On Wed Oct 23 16:35:43 2024 +0000, Rémi Bernon wrote:
They are randomly failing/passing with `fvwm` because of various race conditions with X11 window state/config events, `kwin`/`openbox` behave more consistently and don't trigger them, virtual desktop mode of course don't suffer from these problems. I can remove the todo_wine, but I think they are more often failing on Gitlab for some reason, or make them flaky and hide the issue, what do you prefer? Fwiw https://gitlab.winehq.org/wine/wine/-/merge_requests/6569 is meant to fix these window state/config race conditions. There's still some with input focus which are trickier but less frequent, and this MR is also needed to fix the racy dummy window creation in the test itself.
If the tests pass inconsistently on fvwm, I would assert there's not much point picking one or the other result, since we can't rely on it and need to keep d3d9:d3d9ex disabled anyway until they're passing consistently.
Meanwhile, I do run tests regularly on virtual desktop (and occasionally kwin), both for development and always when reviewing patches.
So, given that the todos hamper review while not really fixing the CI, I'd advocate for removing them.