Magnus Hagander <> writes:
> On Tue, Jan 3, 2017 at 7:33 PM, Tom Lane <> wrote:
>> Somehow the reset is clobbering local configuration on some members?

> I doubt that. I think that was probably never configured, it just didn't
> show up when everything was working.

> I don't know what the buildfarm run, but perhaps it's trying to inject a
> merge commit there or something, and that's why it's failing on not having
> a name?

AFAICS, for a pre-existing branch it'll just do

        # do a checkout in case the work tree has been removed
        # this is harmless if it hasn't
        my @colog = `git checkout . 2>&1`;
        my @pulllog = `git pull 2>&1`;

More reports are coming in now, and it's clear that only some of the
critters are failing.  It sort of looks like the fastest machines
are the unhappy ones, which might mean that the ones that aren't failing
happened to never pull the bad commit, because they were busy rebuilding
other branches while it was there.

> I'm guessing the solution is to reset the 9.2 branch to a point prior to
> the commit and then pull again? Or wouldn't just a rebase work?

Flushing the local mirror would likely be the easiest way out.

                        regards, tom lane

Sent via pgsql-hackers mailing list (
To make changes to your subscription:

Reply via email to