Обсуждение: Regression test horology failure

Attached are regression diffs for 7.4.10, compiled from source on RHEL 3 U6 
(gcc 3.2.3 20030502, glibc-2.3.2-95.37) using:

make distclean
./configure '--with-perl' '--prefix=/usr/local/postgresql-7.4.10'
make && make install && make check

The tests fail for PST/PDT in 2034.

Looking at the buildfarm there is no other RHEL 3 system building the 7.4 
branch at the moment. The same is true for 7.3.12 and 7.4.7 by the way.

Is this a local problem in my glibc/tz libraries? I am not really worried 
because I will upgrade to 8.1 shortly, but understanding the problem would 
be a good thing.

Tom, you are building 7.3 for RedHat, do you see any similar regression 
failures?

Best Regrads,
Michael Paesold

"Michael Paesold" <mpaesold@gmx.at> writes:
> The tests fail for PST/PDT in 2034.

This probably indicates that you've got TZ data reflecting the new US
DST rules.  We have not updated the pre-8.0 regression test results
to deal with that.
        regards, tom lane

Tom Lane schrieb:
> "Michael Paesold" <mpaesold@gmx.at> writes:
> 
>>The tests fail for PST/PDT in 2034.
> 
> 
> This probably indicates that you've got TZ data reflecting the new US
> DST rules.  We have not updated the pre-8.0 regression test results
> to deal with that.

You're right as far as I can tell. 8.1 has the expected output.

The failing tests are different instances of this basic example:

Expected:
Wed Mar 15 08:14:01 2000 PST  +  34 years = Wed Mar 15 08:14:01 2034 PST

Here (and PostgreSQL >= 8.x):
Wed Mar 15 08:14:01 2000 PST  +  34 years = Wed Mar 15 08:14:01 2034 PDT

Still, I don't understand why March 15th should be day light saving 
time. But hey, I don't live in the PST8PDT time zone. ;-)

Thanks for your answer.

Best Regards,
Michael Paesold