On Tue Aug 22 06:37:25 2023 +0000, Stefan Dösinger wrote:
I think the host machine is running multiple containers in parallel. The 50k query test did randomly take more than 50 seconds. At some point we will run into the test timeout (afaik 120s) I am open to removing this test. I am not exactly sure what it is/was good for. Showing that starting and stopping a query doesn't randomly add samples rendered?
It's from 2dc3d2454, for bug 45932. So it's basically a regression test, and as such does have some real value. But if it's going to take a ridiculous amount of time then I think it's just not worth that cost.
Ideally we'd just get rid of it on llvmpipe where it takes too long to render. We don't have a good way to detect llvmpipe, but we could add one.
But if tweaking the numbers is enough for now then we can live with that instead. I just don't think the todo_wine makes sense.