Possibility of accidental merge commits in stable backports
Just wanted to provide a heads up that this Gerrit bug was brought to the attention of the Infra team yesterday: http://code.google.com/p/gerrit/issues/detail?id=1107 In short, be on the lookout for any backport change which is actually the same Git SHA as on the master branch (probably the most obvious sign will be lack of a cherry-pick header in the commit message, since changes to the commit message also change the Git SHA). Currently Gerrit won't protect us from someone proposing a change from one branch to another if the parent commit of that change is already in the former branch, even though it's not (yet) in the latter. -- Jeremy Stanley
In short, be on the lookout for any backport change which is actually the same Git SHA as on the master branch (probably the most obvious sign will be lack of a cherry-pick header in the commit message, since changes to the commit message also change the Git SHA). Currently Gerrit won't protect us from someone proposing a change from one branch to another if the parent commit of that change is already in the former branch, even though it's not (yet) in the latter.
Nice one, thanks for the heads up Jeremy. Any ideas how to check git history for those? But I guess it would have been noticed, master and stable branches are different enough. Just in case, I've added an explicit note[1] about the additional reason cherry-pick -x required, it was implicit by example and stable reviewers have been enforcing it already. Cheers, Alan [1] https://wiki.openstack.org/w/index.php?title=StableBranch&diff=63196&oldid=62527
On 2014-09-20 18:31:15 +0200 (+0200), Alan Pevec wrote:
Nice one, thanks for the heads up Jeremy. Any ideas how to check git history for those?
Not really, unfortunately. Now that I've become aware of this I'm going to be pressing for some sort of solution (hopefully implemented upstream and probably, quite ironically, backported to the release of Gerrit we're running right now).
But I guess it would have been noticed, master and stable branches are different enough. [...]
Yeah, at least that's how the problem came to our attention, courtesy of the stackforge/fuel-main core reviewers. -- Jeremy Stanley
participants (2)
-
Alan Pevec
-
Jeremy Stanley