Uploads with extra data in the .tar.gz rejected unnecessarily

confirmed, got it there too.

Can you post an example of something that breaks please - is it reproducible?
I haven't got much to go on here in terms of testing.

Thanks.

This tarball had failed for me, I had to extract it and recreate in order to upload:

http://edge.launchpad.net/dbusmenu/0.3/0.3.13/+download/libdbusmenu-0.3.13.tar.gz

And I think the tarball Didier had problems with was:

http://ftp.gnome.org/pub/GNOME/sources/gnome-bluetooth/2.31/gnome-bluetooth-2.31.90.tar.gz

On Mon, 13 Sep 2010 19:36:11 -0000, James Westby <email address hidden> wrote:
> Public bug reported:
>
> Hi,
>
> Launchpad suffers from a problem that many of us python programmers who
> work with Debian packages suffered from.
>
> Colin does a good summary of the problem here:
>
> http://www.chiark.greenend.org.uk/ucgi/~cjwatson/blosxom/2009-07-02
> -python-sigpipe.html
>
> and William here:
>
> https://bugs.edge.launchpad.net/ubuntu/+source/bzr-
> builddeb/+bug/399938/comments/7
>
> There are a fair number of packages that have this issue, which presumably hit on the
> Lucid upgrade.
>
> The fix is very simple.

Oh, and it's worth looking at anywhere that you call subprocess, as they
can suffer from the same thing. I can't say that making this change is
always the correct thing to do, but it may prevent some bugs in the
future.

Thanks,

James

Fixed in stable r11862 <http://bazaar.launchpad.net/~launchpad-pqm/launchpad/stable/revision/11862>.

Marking QA-ok because the test suite is a sufficient check of this code path.

We've deployed this to loganberry and the appservers. Does it need a wider deploy to be fixed?

Yes - archiveuploader and archivepublisher use subprocesses, so this needs to go to all soyuz machines (apart from cesium as we're blocked on me fixing that buildd-manager snafu)