Re: [HACKERS] NetBSD/NS32K regression results

Jon Buller <jonb(at)metronet(dot)com> writes:
> After rebuilding and retesting from a CVS update last night, I can
> report that all tests passed with the following exceptions:

Great!

> float8 has some problems with very large and very small numbers.
> datetime and horology have some problems with 'current' and 'epoch'
> which might (or might not) be related to the float8 problems.

I think you guessed right. IIRC, datetime has an underlying
representation of float8, and some of the special values like
"current" are represented as particular very-large or very-small
numbers. So it might break if your machine's float math is not
up to IEEE-float specs. However, you can do an awful lot of stuff
with datetime without ever needing the special values, so this
might not be much of a problem for you.

regards, tom lane