Bug #780: Timestamp shifted by one minute

Started by PostgreSQL Bugs Listover 23 years ago2 messagesbugs
Jump to latest
#1PostgreSQL Bugs List
pgsql-bugs@postgresql.org

Rob Abbot (rra42@yahoo.co.uk) reports a bug with a severity of 1
The lower the number the more severe it is.

Short Description
Timestamp shifted by one minute

Long Description
The output of the code below is:
25/09/02 23:59:00 *NOT* 26/09/02 00:00:00
which is timeshifted backward by 1 minute from the inserted timestamp. In this case the timestamp is shifted backward to the previous day!!!

Details of set-up:
Dell 1650 PowerEdge, 1.13Mhz Pentium III, 512k cache, 512 MB RAM
RedHat Linux 7.1

Result of SELECT version();
"PostgreSQL 7.2.1 on i686-pc-linux-gnu, compiled by GCC 2.96"

Details of build:
(Comments below from my notes)
Installed postgresql 7.2.1 from source:
the default install is to /usr/local/pgsql/
unpacked in rob, got rob/postgresql-7.2.1 as 'rob'
./configure
gmake (takes forever)
gmake check
su to root and get back to /home/rob/postgresql-7.2.1
gmake install
gmake install-all-headers

This bug appears similar to:
http://archives.postgresql.org/pgsql-bugs/2001-04/msg00072.php

I have another similar setup which works perfectly: the binary was built with the same source tree and options, however the differences between the two setups are:
OS: Suse 7.1
Compiler: GCC 2.95.2
CPU: AMD Athlon 533MHz, 256MB RAM

I am also running PostgreSQL on a Compaq Proliant, RedHat 6.0, PostgreSQL 7.0.3 compiled gcc egcs-2.91.66 again this does not show the timestamp problem.

Best wishes,
Rob

Sample Code
create table timetester (timefield timestamp);
insert into timetester values ('2002-09-26 00:00:00.00');
select * from timetester;

No file was uploaded with this report

#2Tom Lane
tgl@sss.pgh.pa.us
In reply to: PostgreSQL Bugs List (#1)
Re: Bug #780: Timestamp shifted by one minute

pgsql-bugs@postgresql.org writes:

Timestamp shifted by one minute [ in locally-built PG ]

Do you have CFLAGS set in your environment, and if so how?

This sounds quite a bit like the known problems with -ffast-math,
though I have not heard this exact complaint before. If you built
with -ffast-math then try without.

regards, tom lane