Rémi Bernon (@rbernon) commented about dlls/mfmediaengine/tests/mfmediaengine.c:
+ desc.SampleDesc.Count = 1; + hr = ID3D11Device_CreateTexture2D(device, &desc, NULL, &texture); + ok(hr == S_OK, "Unexpected hr %#lx.\n", hr); + + if (hr == E_INVALIDARG) + goto skip_transfer; /* Skip when texture can't be created. */ + + res = WaitForSingleObject(notify.ready_event, 500); + ok(!res, "Unexpected res %#lx.\n", res); + + SetRect(&dst_rect, 0, 0, desc.Width, desc.Height); + hr = IMFMediaEngineEx_TransferVideoFrame(notify.media_engine, (IUnknown *)texture, NULL, NULL, NULL); + + todo_wine_if(tested_formats[i].output_todo) + ok(hr == tested_formats[i].exp_hr || broken(hr == tested_formats[i].broken_hr || hr == E_POINTER /* w1064v1507 */ || + hr == MF_E_NO_VIDEO_SAMPLE_AVAILABLE /* Spontaneous error */), "Unexpected hr %#lx.\n", hr); There too it's a bit hard to parse what is expected and if this will actually be tested. There's a lot of possible broken results. Could it be reduced somehow? At least the `broken_hr` seems a bit pointless if it doesn't cover all the broken hrs, maybe it should just be a boolean to accept failures (though it makes the test unlikely to catch anything).
-- https://gitlab.winehq.org/wine/wine/-/merge_requests/1046#note_17231