[openstack-dev] [tripleo] patch abandoment policy
dprince at redhat.com
Mon Mar 27 12:47:59 UTC 2017
On Mon, 2017-03-27 at 13:49 +0200, Flavio Percoco wrote:
> On 24/03/17 17:16 -0400, Dan Prince wrote:
> > On Thu, 2017-03-23 at 16:20 -0600, Alex Schultz wrote:
> > > Hey folks,
> > >
> > > So after looking at the backlog of patches to review across all
> > > of
> > > the
> > > tripleo projects, I noticed we have a bunch of really old stale
> > > patches. I think it's time we address when we can abandon these
> > > stale
> > > patches.
> > >
> > > Please comment on the proposed policy. I know this has
> > > previously
> > > been brought up  but I would like to formalize the policy so
> > > we
> > > can
> > > reduce the backlog of stale patches. If you're wondering what
> > > would
> > > be abandoned by this policy as it currently sits, I have a gerrit
> > > dashboard for you (it excludes diskimage-builder) .
> > I think it is fine to periodically review patches and abandon them
> > if
> > need be. Last time this came up I wasn't in fan of auto-abandoning
> > though. Rather I just made a pass manually and did it in fairly
> > short
> > order. The reason I like the manual approach is a lot of ideas
> > could
> > get lost (or silently ignored) if nobody acts on them manually.
> > Rather then try to automate this would it serve us better to add a
> > link
> > to your Gerrit query in  below to highlight these patches and
> > quickly go through them.
> I used to do this in Glance. I had 2 scripts that ran every week. The
> first one
> would select the patches to abandon and comment on them saying that
> the patches
> would be abandoned in a week. The second script abandoned the patches
> that had
> been flagged to be abandoned that were not updated in a week.
I don't think a week is enough time to react in all cases though. There
could be a really good idea that comes in, gets flagged as abandoned
and then nobody thinks about it again because it got abandoned.
There is sometimes a fine line between automation that helps humans do
their job better... and automation that goes to far. I don't think
TripleO or Glance projects have enough patch volume that it would take
the core team more than an hour to triage patches that need to be
abandoned. We probably don't even need to do this weekly. Once a month,
or once a quarter for that matter would probably be fine I think.
> It was easy to know what patches needed to be checked since these
> script ran w/
> its own user (Glance Bot). I believe this worked pretty well and the
> Glance team
> is now working on a better version of that bot.
> I'd share the scripts I used but they are broken and depend on
> another broken
> library but you get the idea/rule we used in Glance.
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubs
More information about the OpenStack-dev