7.x horology regression test on Solaris buildfarm machines

Tom Lane wrote:

I see that kudu and dragonfly are now failing regression in the 7.3 and
7.4 branches, as a consequence of this patch:
http://archives.postgresql.org/pgsql-committers/2007-02/msg00491.php

Is it reasonable to assume that that machine will soon be patched to
know about the new US DST rules? If not, I suppose we could install
variant horology-solaris-1947_1.out files to allow either result to
succeed, but it seems a bit silly ...

Yeah. I have stopped building 7.4 on hyena.

cheers

andrew

On Tue, 27 Feb 2007, Tom Lane wrote:

I see that kudu and dragonfly are now failing regression in the 7.3 and
7.4 branches, as a consequence of this patch:
http://archives.postgresql.org/pgsql-committers/2007-02/msg00491.php

Is it reasonable to assume that that machine will soon be patched to
know about the new US DST rules? If not, I suppose we could install
variant horology-solaris-1947_1.out files to allow either result to
succeed, but it seems a bit silly ...

I'll look at an upgrade. Eel is failing as well, but surprisingly canary
is not. Canary hasn't had any updates applied, so why isn't it failing as
well:

jurka@netbsd:~$ zdump -v /etc/localtime | grep 2007
/etc/localtime Sun Apr 1 08:59:59 2007 UTC = Sun Apr 1 01:59:59 2007
MST isdst=0
/etc/localtime Sun Apr 1 09:00:00 2007 UTC = Sun Apr 1 03:00:00 2007
MDT isdst=1
/etc/localtime Sun Oct 28 07:59:59 2007 UTC = Sun Oct 28 01:59:59 2007
MDT isdst=1
/etc/localtime Sun Oct 28 08:00:00 2007 UTC = Sun Oct 28 01:00:00 2007
MST isdst=0

Kris Jurka

On Tue, 27 Feb 2007, Kris Jurka wrote:

I'll look at an upgrade. Eel is failing as well, but surprisingly canary is
not. Canary hasn't had any updates applied, so why isn't it failing as well:

Shouldn't all of the buildfarm members be failing either before or after
the patch? That doesn't seem to be the case for any of them.

Kris Jurka

Kris Jurka <books@ejurka.com> writes:

Shouldn't all of the buildfarm members be failing either before or after
the patch? That doesn't seem to be the case for any of them.

No, because for the "standard" regression behavior we have variant
result files both with and without the DST law change: see horology and
horology_1. The issue only comes up for machines that were matching to
horology-no-DST-before-1970.out (which may be only old HPUX) or
horology-solaris-1947.out (which I suppose is only Solaris). The reason
I got around to making the patch now is I recently updated my old HPUX
machine for the law change, and so 7.x started to fail for me ...

regards, tom lane

On Tue, 27 Feb 2007, Tom Lane wrote:

No, because for the "standard" regression behavior we have variant
result files both with and without the DST law change: see horology and
horology_1. The issue only comes up for machines that were matching to
horology-no-DST-before-1970.out (which may be only old HPUX) or
horology-solaris-1947.out (which I suppose is only Solaris). The reason
I got around to making the patch now is I recently updated my old HPUX
machine for the law change, and so 7.x started to fail for me ...

OK, kudu + dragonfly updated. I probably won't get around to fixing eel
(cygwin) for some time though.

Kris Jurka