[PATCH 1/2] d3d10core/tests: Multisampled render targets are zeroed on creation (v2).

[Henri Verbeet]

On 6 January 2016 at 23:56, Stefan Dösinger <stefandoesinger at gmx.at> wrote:
> +    hr = ID3D10Device_CreateTexture2D(device, &desc, NULL, &single);
> +    ok(SUCCEEDED(hr), "Failed to create texture, hr %#x.\n", hr);
> +    ID3D10Device_ResolveSubresource(device, (ID3D10Resource *)single, 0,
> +            (ID3D10Resource *)multi, 0, DXGI_FORMAT_R8G8B8A8_UNORM);
> +
> +    get_texture_readback(single, &rb);
> +    for (y = 0; y < 480; ++y)
> +    {
> +        for (x = 0; x < 640; ++x)
> +        {
> +            color = get_readback_color(&rb, x, y);
> +            if (!compare_color(color, 0x00000000, 0))
> +            {
> +                all_zero = FALSE;
> +                break;
> +            }
> +        }
> +        if (!all_zero)
> +            break;
> +    }
> +    release_texture_readback(&rb);
> +    ok(all_zero, "Got unexpected color 0x%08x, position %ux%u.\n", color, x, y);
> +
This probably works, but depends very much on ResolveSubresource()
working as intended. I think it would be a little more robust to first
clear "single" with a different color. Also, it's not strictly
required, but if you use a swapchain surface to resolve to, it's
easier to later debug the test visually by inserting a Present() call
in the appropriate place.