Raytracing rendering can segfault on MacOS

[mwestphal]

According to the CI, it looks like Raytracing rendering using OSPRay can segfault randomly on MacOS, consider this two runs:

1. https://github.com/f3d-app/f3d/actions/runs/4062370237/jobs/6993347563
2. https://github.com/f3d-app/f3d/actions/runs/4064558126/jobs/6998137126

(1) was run Feb 01, two Raytracing tests failed, one of them failed again on the rerun.
(2) was run Feb 02, two raytracing tests failed, none of them failed again on the rerun.

For both run, cache was used and nothing was rebuilt. macos-latest was probably not updated in the meantime.

So this looks and smells like a flaky tests :/, probably due to memory issues in OSPRay.

Next step is to reproduce this locally on a mac.

[mwestphal]

FYI @Meakk @demarle

[demarle]

Joy - occasional/hard to reproduce test failures - not.

You might also try turning on debug flags for that builder temporarily and see if 1) you can get a stack trace out of it, 2) the debug flags hide the failures via initializing some uninitialized variables to 0. I don't have any great ideas otherwise of top of my head.

[mwestphal]

Thanks, If I identify specific code, I'll point it to you :)

[mwestphal]

I've run 5000 local test without reproducing, Ill need to try a get a stack trace from the ci.

[mwestphal]

Dropping for the 2.0 milestone as this is MacOS specific. If someone reproduce it, then we should investigate. wdyt @Meakk ?

[snoyer]

not sure it helps but here's another random one https://github.com/f3d-app/f3d/actions/runs/4292447814/jobs/7478891761#step:16:565, it passed on the PR before the merge and then it broke on master

[Meakk]

I experienced it again. Should we disable OSPRay tests for macos until it's fixed?

[mwestphal]

I'd say to just rerun the failed test for now