Re: pg_regress cleans up tablespace twice.

On Wed, Jun 17, 2020 at 05:02:31PM +0900, Kyotaro Horiguchi wrote:
> It look good to me as the Windows part. I agree that vcregress.pl
> don't need to parse EXTRA_REGRESS_OPTS by allowing a bit more tight
> bond between the caller sites of pg_regress and pg_regress.

Thanks, applied this part to HEAD then after more testing.

> Chaining to TAP sounds nice as a goal.

I submitted a patch for that, but we had no clear agreements about how
to handle major upgrades, as this involves a somewhat large
refactoring of PostgresNode.pm so as you register a path to the
binaries used by a given node registered.

> As the next step we need to amend GNUmakefile not to cleanup
> tablespace for the four test items. Then somehow treat tablespaces at
> non-default place?

Ah, you mean to not reset testtablespace where that's not necessary in
the tests by reworking the rules?  Yeah, perhaps we could do something
like that.  Not sure yet how to shape that in term of code but if you
have a clear idea, please feel free to submit it.  I think that this
may be better if discussed on a different thread though.
--
Michael