RE: [bugfix]"make installcheck" could not work in PGXS

Hello,

On Fri, Aug 5, 2020 at 10:57:56 +0000, Michael Paquier <michael(at)paquier(dot)xyz> wrote:
> Yeah, that's a known problem. One way to counter that is for example
> to grab the path of pg_regress from pg_config --libdir and set
> $ENV{PG_REGRESS} to it, but that's hacky. So I agree that it would be
> good to do something.

Thank you.
I attach a new patch.

> Not sure that this goes completely to the right direction. It seems
> to me that we should have room to set and use PG_REGRESS also for
> pg_regress_check and pg_regress_installcheck.

I understand that PG_REGRESS is an environment variable for each test program.
So I add a gmake variable PG_REGRESS_PATH.

The followings are other changings.
- Change REGRESS_SHLIB like as PG_REGRESS.
- Replace $(CURDIR)/$(top_builddir) to $(abs_top_builddir).
- Remove setting of environment variable 'top_builddir' in command line for prove_installcheck.
I wonder what I should set to it and then I remove it.
Because top_builddir is used for gmake genellaly not for test programs and PostgreSQL's test framework doesn't use it.
Is it going too far?

Regards
Ryo Matsumura