Currently when our D3D11 device is reinitialized because of hardware updates or removal, we crash. We shouldn't.

Comment on attachment 8532262 [details] [diff] [review] Deal properly with a D3D11 device reset Review of attachment 8532262 [details] [diff] [review]: ----------------------------------------------------------------- ::: gfx/layers/d3d11/CompositorD3D11.cpp @@ +952,5 @@ > 0); > HandleError(hr); > } > + > + if (FAILED(hr)) { please replace this FAILED(hr) by Failed(hr) and remove the two HandleError above. Also you could just do return Succeeded(hr); ::: gfx/thebes/gfxWindowsPlatform.cpp @@ +385,5 @@ > */ > + if (DidRenderingDeviceReset()) { > + mD3D11DeviceInitialized = false; > + mD3D11Device = nullptr; > + mD3D11ContentDevice = nullptr; Shouldn't you null out the dxgi adapter too? (I had to do it in my own version of the device reset patch.

This seems to work properly now for all reset scenarios I've tried. I'm sure I've missed something though :-).

Comment on attachment 8539894 [details] [diff] [review] Deal properly with a D3D11 device reset v2 Review of attachment 8539894 [details] [diff] [review]: ----------------------------------------------------------------- r+ with the assertion below back in place. ::: gfx/layers/d3d11/CompositorD3D11.cpp @@ +1221,5 @@ > return; > } > // XXX - It would be nice to use gfxCriticalError, but it needs to > // be made to work off the main thread first. > + //MOZ_ASSERT(aSeverity != DebugAssert); Please uncomment this line.

https://hg.mozilla.org/mozilla-central/rev/6c673d6a2fb8 https://hg.mozilla.org/mozilla-central/rev/3ca05c4232ab