On 5/24/22 11:12, Alexandre Julliard wrote:
Folks,
The Gitlab experiment seems to be going well. For the last release, more than half the patches were submitted as Gitlab merge requests, so that seems to be working well for us.
There are still improvements we can make, and of course many more Gitlab features that we could start using, but before investing more effort we need to decide whether we want to use it going forward.
So now that you have had a chance to try it, what do you people think? Should we adopt Gitlab as our development platform?
What I like about Gitlab:
* push/fetch model eliminates ML hacks by design (where we give up proper git history when sending an email, only to have hackish reconstruction of it in patch tracker, TestBot and reviewers' tools).
* It provides a platform where having Wine Gecko repo fits Wine infrastructure.
* It gives us a nice place to share WIP branches within Wine infrastructure.
* It seems to have a good potential for automation.
On top of above it makes Alexandre's life easier, so I think it's good enough to continue adopting it. There are some things that could be improved:
* The ability of reviewer to push simple fixups to MRs. Gitlab provides some support for it, but it currently doesn't fit our workflow. For the existing option to work, reviewer would need commit access to the target branch. I worked around that with Huw by giving him access to my tree instead, but that does not scale.
* Related to above, when reviewer pushes to MR, MR creator is used by ML bridge as a sender. It seems like it should use the reviewer instead and Gitlab has that information. Eg. [1] should use Huw as a sender.
* I think that a mailing list dominated by a bot does not give an impression welcoming for general discussion. If we're moving forward with this, I think we should move Gitlab bridge to a separated mailing list. We had wine-patches in the past (although in this case it would contain review comments as well).
Jacek
[1] https://www.winehq.org/pipermail/wine-devel/2022-May/217574.html