On Wed Apr 5 17:33:04 2023 +0000, Alex Henrie wrote:
Thanks for the feedback. Who is the right person to review this code? I thought either Stefan or you would be familiar with the problem, and I'm at a loss as to who else to ask.
I don't know, and at some point it's maybe just a matter of whether this seems to make sense, and testing it. The test indeed passes on AMD, so maybe it's fine?
At the same time you say that you tried on an NVidia machine and the test passes -it doesn't allow sharing with a destination context that's been used? or do you mean it succeeds sharing and fails the `ok`?- how can the application also work then?
If this test is supposed to be reversed, maybe instead of removing it you should reverse its expectations, adding a broken result for all the cases where it fails, and adding a comment about what we chose as the "right" case to make it more explicit.