Re: Improving backend launch time by preloading relcache

Tom Lane wrote:
>
> Hiroshi Inoue <Inoue(at)tpf(dot)co(dot)jp> writes:
> > Anyway there seems no point on changing REINDEX.
> > The only thing I have to do is to remove the needless
> > check in tcop/utility.c as soon as 7.2 is released.
>
> I don't believe it's needless, and I suggest you not remove it,
> until we do something about making the pg_internal unlink happen
> at the right time. With the unlink where it is, I think it's quite
> unsafe to reindex system indexes in a live database.

Currently there are just a few relations info kept in
pg_internal.init and they are all nailed. I'm not
allowing REINDEX for nailed relations though there's
a #ifdef'd trial implementation. I'm intending the
change for 7.2.1 not 7.3. If it isn't allowed in 7.2.x
I would strongly object to the 7.2 release itself.

regards,
Hiroshi Inoue