[WINED3D] QueryInterface corrections

[Ivan Gyurdiev]

H. Verbeet wrote:
> On 25/04/06, Ivan Gyurdiev <ivg2 at cornell.edu> wrote:
>> H. Verbeet wrote:
>> > On 25/04/06, Ivan Gyurdiev <ivg2 at cornell.edu> wrote:
>> >
>> >> (2) write a NULL argument on no match.
>> >>
>> >> This patch corrects those issues, and also changes GetContainer() for
>> >> surfaces and volumes to use the return value as MSDN specifies
>> >> (WINED3D_INVALIDCALL), not E_NOINTERFACE on failure.
>> >>
>> > Did you verify this on Windows? MSDN tends to lie about these kind 
>> of things.
>> >
>> No, I'll admit I did not (because I don't have Windows).
>> If somebody could check I'd appreciate it.
>
> I was under the impression that patches are supposed to be verified to
> be correct before being submitted.
You are right. Since it seems I can't rely on the spec to be correct, 
and I don't have Windows, in the future I'll only send no-op patches, or 
patches that I can observe the effect of....

> - At least for d3d9, GetContainer returns E_NOINTERFACE on Windows. I
> verified that when I rewrote part of the function some time ago, and I
> just did again. I originally intended to add a test for that, but
> well, guess I forgot about it.
Can you add a comment to the code when the spec is verified to be wrong.
Otherwise programmers like me will make a mistake and assume the spec to 
be valid.