Code review comment for lp:~gagern/bzr/encodingSafeTests

Making the test result output robust against dodgy bytestrings is a worthy goal, but there are issues with the specifics here. Most generally, I think testtools is the place to start.

There are three things this patch is making 'safe':

* Test descriptions, which come from test ids
Always ascii, as noted by John last time around:
<https://lists.ubuntu.com/archives/bazaar/2008q3/045619.html>

* Exception instances from testtools
Already been run through an encoding and decoding process. There *are* issues like bug 501166 that need fixing upstream.

* Diffs where one text is a (non-ascii) str and the other is unicode
This is undecidable as the function does not know the encoding of the bytestring. Just escaping both inputs can result in a bogus pass, asserting the types are the same would catch badly written tests.