[QGIS-Developer] 32 by 3.2?

[Nyall Dawson]

Hi all,

It's no surprise to anyone familiar with the QGIS project that we've
got an issue with the Pull Request queue. It's been slowly growing
over time, recently hitting over 150 open requests! It's a bit of an
embarrassment to the project (some of these PRs have been open for
years!), and is likely causing us to lose new contributors and code.

The usual magic QGIS coding pixies did some work lately and squashed
the queue back below 100 requests. But the remaining ones are all the
difficult, unfinished or orphaned PRs...

PR reviewing is hard. Not everyone can review every open PR due to
different familiarity with areas of the codebase. (Which is why I
don't think a funding grant to cover this will ever work
successfully). And no-one wants to be the 'bad guy" who closes an
unmerged PR representing someone else's hard work.

So I propose a "32 by 3.2" sprint, where we ALL collaboratively aim to
reduce the PR queue to <32 open requests before 3.2 release.

I think we could achieve this by:

1. Adopting a hard-line approach to the older, orphaned PRs. Even if
they have some value or reflect real issues, if no-one is interested
in cleaning up the request to get it merge ready then we close it.

2. Adopt a "open-one, close-one" guideline for core committers. Heck,
I think every core committer has at least 1 or 2 open PRs representing
various experiments and WIP in unfinished states. These should either
be finished off, or closed and re-opened when the work is actually
ready to go. And for test PRs which are "for comment only" I'd suggest
a QEP is more likely to get better feedback and is the more
appropriate place for this discussion of this nature.

3. Closing orphaned or risky PRs which are targeted to 2.18 and which
have been fixed in master branch.

4. Sharing the hard work so that the magic pixies don't lose their
magic powers :)

Thoughts?

Nyall