Re: wined3d: Set wrapmode for cubemags to clamp regardeless of the sampler state

[Stefan Dösinger]

Am Samstag 10 März 2007 15:02 schrieb Fabian Bieler:
> This Patch forces GL_CLAMP_TO_EDGE on all cubemaps.
>
> I tested this on XP with some Radeon X1???.
>
> Also, this fixes some renderissues Half-Life 2.

Say, does that fix the strange quad boarders like in this screenshot? 
http://bugs.winehq.org/attachment.cgi?id=5297&action=view (Eve Online). It 
sounds like it would do that, I'm very impressed :-)

How did you test that? It sounds like a test case using the visual test 
framework is worth an investigation. I can try to write a test for that.

wined3d: Set wrapmode for cubemags to clamp regardeless of the sampler state

[Fabian Bieler]

> Am Samstag 10 März 2007 15:02 schrieb Fabian Bieler:
> > This Patch forces GL_CLAMP_TO_EDGE on all cubemaps.
> >
> > I tested this on XP with some Radeon X1???.
> >
> > Also, this fixes some renderissues Half-Life 2.
>
> Say, does that fix the strange quad boarders like in this screenshot? 
> http://bugs.winehq.org/attachment.cgi?id=5297&action=view (Eve Online). It
> sounds like it would do that, I'm very impressed :-)

I don't have Eve Online, but the screenshot looks like a glitch with the flashlight in Half-Life 2 that this patch fixes.
> How did you test that? It sounds like a test case using the visual test 
> framework is worth an investigation. I can try to write a test for that.

I adapted a Directx9 HLSL tutorial to display a black cubemap with one white line on one edge of one side and visually (by hand or better to say by eye) confirmed that the line is not "mirrored" on the opposite edge.

wined3d: Set wrapmode for cubemags to clamp regardeless of the sampler state

[Stefan Dösinger]

> I adapted a Directx9 HLSL tutorial to display a black cubemap with one
> white line on one edge of one side and visually (by hand or better to say
> by eye) confirmed that the line is not "mirrored" on the opposite edge.

I'd say the patch is fine. Anything else but clamp to edge does not make sense 
for cube maps anyway, considering how the location of the pixel is 
calculated[1]. The only way I can imagine that a texture coordinate gets > 
1.0 or < -1.0 is due to filtering.

So the patch has my blessing to go in, if any regressions occur then we can 
always add a test case. But I cannot think of any good way to test this, so I 
prefer to wait for an application to show up what can go wrong :-)

[1]... 
http://oss.sgi.com/projects/ogl-sample/registry/ARB/texture_cube_map.txt

wined3d: Set wrapmode for cubemags to clamp regardeless of the sampler state

[H. Verbeet]

On 11/03/07, Stefan Dösinger  wrote:
> But I cannot think of any good way to test this, so I
> prefer to wait for an application to show up what can go wrong :-)

I've written a test for this, I'll send it tomorrow together with some
other patches.

wined3d: Set wrapmode for cubemags to clamp regardeless of the sampler state

[H. Verbeet]

On 12/03/07, H. Verbeet  wrote:
> I've written a test for this, I'll send it tomorrow together with some
> other patches.

Since this patch doesn't appear to have been applied yet, I can't
submit my test yet. I've attached it to this mail instead.