[Qgis-psc] Official PSC call on pull request policies

Nyall Dawson nyall.dawson at gmail.com
Sun Feb 25 15:21:05 PST 2024


Hi PSC,

I would love for an official call to be made on which of two
conflicting pull request queue management policies should be adopted
by QGIS.

There are currently two proposals, and the lack of a formal policy is
causing confusion/conflict in how pull requests are managed.

Policy #1: https://github.com/qgis/QGIS/pull/56062

In short, Sandro proposes that the pull request queue be an open queue
of ALL work happening everywhere, in any state of completeness. Pull
requests are permitted for semi-complete work, and for long-term
(including multi-year) projects which are not yet ready for review or
merge. The justification here is that having this work open in the
queue makes it widely visible and so that other developers are aware
of ongoing work across the community.

Currently, these pull requests will be auto-closed by stalebot due to
the lack of activity on the ticket. Sandro's proposal is to disable
stalebot handing of draft / WIP pull requests, and effectively to
formalise that the queue is a valid place for work of this nature and
status.

(@strk please expand here if you feel I haven't summarised your point
of view correctly!)

Policy #2: https://github.com/qgis/QGIS/pull/56523

In this PR I propose to set a formal policy that draft and WIP pull
requests are NOT suitable for opening against the QGIS repository.

My justification is that we have a long-standing issue with
maintainability of the pull request queue, and anything which
decreases the signal-to-noise ratio on open tickets is undesirable.
When the queue includes work which is not ready for review, then it
becomes very tricky to work out the actual status of pull requests and
which ones should be focused on during review time. (Effectively right
now we have a situation where any pull request which is pushed on the
2nd page of requests will basically NEVER get reviewed, as there is a
constant stream of ready-for-review work flowing into the first page
and the signal-to-noise ratio of ready-for-review/merge PRs on
subsequent pages is extremely low). I do not believe it is fair for
submissions like https://github.com/qgis/QGIS/pull/55172 or
https://github.com/qgis/QGIS/pull/55293 where reviews take SUCH a long
time, and it is my belief that by keeping the queue as small as
possible and avoiding WIP/draft work we will increase the likelihood
that PRs like these can be reviewed more quickly in future.

Please note that there is considerable discussion on
https://github.com/qgis/QGIS/pull/56062 already which should be read
when reviewing this decision.

Can I ask that PSC choose one of these two policies to formally adopt
so that there is no misunderstanding or conflict in future?

Thanks in advance!
Nyall


More information about the QGIS-PSC mailing list