What to do for the d3d11:d3d11 crash on cw-gtx560?

[Francois Gouget]

> From bug https://bugs.winehq.org//show_bug.cgi?id=51103

The commit adding more SwapDeviceContextState() tests (ccffc06c6df3) is 
causing d3d11:d3d11 to crash on the GTX 560 Nvidia card:

https://test.winehq.org/data/patterns.html#d3d11:d3d11

It turns out that the crash is happening in the final 
ID3D11Device1_Release(device) call in test_device_context_state().

So that's probably a reference counting bug in the Nvidia driver but it 
still happens on the latest driver for that GeForce GTX 560 card : 
391.35.

So what should we do?
Drop the final ID3D11Device1_Release(device) call?
Drop the extra tests?
Something else?

[Henri Verbeet]

On Sat, 3 Jul 2021 at 11:30, Francois Gouget  wrote:
> From bug https://bugs.winehq.org//show_bug.cgi?id=51103
>
> The commit adding more SwapDeviceContextState() tests (ccffc06c6df3) is
> causing d3d11:d3d11 to crash on the GTX 560 Nvidia card:
>
> https://test.winehq.org/data/patterns.html#d3d11:d3d11
>
> It turns out that the crash is happening in the final
> ID3D11Device1_Release(device) call in test_device_context_state().
>
> So that's probably a reference counting bug in the Nvidia driver but it
> still happens on the latest driver for that GeForce GTX 560 card :
> 391.35.
>
> So what should we do?
> Drop the final ID3D11Device1_Release(device) call?
> Drop the extra tests?
> Something else?

We probably don't want to drop all of the extra tests, but if there's
a particular operation that's problematic, it seems fine to drop that
part of the test, or at least replace it with a comment.