geode-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Robert Houghton <rhough...@pivotal.io>
Subject Re: [DISCUSS] is overriding a PR check ever justified?
Date Wed, 30 Oct 2019 23:32:41 GMT
Any committer has the 'status' permission on apache/geode.git. Some API
tokens have it as well. Its curl + github API reasoning to do this.

On Wed, Oct 30, 2019 at 2:02 PM Jason Huynh <jhuynh@pivotal.io> wrote:

> If we are going to allow overrides, then maybe what Owen is describing
> should occur.  Make a request on the dev list and explain the reasoning.
>
> I don't think this has been done and a few have already been overridden.
>
> Also who has the capability to override and knows how.  How is that
> determined?
>
> On Wed, Oct 30, 2019 at 1:59 PM Owen Nichols <onichols@pivotal.io> wrote:
>
> > > How do you override a check, anyway?
> >
> > Much like asking for jira permissions, wiki permissions, etc, just ask on
> > the dev list ;)
> >
> > Presumably this type of request would be made as a “last resort”
> following
> > a dev list discussion wherein all other reasonable options had been
> > exhausted (reworking or splitting up the PR, pushing empty commits,
> > rebasing the PR, etc)
> >
> > > On Oct 30, 2019, at 1:42 PM, Dan Smith <dsmith@pivotal.io> wrote:
> > >
> > > +1 for allowing overrides. I think we should avoid backing ourselves
> > into a
> > > corner where we can't get anything into develop without talking to
> apache
> > > infra. Some infrastructure things we can't even fix without pushing a
> > > change develop!
> > >
> > > How do you override a check, anyway?
> > >
> > > -Dan
> > >
> > > On Wed, Oct 30, 2019 at 12:58 PM Donal Evans <doevans@pivotal.io>
> wrote:
> > >
> > >> -1 to overriding from me.
> > >>
> > >> The question I have here is what's the rush? Is anything ever so
> > >> time-sensitive that you can't even wait the 15 minutes it takes for it
> > to
> > >> build and run unit tests? If some infrastructure problem is preventing
> > >> builds or tests from completing then that should be fixed before any
> new
> > >> changes are added, otherwise what's the point in even having the pre
> > >> check-in process?
> > >>
> > >> -Donal
> > >>
> > >> On Wed, Oct 30, 2019 at 11:44 AM Nabarun Nag <nnag@apache.org> wrote:
> > >>
> > >>> @Aaron
> > >>> It's okay to wait for at least the build, and unit tests to complete,
> > to
> > >>> cover all the bases. [There may have been commits in between which
> may
> > >>> result in failure because of the revert]  And it's not hard to get
a
> PR
> > >>> approval.
> > >>>
> > >>> -1 on overriding. If the infrastructure is down, which is the test
> > >>> framework designed to ensure that we are not checking in unwanted
> > changes
> > >>> into Apache Geode, wait for the infrastructure to be up, get your
> > changes
> > >>> verified, get the review from a fellow committer and then check-in
> your
> > >>> changes.
> > >>>
> > >>> I still don't understand why will anyone not wait for unit tests and
> > >> build
> > >>> to be successful.
> > >>>
> > >>> Regards
> > >>> Nabarun Nag
> > >>>
> > >>> On Wed, Oct 30, 2019 at 11:32 AM Aaron Lindsey <alindsey@pivotal.io>
> > >>> wrote:
> > >>>
> > >>>> One case when it might be acceptable to overrule a PR check is
> > >> reverting
> > >>> a
> > >>>> commit. Before the branch protection was enabled, a committer could
> > >>> revert
> > >>>> a commit without a PR. Now that PRs are mandatory, we have to wait
> for
> > >>> the
> > >>>> checks to run in order to revert a commit. Usually we are reverting
> a
> > >>>> commit because it's causing problems, so I think overruling the
PR
> > >> checks
> > >>>> may be acceptable in that case.
> > >>>>
> > >>>> - Aaron
> > >>>>
> > >>>>
> > >>>> On Wed, Oct 30, 2019 at 11:11 AM Owen Nichols <onichols@pivotal.io>
> > >>> wrote:
> > >>>>
> > >>>>> Our new branch-protection rules can sometimes lead to unexpected
> > >>>> obstacles
> > >>>>> when infrastructure issues impede the intended process.  Should
we
> > >>>> discuss
> > >>>>> such cases as they come up, and should overruling the result
of a
> PR
> > >>>> check
> > >>>>> ever be an option on the table?
> > >>>>>
> > >>>>> -Owen
> > >>>>
> > >>>
> > >>
> >
> >
>

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