Analysis of current ppc64el test failures

[Martin Pitt]

Hey Martin,

Martin Packman [2014-02-26 15:39 +0000]:
> On 26/02/2014, Martin Pitt <martin.pitt at ubuntu.com> wrote:
> >  bzr
> 
> Two issues:
> 
>   File ".../bzrlib/tests/__init__.py", line 2862, in makeAndChdirToTestDir
>     os.mkdir(self.test_base_dir)
> OSError: [Errno 36] File name too long:
> '/tmp/adt-downtmp/...<reallylongfilename>...'
> 
> Does the container have a 256 char limit on paths or something? What
> does a script along the lines of the following return:

It says "Max path length 4096". But we run into a similar error in
the coreutils test, it does look LXC related somehow.

> The second issue seems to be a lack of default certs inside the
> container, does /etc/ssl/certs/ca-certificates.crt exist? Would be
> need to add some package as a test dep to insure it does?

It doesn't exist; bzr's debian/tests/control needs to pull that in as
a test dependency; or, if that's not test specific but bzr itself
needs it, then bzr needs to grow a dependency to it.

> >  juju-core
> 
> I don't understand the output here, the latest log seems to fail on
> the initial setup before getting to juju:
> 
> Errors were encountered while processing:
>  /var/cache/apt/archives/python3.3-minimal_3.3.5~rc1-1_ppc64el.deb
> E: Sub-process /usr/bin/dpkg returned an error code (1)
> adt-run: unexpected error: testbed setup commands failed with status 100

Right, there were a few failed runs with using aufs as overlay file
system, which caused this. So I disabled aufs again and now use
cloning on ppc64el; much slower, but avoids the various bugs that we
see with aufs and overlayfs there.

> However I guess if mongodb isn't working, juju-core is stuck for now anyway.

Correct, that's the problem. Look at the previous log at
https://jenkins.qa.ubuntu.com/job/trusty-adt-juju-core-ppc64el/6/console

Thanks,

Martin

-- 
Martin Pitt                        | http://www.piware.de
Ubuntu Developer (www.ubuntu.com)  | Debian Developer  (www.debian.org)