Neil Conway <neilc@samurai.com> writes:
> Bruce Momjian wrote:
>> Add pg_postmaster_start_time() function.
> You should have bumped the catversion, no?
Probably, but it's not essential since nothing depends on that pg_proc
entry being there. If, for instance, he'd added a regression test
using it then a catversion bump would be required IMHO.
I had a similar situation a few days ago when I was taking out the unary
% and ^ operators: a catversion bump could have been called for, but I
felt it was OK to not do it. Running with the old pg_operator contents
wouldn't have any ill effects.
Part of the reason for being lazy here is that I'm quite sure there will
be more catversion bumps before 8.1 goes final; so the only question is
whether anyone's development installation is likely to fail if they run
an updated postmaster without initdb'ing. If I thought these might be
the last catalog changes before a release, I'd want a catversion bump
so we could be certain everyone running a release postmaster has release
catalogs.
regards, tom lane