Any commit that is present under refs/heads/* is considered already reviewed
Any commit that is present under refs/heads/* is considered already reviewedYes, I understand this, and this is the way I want it to be. I just don't want commits to be able to land there without going through review. And any user with "Create Reference" privileges somewhere under refs/heads/* can bypass this. So if I need to ensure that everything goes through code review, I cannot have "Create Reference" on any user, am I correct?
--
--
To unsubscribe, email repo-discuss...@googlegroups.com
More info at http://groups.google.com/group/repo-discuss?hl=en
---
You received this message because you are subscribed to the Google Groups "Repo and Gerrit Discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to repo-discuss...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
Ideally Gerrit could have a private space under refs/heads/* which it considered as ineligible for review (e.g. refs/heasd/unreviewed/*) so that it wouldn't need to have this as a workaround - but at the moment this is not the case.
A commit validation plugin can enforce what people can do over a Git push command.The Create Reference would then apply for the GUI whilst the plugin can actually enforce the review path over a Git push.What do you think ?
I think what you want is the ability to restrict the create reference permission to only allow creation with a head that has already been reviewed? Both in the cases where it is pushed or created through the UI. Presumably this would just require a check that the reference is reachable from one of the heads? (Or something similar).
This would also be interesting to us as it is the reason why we are hesitant about allowing people to create branches.
Thomas
--
I think what you want is the ability to restrict the create reference permission to only allow creation with a head that has already been reviewed? Both in the cases where it is pushed or created through the UI. Presumably this would just require a check that the reference is reachable from one of the heads? (Or something similar).
This would also be interesting to us as it is the reason why we are hesitant about allowing people to create branches.
Thomas
From: repo-d...@googlegroups.com [mailto:repo-d...@googlegroups.com] On Behalf Of Tomas Hellberg
Sent: 29 October 2013 12:28
To: repo-d...@googlegroups.com
Cc: Tomas Hellberg
Subject: Re: Preventing bypass of review
A commit validation plugin can enforce what people can do over a Git push command.
The Create Reference would then apply for the GUI whilst the plugin can actually enforce the review path over a Git push.
What do you think ?
Yes, possibly. But I just tested to create a branch from the web interface under refs/heads/* and gave the SHA-1 of something under refs/changes/* (not yet submitted) which got accepted as a new branch. So that's still a problem.
--
--
To unsubscribe, email repo-discus...@googlegroups.com
To unsubscribe, email repo-discuss...@googlegroups.com
I’d be inclined to say it should possibly be core functionality as once a project has been setup this is normally the permissions that you want.
Otherwise a plugin/extension would be fantastic.
Thomas
On Oct 29, 2013 7:16 AM, "Alex Blewitt" <alex.b...@gmail.com> wrote:
>
> On 29 Oct 2013, at 11:54, Tomas Hellberg wrote:
>
> > I need help with figuring out how to configure Gerrit. Let me try to explain what I need:
> >
> > 1. I need to enforce that all commits go through review
> > 2. I want developers to be able to create branches under refs/heads/user/${username}/* (or some other namespace) so I grant them "Create Reference" on that reference
>
> Any commit that is present under refs/heads/* is considered already reviewed, so these can be pushed/submitted to any other branch without going through a review process.
This sounds like a bug to me. What if I want a 'next' branch which is reviewed by one team, and a 'master' branch which is reviewed by a different one?
Phil
P.s. sorry for the thread resurrection.
Phil
P.s. sorry for the thread resurrection.
--