Closed traversaro closed 2 years ago
It seems that there is some sort of cross-talking, so even if after one round I disable all failing test, and the following round more test fail. So probably for the moment it is a better idea just to run a few specified tests.
Ok, by running just the INTEGRATION_transport_msg_count
I reproduced the problem that we have in https://github.com/conda-forge/gazebo-feedstock/pull/152 .
Apparently the problem in INTEGRATION_transport_msg_count
on macOS will be solved by https://github.com/gazebosim/gazebo-classic/pull/3269 , but even with that fix the test is not passing in CI for some reason. Anyhow, we can close the PR for now.
In https://github.com/conda-forge/gazebo-feedstock/issues/148 we debugged quite a tricky issue that only appeared at runtime. To avoid having problems like this in the future, I tried to enable tests (skipping the one that are currently failing) in the conda-forge job.
As the conda-forge CI job is already quite slow (~1 hour and 15 minutes), we can also think of changing the job to run the tests only on a given schedule, but for the time being I just implemented the simplest solution.