On Sun, Jul 12, 2020 at 09:25:07PM +0900, Michael Paquier wrote:
> longfin has been failing here:
> https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=longfin&dt=2020-07-12%2011%3A49%3A08
>
> The buildfarm logs include no output, but the problem comes from
> -DENFORCE_REGRESSION_TEST_NAME_RESTRICTIONS, where replication origins
> should be prefixed by "regress_". I did not know that this rule was
> applied here.
Sorry for the delay. jacana has been complaining here:
https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=jacana&dt=2020-07-12%2018%3A02%3A48
Visily switching to tstzrange() is not a brilliant idea, as I suspect
here that the problem is caused by the timestamp comparison in the
upper bound: using the internal range checks, the upper bound is
excluded, so if the follow-up query's now() matches the commit
timestamp of the previous transaction, the comparison fails.
--
Michael