While the test VM was functional I was running locally as many tests as possible including Postgres, Druid, etc.
The current instructions for preparing a release do ask to ensure that tests succeed and optionally ask for extra tests. If more people believe that the release manager should not run additional tests then we should update the instructions on the website. Personally, I don't think it is a huge overhead and it adds some additional variety to the testing environment. Btw it would be great if one day we manage to incorporate all tests in the CI. Best, Stamatis On Sun, Mar 29, 2020, 6:35 PM Vladimir Sitnikov <[email protected]> wrote: > >sharing since there may be people who think that ./gradlew prepareVote > does > >everything. > > Frankly speaking, I think our CI is much stronger than the test developers > run on their machines. > For instance: do you expect people to run `slow tests` during prepareVote? > Do you expect people to run PostgreSQL tests as a part of prepareVote? > > That is why I think it makes no sense running the tests during prepareVote. > > At the end of the day, the release candidates (==source release) is what > everyone tries to download and test. > The jar files are reproducible. In other words, Nexus artifacts will be > exactly the same when built from the source release. > > So this boils down to "please refrain from building a release candidate > from a commit you have never pushed through CI". > > >In that sense adding test task as prerequisite of prepare would be merely > a > >reminder. > > That is true. > > Vladimir >
