strange time differences - Mailing list pgsql-bugs

I mentioned the problem described below in the psql-general list, but
didn't get any reply. Maybe someone has already posted it to you, but
anyway, this seems the right place for may-be-bugs, isn't it?

============================================================================
                        POSTGRESQL BUG REPORT TEMPLATE
============================================================================


Your name        : Ulf Mehlig
Your email address    : umehlig@uni-bremen.de


System Configuration
- ---------------------
  Architecture (example: Intel Pentium)      : intel Pentium II

  Operating System (example: Linux 2.0.26 ELF)     : Linux 2.0.35 ELF,
                                                  DLD  5.4 Distribution

  PostgreSQL version (example: PostgreSQL-6.4)  : PostgreSQL-6.4 (and
                                                  6.3.2, too)

  Compiler used (example:  gcc 2.8.0)        : gcc 2.7.2.3


Please enter a FULL description of your problem:
- ------------------------------------------------

Subtracting one datetime value from another with `time' parts beeing
both 00:00 h gives a reasonable time difference, but if the second
datetime's `time' part is 00:01 h (or even 00:00:01, or something
else), there is a difference to the previous calculated value of about
7 days. Seems to be restricted to large time spans.


Please describe a way to repeat the problem.   Please try to provide a
concise reproducible example, if at all possible:
- ----------------------------------------------------------------------

   => select date_part ('epoch', age(datetime('29.02.2000','00:00'),
                                     datetime('29.02.1968','00:00')));
    date_part
   ----------
   1009843200
       ^

   =>  select date_part ('epoch', age(datetime('29.02.2000','00:00'),
                                      datetime('29.02.1968','00:01')));
    date_part
   ----------
   1009303140
       ^
Difference is  60 seconds + 54000 secs for free!   ;-)

This calculation is correct:

   =>  select date_part ('epoch', age(datetime('01.01.1998', '00:00'),
                                      datetime('01.01.1998', '00:01')));
date_part
- ---------
      -60
(1 row)

... but it's not a year-2000-problem nor victim of 29th of February,
is it?!

   => select date_part ('epoch', age(datetime('01.01.1998', '00:00'),
                                     datetime('01.01.1968', '00:00')));
date_part
- ---------
946728000
(1 row)

   => select date_part ('epoch', age(datetime('01.01.1998', '00:00'),
                                     datetime('01.01.1968', '00:01')));
date_part
- ---------
946360740
(1 row)

(only four days gained.)


If you know how this problem might be fixed, list the solution below:
- ---------------------------------------------------------------------

no idea ;-) Do you (or some library) handle dates internally as
float values?

Best regards,
Ulf

- --
======================================================================
 %%%%%            Ulf Mehlig              <umehlig@zmt.uni-bremen.de>
   %%%%!%%%       Projekt "MADAM"         <umehlig@uni-bremen.de>
%%%% %!% %%%%     ----------------------------------------------------
 ---| %%%         MADAM:  MAngrove    |  Center for Tropical Marine
    ||--%!%              Dynamics     |  Biology
    ||                  And           |  Fahrenheitstrasse 1
 _ /||\_/\_            Management     |
/  /    \  \ ~~~~~~~~~~~~~~~~~        |  28359 Bremen/Germany
  ~~~~~~~~~~~~~~~~~~~~