On 12/1/17, 2:03 PM, "Robert Haas" <robertmhaas@gmail.com> wrote:
> Thanks. I think this looks fine now, except that (1) it needs a
> pgindent run and (2) I vote for putting the test case back. Michael
> thought the test case was too much because this is so obscure, but I
> think that's exactly why it needs a test case. Otherwise, somebody a
> few years from now may not even be able to figure out how to hit this
> message, and if it gets broken, we won't know. This code seems to be
> fairly easy to break in subtle ways, so I think more test coverage is
> good.
Makes sense. I ran pgindent and re-added the test case for v6 of the
patch.
Nathan