[Spice-devel] SPICE: changing the merge rules - a proposal

[Marc-André Lureau]

Hi

On Fri, May 15, 2020 at 7:06 PM Francesco Giudici <fgiudici at redhat.com>
wrote:

> Hi,
>    the community around the SPICE project always tried to follow one
> fundamental, implicit rule for accepting code contributions to the
> project: every merge request (beside trivial patches) should be reviewed
> and acked at least by one before getting merged.
> While everyone agrees with this fundamental rule, the actual status of
> some SPICE projects makes the rule impractical to let the project move
> forward.
>

I wasn't aware of a maintenance problem. Perhaps we should first list the
projects that have maintenance issues & discuss our options, before
changing the common rule.

Let's consider the spice/spice project as an example: the number of
> contributions is very low, both on the commit side (only 4 different
> contributors with more than 1 commit from the beginning of the year, and
> a single contributor with 90% of commits) and on the review side (in the
> last 40 merge requests before the C++ switch one, 21 had no comments).
>

You are omitting the passive reviewers. I consider myself as one of them.
If you need people to be more proactive, you could at least reach me &
probably others past contributors.

The x11spice project is another example: we have only 4 contributors
> from the beginning of the year (and a single contributor holding 70% of
> the commits) and the reviews on the gitlab merge requests have been
> provided by two people only, each one reviewing the merge requests of
> the other.
>

As projects become more specific/marginal, it's clear that the number of
maintainers is lower. Yet, we have those projects under the same umbrella,
and I don't think they should be treated differently. 2 active
developers/maintainers can be very healthy, I would say. So do we have a
maintenance issue with x11spice? Do we want to move the project out of the
"Spice-space" projects for ex? What is the problem exactly?

For the sake of having the projects being able to move forward with a
> reduced number of contributors/reviewers, the proposal is to *allow* a
> maintainer to merge a Merge Request without an explicit ack if the three
> following conditions are met:
> 1) The Merge Request has been pending for at least 3 weeks without
> getting new comments
> 2) The Merge Request submitter has kept asking a review on a weekly basis
> 3) There are no pending nacks on the Merge Request
>
>
It's hard to define a delay to bypass a reviewing & consensus rule. In
general, it should really be frowned upon. There is clearly more than one
person interested and using Spice. If the issue is important, it should be
fairly easy to get someone else to look at the issue in a timely manner. If
not, there are probably more important/interesting things to do to get
other interested.

If Spice is good enough for our users and interested parties, then it may
be risky to change it in wild directions without their approval.

But 3 weeks is way too short. You could have more important work, family
issue, get sick and go on holiday, all happening each after the other. If
we need you to review some code, because you have the best experience, we
should wait. If really we want a delay, I would argue waiting 3 months
minimum (there might be exceptional circumstances, but they are exception,
not the rule). And during that time, the contributor should have attempted
multiple time to involve people, by different means (at least ML, irc and
gitlab issue+MR - eventually try a conf call to explain the motivation,
present the work differently etc, complain about the Spice project laziness
on public blog if need be etc).


Note that having patches reviewed would still be the preferred way. If
> at any time the number of contributors would raise again, we can switch
> back to the mandatory review rule. Until then the priority is to allow
> the project to move forward.
>
> What do you think? Please share your thoughts and/or contribute with
> your own ideas.
>

Thanks, but I think the trivial rule is already very subtle and generally
disapproved (fwiw, I am still in favor of a subjective trivial rule), so I
don't think this proposal would work.

We have to grow a community, by convincing people and doing interesting
work. Not by doing personal thing, and then leave the pieces behind,
because we did it alone and didn't gather others.

Perhaps Spice is doing the job. Perhaps Spice needs to define new
objectives. These are interesting topics that I believe people would want
to discuss and participate. If not, we are doing it wrong.

thanks

-- 
Marc-André Lureau
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/spice-devel/attachments/20200515/143c9262/attachment.htm>