pcre.c

ncbi源码

      if ((cc[1] << 8) + cc[2] != (cc[3] << 8) + cc[4]) return -1;
      branchlength += (cc[1] << 8) + cc[2];
      cc += 5;
      break;

      default:
      branchlength++;
      }
    break;

    /* Anything else is variable length */

    default:
    return -1;
    }
  }
/* Control never gets here */
}




/*************************************************
*           Check for POSIX class syntax         *
*************************************************/

/* This function is called when the sequence "[:" or "[." or "[=" is
encountered in a character class. It checks whether this is followed by an
optional ^ and then a sequence of letters, terminated by a matching ":]" or
".]" or "=]".

Argument:
  ptr      pointer to the initial [
  endptr   where to return the end pointer
  cd       pointer to compile data

Returns:   TRUE or FALSE
*/

static BOOL
check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)
{
int terminator;          /* Don't combine these lines; the Solaris cc */
terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
if (*(++ptr) == '^') ptr++;
while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;
if (*ptr == terminator && ptr[1] == ']')
  {
  *endptr = ptr;
  return TRUE;
  }
return FALSE;
}




/*************************************************
*          Check POSIX class name                *
*************************************************/

/* This function is called to check the name given in a POSIX-style class entry
such as [:alnum:].

Arguments:
  ptr        points to the first letter
  len        the length of the name

Returns:     a value representing the name, or -1 if unknown
*/

static int
check_posix_name(const uschar *ptr, int len)
{
register int yield = 0;
while (posix_name_lengths[yield] != 0)
  {
  if (len == posix_name_lengths[yield] &&
    strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;
  yield++;
  }
return -1;
}




/*************************************************
*           Compile one branch                   *
*************************************************/

/* Scan the pattern, compiling it into the code vector.

Arguments:
  options      the option bits
  brackets     points to number of extracting brackets used
  code         points to the pointer to the current code point
  ptrptr       points to the current pattern pointer
  errorptr     points to pointer to error message
  optchanged   set to the value of the last OP_OPT item compiled
  reqchar      set to the last literal character required, else -1
  countlits    set to count of mandatory literal characters
  cd           contains pointers to tables

Returns:       TRUE on success
               FALSE, with *errorptr set on error
*/

static BOOL
compile_branch(int options, int *brackets, uschar **codeptr,
  const uschar **ptrptr, const char **errorptr, int *optchanged,
  int *reqchar, int *countlits, compile_data *cd)
{
int repeat_type, op_type;
int repeat_min, repeat_max;
int bravalue, length;
int greedy_default, greedy_non_default;
int prevreqchar;
int condcount = 0;
int subcountlits = 0;
register int c;
register uschar *code = *codeptr;
uschar *tempcode;
const uschar *ptr = *ptrptr;
const uschar *tempptr;
uschar *previous = NULL;
uschar class[32];

/* Set up the default and non-default settings for greediness */

greedy_default = ((options & PCRE_UNGREEDY) != 0);
greedy_non_default = greedy_default ^ 1;

/* Initialize no required char, and count of literals */

*reqchar = prevreqchar = -1;
*countlits = 0;

/* Switch on next character until the end of the branch */

for (;; ptr++)
  {
  BOOL negate_class;
  int class_charcount;
  int class_lastchar;
  int newoptions;
  int skipbytes;
  int subreqchar;

  c = *ptr;
  if ((options & PCRE_EXTENDED) != 0)
    {
    if ((cd->ctypes[c] & ctype_space) != 0) continue;
    if (c == '#')
      {
      /* The space before the ; is to avoid a warning on a silly compiler
      on the Macintosh. */
      while ((c = *(++ptr)) != 0 && c != NEWLINE) ;
      continue;
      }
    }

  switch(c)
    {
    /* The branch terminates at end of string, |, or ). */

    case 0:
    case '|':
    case ')':
    *codeptr = code;
    *ptrptr = ptr;
    return TRUE;

    /* Handle single-character metacharacters */

    case '^':
    previous = NULL;
    *code++ = OP_CIRC;
    break;

    case '$':
    previous = NULL;
    *code++ = OP_DOLL;
    break;

    case '.':
    previous = code;
    *code++ = OP_ANY;
    break;

    /* Character classes. These always build a 32-byte bitmap of the permitted
    characters, except in the special case where there is only one character.
    For negated classes, we build the map as usual, then invert it at the end.
    */

    case '[':
    previous = code;
    *code++ = OP_CLASS;

    /* If the first character is '^', set the negation flag and skip it. */

    if ((c = *(++ptr)) == '^')
      {
      negate_class = TRUE;
      c = *(++ptr);
      }
    else negate_class = FALSE;

    /* Keep a count of chars so that we can optimize the case of just a single
    character. */

    class_charcount = 0;
    class_lastchar = -1;

    /* Initialize the 32-char bit map to all zeros. We have to build the
    map in a temporary bit of store, in case the class contains only 1
    character, because in that case the compiled code doesn't use the
    bit map. */

    memset(class, 0, 32 * sizeof(uschar));

    /* Process characters until ] is reached. By writing this as a "do" it
    means that an initial ] is taken as a data character. */

    do
      {
      if (c == 0)
        {
        *errorptr = ERR6;
        goto FAILED;
        }

