On Sat, Jan 25, 2020 at 3:39 PM Melanie Plageman <melanieplage...@gmail.com> wrote: > So, I think from a code review perspective the code in the patches > LGTM. > As for the EXPLAIN ANALYZE tests--I don't see that many of them in > regress, so maybe that's because they aren't normally very useful. In > this case, it would only be to protect against regressions in printing > the leader instrumentation, I think. > The problem with that is, even with all the non-deterministic info > redacted, if the leader doesn't participate (which is not guaranteed), > then its stats wouldn't be printed at all and that would cause an > incorrectly failing test case...okay I just talked myself out of the > usefulness of testing this. > So, I would move it to "ready for committer", but, since it is not > applying cleanly, I have changed the status to "waiting on author".
Hi Melanie, Thanks for the reviews! I think I'm going to abandon 0002 for now, because that stuff is being refactored independently over here, so rebasing would be futile: https://www.postgresql.org/message-id/flat/CAOtHd0AvAA8CLB9Xz0wnxu1U%3DzJCKrr1r4QwwXi_kcQsHDVU%3DQ%40mail.gmail.com On that basis, I've set it to ready for committer (meaning 0001 only). Thanks for the rebase. I'll let that sit for a couple of days and see if anything conflicting comes out of that other thread. It's a fair complaint that we lack tests that show the new output; I'll think about adding one too.
