Stefan Dösinger (@stefan) commented about dlls/mfmediaengine/tests/mfmediaengine.c:
+ hr = ID3D11Device_CreateTexture2D(d3d11_dev, &texture_desc, NULL, &video_texture); + ok(hr == S_OK, "Unexpected hr %#lx.\n", hr); + + hr = ID3D11Device_CreateShaderResourceView(d3d11_dev, (ID3D11Resource *)video_texture, NULL, &d3d11_srv); + ok(hr == S_OK, "Unexpected hr %#lx.\n", hr); + + rasterizer_desc.CullMode = D3D11_CULL_NONE; + rasterizer_desc.FillMode = D3D11_FILL_SOLID; + hr = ID3D11Device_CreateRasterizerState(d3d11_dev, &rasterizer_desc, &d3d11_rs); + ok(hr == S_OK, "Unexpected hr %#lx.\n", hr); + + /* Media engine init */ + notify = create_callback(); + ok(!!notify, "notify was %p.\n", notify); + + engine = create_media_engine(¬ify->IMFMediaEngineNotify_iface, (IUnknown *)d3d11_dev, DXGI_FORMAT_R32G32B32_FLOAT); I know the RGB32F triggered writing this test, but I think a basic test with a known-good video format is worth having too. Consider using RGBA8 here and add a separate test for weird formats.
Is there a way to prove that patch 4 is correct, i.e., can you query the actual video format MF gives you somehow? -- https://gitlab.winehq.org/wine/wine/-/merge_requests/1046#note_10529