      /* Handle POSIX class names. Perl allows a negation extension of the
      form [:^name]. A square bracket that doesn't match the syntax is
      treated as a literal. We also recognize the POSIX constructions
      [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
      5.6 does. */

      if (c == '[' &&
          (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
          check_posix_syntax(ptr, &tempptr, cd))
        {
        BOOL local_negate = FALSE;
        int posix_class, i;
        register const uschar *cbits = cd->cbits;

        if (ptr[1] != ':')
          {
          *errorptr = ERR31;
          goto FAILED;
          }

        ptr += 2;
        if (*ptr == '^')
          {
          local_negate = TRUE;
          ptr++;
          }

        posix_class = check_posix_name(ptr, tempptr - ptr);
        if (posix_class < 0)
          {
          *errorptr = ERR30;
          goto FAILED;
          }

        /* If matching is caseless, upper and lower are converted to
        alpha. This relies on the fact that the class table starts with
        alpha, lower, upper as the first 3 entries. */

        if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
          posix_class = 0;

        /* Or into the map we are building up to 3 of the static class
        tables, or their negations. */

        posix_class *= 3;
        for (i = 0; i < 3; i++)
          {
          int taboffset = posix_class_maps[posix_class + i];
          if (taboffset < 0) break;
          if (local_negate)
            for (c = 0; c < 32; c++) class[c] |= ~cbits[c+taboffset];
          else
            for (c = 0; c < 32; c++) class[c] |= cbits[c+taboffset];
          }

        ptr = tempptr + 1;
        class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
        continue;
        }

      /* Backslash may introduce a single character, or it may introduce one
      of the specials, which just set a flag. Escaped items are checked for
      validity in the pre-compiling pass. The sequence \b is a special case.
      Inside a class (and only there) it is treated as backspace. Elsewhere
      it marks a word boundary. Other escapes have preset maps ready to
      or into the one we are building. We assume they have more than one
      character in them, so set class_count bigger than one. */

      if (c == '\\')
        {
        c = check_escape(&ptr, errorptr, *brackets, options, TRUE, cd);
        if (-c == ESC_b) c = '\b';
        else if (c < 0)
          {
          register const uschar *cbits = cd->cbits;
          class_charcount = 10;
          switch (-c)
            {
            case ESC_d:
            for (c = 0; c < 32; c++) class[c] |= cbits[c+cbit_digit];
            continue;

            case ESC_D:
            for (c = 0; c < 32; c++) class[c] |= ~cbits[c+cbit_digit];
            continue;

            case ESC_w:
            for (c = 0; c < 32; c++) class[c] |= cbits[c+cbit_word];
            continue;

            case ESC_W:
            for (c = 0; c < 32; c++) class[c] |= ~cbits[c+cbit_word];
            continue;

            case ESC_s:
            for (c = 0; c < 32; c++) class[c] |= cbits[c+cbit_space];
            continue;

            case ESC_S:
            for (c = 0; c < 32; c++) class[c] |= ~cbits[c+cbit_space];
            continue;

            default:
            *errorptr = ERR7;
            goto FAILED;
            }
          }

        /* Fall through if single character, but don't at present allow
        chars > 255 in UTF-8 mode. */

#ifdef SUPPORT_UTF8
        if (c > 255)
          {
          *errorptr = ERR33;
          goto FAILED;
          }
#endif
        }

      /* A single character may be followed by '-' to form a range. However,
      Perl does not permit ']' to be the end of the range. A '-' character
      here is treated as a literal. */

      if (ptr[1] == '-' && ptr[2] != ']')
        {
        int d;
        ptr += 2;
        d = *ptr;

        if (d == 0)
          {
          *errorptr = ERR6;
          goto FAILED;
          }

        /* The second part of a range can be a single-character escape, but
        not any of the other escapes. Perl 5.6 treats a hyphen as a literal
        in such circumstances. */

        if (d == '\\')
          {
          const uschar *oldptr = ptr;
          d = check_escape(&ptr, errorptr, *brackets, options, TRUE, cd);

#ifdef SUPPORT_UTF8
          if (d > 255)
            {
            *errorptr = ERR33;
            goto FAILED;
            }
#endif
          /* \b is backslash; any other special means the '-' was literal */

          if (d < 0)
            {
            if (d == -ESC_b) d = '\b'; else
              {
              ptr = oldptr - 2;
              goto SINGLE_CHARACTER;  /* A few lines below */
              }
            }
          }

        if (d < c)
          {
          *errorptr = ERR8;
          goto FAILED;
          }

        for (; c <= d; c++)
          {
          class[c/8] |= (1 << (c&7));
          if ((options & PCRE_CASELESS) != 0)
            {
            int uc = cd->fcc[c];           /* flip case */
            class[uc/8] |= (1 << (uc&7));
            }
          class_charcount++;                /* in case a one-char range */
          class_lastchar = c;
          }
        continue;   /* Go get the next char in the class */
        }

      /* Handle a lone single character - we can get here for a normal
      non-escape char, or after \ that introduces a single character. */

      SINGLE_CHARACTER:

      class [c/8] |= (1 << (c&7));
      if ((options & PCRE_CASELESS) != 0)
        {
        c = cd->fcc[c];   /* flip case */
        class[c/8] |= (1 << (c&7));
        }
      class_charcount++;
      class_lastchar = c;
      }

    /* Loop until ']' reached; the check for end of string happens inside the
    loop. This "while" is the end of the "do" above. */

    while ((c = *(++ptr)) != ']');

    /* If class_charcount is 1 and class_lastchar is not negative, we saw
    precisely one character. This doesn't need the whole 32-byte bit map.
    We turn it into a 1-character OP_CHAR if it's positive, or OP_NOT if
    it's negative. */

    if (class_charcount == 1 && class_lastchar >= 0)
      {
      if (negate_class)
        {