Buildfarm member Narwhal: Python 2.5/8.1

On 4/17/07, Dave Page <dpage@postgresql.org> wrote:

New buildfarm member Narwhal is failing the PL regression tests for
Python on REL8_1_STABLE. This appears to be because it's running Python
2.5 (the causes being a deprecated module - whrandom - and some changed
messages).

I also suggest backporting following fix to 8.2 and 8.1:

http://archives.postgresql.org/pgsql-hackers/2007-04/msg00127.php

because otherwise python2.5 on 64bit platform will crash.

--
marko

Dave Page wrote:

New buildfarm member Narwhal is failing the PL regression tests for
Python on REL8_1_STABLE. This appears to be because it's running
Python 2.5 (the causes being a deprecated module - whrandom - and some
changed messages).

The former problem was fixed by Peter, and the latter by Tom but both
only for 8.2+`- any reason this wasn't backported to 8.1? I couldn't
find anything in the archives.

http://www.pgbuildfarm.org/cgi-bin/show_stage_log.pl?nm=narwhal&amp;dt=2007-04-17%20085153&amp;stg=pl-install-check

The question in my mind is this: how much do we back-patch to cover new
and incompatible releases of software we depend on? Python 2.5 was
released on 19 Sept 2006, long after Postgres 8.1. I guess you could
make a case to say that we should back-patch to the release immediately
before the library change.

(BTW, it is possible to include python only in certain branches in your
buildfarm client - examples are in the config file).

cheers

andrew

Andrew Dunstan wrote:

The question in my mind is this: how much do we back-patch to cover new
and incompatible releases of software we depend on? Python 2.5 was
released on 19 Sept 2006, long after Postgres 8.1. I guess you could
make a case to say that we should back-patch to the release immediately
before the library change.

I guess that depends on the invasiveness - in this case it's a couple of
simple updates to the regression tests so I think it's probably worth doing.

(BTW, it is possible to include python only in certain branches in your
buildfarm client - examples are in the config file).

Yeah, I already use that functionality to handle the features that we've
added over past releases to the windows port (ldap, thread safety etc).
Handy :-)

/D

Dave Page <dpage@postgresql.org> writes:

Andrew Dunstan wrote:

The question in my mind is this: how much do we back-patch to cover new
and incompatible releases of software we depend on?

I guess that depends on the invasiveness - in this case it's a couple of
simple updates to the regression tests so I think it's probably worth doing.

It's not just the regression tests; there are at least two rounds of
patches in the C code --- plpython.c r1.90, r1.97, maybe r1.100.
Only the first of these has seen any testing "in the wild".

Another objection to patching 8.1 is why stop there ... why not 8.0,
etc?

regards, tom lane

Tom Lane wrote:

Dave Page <dpage@postgresql.org> writes:

Andrew Dunstan wrote:

The question in my mind is this: how much do we back-patch to cover new
and incompatible releases of software we depend on?

I guess that depends on the invasiveness - in this case it's a couple of
simple updates to the regression tests so I think it's probably worth doing.

It's not just the regression tests; there are at least two rounds of
patches in the C code --- plpython.c r1.90, r1.97, maybe r1.100.
Only the first of these has seen any testing "in the wild".

Ahh - missed that bit.

Another objection to patching 8.1 is why stop there ... why not 8.0,
etc?

8.0 didn't have the PL regression tests and as it appeared to be a
regression test issue...

I'll disable python on < 8.2.

Regards, Dave.