2161      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2162  #else  #else
2163      othercase = NOTACHAR;      othercase = NOTACHAR;
2164  #endif  #endif
# Line 2113  if (next >= 0) switch(op_code) Line 2171  if (next >= 0) switch(op_code)
2171    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2172    
2173    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2174    if (item == next) return TRUE;    if (item == next) return TRUE;
2175    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2176  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 2122  if (next >= 0) switch(op_code) Line 2179  if (next >= 0) switch(op_code)
2179      unsigned int othercase;      unsigned int othercase;
2180      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2181  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2182      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2183  #else  #else
2184      othercase = NOTACHAR;      othercase = NOTACHAR;
2185  #endif  #endif
# Line 2376  uschar classbits[32]; Line 2433  uschar classbits[32];
2433  BOOL class_utf8;  BOOL class_utf8;
2434  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2435  uschar *class_utf8data;  uschar *class_utf8data;
2436    uschar *class_utf8data_base;
2437  uschar utf8_char[6];  uschar utf8_char[6];
2438  #else  #else
2439  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2615  for (;; ptr++) Line 2673  for (;; ptr++)
2673      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2674      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2675      previous = code;      previous = code;
2676      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2677      break;      break;
2678    
2679    
# Line 2630  for (;; ptr++) Line 2688  for (;; ptr++)
2688      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,
2689      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2690      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.
2691      */  
2692        In JavaScript compatibility mode, an isolated ']' causes an error. In
2693        default (Perl) mode, it is treated as a data character. */
2694    
2695        case ']':
2696        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2697          {
2698          *errorcodeptr = ERR64;
2699          goto FAILED;
2700          }
2701        goto NORMAL_CHAR;
2702    
2703      case '[':      case '[':
2704      previous = code;      previous = code;
# Line 2664  for (;; ptr++) Line 2732  for (;; ptr++)
2732        else break;        else break;
2733        }        }
2734    
2735        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2736        an initial ']' is taken as a data character -- the code below handles
2737        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2738        [^] must match any character, so generate OP_ALLANY. */
2739    
2740        if (c ==']' && (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2741          {
2742          *code++ = negate_class? OP_ALLANY : OP_FAIL;
2743          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2744          zerofirstbyte = firstbyte;
2745          break;
2746          }
2747    
2748      /* 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
2749      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
2750      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2687  for (;; ptr++) Line 2768  for (;; ptr++)
2768  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2769      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2770      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2771        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2772  #endif  #endif
2773    
2774      /* 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 2702  for (;; ptr++) Line 2784  for (;; ptr++)
2784          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2785          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2786          }          }
2787    
2788          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2789          data and reset the pointer. This is so that very large classes that
2790          contain a zillion UTF-8 characters no longer overwrite the work space
2791          (which is on the stack). */
2792    
2793          if (lengthptr != NULL)
2794            {
2795            *lengthptr += class_utf8data - class_utf8data_base;
2796            class_utf8data = class_utf8data_base;
2797            }
2798    
2799  #endif  #endif
2800    
2801        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
# Line 3251  for (;; ptr++) Line 3345  for (;; ptr++)
3345          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3346            {            {
3347            unsigned int othercase;            unsigned int othercase;
3348            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3349              {              {
3350              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3351              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3806  we set the flag only if there is a liter Line 3900  we set the flag only if there is a liter
3900    
3901        if (repeat_min == 0)        if (repeat_min == 0)
3902          {          {
3903          /* 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
3904          altogether. */          output altogether, like this:
3905    
3906          if (repeat_max == 0)          ** if (repeat_max == 0)
3907            {          **   {
3908            code = previous;          **   code = previous;
3909            goto END_REPEAT;          **   goto END_REPEAT;
3910            }          **   }
3911    
3912          /* 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
3913          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
3914          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
3915          any internal or forward referenced group, because the offset is from          groups are referenced, we cannot do this selectively.
3916          the start of the whole regex. Temporarily terminate the pattern while  
3917          doing this. */          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
3918            and do no more at this point. However, we do need to adjust any
3919            OP_RECURSE calls inside the group that refer to the group itself or any
3920            internal or forward referenced group, because the offset is from the
3921            start of the whole regex. Temporarily terminate the pattern while doing
3922            this. */
3923    
3924          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
3925            {            {
3926            *code = OP_END;            *code = OP_END;
