Bruce Momjian <pgman@candle.pha.pa.us> writes:
> Michael Fuhr wrote:
>> I'd guess this is due to the 32-bitness of abstime. Those timestamps
>> are around the min and max values of a 32-bit timestamp based on the
>> traditional Unix epoch.
> Yea, I see the same thing in 8.0.X. I don't think abstime should be
> used in that date range, timestamp is a better solution.
It's still a bug though; if the value is out of range, abstimein should
reject it, not misconvert it.
regards, tom lane