[QGIS-Developer] A plea: more volunteers needed for reviewing backports

[Matthias Kuhn]

On Fri, Apr 23, 2021 at 2:00 AM Nyall Dawson <nyall.dawson at gmail.com> wrote:

> On Fri, 9 Apr 2021 at 19:10, Matthias Kuhn <matthias at opengis.ch> wrote:
> >
> > Hi,
> >
> > Thanks for the continued effort in reviewing. It's one of the
> not-so-visible-highly-technically-and-socially-qualified-people-needed-but-very-important
> aspects of (open source) software development.
> >
> > I try to go over the PR queue every now and then and
> merge/comment/review what's possible, but these days I find less time than
> in the past. While at the same time it seems pull request activity
> increases it seems.
> > I'll make sure we raise this topic at an upcoming OPENGIS.ch meeting and
> encourage people to participate.
> >
> > Additionally to what has been said already, backports are easy to forget
> / ignore by the original author because they have been opened by a bot and
> as an author you won't receive notifications by reviewers or the stale bot.
> I wonder if we could get the original author more involved if we'd mention
> them in the backport comment and increase their responsibility for this
> part. This might be worth some automation.
>
> I definitely think this would help. The mentality at the moment is
> predominantly that "as soon as the original PR is merged, the
> responsibility is no longer mine" and that "someone else" will handle
> the backports. In part this is due to how easy the bot has made
> backporting.. we've now all got the mentality that the bot will handle
> everything for us, but that's not the case in reality. I think
> tweaking the backport message so that the author is mentioned will
> help here, as at least they'll get notified if we close the backport
> due to merge conflicts/etc. I'd also love to see the bot fixed so that
> commits are cherry-picked and the ORIGINAL author name is attributed
> to the commit. I think this is very important for accountability...
> currently all the backport commits are anonymous and attributed to the
> bot only, which makes it hard to tell who is responsible for the
> change. I wish we could keep the original author here, as I think this
> helps increase burden of responsiblity for that author... A poor
> quality or buggy backport will directly reflect on their reputation,
> so they are more likely to self-police backport PRs and ensure they
> are suitable for merge. (At least, I hope so).
>

Points addressed in an update to the backport bot. Merge commits are
"unpacked", responsibility given to the original author. (Rebase merges
still not supported).
I'm not super confident about everything I did here, if the backport bot
behaves weird in the next few days, please let me know.


>
> The other really painful thing with backport bot is that we have to
> manually close and reopen ALL automated backports in order for the
> tests to run. It's a minor thing, but definitely contributes to the
> "chore" and drudgery of maintaining the PR list. Especially because
> only a few have permission to do this, and unless the original
> contributor has merge rights they can't even close/open their own
> backports to help speed things along.
>

Not sure what's happening, any idea why?


>
> Don't get me wrong - backport bot *is* great, and has simplified work
> a lot. But with a bit more investment and refinement it could be
> incredible and save me substantially more time!
>
> > Something else is that for example for me, the process to decide which
> backports get into pending backports was not obvious at first (only the LTR
> or also LR? At which stage of the LTR? How exactly do they get in there?),
> I'm sure for other people there are other parts of the process that are not
> immediately clear. I think documenting the review process could help here
> (the suggestions written by Nyall above for a lower entry barrier into the
> reviewer process would already be worth mentioning).
>
> Right, we should definitely document this better. I'd suggest you and
> I get together sometime to do this, as we've been appointed this
> responsibility by PSC already. Can you ping me off list so we can
> arrange this?
>

Done

Matthias
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/qgis-developer/attachments/20210426/a54214b5/attachment.html>