Hello.

This is just a little cleanup of the code, removing one addition and
one substraction from the g_date_update_dmy routine.

One subtle point: when compiled with a non-optimizing compiler, this code
runs a tiny bit slower on Pentium+ machines, but faster on a 486, it seems
to be an alignment problem. When compiled with an optimizing compiler, the
new code is about 10% faster on a P-II, not tested on other machines.

The question remeans whether this is relevant, it doesn't seem to be
likely that this code will be called in a tight loop.

It was tested by me, using all dates from 1/1/1 to 24/1/27380, and it seems
to be working flawlessly.

It should make this code more legible and understandable, though it remains
a bit of black magic when you look at it the first time.

Regards,

Jasper Spaans <spaans@vvtp.tn.tudelft.nl>
