+1 maintenance report

[Adrien Nader]

On Mon, Feb 20, 2023, Steve Langasek wrote:
> On Mon, Feb 20, 2023 at 03:51:50PM +0100, Adrien Nader wrote:
> > Due to the size of the queues, I decided to start with the oldest
> > issues.
> 
> Out of interest, how did you identify these "oldest issues"?  I happened to
> spend some time on proposed-migration on Thursday, and all of the packages I
> worked on are older than the oldest package on your list, AFAICS.  (The
> *youngest* package I worked on was node-object-inspect, which is currently
> 208 days old.)
> 

Ah, right, "oldest issues" was how I approached this initially but most
of these were either node ones which I have no experience with, or
blocked on purpose, or packages for which I didn't spot a path forward
relatively quickly. Since that was my first +1 maintenance, I didn't
want to be stuck the whole week on one or two packages which were still
problematic even though they had most likely been looked by others before.

> > # sparse
> > I'm under the impression that upstream doesn't run their own testsuite
> > at the moment.
> > For instance, there is a diagnostic which has been changed in 2019 in
> > 2094267c7d36d8696897c509558fc63e8a695765 and new testcases have been
> > added with that diagnostic but older ones have not been updated
> > accordingly (the one failing because of that change has been created in
> > 2018 and hasn't been changed since then).
> > All in all this seemed to much for a +1 shift and I didn't act on the
> > package.
> > 
> > # blurhash-python
> > I tried to reproduce the issue which occured on armhf but I didn't
> > manage to. There were actually two different issues and that made me
> > think the issues were unrelated to this package. Due to the size of the
> > queues, I didn't ask for re-triggers.
> 
> This package has since migrated.  I guess someone did retrigger and it
> passed.
> 
> Please don't be shy about retriggering because of the queues; the tests get
> run eventually...

My main issue was the length of the queues and how long it would take to
get some feedback.

Overall I asked Lukas to re-trigger two dozens of tests (I thought that
was more but I just counted again) but for blurhash-python, it felt like
overloading the queues without being able to look at the results. I also
identified these as potentially fixed by retries early in the week but
since I probably wouldn't get results during the week, I chose to
analyse them more first.

I think I would have acted very differently had I known roughly when the
tests would have run. I jokingly said to Lukas that we would get the
results in two weeks, i.e. long enough that I'm done with the +1 shift
and that we forget about them.

Nonetheless, I'll probably err on the other side next time and begin
with retrying in similar situations and look again near the end of the
week.

-- 
Adrien