Re: [COMMITTERS] pgsql: Improve pg_dump regression tests and code coverage

Robert,

* Robert Haas (robertmhaas(at)gmail(dot)com) wrote:
> On Mon, Mar 20, 2017 at 8:33 AM, Stephen Frost <sfrost(at)snowman(dot)net> wrote:
> > * Robert Haas (robertmhaas(at)gmail(dot)com) wrote:
> >> So was this 3340 line patch posted or discussed anyplace before it got
> >> committed?
> >
> > I've mentioned a few times that I'm working on improving pg_dump
> > regression tests and code coverage, which is what these were. I'm a bit
> > surprised that it's, apparently, a surprise to anyone or that strictly
> > adding regression tests in the existing framework deserves very much
> > discussion.
>
> I'm not saying it does. I'm saying that it's polite, and expected, to
> post patches and ask for opinions before committing things.

While I certainly agree with that when it comes to new features, changes
in work-flow, bug fixes and other things, I'm really not sure that
requiring posting to the list and waiting for responses every time
someone wants to add some regression tests is a useful way to spend
time. While the patch looked big, a lot of that is just that the
current structure requires listing out every 'like' and 'unlike' set for
each test, which adds up.

In this particular case, I've been discussing these pg_dump regression
tests for months, as I've been going through fixing bugs found by them
and back-patching them. I had time over this weekend to watch the
buildfarm and make sure that it didn't explode (in which case, I would
have reverted the patch immediately, of course). I would have preferred
to commit these new tests in a more fine-grained fashion, but I kept
finding issues, which meant that commiting them earlier would have just
turned the buildfarm red, which wouldn't have been beneficial to anyone.

I'm quite pleased to see that, for the most part, the tests have been
successful on the buildfarm.

Thanks!

Stephen