Last modified: 2013-04-14 16:55:49 UTC
See this patchset: https://gerrit.wikimedia.org/r/#/c/59051/2 I had originally uploaded its dependent (55450), then split out the patchset in question as an antecedent. Zuul rolled over and died while checking out the new intermediate patch, with an error that the treeish of 55450 (b7c0...) does not exist. There is no dependency in that direction, so what gives.
The failed merge is https://integration.wikimedia.org/ci/job/mwext-CentralNotice-merge/538/console which happened at 06:11:40 UTC. That job was triggered with: ZUUL_REF refs/zuul/master/Z541ad16acc6a42518c2ba95a19020a5e ZUUL_COMMIT 4fe5839b9d3f2720c3864a257df5c0bb782861d5 But it tried to checkout b7c0a44e386e0bf296c17c3de4294d43d61a719a which is patchset 13 of https://gerrit.wikimedia.org/r/#/c/55450/ which was there at 06:10 roughly. So that is the race condition described in bug 46723: ---------------------------------------------- The branch specifier in the git plugin is set to ZUUL_BRANCH which is 'master'. In the git plugin (at git-plugin/src/main/java/hudson/plugins/git/util/DefaultBuildChooser.java ), the getCandidateRevisions() will recognize whether the branch looks like a sha1 (if it matches /[0-9a-f]{6,40}/) and in such a case will create a detached branch using that commit. Seems the Jenkins job macro should then use ZUUL_COMMIT as a branch specifier. ---------------------------------------------- I have deployed the fix on mediawiki-core jobs and will populate it for extensions on monday. Meanwhile this bug report is valid but an exact duplicate of bug 46723 :-) *** This bug has been marked as a duplicate of bug 46723 ***
And the fix is https://gerrit.wikimedia.org/r/#/c/58865/ which I will deploy on monday :-D
nice one!
Yeah that is a nice bug. I cant believe I haven't noticed it since we have switched to Zuul to trigger the job (that was six months ago).