Farid Zaripov-2 wrote:
>
> The 0.fnmatch test fails on MSVC and ICC/Windows with two assertions:
>
> --------------------------
> 291. rw_fnmatch("[!a]", "", 0) ==> 0, expected (native:own) 1:1
> 301. rw_fnmatch("[!ab]", "", 0) ==> 0, expected (native:own) 1:1
> --------------------------
>
> The corresponding failed lines:
>
> --------------------------
> TEST (1, "[!a]", "");
> TEST (1, "[!ab]", "");
> --------------------------
>
> I don't see platform specific code in rw_fnmatch() or in
> _rw_bracketmatch() and I can't understand how this test is
> fails on Windows and succeeds on another platforms?
>
> Farid.
>
>
The problem is that the `next' pointer [in rw_fnmatch] is incremented past
the end of string...
case '[':
if (esc) {
if (*pc != *next)
return 1;
esc = false;
}
else {
pc = _rw_bracketmatch (pc + 1, UChar (*next), arg); // *next
is null here
if (0 == pc)
return 1;
}
++next; // next is now one byte past null
break;
Then dereferenced in the next iteration...
case '\0':
return *pc == *next ? 0 : 1; // *next is past end of string
On other platforms, the string may have multiple null bytes, but on this one
it does not.
Travis
--
View this message in context:
http://www.nabble.com/0.fnmatch-failures-on-MSVC-and-ICC-Windows-tp15141900p15154035.html
Sent from the stdcxx-dev mailing list archive at Nabble.com.
