On Tue, Sep 04, 2018 at 01:41:53PM -0400, Tom Lane wrote:
> It's really accidental that $(pg_regress_clean_files) happens to be a
> superset of what the TAP tests need to have cleaned; we shouldn't build
> that assumption in further.
Fine for me.
> If we're gonna do anything here, I think it'd be better to invent some new
> symbol like HAVE_TAP_TESTS for calling Makefiles to define, then move all
> the support into pgxs.mk, including the prove_[install]check rules.
Okay, I don't want to create a dependency between REGRESS and
HAVE_TAP_TESTS either, but modules specifying both should be able to
trigger both regressions and tap tests. So I would be inclined to
create two new rules, say check-regress and installcheck-regress, which
are invoked if check is called, and trigger pg_regress stuff. Then add
on top of it the existing prove-check and prove-installcheck. What do
you think? check and installcheck become this way the centralized place
for all types of test suites.
This would clean up src/test/modules/test_pg_dump/Makefile as well.
--
Michael