Hi Michael, that does sound like a frustrating experience. Thanks for saying something — it is harder to track how often this is occurring if people don’t report the issue on the dev list.
Geode’s PR pipeline uses https://hub.docker.com/r/teliaoss/github-pr-resource <https://hub.docker.com/r/teliaoss/github-pr-resource> to detect changes to PRs against Geode and trigger PR checks. We don’t know why, but sometimes it misses things. Some options are: 1) just deal with it 2) find a different plugin that is more reliable, if such a thing exists? 3) dive into the inner workings of the Telia project and see if we can fix their bug for them? If you see your PR checks stuck in “expecting” state and they don’t proceed to “pending” within a few minutes, you can push another commit (e.g. an empty commit) to try again. It’s unusual to have to do this more than once, but statistically about 1% of the time it could be necessary (based on a wild guess that Telia might be missing up to 10% of all PR commits). Missing PR checks can also be a symptom of a merge conflict, although this is less common. You may notice that the LGTM checks and/or GitHub itself sometimes report that there is a conflict, but Telia does its own merge that sometimes sees conflicts even when GitHub believes they could be auto-resolved. Pushing an empty commit will force Telia to try again to merge the latest develop, or better yet, it never hurts to merge the latest develop yourself. Does someone want to volunteer to look into 2) or 3)? Is there another solution (I assume you weren’t suggesting getting rid of required checks entirely)? -Owen > On Dec 13, 2019, at 11:30 AM, Michael Oleske <mole...@pivotal.io> wrote: > > Hi Geode Dev! > > This PR https://github.com/apache/geode/pull/4406 is a change to the > readme. However it took 3 empty commits to get it to go green enough to be > allowed to be merged. This seems odd, especially with just a readme > change. Is there something going with how CI works for PRs? This was an > incredibly frustrating experience. > > Thanks! > -michael
