On Mon, Aug 19, 2019 at 11:02:42AM -0400, Tom Lane wrote:
> Michael Paquier <michael@paquier.xyz> writes:
>> I have been looking at the isolation tests, and we have in some specs
>> steps which are defined but not used in any permutations.
>
> Hmm, might any of those represent actual bugs? Or are they just
> leftovers from test development?
I cannot yet enter the minds of each test author back this much in
time, but I think that's a mix of both. When working on a new
isolation spec, I personally tend to do a lot of copy-pasting of the
same queries for multiple sessions and then manipulate the
permutations to produce a set of useful tests. It is rather easy to
forget to remove some steps when doing that. I guess that's what
happened with tuplelock-upgrade, insert-conflict-do-update* and
freeze-the-dead.
--
Michael