On Wed, Dec 19, 2007 at 08:14:17PM -0500, Tom Lane wrote:
> Richard Huxton <dev@archonet.com> writes:
> > I'm not sure that (CURRENT_DATE AT TIME ZONE 'UTC') does what you think
> > it does. Try setting your timezone to various offsets and exploring.
>
> In fact, I think it's adjusting in exactly the wrong direction.
>
> I get the right number from
>
> regression=# select date_part('epoch', 'today'::timestamp at time zone 'UTC');
> date_part
> ------------
> 1198022400
> (1 row)
>
> and the wrong one from
>
> regression=# select date_part('epoch', 'today'::timestamptz at time zone 'UTC');
> date_part
> ------------
> 1198058400
> (1 row)
>
> and I think the locution with CURRENT_DATE is equivalent to the second
> case because timestamptz is the preferred type to promote date to.
Does that mean it's a postgresql bug?