Dear Michael, Paul, Mutaamba,
> FWIW, I had my doubts at the beginning on the thread about the
> use-case, but after re-reading the whole thing I am going to side with
> Robert's opinion that if a fix to authorize some of the operations is
> simple, then let's just authorize the case(s) and have the fix.
OK, thanks for the clarification. Let's focus on 0001 now.
> Echoing with Paul, what I find as critically lacking from the proposed
> patches are regression tests to check and validate the behaviors you
> are looking for in the patches. Please add some in the shape of perl
> commands that use direct --single commands of postgres. We have cases
> that do so currently in the tree: sepgsql and shm. I don't see a
> reason why we could not do that as well here with run_log(). It is
> also possible to pass a -c exit_on_error=true to force failures.
>
> Not having a test for the single-user case of a slot drop is of course
> something that we are lacking now. As we are playing with this area
> of the code, let's add something for this case as well.
I added a first version of the test. It could work on my env (Linux).
Since I cannot come up the appropriate place, I introduced new test file in
recovery test. creating/dropping/advancing a slot, and doing a logical decoding
is tested. Since standby mode is not supported by the single-user instance,
I did not test the slot synchronization. I have not known that.
0002 was updated accordingly just in case. In the patch some functions were expected
to fail. Also, a check for pg_sync_replication_slots() was removed because we cannot call
the function in the first place.
IIUC, 0001 can be backpatched for all supported branches. I will create for them
after patch would be a good shape.
Best regards,
Hayato Kuroda
FUJITSU LIMITED
