micro release exception for bzr

Martin Pool mbp at canonical.com
Wed Sep 22 04:09:35 BST 2010

On 22 September 2010 03:05, Martin Pitt <martin.pitt at ubuntu.com> wrote:
> Hello Martin,
> Martin Pool [2010-09-20 12:08 +1000]:
>> Thanks, everyone.  Since there do not seem to be any objections, would
>> one of you please add bzr to
>> <https://wiki.ubuntu.com/StableReleaseUpdates/MicroReleaseExceptions>?
> The request was approved by the TB today, so I added it.
> The permission is coupled to those conditions:
>  * The test suite gets re-enabled during package builds.
>  * SRU verification should include tunning the tests in an installed system.
> For the latter, can you please give us some instructions how to do
> that? Is it enough to run "bzr selftest" or should this also add some
> options for plugins, etc.? It should be noted that this currently
> fails, so for lucid/maverick SRUs we should probably just compare the
> test result output.

'bzr selftest' should be enough.  If you have plugins installed other
than from the bzr package, you might 'bzr --no-plugins selftest'.  If
it ever fails, please file a bug against upstream bzr.  I just added
<https://bugs.edge.launchpad.net/bzr/+bug/644855> which is probably
pretty shallow: a check that we're not touching things out of the
source tree is wrong if we actually are specifically trying to test
the installed copy.

> However, it would be nice if the test would fully succeed. I assume
> you are developing bzr on Ubuntu, do you have any test failures
> locally?

In short, no.  There are some lurking very-intermittent
probably-gc-related failures, but keeping the suite clean is very
important to us.

> That'd be nice. It would also be great if we could enable it for
> lucid/maverick SRUs as well? Is it easy/possible to suppress tests
> which can't succeed in the buildd environment, such as the ones
> involving the network?

We should only need networking to localhost for testing.  Any test
that depends on something external that might not be present (eg a
library or a non-ascii locale) ought to cleanly skip if it's not
present.  If not, it's a bug, and again please report it.


More information about the technical-board mailing list