Re: Guix Days: Patch flow discussion

[Suhail <suhail@bayesians.ca>]

Tomas Volf <~@wolfsden.cz> writes:

> In ideal world, there would always be *some* reaction from the
> project, even if that reaction would be "we do not want this change".

Agreed.  A reduction in the time between a patch (or patch revision)
submission and a review/reaction would be a positive change in my
opinion.

> Even if it would be just an auto-close (for the "contributor can't
> work effectively..." case).

Are there current instances of "contributor can't work effectively"?  If
not, I propose that decisions concerning those situations be deferred
till we have actual instances to guide us.

> Or, to put it in a different way: The problem is not that too few patches get
> merged.  The problem is that too few patches get reviewed.

Agreed.

I believe the below steps would help with the current situation, and
perhaps both should be pursued (among possibly others).

1. It would help to eliminate the need for review in certain kinds of
   patches.  Some version upgrades for a package $foo where there exists
   no package $bar that depends on $foo may fall in this category.  More
   generally, some "known to be safe with reasonable confidence" changes
   may be safe to be auto-committed without needing manual review.
   Recognizing such changes could be challenging, but we don't have to
   correctly label all such changes in order for this to be helpful.

2. It would help if it were easier for the community to be able to identify the next
   steps, given a patch submission.  It might help to (more?) clearly
   differentiate between the below states:
   - patch *needs review* (automatically set)
   - patch *needs revision* (set explicitly by the reviewer, say, by
     using a specific keyword in their email)
   - patch *needs followup review* (automatically set if there's a
     revised patch)
   - patch has a *satisfied reviewer* (set explicitly by the reviewer,
     say, by using a specific keyword in their email)

-- 
Suhail