Re: Undocumented changes to Debian source packages
Bill Allombert <allomber@math.u-bordeaux.fr> writes:
> Hello debian-amd64,
>
> I wanted to check whether my packages worked on amd64 and after much
> puzzling because I could not rebuild packages available on alioth.
> I finally get across the following discrepancy between Debian source
> package and source packages available on debian-amd64.alioth.debian.org:
Here is our scapegoat (from the wanna-build db):
unstable main amd64 pari 2.1.5-5 broken 2004-06-15.10:45:18 Andreas Jochens <aj@andaco.de>
The package has been flaged as being broken in wanna-build for month
now.
A good overview of the buildd status for your package is available
from:
http://people.debian.org/~igloo/status.php?packages=pari&arches=
That there actualy is a deb in the pure64 repository without an
incremented version and without patch in the BTS is regrettable but an
oversight from one of the earlier build cycles. We did rebuild all of
sid when we changed the archive software and when alioth destroyed our
database but this deb must have remained across the rebuilds.
Regrettably alioth is always overloaded and running a consistency
check over the full archive is very time consuming and would probably
create a lot of screaming svn users. Luckily it seems that alioth will
move to a new system soon and amd64 stay on the old one on its
own. Once that happens we would be free to run such tests regularly.
> Here a diff /in extenso/ of pari_2.1.5-5.diff.gz from the Debian mirrors
> and the one available at
> <http://debian-amd64.alioth.debian.org/pure64/pool/unstable/main/source/p/pari/pari_2.1.5-5.diff.gz>
>
> %zdiff -u pari_2.1.5-5.diff.gz pari_2.1.5-5.diff.gz.pure64
> --- - Thu Nov 25 17:58:43 2004
> +++ /tmp/pari_21.5-5.diff.gz.pure64.12076 Thu Nov 25 17:58:43 2004
> @@ -587,7 +587,7 @@
> + test -f pic-lib || $(MAKE) -f debian/rules build-nonpic-lib
> + $(MAKE) gp
> +ifeq (,$(findstring notest,$(DEB_BUILD_OPTIONS)))
> -+ $(MAKE) dobench
> ++# $(MAKE) dobench
> +endif
> + touch build-arch-stamp
> +
>
> I am not terribly happy about that. Whoever made that change should at
> least add a comment or a changelog entry, etc. explaining what is the
> problem and how the patch fix it. Bumping the version and contacting me
> would be a plus.
We usualy do this, using 2.1.5-5.0.0.1.pure64 or 2.1.5-5.0.0.amd64.1
or similar as well as sending the patch to the BTS. Might have been
one of the earlier packages Andreas did build.
> Since I cannot know who made that change, I post here my comment on that
> patch:
You could have looked at the changes file (in the same subdir) to see:
Maintainer: Andreas Jochens <aj@andaco.de>
as well as the gpg signature on it. Blame him. :)
> This patch is ineffective: it only hides the fact that pari-gp is broken
> on that platform by disabling the test-suite
>
> I am not terribly happy about that either. I don't think disabling
> test-suites to build more packages is the way to go.
>
> I plan to write a proper patch to really solve this issue.
> (some int* is incorrectly used as a long*). However, have I been
> notified sooner the fix would probably already in the archive.
Strange that this doesn't show up on the other 64bit
architectures.
We are looking forward to a fix.
> Anyway, good luck to the Debian amd64 port effort.
>
> Cheers,
> --
> Bill. <ballombe@debian.org>
>
> Imagine a large red swirl here.
MfG
Goswin
Reply to: