On 6/15/20 2:23 AM, Michael Paquier wrote:
> On Sat, Jun 13, 2020 at 03:00:54PM +0900, Michael Paquier wrote:
>> My take would be to actually enforce that as a requirement for 14~ if
>> that works reliably, and of course not backpatch that change as that's
>> clearly an improvement and not a bug fix. It would be good to check
>> the status of each buildfarm member first though. And I would need to
>> also check my own stuff to begin with..
> So, I have been looking at that. And indeed as Peter said we are
> visibly missing one call to perl2host in 010_pg_basebackup.pl.
>
> Another thing I spotted is that Win32::Symlink does not allow to
> detect properly if a path is a symlink using -l, causing one of the
> tests of pg_basebackup to fail when checking if a tablespace path has
> been updted. It would be good to get more people to test this patch
> with different environments than mine. I am also adding Andrew
> Dunstan in CC as the owner of the buildfarm animals running currently
> TAP tests for confirmation about the presence of Win32::Symlink
> there as I am afraid it would cause failures: drongo, fairywen,
> jacana and bowerbird.
Not one of them has it.
I think we'll need a dynamic test for its presence rather than just
assuming it's there. (Use require in an eval for this).
However, since all of them would currently fail we wouldn't actually
have any test coverage. I could see about installing it on one or two
animals (jacana would be a problem, it's using a very old and limited
perl to run TAP tests.)
cheers
andrew
--
Andrew Dunstan https://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services