From: Tom Lane <tgl@sss.pgh.pa.us>
To: Thomas Munro <thomas.munro@enterprisedb.com>
cc: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Subject: Re: Instability in parallel regression tests
In-reply-to: 
 <CAEepm=11VHs2HhcUWzD47M4vFooUc+4io3a0iYBK2Y+=VfaGdA@mail.gmail.com>
References: <4627.1521070268@sss.pgh.pa.us>
 <CAEepm=1vD0XO9ZzLjcY1CkkKAezLaQga0bBiUvfH59J_EDmrVw@mail.gmail.com>
 <16738.1521081130@sss.pgh.pa.us>
 <CAEepm=11VHs2HhcUWzD47M4vFooUc+4io3a0iYBK2Y+=VfaGdA@mail.gmail.com>
Comments: In-reply-to Thomas Munro <thomas.munro@enterprisedb.com>
	message dated "Thu, 15 Mar 2018 15:42:02 +1300"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <17541.1521082274.1@sss.pgh.pa.us>
Date: Wed, 14 Mar 2018 22:51:14 -0400
Message-ID: <17542.1521082274@sss.pgh.pa.us>
Precedence: bulk

Thomas Munro <thomas.munro@enterprisedb.com> writes:
> On Thu, Mar 15, 2018 at 3:32 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> ... I manually filtered
>> out a bunch of non-problems, in particular discarding names that are in
>> per-test schemas; I think it's all right to allow tests that are taking
>> that precaution to do what they like name-wise.

> What if we always did that?  That is, create a schema with a name
> corresponding to the .sql filename and make it default, as a
> convention?  That might be a smaller and more localised change than
> renaming all these objects.  It would also provide a convenient way to
> drop everything wholesale at the end.

There's a lot of objects that we *want* propagated from earlier tests to
later ones, and/or left around to help with pg_dump testing based on the
final state of the regression database.  I don't think that just dropping
them all would be an improvement.

It might be possible to identify a set of tests that set up persistent
objects as opposed to ones that don't, but it wouldn't be a trivial
task.

			regards, tom lane