I agree with your comments.

Ill extend this patch with more tests, like non-1channel textures and other corner cases.
Also I will detail more on the logic behind the arguments used for the test.

Regarding the D3DPTEXTURECAPS_POW2, I just wanted to be sure those textures are supported,
but if you think is not needed ill remove it in the next version.

And I can try to draft a DF16/DF24 tests. However I did not think it was necessary
since the depth formats are very similar to other depth formats in wine.
And there is 1:1 mapping between those depth formats. (DF16 -> D16, DF24 -> D24S8)
(unless there is a difference in precision or int/float used for storage,
but that is not detailed in the spec)

El jue., 22 nov. 2018 a las 15:01, Stefan Dösinger (<stefandoesinger@gmail.com>) escribió:
Hi,

Can you add a test showing that FETCH4 is disabled when the bound
texture is not a single-channel texture? I sense that there might be
applications that enable FETCH4 and then "forget" to disable it when
they render from an ARGB texture.

A test for DF16 and DF24 would be great as well, but it doesn't belong
in this patch and should be a separate one. It's probably best to extend
intz_test.

> +    if (caps.TextureCaps & D3DPTEXTURECAPS_POW2)
> +    {
> +        skip("No unconditional NP2 texture support, skipping FETCH4 test.\n");
> +        IDirect3DDevice9_Release(device);
> +        goto done;
> +    }
Do you still need this?

> +    hr = IDirect3DDevice9_SetSamplerState(device, 0, D3DSAMP_MAGFILTER, D3DTEXF_LINEAR);
> +    ok(SUCCEEDED(hr), "SetSamplerState failed, hr %#x.\n", hr);
> +    hr = IDirect3DDevice9_SetSamplerState(device, 0, D3DSAMP_MINFILTER, D3DTEXF_LINEAR);
> +    ok(SUCCEEDED(hr), "SetSamplerState failed, hr %#x.\n", hr);
> +    hr = IDirect3DDevice9_SetSamplerState(device, 0, D3DSAMP_MIPFILTER, D3DTEXF_LINEAR);
> +    ok(SUCCEEDED(hr), "SetSamplerState failed, hr %#x.\n", hr);
This contradicts AMD's description, but I assume you placed them here to
show that their docs are wrong and FETCH4 is used regardless of the
filter. I think a comment would be useful.