Michael Paquier <michael(dot)paquier(at)gmail(dot)com> writes:
> And attached is the patch. I arrived at the conclusion that the
> assertion being broken here just needs to be relaxed a bit so as it
> understands that this code path can be taken by a matview WITH NO
> DATA. I have bundled a regression test as well. This should be
> backpatched down to 9.3.
Pushed. The Assert was still a bit too strong: as you had it, it
still crashed on CREATE TABLE AS ... WITH NO DATA.
regards, tom lane