On Wed, Feb 26, 2020 at 06:02:22PM +0100, Bernd Helmle wrote:
> My feeling is that in the case we cannot successfully resolve a
> tablespace location from pg_tblspc, we should error out, but i could
> imagine that people would like to have just a warning instead.
Thanks, this patch is much cleaner in its approach, and I don't have
much to say about it except that the error message for lstat() should
be more consistent with the one above in scan_directory(). The
version for v11 has required a bit of rework, but nothing huge
either.
> I've updated the TAP test for pg_checksums by adding a dummy
> subdirectory into the tablespace directory already created for the
> corrupted relfilenode test, containing a file to process in case an
> unpatched pg_checksums is run. With the patch attached, these
> directories simply won't be considered to check.
What you have here is much more simple than the original proposal, so
I kept it. Applied and back-patched down to 11.
--
Michael