No Parallel testing on PostgreSQL 9.6 or Higher

[theory]

I noticed because the partitions test fails on 9.6, even though it should be added to EXCLUDE_TEST_FILES in the Makefile. But the tests were also serial; when I tested on 9.5, the tests ran in parallel (way faster!). But they do not run in parallel for me on 9.6 - 14. Shouldn't they?

And if there are situations where tests should not or cannot run in parallel, we should be sure that the proper tests are skipped for them, too.

I think we should get this straightened out before releasing v1.2.0, which we should try to do ASAP now that Postgres 14 is out and tests fail (see #278).

[nasbyj]

If it won't run in parallel I'd wager it's because it's using the same object names as some other test.

[theory]

Object names?

[nasbyj]

Most likely table names duplicated between tests, but could be functions or other things as well.

[theory]

It runs in parallel fine on older versions.

[theory]

Fixed the issue with parallel test on my Mac, but this still fails on 9.6:

make installcheck PARALLEL_CONN=1

It's odd, because test/build/run.sch looks like this:

ignore: partitions
test: aretap
test: check
test: cmpok
test: coltap
test: do_tap
test: enumtap
test: extension
test: fktap
test: functap
test: hastap
test: index
test: inheritance
test: istap
test: matching
test: moretap
test: ownership
test: partitions
test: performs_ok
test: performs_within
test: pg73
test: pktap
test: policy
test: privs
test: resultset
test: roletap
test: ruletap
test: runjusttests
test: runnotests
test: runtests
test: throwtap
test: todotap
test: trigger
test: unique
test: usergroup
test: util
test: valueset

Note that partitions appears twice. When I build it without PARALLEL_CONN=1, it looks like this:

test: coltap
test: hastap
ignore: partitions
test: aretap check cmpok do_tap enumtap extension fktap functap index inheritance istap matching moretap ownership performs_ok performs_within pg73 pktap policy privs resultset roletap ruletap runjusttests runnotests runtests throwtap todotap trigger unique usergroup util valueset

So it appears only once. Something isn't filtering it out.