Richard Huxton <dev(at)archonet(dot)com> writes:
> Yep - I'm guessing it's an issue with an array (group membership?) in
> pg_user.
No, I'll bet lunch that Alvaro fingered the correct culprit: either
addition or subtraction of that darn hierarchical-query patch. It
changes the on-disk representation of view rules in a way that produces
exactly this failure mode.
One of the disadvantages of open source is that you can't prevent people
from breaking your software :-(
regards, tom lane