CI picked up regression with upgrades today, and Curtis filed the following blocking bug:
<https://bugs.launchpad.net/juju-core/+bug/1489896> This was a nice simple case as there was only one change under test, and the failure also made it clear the changed code was the issue. But no one jumped on reverting the change when I poked in the channel. So, here's what I did to unblock master as a reference for future - any dev can do this. First, pull master, look at the log, and find the change implicated in the bug: commit 5656861f7e3f2516aaf1a067d9fee7c876a3de15 Merge: 716cafd d260119 Author: Juju bot <[email protected]> Date: Fri Aug 28 06:06:51 2015 +0200 Merge pull request #2871 from waigani/last-conn Make a new branch from master for the revert: $ git checkout -b revert_pr2871 Switched to a new branch 'revert_pr2871' Revert that merged revision: $ git revert -m 1 5656861f7e3f2516aaf1a067d9fee7c876a3de15 This will also commit, and come up with a prefilled message, which I tend to amend by making the title a bit less verbose and add a mention of the regression bug being fixed. In our case with juju, the mainline revision will always be the left-hand parent, so `-m 1` is what you want. Look down further in the log to confirm, the left merge sha prefix should be the previous juju bot commit. Finally push the branch up and propose it for merging: $ git push origin revert_pr2871 Afterwards, it's helpful to reopen the bug the original commit was fixing if one exists, or just mail the list/author otherwise. The whole process, including eyeballing the diff to confirm the right stuff was reverted, takes less than ten minutes, so there's no reason to leave a branch in a broken state. Martin -- Juju-dev mailing list [email protected] Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/juju-dev
