"anarazel@anarazel.de" <andres@anarazel.de> writes:
> 9.2 represents inisdead as live && !ready, doesn't it? So just looking at indislive will include about to be dropped
orpartially dropped indexes?
Ooooh ... you're right, in 9.2 (only) we need to check both indisvalid
and indisready (cf IndexIsValid() macro in the different branches).
So that's actually a bug in the committed pg_dump patch, too. I'll fix
that shortly.
regards, tom lane