Re: [HACKERS] Re: [GENERAL] date_part bug?

> Not all glibc2: I run glibc2 Debian Linux and do not see this problem.
> I was wondering if it was libc5 that was giving trouble...

Not on my libc5-only machine (where I developed the code). We'll need to
get a reproducible case to be able to track it down. I'm guessing that
we are seeing float->int rounding problems, though I don't know why this
test query should show different results for the two columns.

What glibc2, compiler, and optimization level are you using? If you are
using anything above -O2 try backing down to that; if you are already
there then try -O0 and tell us what changes.

- Tom

postgres=> create table tmp (v1 date, v2 datetime);
CREATE
postgres=> insert into tmp values ('06-01-1999', '06-01-1999');
INSERT 901482 1
postgres=> select date_part('month', v1) as m1, date_part('month', v2)
as m2 from tmp;
m1|m2
--+--
6| 6
(1 row)