[Proposal] Maintainer Approvals and SKARA

[Andrew Hughes]

On 14:05 Thu 24 Mar     , Severin Gehwolf wrote:

snip...

> 
> My worry would be that "Fix Request" comments would be spread across
> different PRs if a single bug gets backported to, many update releases.
> JDK 18u, 17u, 11u and 8u, for example. It would move the "approval"
> conversation to the source control system (and per backport PR) rather
> than the bug database. From a maintenance perspective it makes sense to
> have all the info related to backports in one place (on the bug). If
> something gets requested for 8u, say, I can see immediately in the bug
> database that it has been backported to later releases too and can see
> the approval flags (and who approved it). Moving this to skare makes
> the info much more dispersed.

I see this as an advantage. It doesn't make sense to me for the
approval conversation to be separated from the rest of the PR process.
It's even more of a problem when the patch submitter can't comment on
JBS, so there is a level of indirection between the person requesting
approval on the bug and the actual author of the patch.

This came to mind again today when I found myself effectively having a
conversation with myself for
https://bugs.openjdk.java.net/browse/JDK-8255239

Do you really find the fix request comments that useful? I usually
find I have to dig out the original patch review if I need anything
more detailed than whether it went in cleanly or not.

For me, I find more often that they create noise on the bug which
makes it harder for me to find the original diagnosis of the issue
itself.

As to the approval labels (I assume that's what you mean by flags),
they would still be set, but by the SKARA bot, so e.g. filters would
continue to work.  Even without them, the links to backport issues
already demonstrate whether the fix is already in another release.

> 
> Besides, we've had the issue of update project committers being able to
> integrate without formal approval in JBS before. It's only much more
> visible now (which is a good thing). I.e. the problem doesn't change
> with Skara. The difference really is bot guidance. In general, the
> process is fairly well understood and I don't anticipate this to be a
> huge problem in practise.

As I mentioned in my original e-mail, it's less an issue of technical
possibility than ease and the bot guidance.

Yes, someone could commit a bug without approval under Mercurial, but
it was a bit more work to go back to the patch, check it still
applied and worked (no easy branches or rebasing with Mercurial),
and then push it yourself. Even with approval, I always felt a bit
of trepidation doing the actual push.

With SKARA, all it takes is for someone to type '/integrate'. This is
why I have been holding back a formal review (as recognised by the
bot) until the bug is approved, but another reviewer may have done
that instead. The bot currently encourages someone to push without
approval.

8u on GitHub already seems to be attracting more attention than we've
had over the last six months. I'm not sure anyone who can open such a
PR is going to be aware of every aspect of the process, especially
when it contradicts what they are being told by the bot.

> 
> Thanks,
> Severin
> 

Thanks,
-- 
Andrew :)
Pronouns: he / him or they / them
Senior Free Java Software Engineer
OpenJDK Package Owner
Red Hat, Inc. (http://www.redhat.com)

PGP Key: ed25519/0xCFDA0F9B35964222 (hkp://keys.gnupg.net)
Fingerprint = 5132 579D D154 0ED2 3E04  C5A0 CFDA 0F9B 3596 4222