fix for fnmatch
Bruno Haible
haible@ilog.fr
Fri Jul 28 06:41:00 GMT 2000
Hi,
There are three problems with the function is_char_class (used for the
wide-char version) in posix/fnmatch.c:
- The _LIBC version of this function will silently overrun its stack
allocated buffer if a string longer than CHAR_CLASS_MAX_LENGTH + 1
wide characters is passed in.
- The _LIBC version of this function will always return 0 if the
string is not too long, because the (*wcs++ != L'\0') test can never
stop the loop.
- The non-_LIBC version of this function is more complicated than
necessary. Outside glibc you cannot assume ASCII and UCS-4 encodings,
but SUSV2 guarantees that btowc(x) == x for all characters from the
portable character set (i.e. all printable ASCII characters minus
'$', '`', '@').
2000-07-27 Bruno Haible <haible@clisp.cons.org>
* posix/fnmatch.c (is_char_class): Fix logic. Avoid buffer overrun.
Use the optimized version of this function outside glibc as well.
*** glibc-20000724/posix/fnmatch.c.bak Thu Jul 6 17:36:53 2000
--- glibc-20000724/posix/fnmatch.c Fri Jul 28 02:33:00 2000
***************
*** 224,236 ****
# define SUFFIX WC
# define WIDE_CHAR_VERSION 1
-
# undef IS_CHAR_CLASS
- # ifdef _LIBC
/* We have to convert the wide character string in a multibyte string. But
! we know that the character class names are ASCII strings and since the
! internal wide character encoding is UCS4 we can use a simplified method
! to convert the string to a multibyte character string. */
static wctype_t
is_char_class (const wchar_t *wcs)
{
--- 224,236 ----
# define SUFFIX WC
# define WIDE_CHAR_VERSION 1
# undef IS_CHAR_CLASS
/* We have to convert the wide character string in a multibyte string. But
! we know that the character class names consist of alphanumeric characters
! from the portable character set, and since the wide character encoding
! for a member of the portable character set is the same code point as
! its single-byte encoding, we can use a simplified method to convert the
! string to a multibyte character string. */
static wctype_t
is_char_class (const wchar_t *wcs)
{
***************
*** 239,284 ****
do
{
! if (*wcs < 0x20 || *wcs >= 0x7f)
! return 0;
! *cp++ = (char) *wcs;
}
! while (*wcs++ != L'\0');
return __wctype (s);
- }
# else
- /* Since we cannot assume anything about the internal encoding we have to
- convert the string back to multibyte representation the hard way. */
- static wctype_t
- is_char_class (const wchar_t *wcs)
- {
- mbstate_t ps;
- const wchar_t *pwc;
- char *s;
- size_t n;
-
- memset (&ps, '\0', sizeof (ps));
-
- pwc = wcs;
- n = wcsrtombs (NULL, &pwc, 0, &ps);
- if (n == (size_t) -1)
- /* Something went wrong. */
- return 0;
-
- s = alloca (n + 1);
- assert (mbsinit (&ps));
- pwc = wcs;
- (void) wcsrtombs (s, &pwc, n + 1, &ps);
-
return wctype (s);
- }
# endif
# define IS_CHAR_CLASS(string) is_char_class (string)
# include "fnmatch_loop.c"
# endif
int
fnmatch (pattern, string, flags)
--- 239,299 ----
do
{
! /* Test for a printable character from the portable character set. */
! # ifdef _LIBC
! if (*wcs < 0x20 || *wcs > 0x7e
! || *wcs == 0x24 || *wcs == 0x40 || *wcs == 0x60)
! return (wctype_t) 0;
! # else
! switch (*wcs)
! {
! case L' ': case L'!': case L'"': case L'#': case L'%':
! case L'&': case L'\'': case L'(': case L')': case L'*':
! case L'+': case L',': case L'-': case L'.': case L'/':
! case L'0': case L'1': case L'2': case L'3': case L'4':
! case L'5': case L'6': case L'7': case L'8': case L'9':
! case L':': case L';': case L'<': case L'=': case L'>':
! case L'?':
! case L'A': case L'B': case L'C': case L'D': case L'E':
! case L'F': case L'G': case L'H': case L'I': case L'J':
! case L'K': case L'L': case L'M': case L'N': case L'O':
! case L'P': case L'Q': case L'R': case L'S': case L'T':
! case L'U': case L'V': case L'W': case L'X': case L'Y':
! case L'Z':
! case L'[': case L'\\': case L']': case L'^': case L'_':
! case L'a': case L'b': case L'c': case L'd': case L'e':
! case L'f': case L'g': case L'h': case L'i': case L'j':
! case L'k': case L'l': case L'm': case L'n': case L'o':
! case L'p': case L'q': case L'r': case L's': case L't':
! case L'u': case L'v': case L'w': case L'x': case L'y':
! case L'z': case L'{': case L'|': case L'}': case L'~':
! break;
! default:
! return (wctype_t) 0;
! }
! # endif
!
! /* Avoid overrunning the buffer. */
! if (cp == s + CHAR_CLASS_MAX_LENGTH)
! return (wctype_t) 0;
! *cp++ = (char) *wcs++;
}
! while (*wcs != L'\0');
+ *cp = '\0';
+
+ # ifdef _LIBC
return __wctype (s);
# else
return wctype (s);
# endif
+ }
# define IS_CHAR_CLASS(string) is_char_class (string)
# include "fnmatch_loop.c"
# endif
+
int
fnmatch (pattern, string, flags)
More information about the Libc-alpha
mailing list