Re: Improving test coverage of extensions with pg_dump

On 03/07/2015 02:34 PM, Michael Paquier wrote:
> On Tue, Mar 3, 2015 at 2:40 PM, Michael Paquier
> <michael(dot)paquier(at)gmail(dot)com> wrote:
>> Those patches are really simple, but then perhaps there are better or
>> simpler ways than what is attached, so feel free to comment if you
>> have any ideas.
>
> Attached are new patches somewhat based on the comments provided by
> Peter Eisentraut
> (http://www.postgresql.org/message-id/54F62C3F.8070702@gmx.net).
> - 0001 makes prove_check install the contents of the path located at
> $(CURDIR)/t/extra if present, which would be the path that test
> developers could use to store extensions, plugins or modules that
> would then be usable by a given set of TAP tests as they are installed
> in the temporary installation before launching the tests.
> - 0002 is the test for pg_dump checking extensions containing FK
> tables, this time integrated as a TAP test in src/bin/pg_dump, with
> the extension in src/bin/pg_dump/t/extra.
> IMO, this approach is scalable enough thinking long-term. And I think
> that the location of those extra extensions is fine in t/ as an
> hardcoded subfolder (any fresh idea being of course welcome) as this
> makes the stuff in extra/ dedicated to only on set of TAP tests, and
> it is even possible to set multiple extensions/modules.

Hmm. I think it'd be better to put the tables_fk extension into
src/test/modules, and the test case under src/test/tables_fk/t/. I'm
inclined to think of this as a test case for an extension that contains
a table, which includes testing that pg_dump/restore works, rather than
as a test of pg_dump itself.

- Heikki