Closed azeey closed 2 years ago
Thanks for reporting, the test only fails in foxy and not in Nightly or Rolling jobs is that correct?
There was an oberhaul of the tests structure to remove the use of the daemon and make these tests less flaky https://github.com/ros2/sros2/pull/214, it was not backported to foxy because it was not considered a critical bug as this had no impact on the resulting debs.
Thanks for reporting, the test only fails in foxy and not in Nightly or Rolling jobs is that correct?
It seems like this issue may have combined two separate sets of failures. The issue described is present in Foxy on Linux. The recent nightly and rolling builds don't seem to exhibit this failure.
it was not backported to foxy because it was not considered a critical bug as this had no impact on the resulting debs.
If it's a non-breaking change that can be backported doing so would increase the confidence when testing and producing Foxy patch releases.
Separately on macOS with rmw_connext_cpp there has been an sros2 test failure in two of the last three nightlies
I can open a separate issue for that if preferred.
sros2.test.sros2.commands.security.verbs.test_generate_policy_no_nodes.test_generate_policy_no_nodes
has also failed on 3 out of 4 of the most recent Windows Debug nightlies.
This continues to be an issue on Windows debug: https://ci.ros2.org/job/nightly_win_deb/1921/testReport/sros2.test.sros2.commands.security.verbs/test_generate_policy_no_nodes/test_generate_policy_no_nodes/history/
And it's also occurring on Linux foxy CI: https://build.ros2.org/job/Fci__nightly-fastrtps_ubuntu_focal_amd64/256/testReport/sros2.test.sros2.commands.security.verbs/test_generate_policy/test_generate_policy_topics/history/
Foxy flakyness:
I can open a separate issue for that if preferred.
Yes please let's open another issue for the nightly flakiness so that we dont mix the code versions to modify.
Separately on macOS with rmw_connext_cpp there has been an sros2 test failure in two of the last three nightlies
https://ci.ros2.org/view/nightly/job/nightly_osx_release/1981/testReport/junit/ros2param.ros2param.test/test_verb_list/test_verb_list/
https://ci.ros2.org/view/nightly/job/nightly_osx_release/1983/testReport/junit/ros2param.ros2param.test/test_verb_list/test_verb_list/
Wrong links ? these seem to be for ros2param
These tests have been causing issues for a long time as they rely on the ROS Graph to be clean. This seemed to be improved after #214 but from what I read here there seem to still be an issue on Windows Debug. Are there processes not terminating properly on these jobs ?
This continues to be an issue on Windows debug
These look like hanging issues. From https://github.com/ros2/sros2/pull/214#issuecomment-644382734 and looking at the job output, seems that most(/all?) ros2 cli related tests are disabled on windows. So maybe it's an issue from the tooling below the code in this repository? Should all ros2cli based tests just be disabled for Windows?
@azeey did you have a chance to confirm if #254 fixes this issue ?
There's a similar issue with the new rmw_connextdds
, though the root cause might be different:
On that note, https://github.com/ros2/sros2/pull/260 may help. Launch related hangings may also have been solved in Galactic since https://github.com/ros2/launch/pull/476. We haven't re-enabled tests on Windows though (still too flaky), so take it with a grain of salt.
I've merged #254. But also, macOS is no longer a supported platform for Foxy. So either way, I think we can close this ticket.
Bug report
Required Info:
Steps to reproduce issue
test_generate_policy
ortest_generate_service
tests fail consistently if ros2 daemon is running. On CI machines, the the tests are flaky, but fail most of the time (eg. history on Focal)Expected behavior
All tests pass.
Actual behavior
test_generate_policy_topics
fails withand
test_generate_policy_services
fails withAdditional information
Adding a small sleep right before calling
cli.main
in both tests fixes the tests for me locally. Alternatively, passing--no-daemon
togenerate_policy
fixes the issue as well, but I'm not sure which one, if either, is a good option.