Re: [PATCH v3 0/2] MR2821: wined3d: Improve d3d8 compatibility of minification and magnification filters.

[Stefan Dösinger]

Stefan Dösinger (@stefan) commented about dlls/d3d8/tests/device.c:
> +        BOOL has_texture;
> +        DWORD filters[7];
> +    }
> +    tests[] =
> +    {
> +        {
> +            TRUE,
> +            {
> +                D3DTEXF_NONE,
> +                D3DTEXF_POINT,
> +                D3DTEXF_LINEAR,
> +                D3DTEXF_ANISOTROPIC,
> +                D3DTEXF_FLATCUBIC,
> +                D3DTEXF_GAUSSIANCUBIC,
> +                0xdeadbeef,
> +            },
> This feels awkward, the filter enum is the same in both cases.

I think the test[] array can be removed entirely, or one filter[] array be used for both. (At which point tests[] becomes a relatively pointless array with TRUE and FALSE in it).

I am not married to testing the value 0xdeadbeef, D3DTEXF_GAUSSIANCUBIC + 1 (6) would work just as well imo. That'd allow you to do for (mip = 0; mip <= D3DTEXF_GAUSSIANCUBIC + 1; ++mip).

To answer your question re texture formats that can't be filtered: To my knowledge that doesn't exist in d3d8, and there is no equivalent to D3DUSAGE_QUERY_FILTER in d3d8 - this value is used to check for filtering capability in IDirect3D9::CheckDeviceFormat.

Hypothetically, a volume texture could have different filter capabilities than a 2D texture. This seems to be the case even on my r200 - it supports mip filters (D3DPTFILTERCAPS_MIPFPOINT, D3DPTFILTERCAPS_MIPFLINEAR) for 2D, but not cube or 3D textures. I highly doubt that it complains about this in ValidateTexture though. I'll give it a try just in case.

-- 
https://gitlab.winehq.org/wine/wine/-/merge_requests/2821#note_33093