Alvaro Herrera <alvherre@dcc.uchile.cl> writes:
> Given that recompiling and adding the new tuple into pg_proc by means of
> a simple INSERT(*) should be enough, it's hardly necessary to force an
> initdb.
> (*) It _can_ be done, right?
I think so. The critical point is that interval_scale is not in the
compiled-in table of known internal functions in 7.3.2. A recompile
with the added pg_proc.h line should get it in there, and then you could
manually add the pg_proc entry without actually doing initdb. But I
haven't tested it to be sure there are no gotchas.
regards, tom lane