Hi everyone, The test which is consistently failing is *stmgr_unittest::test_back_pressure_instance* (*StMgr, test_back_pressure_instance*). It appears to be a deadlock or livelock situation that can be alleviated with more cores or multiple runs (we get lucky with the thread interleaving). I will create a GH issue with my findings today.
As far as CI goes, I think a combination of a provider (TravisCI or Jenkins) with Artifactory would be ideal. I would choose the CI provider who can give us more cores for each build. It is also essential that people actively working on PRs have the ability to restart/retrigger a CI build without the need to push more code. On Fri, Nov 12, 2021 at 11:04 AM Josh Fischer <[email protected]> wrote: > Hi All, > > I participated in a discussion with some people from the infra team about > our issues with TravisCI. I've asked Gavin to look into what it would take > to host a build cache for Heron to make our builds faster as well as > options to remove Travis CI. Some options for us to consider are > > - Jenkins > - Artifiactory Pipelines > - GithubActions > > From my perspective, infra is suggesting we use either Artifactory > pipelines or Jenkins. Any thoughts? >
