Andres Freund <andres@anarazel.de> writes: > I don't quite understand how this "visibly" broke between 9.5.15 and > 9.5.19 as the op reports? Haven't dug into it yet though.
I'm confused about that too, since in my hands the submitted test case fails in both of those branches. I've also been able to adapt it to fail in v10/v11. Will commit in a moment.