Merge lp://staging/~barry/ubuntu-system-image/test-fixes into lp://staging/ubuntu-system-image/server

What does this mean, that python 2.7 "does not have" tar.xz files? We are using python2.7 in production, and we must support tar.xz in production (both as an import source and as a generation target). If the test as written can't test tar.xz support correctly on python2.7, then the test should be fixed, not skipped.

looking at the test code, I see that the test relies on xz support in tarfile:

            # Check that for touch and pd the android hacks are executed
            target_obj = tarfile.open(
                os.path.join(self.config.publish_path, "pool",
                             "ubuntu-HASH.tar.xz"), "r:xz")

This appears to be newly-introduced behavior as part of Lukasz's latest change. However, your change disables the complete test_generate_file_cdimage_ubuntu() test, which is a very extensive test that tests many other aspects of the generator. We cannot reasonably disable this test as a whole on python2.7.

On Sep 28, 2015, at 11:23 PM, Steve Langasek wrote:

>We are using python2.7 in production

/me has a sad, but also a question: does anything or anyone actually run the
server's test suite? While the .xz test failure may be new, the PEP 8
failures have probably been there a while.

In any case, I pushed an update that avoids the direct use of xz in the
tarfile.open() and removes the skip.

The changes in the test code looks good, branch is passing all tests here. As for ignoring the certain PEP8 warnings - well, as long as you're ok with that, then I'm ok with that as well ;)

On Tue, Sep 29, 2015 at 01:42:33AM -0000, Barry Warsaw wrote:
> On Sep 28, 2015, at 11:23 PM, Steve Langasek wrote:

> >We are using python2.7 in production

> /me has a sad, but also a question: does anything or anyone actually run the
> server's test suite? While the .xz test failure may be new, the PEP 8
> failures have probably been there a while.

Well, the test suite ought to be run when people are committing changes to
the server ;) I don't know the last time I ran it myself.

> In any case, I pushed an update that avoids the direct use of xz in the
> tarfile.open() and removes the skip.

Thanks!