hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Misty Linville <mi...@apache.org>
Subject Re: [DISCUSS] Move to github flow (WAS subtopic of "Rough notes from dev meetup, day after hbaseconasia 2018, saturday morning")
Date Tue, 21 Aug 2018 16:14:36 GMT
Is GitHub integration going to make patch review more palatable and
efficient for reviewers? Maybe we can increase our roster of regular
reviewers by using GitHub.

We can make a bot that looks for things in PR comment text. That way you
could add the JIRA number in a way the bit can pick it up and use it. Maybe
the bot could even attach the patch to the JIRA automatically on new
commits. Kubernetes uses something called Prow and something else called
Blunderbuss for this type of thing.

On Tue, Aug 21, 2018, 9:03 AM Josh Elser <elserj@apache.org> wrote:

> Summarizing my feelings: A first-step might be to inch towards some
> middle ground where we allow code-review via PRs, but we still require a
> Jira issue and a patch to trigger QA (and avoid authorship issues,
> release note issues, etc) to "gate" the commit.
>
> That gives us clear next steps:
> 1) Once QA works for PRs automatically, we can remove patch step
> 2) Once Yetus doc-maker can pull from GH, we can skip Jira issues
> ...
>
> There's some precedence here with already allowing reviewboard and
> phabricator (in my mind). I would be OK with doing code-review on GH and
> would personally prefer it over all other tools at our disposal.
>
> (some things inline too)
>
> On 8/21/18 12:30 AM, Sean Busbey wrote:
> > A couple of other contribution concerns:
> >
> > a) Yetus Precommit works with github PRs, but the ASF Jenkins admin
> > job that sends requests over to our job that runs the precommit tests
> > does not. It's a minor issue (in that collectively we know what has to
> > change), but someone will have to do the work.
>
> Thanks, Sean. I expected something like this to be an immediate blocker.
>
> > b) We've just started getting better release notes together by using
> > Yetus Release Doc Maker. AFAIK it currently only works off of JIRA. I
> > know we haven't said anything about not having a JIRA associated with
> > changes just because we support github PRs, but it'll be around the
> > corner because we'll be duplicating work for those PRs. The two may
> > very well stay side-by-site for those who don't have/want GitHub
> > accounts, but if anything that makes the situation for "how do I
> > gather release notes" more complicated.
>
> Agreed, that would be the next thing to come down the pipe.
>
> Like point-A, I would guess that this is something we can fix by
> expanding Yetus release-doc maker, but requires someone to do it.
>
> > c) Are more casual PRs a boon? In addition to HBase I spend time on an
> > open source project that relies exclusively on GitHub tooling and I
> > lurk on several others. One thing I've noticed is that while the
> > number of casual PRs is certainly higher they tend to be "drop off
> > PRs"; the engagement for follow up is much lower. Many folks who get
> > that PR up on GitHub then don't come back to address requests from
> > reviewers. We'll have to pick one or more of closing unresponsive PRs,
> > more proactively having committers "fix up" contributions, providing
> > more feedback as "follow-on work" instead of something that gets done
> > during the review. I personally would favor closing unresponsive PRs
> > because it has the least overhead for our already sparse reviewing
> > bandwidth.
>
> That's a good point. I don't have strong feelings on how to interpret
> these.
>
> > d) All this said, we don't need to move to gitbox to accept PRs. We
> > can do anything today that we could do after moving to gitbox except
> > have committers merge the PR directly from the GitHub interface.
> > That's not easier for contributors, that's easier for committers. I'm
> > definitely not saying this is a bad thing. I do a non-trivial amount
> > of reviews from my phone and the github UI is definitely worlds
> > better.
>
> Agreed. I think these are two separate issues.
>
> > On Mon, Aug 20, 2018 at 10:08 PM, Stack <stack@duboce.net> wrote:
> >> This came up at the recent devs meeting: could we move to github flow
> >> committing to Apache HBase? Do folks want this? If so, what would it
> take?
> >> What would it look like?
> >>
> >> The new gitbox repos at apache allow contribution back into apache via
> >> github tooling: PRs can be merged into apache repos with a click of a
> >> button, github-based comments can show as comments in apache JIRA. The
> new
> >> hbase-operator-tools and hbase-connector repos are gitbox based. We can
> run
> >> experiments there with fear of damage to the core.
> >>
> >> The justification is that if our project supported PRs and contribution
> via
> >> github, we could glean more contributors.
> >>
> >> Below I repeat two follow-on comments taken from the "Rough notes from
> dev
> >> meetup, day after hbaseconasia 2018, saturday morning" thread by way of
> a
> >> kickstart:
> >>
> >>  From our Josh Elser:
> >>
> >>> This [supporting PRs] is something the PMC should take to heart. If we
> >> are excluding
> >>> contributions because of how we choose accept them, we're limiting our
> own
> >>> growth. Do we have technical reasons (e.g. PreCommit) which we cannot
> >> accept
> >>> PR's or is it just because "we do patches because we do patches"?
> >>>
> >>
> >> By our Sean:
> >>
> >> "I don't want to bog down this thread, but there are a ton of
> >> unanswered questions for allowing github PRs.
> >>
> >> "The biggest one for me is that JIRA is currently our best hope for an
> >> authoritative place for authorship information. If we're taking PRs
> >> from folks who have GitHub accounts but find ASF JIRA accounts too
> >> burdensome, what are we putting for the author in JIRA? Am I going to
> >> have to look in JIRA before a certain date and in Git after? Or in Git
> >> only if JIRA is set to some "HBase Contributor from GitHub" account?"
> >>
> >> Thanks,
> >> St.Ack
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message