Gisle Aas wrote:
> Perl use the UTF8_ALLOW_ANYUV mask in functions that should not be
> restricted to only the valid Unicode code points. For some reason
> this mask currently include the UTF8_ALLOW_LONG flag. This seems
> totally wrong as there can't be a good reason to allow overlong
> sequences just because we don't want to restrict the valid values.
>
> Perl's ord() function is for instance perfectly happy with an overlong NUL:
>
> $ perl -MEncode -wle '$a = "\xe0\x80\x80";Encode::_utf8_on($a);print ord($a)'
> 0
>
> This patch fixes this problem:
Thanks, applied as change #23632 to bleadperl (although I'm not sure I
fully understand all the implications.)
> --- utf8.h.cur 2004-12-06 11:16:52.176181667 +0100
> +++ utf8.h 2004-12-06 11:17:16.672129909 +0100
> @@ -183,8 +183,7 @@
> #define UTF8_ALLOW_FFFF 0x0040 /* Allows also FFFE. */
> #define UTF8_ALLOW_LONG 0x0080
> #define UTF8_ALLOW_ANYUV (UTF8_ALLOW_EMPTY|UTF8_ALLOW_FE_FF|\
> - UTF8_ALLOW_SURROGATE|\
> - UTF8_ALLOW_FFFF|UTF8_ALLOW_LONG)
> + UTF8_ALLOW_SURROGATE|UTF8_ALLOW_FFFF)
> #define UTF8_ALLOW_ANY 0x00FF
> #define UTF8_CHECK_ONLY 0x0200
>
>
>
> With this patch the example above outputs:
>
> $ perl -MEncode -wle '$a = "\xe0\x80\x80";Encode::_utf8_on($a);print ord($a)'
> Malformed UTF-8 character (3 bytes, need 1, after start byte 0xe0) in ord at
> -e line 1.
> 0
Could you turn this into a regression test ?
--
You probably wouldn't have expected a communist to have a dog named Harpo.
-- Malcolm Lowry, Under the Volcano
