@Everyone After some investigation with Kirk, it appears that the newer Concourse PR resource prioritizes commit time as part of its identification of "newer" commits on a given PR's branch. As a result, you can accidentally paint yourself into a corner with git rebase if your (actually newer) Commit Time does not appear more recent than your previous commit's Commit Time. This can particularly be an issue coming from a place where we have pushed empty commits to kick the old PR resource, since (a) empty commits are often dropped during a rebase, and (b)
The take away is this: *If your precheckin isn't firing, you might have accidentally have a HEAD that is older than the previous precheckin run.* This generally shouldn't be an issue, but "generally shouldn't be" isn't the same as "won't be". If it happens to you: We are referring here to the Commit Time, not the Author Time. Use git log --format=fuller to get all of your commit metadata. You can find Concourse's view of your PR's history in the resource page [1]. Search for your PR number there. If the time shown is newer than your PR's HEAD, you need to get a newer commit to your HEAD. Your options are: * Merge with the develop branch. * Push an empty commit. * Amend the commit of your HEAD to update the Commit Time, then force your push. I advocate them in that order, but understand there are times where you want a particular git history. Good luck out there. Imagination is Change. ~Patrick [1] https://concourse.apachegeode-ci.info/teams/main/pipelines/apache-develop-pr/resources/geode On Tue, Nov 13, 2018 at 9:31 AM, Kirk Lund <kl...@apache.org> wrote: > Just in case you don't believe me and want to double-check my answer: > > My PR which says there are NO CONFLICTS: > https://github.com/apache/geode/pull/2778 > > Thanks, > Kirk > > On Tue, Nov 13, 2018 at 9:31 AM, Kirk Lund <kl...@apache.org> wrote: > > > Yes github says it merges cleanly (still does in fact): > > > > This branch has no conflicts with the base branch when rebasingRebase and > > merge can be performed automatically. > > > > The previous revision DID have a merge conflict which is exactly why I > > rebased on develop, resolved conflicts and then pushed to my fork. > > Rerunning the tests against my latest push is what seems to be failing. > > > > So I'll change my question since some folks seem convinced that I'm > > causing the problem: > > > > Why does concourse not run tests for my PR after I push changes that > > resolve a conflict? > > > > Thanks, > > Kirk > > > > On Mon, Nov 12, 2018 at 5:26 PM, Robert Houghton <rhough...@pivotal.io> > > wrote: > > > >> Does github say it merges cleanly? I thought their messaging was pretty > >> clear > >> > >> On Nov 12, 2018 17:19, "Alexander Murmann" <amurm...@pivotal.io> wrote: > >> > >> I wonder if we somehow could find a way to make the error message > clearer. > >> Since it didn't merge cleanly, there is action I need to take as the > >> committer. However, I would never know that from the error message. > >> > >> On Mon, Nov 12, 2018 at 4:34 PM Patrick Rhomberg <prhomb...@pivotal.io> > >> wrote: > >> > >> > >> > See also the previous time this occurred. > >> > > >> > https://markmail.org/thread/apwupkbkg3qipygo > >> > > >> > On Mon, Nov 12, 2018 at 4:18 PM, Owen Nichols <onich...@pivotal.io> > >> wrote: > >> > > >> > > This error usually means it didn’t merge cleanly, not that concourse > >> is > >> > > “down" > >> > > > >> > > > On Nov 12, 2018, at 3:58 PM, Kirk Lund <kl...@apache.org> wrote: > >> > > > > >> > > > I'm trying to get my PR to run tests, but the latest concourse > jobs > >> are > >> > > > failing with: > >> > > > > >> > > > rsync_code_down-OpenJDK8 > >> > > > missing inputs: geode, instance-data > >> > > > archive-results-OpenJDK8 > >> > > > missing inputs: concourse-metadata-resource, geode, geode-results > >> > > > delete_instance-OpenJDK8 > >> > > > missing inputs: geode, instance-data > >> > > > > >> > > > Anyone know what that means? > >> > > > > >> > > > Thanks, > >> > > > Kirk > >> > > > >> > > > >> > > >> > > > > >