CI: RDF loader test is flaky

[henrygerardmoore]

Description

The RDF Loader appears to fail in CI for no reason on one commit and then succeeds on the other where only a comment was removed.

Your environment

• ROS Distro: Humble
• OS Version: Ubuntu 22.04
• Built from source on the remove_boost branch of my fork (gh pr checkout 1331 to see)

Steps to reproduce

Only encountered this once, so it is difficult to say how to reproduce it. Possibly just repeated runs of CI will do it.

Expected behaviour

Test should pass consistently

Actual behaviour

Test appears to fail inconsistently

Backtrace or Console output

CI for failing test CI for passing test

[mikeferguson]

I notice in the logs "terminate called without an active exception" - which usually means that a std::thread went out of scope while not joined/stopped - I notice we have a std::thread in this test - but I don't actually see how it could go out of scope...

I fixed an instance of this in https://github.com/ros-planning/moveit2/pull/1384

[henningkayser]

I haven't seen this one lately. Are we still running into this? Maybe it really was an unrelated threading issue somewhere...

[vatanaksoytezer]

I think I saw this like a few months ago, but haven't seen lately.

[henningkayser]

closing for now