Re: A test for replay of regression tests

вт, 8 июн. 2021 г. в 02:25, Thomas Munro <thomas.munro@gmail.com>:

Ok, here's a new version incorporating feedback so far.

1.  Invoke pg_regress directly (no make).

2.  Use PG_TEST_EXTRA="wal_consistency_checking" as a way to opt in to
the more expensive test.

3.  Use parallel schedule rather than serial.  It's faster but also
the non-determinism might discover more things.  This required
changing the TAP test max_connections setting from 10 to 25.

4.  Remove some extraneous print statements and
check-if-data-is-replicated-using-SELECT tests that are technically
not needed (I had copied those from 001_stream_rep.pl).

Thank you for working on this test set!

I was especially glad to see the skip-tests option for pg_regress. I think it will become a very handy tool for hackers.

To try the patch I had to resolve a few merge conflicts, see a rebased version in attachments.

>       auth_extra       => [ '--create-role', 'repl_role' ]);

This line and the comment above it look like some copy-paste artifacts. Did I get it right? If so, I suggest removing them.

Other than that, the patch looks good to me.