Originally posted by: nik@zurich.ibm.com

We have been using

TZ=MET-01MDT-02,M3.5.0/02:00:00,M10.5.0/03:00:00

successfully for many years now to specify the timezone valid in Zurich (and most of Central Europe).

We upgraded a lot of machines to AIX 6100-07-01-1141 a few days ago. Today, we realized that local time information as displayed by xclock (and many other applications including sendmail and syslog) had changed from Daylight Savings Time to normal during the night. According to our TZ variable, this should have happened on the last Sunday in October, i.e. 5 days later.

Here are a few related observations:

1. The problem DOES NOT show in the 'date' command.

2. The problem DOES NOT occur with TZ=Europe/Zurich.
However, that setting does not work on AIX 5.
On AIX 6, it works with 'date', but not with strftime().

3. The problem DID NOT occur on machines that are
still on AIX 6100-06-05-1115.

4. The problem DID NOT occur on machines that were
upgraded to AIX 5300-12-05-1140.

5. The problem DID occur on a machine that was
upgraded to AIX 7100-01-01-1141.

Experiments with a program that uses strftime() showed that, on AIX 6100-07-01-1141, the M10.5.0 part in our TZ variable is somehow misinterpreted to mean the last Tuesday of the month instead of the last Sunday. Using M10.5.1 instead, the time change happens on the last Wednesday, and using M10.5.5, it happens on the last Sunday (though, in that case, it should happen on the last Friday). On AIX 6100-07-01-1141, the value returned by the %Z format specifier to strftime() changes from MDT to MET when the hour returned by %H switches back from 03 to 02. This makes at least some sense, and it keeps the human-readable date unique if %Z is included. On 7100-01-01-1141, things are even worse, because %Z always returns MDT, regardless of whether the %H value does or does not correspond to Daylight Savings Time.

I wonder whether this bug is already known, and whether a fix is available.
#AIX-Forum

Originally posted by: SystemAdmin

hi nik,

given the fact that these TLs were released last week, you get what you deserve ;-)

I noticed with tests on aix 7.1TL1SP1 that errorlog entries are timestamped 1 hour too early, date output is ok, TZ similar to yours (tried and ok on all previous aixes up to this).

time for a pmr I guess
#AIX-Forum

Originally posted by: SystemAdmin

TZ=NFT-1DFT,M3.5.0/02:00:00,M10.5.0/03:00:00
to make it complete
#AIX-Forum

Originally posted by: j.gann

it's a confirmed defect. i've been informed it will be fixed in tl1sp2 due around end of april 2012 (after dst start...)
#AIX-Forum

Originally posted by: shehua

I received and applied eFix IV09580 about an hour ago. Everything looks OK now, ld is working as expected (with usual warnings about TOC size). Patches now applied succesfully to Oracle software.

---

http://www.girls-fitness.com/page/3/
#AIX-Forum

Originally posted by: nik@zurich.ibm.com

It is indeed fixed in AIX 7100-01-03-1150. It is NOT fixed in AIX 6100-07-03-1207. Embarrassing.
#AIX-Forum

Originally posted by: j.gann

the bug has likely been introduced when making "library changes to comply with the single unix specification V4 (SUS4)" which was mentioned in the release notes of the december TLs. one more noticable change in this area is that you can allow non-root users to use chown() now...
#AIX-Forum

Originally posted by: nik@zurich.ibm.com

I received the following alert:

Possible Action Required: System time may not change properly at DST start/end dates on AIX 7.1 and AIX 6.1

http://www-01.ibm.com/support/docview.wss?uid=isg3T1013017&myns=paix61&mync=E
#AIX-Forum