Merge lp://staging/~freyes/charm-helpers/fix-test-ipv6 into lp://staging/charm-helpers

universal_newlines=True is to normalise newlines between mac, windows and *nix; we're only dealing with *nix machines here, and it indirectly papers over the cracks between Py2 and Py using a IOTextWrapper with default settings.

The 'problem' is that check_output() returns a bytestring in Py3 vs a str on Py2 which is what the decode is for, and the bug is in the test.

I'd prefer to change the test to pass a bytestring if it's Py3 as the code currently works OR (perhaps even better) just make Py2 a bit looser by instead doing:

   return b"net.ipv6.conf.all.disable_ipv6 = 1" in result

as Python2 will paper over the cracks by doing an implicit conversion: e.g. in ipython (py2):

In [4]: s = "hello there\n\n123 and som"

In [5]: b"123" in s
Out[5]: True

Obviously, the test would also need to pass a b"""...""" and b"" (the two mock outputs), and the .decode('UTF-8') would be removed in the original function.

This would work for both py2 and py3. Thoughts?

Alex, I'm OK with changing the patch to the approach you suggest, but universal_newlines=True is a widespread approach to deal with py2 vs py3, if you think that it's worth to take a different method to deal with this, I will do the change.

$ find charmhelpers/ -name "*.py" -exec grep universal_newlines {} \; | wc -l
13

Felipe, that's interesting. I don't particularly want to create work, so on this basis, I'm happy to go with the universal_newlines as its prevalent in charmhelpers. I guess a one line comment to indicate why it is being used 'might' be useful for the future? Thanks for sorting this out. The 'bug' really shouldn't have got merged, but I won't go looking for the culprit :)

LGTM