> How then, am I supposed to explain to a web interface user that when they just
> entered 5:00:00, it's going to sometimes show up to 4:59:60?
Oh, I see. I thought you were doing some interval math on it. Yes,
your right, it should be taken as being 5:00:00.
> I'm entering an exact timestamp, that being 5:00:00. Regardless of how many
> seconds you claim were in the former minute, it should not subtract a second
> from my entry, because 5:00:00 by your definition would mean 4:59 and 61
> seconds.
Either way, I've been unable to reproduce it with either 7.2 or 7.3 --
nor do I see any notes about that feature having been removed or carried
over to current releases -- no regression tests for it in 7.3.
Could you provide a complete test case, or confirm that it does what you
expect in 7.3?
7.2.2:
iqdb=# select '0001-01-01 4:59:60'::timestamptz;
ERROR: Bad timestamp external representation '0001-01-01 4:59:60'
--
Rod Taylor