FCP 20190401-ci_policy: CI policy
Brooks Davis
brooks at freebsd.org
Fri Aug 30 12:44:23 UTC 2019
On Thu, Aug 29, 2019 at 08:05:06PM -0400, Ed Maste wrote:
> On Thu, 29 Aug 2019 at 15:05, Rodney W. Grimes
> <freebsd-rwg at gndrsh.dnsmgr.net> wrote:
> >
> > Here in lies one of the fundemental problems, this view by some that
> > a "revert commit is something bad, it is kind of punishment". That is
> > not true. Reverts are GREAT things, they allow the tree to be returned
> > to a known state, usually quicly. The original commit is STILL IN SVN,
> > and a bad revert can guess what.. be reverted!.
>
> Let me echo Rod here. I'm also very happy that this statement was made
> by one of the original FreeBSD committers.
>
> Reverting a change is not an insult, not a punishment, not something
> bad - it's simply an acknowledgement that some aspect of the change
> didn't meet expectations.
We should be considerably more willing to revert changes that break
things. I agree with Warner that it's worth pinging the developer in
question, but if they don't respond in in a couple hours, I see no
value in waiting further. At that point any loss of productivity is on
them and they have already wasted considerable productivity for other
committers.
All that being said, the LLVM project's aggressive use of reverts is
sometimes enormously costly to downstreams with large changes[0] and we
should avoid swinging too far toward reverts. The issue comes about
when a non-trivial patch conflicts with a local change. When such a
patch is reverted that's another conflict and sometimes you end up
making/unmaking the same fixes multiple times when a change causes
problems in an environment that developer can't test directly.
Some balance is required, but I think we should be considerably more
willing to revert changes where a fix isn't immediately available.
-- Brooks
[0] On the CHERI project we're almost certainly the largest public
downstream for LLVM to the point that they validated the git monorepo
migration tools on our tree(s).
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 455 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/freebsd-fcp/attachments/20190830/278c2482/attachment.sig>
More information about the freebsd-fcp
mailing list