Re: [PATCH v2 0/2] MR8531: mfmediaengine: Support video output in format R10G10B10A2.

[Nikolay Sivov (＠nsivov)]

Nikolay Sivov (@nsivov) commented about dlls/mfmediaengine/tests/mfmediaengine.c:
>  ok(!!map_desc.pData, "got pData %p\n", map_desc.pData);
>  ok(map_desc.DepthPitch == 16384, "got DepthPitch %u\n", map_desc.DepthPitch);
>  ok(map_desc.RowPitch == desc.Width * 4, "got RowPitch %u\n", map_desc.RowPitch);
> - res = check_rgb32_data(L"rgb32frame.bmp", map_desc.pData, map_desc.RowPitch * desc.Height, &dst_rect);
> - res = check_rgb32_data(expected_frame, map_desc.pData, map_desc.RowPitch * desc.Height, &dst_rect);
> - /* R10G10B10A2 equivalents in openGL and Vulkan are not binary compatible with R10G10B10A2 */
> - todo_wine_if(format == DXGI_FORMAT_R10G10B10A2_UNORM)
> ok(res == 0, "Unexpected %lu%% diff\n", res);

Does that mean it gets different "raw" data when accessed directly e.g. mapped, but rendering is still same and correct? This todo means we are not really testing this format on Wine.

I'm not sure we really need to compare full frame, same for existing rgb32frame.bmp resource. It should be enough to read back from output texture in a few locations, like d3d11/d2d tests do.