Linux Opensplice jobs are flaky for test_publisher_subscriber_{Builtins,Primitives,Empty}

[clalancette]

Linux Opensplice jobs sometimes fail the test_publisher_subscriber_{Builtins,Primitives,Empty} tests, like in https://ci.ros2.org/view/nightly/job/nightly_linux-aarch64_ospl_release/15/ . I think this is an ongoing flakiness, but we should probably look into it eventually.

[clalancette]

Last night they also failed for all OpenSplice jobs on test_requester_replier_{Empty,Primitives}: https://ci.ros2.org/view/nightly/job/nightly_linux-aarch64_ospl_release/21. It may or may not be related.

[clalancette]

Another example: https://ci.ros2.org/view/nightly/job/nightly_linux-aarch64_ospl_release/22/testReport/junit/test_communication.build.test_communication/test_requester_replier__Empty__rclpy__rmw_opensplice_cpp_Release/test_requester_replier/

[mikaelarguedas]

The original issue happens only with opensplice talking to fastrtps and was originally described at ros2/ros2#476 (comment)

As the service tests are consistently failing (on amd64 as well) and it started 2 days ago, I think this is a separate issue that should be tracked in another ticket with the list of potential culprit PRs for the day before the failure appeared

[dhood]

the failing empty requester-replier opensplice tests were resolved by ros2/rclpy#194 e.g. https://ci.ros2.org/view/nightly/job/nightly_linux_ospl_release/25/

[clalancette]

I haven't seen this in over 2 weeks, so closing.