3927            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3928            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3929            code++;            code++;
3930              if (repeat_max == 0)
3931                {
3932                *previous++ = OP_SKIPZERO;
3933                goto END_REPEAT;
3934                }
3935            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
3936            }            }
3937    
# Line 4022  we set the flag only if there is a liter Line 4126  we set the flag only if there is a liter
4126          }          }
4127        }        }
4128    
4129        /* If previous is OP_FAIL, it was generated by an empty class [] in
4130        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4131        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4132        error above. We can just ignore the repeat in JS case. */
4133    
4134        else if (*previous == OP_FAIL) goto END_REPEAT;
4135    
4136      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4137    
4138      else      else
# Line 4111  we set the flag only if there is a liter Line 4222  we set the flag only if there is a liter
4222        const char *vn = verbnames;        const char *vn = verbnames;
4223        const uschar *name = ++ptr;        const uschar *name = ++ptr;
4224        previous = NULL;        previous = NULL;
4225        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4226        if (*ptr == ':')        if (*ptr == ':')
4227          {          {
4228          *errorcodeptr = ERR59;   /* Not supported */          *errorcodeptr = ERR59;   /* Not supported */
# Line 4308  we set the flag only if there is a liter Line 4419  we set the flag only if there is a liter
4419    
4420          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4421    
4422          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(ptr, cd, name, namelen,
4423                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4424            {            {
4425            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
# Line 4554  we set the flag only if there is a liter Line 4665  we set the flag only if there is a liter
4665          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4666          through from the Perl recursion syntax (?&name). We also come here from          through from the Perl recursion syntax (?&name). We also come here from
4667          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}
4668          .NET syntax. */          .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4669    
4670          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4671          name = ++ptr;          name = ++ptr;
# Line 4605  we set the flag only if there is a liter Line 4716  we set the flag only if there is a liter
4716              recno = GET2(slot, 0);              recno = GET2(slot, 0);
4717              }              }
4718            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
4719                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(ptr, cd, name, namelen,
4720                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
4721              {              {
4722              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4632  we set the flag only if there is a liter Line 4743  we set the flag only if there is a liter
4743          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case '5': case '6': case '7': case '8': case '9':   /* subroutine */
4744            {            {
4745            const uschar *called;            const uschar *called;
4746              terminator = ')';
4747    
4748              /* Come here from the \g<...> and \g'...' code (Oniguruma
4749              compatibility). However, the syntax has been checked to ensure that
4750              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
4751              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
4752              ever be taken. */
4753    
4754              HANDLE_NUMERICAL_RECURSION:
4755    
4756            if ((refsign = *ptr) == '+')            if ((refsign = *ptr) == '+')
4757              {              {
# Line 4653  we set the flag only if there is a liter Line 4773  we set the flag only if there is a liter
4773            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
4774              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - '0';
4775    
4776            if (*ptr != ')')            if (*ptr != terminator)
4777              {              {
4778              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
4779              goto FAILED;              goto FAILED;
# Line 4706  we set the flag only if there is a liter Line 4826  we set the flag only if there is a liter
4826    
4827              if (called == NULL)              if (called == NULL)
4828                {                {
4829                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(ptr, cd, NULL, recno,
4830                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
4831                  {                  {
4832                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
4833                  goto FAILED;                  goto FAILED;
# Line 4806  we set the flag only if there is a liter Line 4926  we set the flag only if there is a liter
4926          both phases.          both phases.
4927    
4928          If we are not at the pattern start, compile code to change the ims          If we are not at the pattern start, compile code to change the ims
4929          options if this setting actually changes any of them. We also pass the          options if this setting actually changes any of them, and reset the
4930          new setting back so that it can be put at the start of any following          greedy defaults and the case value for firstbyte and reqbyte. */
         branches, and when this group ends (if we are in a group), a resetting  
         item can be compiled. */  
4931    
4932          if (*ptr == ')')          if (*ptr == ')')
4933            {            {
# Line 4817  we set the flag only if there is a liter Line 4935  we set the flag only if there is a liter
4935                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
4936              {              {
4937              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
4938              }              }
4939           else           else
4940              {              {
# Line 4826  we set the flag only if there is a liter Line 4943  we set the flag only if there is a liter
4943                *code++ = OP_OPT;                *code++ = OP_OPT;
4944                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
4945                }                }
   
             /* Change options at this level, and pass them back for use  
             in subsequent branches. Reset the greedy defaults and the case  
             value for firstbyte and reqbyte. */  
   
             *optionsptr = options = newoptions;  
4946              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
4947              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
4948              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
4949              }              }
4950    
4951              /* Change options at this level, and pass them back for use
4952              in subsequent branches. When not at the start of the pattern, this
4953              information is also necessary so that a resetting item can be
4954              compiled at the end of a group (if we are in a group). */
4955    
4956              *optionsptr = options = newoptions;
4957            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
4958            continue;              /* It is complete */            continue;              /* It is complete */
4959            }            }
# Line 5077  we set the flag only if there is a liter Line 5194  we set the flag only if there is a liter
5194        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5195        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5196    
5197          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5198          is a subroutine call by number (Oniguruma syntax). In fact, the value
5199          -ESC_g is returned only for these cases. So we don't need to check for <
5200          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5201          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5202          that is a synonym for a named back reference). */
5203    
5204          if (-c == ESC_g)
5205            {
5206            const uschar *p;
5207            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5208            terminator = (*(++ptr) == '<')? '>' : '\'';
5209    
5210            /* These two statements stop the compiler for warning about possibly
5211            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5212            fact, because we actually check for a number below, the paths that
5213            would actually be in error are never taken. */
5214    
5215            skipbytes = 0;
5216            reset_bracount = FALSE;
5217    
5218            /* Test for a name */
5219    
5220            if (ptr[1] != '+' && ptr[1] != '-')
5221              {
5222              BOOL isnumber = TRUE;
5223              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5224                {
5225                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5226                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5227                }
5228              if (*p != terminator)
5229                {
5230                *errorcodeptr = ERR57;
5231                break;
5232                }
5233              if (isnumber)
5234                {
5235                ptr++;
5236                goto HANDLE_NUMERICAL_RECURSION;
5237                }
5238              is_recurse = TRUE;
5239              goto NAMED_REF_OR_RECURSE;
5240              }
5241    
5242            /* Test a signed number in angle brackets or quotes. */
5243    
5244            p = ptr + 2;
5245            while ((digitab[*p] & ctype_digit) != 0) p++;
5246            if (*p != terminator)
5247              {
5248              *errorcodeptr = ERR57;
5249              break;
5250              }
5251            ptr++;
5252            goto HANDLE_NUMERICAL_RECURSION;
5253            }
5254    
5255        /* \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).
5256        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5257    
# Line 5583  do { Line 5758  do {
5758       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;
5759       }       }
5760    
5761     /* .* 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
5762     are or may be referenced. */     it isn't in brackets that are or may be referenced. */
5763    
5764     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
5765               op == OP_TYPEPOSSTAR) &&               op == OP_TYPEPOSSTAR))
             (*options & PCRE_DOTALL) != 0)  
5766       {       {
5767       if (scode[1] != OP_ANY || (bracket_map & backref_map) != 0) return FALSE;       if (scode[1] != OP_ALLANY || (bracket_map & backref_map) != 0)
5768           return FALSE;
5769       }       }
5770    
5771     /* Check for explicit anchoring */     /* Check for explicit anchoring */
# Line 5635  do { Line 5810  do {
5810     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
5811       NULL, 0, FALSE);       NULL, 0, FALSE);
5812     register int op = *scode;     register int op = *scode;
5813    
5814       /* If we are at the start of a conditional group, skip over the condition.
5815       before inspecting the first opcode after the condition. */
5816    
5817       if (op == OP_COND)
5818         {
5819         scode += 1 + LINK_SIZE;
5820         switch (*scode)
5821           {
5822           case OP_CREF:
5823           case OP_RREF:
5824           scode += 3;
5825           break;
5826    
5827           case OP_DEF:
5828           scode += 1;
5829           break;
5830    
5831           default:     /* Assertion */
5832           do scode += GET(scode, 1); while (*scode == OP_ALT);
5833           break;
5834           }
5835    
5836         scode = first_significant_code(scode, NULL, 0, FALSE);
5837         op = *scode;
5838         }
5839    
5840     /* Non-capturing brackets */     /* Non-capturing brackets */
5841    
# Line 5654  do { Line 5855  do {
5855    
5856     /* Other brackets */     /* Other brackets */
5857    
5858     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE)
5859       { if (!is_startline(scode, bracket_map, backref_map)) return FALSE; }       {
5860         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
5861         }
5862    
5863     /* .* means "start at start or after \n" if it isn't in brackets that     /* .* means "start at start or after \n" if it isn't in brackets that
5864     may be referenced. */     may be referenced. */
# Line 5772  Returns:        pointer to compiled data Line 5975  Returns:        pointer to compiled data
5975                  with errorptr and erroroffset set                  with errorptr and erroroffset set
5976  */  */
5977    
5978  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
5979  pcre_compile(const char *pattern, int options, const char **errorptr,  pcre_compile(const char *pattern, int options, const char **errorptr,
5980    int *erroroffset, const unsigned char *tables)    int *erroroffset, const unsigned char *tables)
5981  {  {
# Line 5780  return pcre_compile2(pattern, options, N Line 5983  return pcre_compile2(pattern, options, N
5983  }  }
5984    
5985    
5986  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
5987  pcre_compile2(const char *pattern, int options, int *errorcodeptr,  pcre_compile2(const char *pattern, int options, int *errorcodeptr,
5988    const char **errorptr, int *erroroffset, const unsigned char *tables)    const char **errorptr, int *erroroffset, const unsigned char *tables)
5989  {  {
# Line 5807  to fill in forward references to subpatt Line 6010  to fill in forward references to subpatt
6010    
6011  uschar cworkspace[COMPILE_WORK_SIZE];  uschar cworkspace[COMPILE_WORK_SIZE];
6012    
   
6013  /* Set this early so that early errors get offset 0. */  /* Set this early so that early errors get offset 0. */
6014    
6015  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern;

Legend:
Removed from v.295  
changed lines
  Added in v.380

  ViewVC Help
Powered by ViewVC 1.1.5