Contributors mailing list archives


Re: Proposal for new workflow, incorporating "Optimistic Merging"

Holger Brunn
- 07/06/2016 22:45:24
> The goal would be for a few volunteers to drive an *experiment* with
> optimistic merging, in a couple of selected repos. It should keep having
> stable versions with peer reviews and mandatory passing tests. Quality of
> published modules must not be compromised in the process.

I just talked to one of the contributors of OpenUpgrade, and we agreed this 
repo might be a good testbed: Here, it's in the nature of the topic that a 
contribution only covers some specific cases (you often only migrate the parts 
of a module you actually use), there are outrageous times for PRs to get 
merged because we're way too few people, and things are not helped by the fact 
that you need in-depth knowledge of Odoo n, Odoo n+1, the code of both 
versions, and a deep functional understanding of the domain to say something 
useful about the code in question.

So what we talked about was:

- introduce rigid tests
- everything that doesn't fail the tests can me merged more or less at once

This will help us with the for this repo quite notorious problem that the 
migration scripts for module X are partially done, but given there's a pending 
PR for months, other people are discouraged from adding fixes because it clumsy 
to PR on the PR's branch.

Would the proponents of optimistic merging maybe be willing to write out their 
exact vision on ? I plan to start 
with the testing part in the coming days (we need that independently of using 
this approach or any other). Another beauty of this repo is that there's no 
real issue of deployment here. It seems to me that many people in this thread 
tie the question of the merge process to the process of deployment, which 
complicates stuff unnecessarily in my opinion, because they are pretty much 
disparate topics.

PS: I still think for most repos this is not the solution. In my opinion, the 
solution is rather some kind of karma/kudo system where you need to have done 
reviews before your stuff is reviewed and subsequently merged. Sounds simple, 
but I fear it's hard to come up with a measurement of what a 'sensible review' 
is. That put aside, it would make reviewing part of the submitting process, 
and this way give us a lot more reviewers. And as soon as you've done a few 
reviews, you'll notice how much you learn from it and do it anyways 
afterwards. If you want to talk about this, please start a new thread in order 
not to hijack this one.

Therp - Maatwerk in open ontwikkeling

Holger Brunn - Ontwerp en implementatie