On Sat, 2008-10-04 at 21:48 +0200, Thomas Johansson wrote:
<snip>
On centos 4.6
test000:/% date -d "2008-10-25 +1 days" "+%Y-%m-%d" 2008-10-26 test000:/% date -d "2008-10-26 +1 days" "+%Y-%m-%d" 2008-10-26 test000:/% uname -a Linux test000... 2.6.9-67.ELsmp #1 SMP Fri Nov 16 12:48:03 EST 2007 i686 i686 i386 GNU/Linux
On centos 4.7
test001:/% date -d "2008-10-25 +1 days" "+%Y-%m-%d" 2008-10-26 test001:/% date -d "2008-10-26 +1 days" "+%Y-%m-%d" 2008-10-26
I just typed the above in on the console on my 4.7 node. WFM.
test001:/% uname -a Linux test001... 2.6.9-78.0.1.ELsmp #1 SMP Tue Aug 5 11:02:47 EDT 2008 i686 athlon i386 GNU/Linux
My kernel doesn't have the "smp", but otherwise matches.
<snip>
Another posted the possibility of TZ issues. Seems unlikely to me since you are providing a date string that does not require the system time zone information. It just converts to the binary offset value from the epoch, adds the binary equivalent of a day in seconds (IIRC) and formats the results as a string.
Maybe some library (like the ones used by asctime, ctime, ...) which are section 3 system calls (those used buy programs like date) are out of sync? Have you tried the rpm --verify to see if any discrepancies are reported?
HTH