Andrew Dunstan <and...@dunslane.net> writes:
> On 10/14/21 12:15 AM, Tom Lane wrote:
>> Looks like a transient/phase of the moon issue to me.
> Bowerbird is having similar issues, so I don't think this is just a
> transient.
Yeah, I noticed that too today, and poked around a bit. But I don't
see what this test is doing differently from other tests that
bowerbird is running successfully. It's failing while trying to crank
up a replica using init_from_backup, which has lots of precedent.
I do see that bowerbird is skipping some comparable tests due
to using "--skip-steps misc-check". But it's not skipping,
eg, pg_rewind's 008_min_recovery_point.pl; and the setup steps
in that sure look just the same. What's different?
(BTW, I wondered if PostgresNode->new's own_host hack could fix this.
But AFAICS that's dead code, with no existing test using it. Seems
like we should nuke it.)
regards, tom lane
