On Wed, Mar 01, 2023 at 10:53:59PM -0800, Nathan Bossart wrote:
> I don't feel a strong need for that, especially now that we aren't
> modifying params anymore.
That was mostly OK for me, so applied after tweaking a couple of
places in the tests (extra explanations, for one), the comments and
the code.
The part that improved the tests of PROCESS_TOAST was useful on its
own, so I have done that separately as 46d490a. Another thing that I
have found incorrect is the need for two pg_stat_reset() calls that
would reflect on the whole database, in the middle of a test running
in parallel of other things. As far as I understand, you have added
that to provide fresh data after a single command while relying on
vactst, but it is possible to get the same amount of coverage by
relying on cumulated counts, and that gets even more solid with all
these commands run on an independent table.
--
Michael
