> I was expecting it to round off ... I think there's something wrong with
> the arithmetic in AdjustIntervalForTypmod.
Yup. I've now updated the lookup tables used for the calculation for the
--disable-integer-datetimes case. The --enable-integer-datetimes case
was already calculated correctly. Regression tests don't pass, but that
is a locale problem.
Thanks for noticing the problem.
- Thomas