On Tue, Nov 27, 2018 at 04:26:45PM +0100, Michael Banck wrote:
> Oh, I kinda followed that thread a bit, but I think that patch fixes
> things more by matter of moving code around, at least I haven't noticed
> that tablespaces were explicitly claimed to be fixed in that thread.
That may have been an unconscious move ;)
It seems to me the root issue is that the original implementation of
pg_verify_checksums naively assumed that files can be skipped without
first checking at their types, which I got confused by in d55241af.
You should definitely be mentioned as this reporter anyway, and get
author credits for the additional test.
> As pg_verify_checksums appears to be broken with respect to tablespaces
> in REL_11_STABLE (so I think 11.1, but not 11.0) as well, I think this
> merits a short-term minimal invasive fix (i.e. the patch you posted,
> just that there's no TAP testsuite for pg_verify_checksums in v11
> unfortunately) on its own, regardless of the wider changes proposed in
> the other thread.
Yes, let's do so rather sooner than later if there are no objections
from others. I am going to ping the other thread about what's discussed
here additionally in case folks missed what you reported. Fixing the
temp file issue which has been reported by Andres first is an additional
bonus.
--
Michael