Add installation and test support for python3.

[ryneeverett]

This is a rebase and update of @wfranzini's work in #387. We had a remarkable number of merge conflicts in such a short time.

This is great progress on #25, but I think we ought to do significant manual testing before closing that issue and declaring python3 compatibility 'done'.

[gdetrez]

I give this a try and I can confirm that pip-insallation works as do some simple tests I ran with gitlab and trello. 👍

Github seems to work with a plain text token in bugwarriorrc but when using an oracle, I got the following crash:

Worker for [bnfc] failed: Can't convert 'bytes' object to str implicitly
Traceback (most recent call last):
  File "/home/gregoire/QA/bugwarrior/ENV-3.5/lib/python3.5/site-packages/bugwarrior/services/__init__.py", line 516, in _aggregate_issues
    service = get_service(service_name)(conf, main_section, target)
  File "/home/gregoire/QA/bugwarrior/ENV-3.5/lib/python3.5/site-packages/bugwarrior/services/github.py", line 234, in __init__
    self.client = GithubClient(auth)
  File "/home/gregoire/QA/bugwarrior/ENV-3.5/lib/python3.5/site-packages/bugwarrior/services/github.py", line 20, in __init__
    authorization = 'token ' + self.auth['token']
TypeError: Can't convert 'bytes' object to str implicitly

I'm using the eval oracle and I'm guessing that the result of the executed command is returned as bytes in self.auth['token'], whereas the 'token' constant in github.py is a unicode str in python 3.

[ryneeverett]

@gdetrez I'm not surprised there are python3 bugs. As long as there aren't python2 regressions though, I'd like to go ahead and merge this and deal with python3 bugs that the test suite doesn't catch in future PR's. (And add the missing tests when feasible.) Any objections to merging as is?

[wfranzini]

@ryneeverett the following code added to test_github.py catch the bug:

class TestGithubIssueToken(TestGithubIssue):
    SERVICE_CONFIG = {
        'github.login': 'arbitrary_login',
        'github.token': '@oracle:eval:python -c "print(\'secret\')"',
        'github.username': 'arbitrary_username',
    }

Maybe this approach can be improved and expanded to the testsuite using parameterized test.

[gdetrez]

@ryneeverett I don't have an objection to merging this now and fixing issues later. But I think @ralphbean wanted to make a new release (#339) in which case it might be safer to merge this PR after. But #339 haven't been updated since July so I don't know if it is still happening.

[ryneeverett]

@gdetrez Agreed. I'll give @ralphbean a few days to respond, but if we don't hear anything I for a bit I may go ahead and merge. It would be a lot simpler to make new PR's 2-to-3 compatible than to keep fixing and rebasing this branch.

[ralphbean]

But #339 haven't been updated since July so I don't know if it is still happening.

Yeah, sorry I've been MIA.

I'm getting some py2 regressions with this:

Python 2.7.12 (default, Sep 29 2016, 12:52:02) 
[GCC 6.2.1 20160916 (Red Hat 6.2.1-2)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> from future import standard_library
>>> standard_library.install_aliases()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/threebean/.virtualenvs/bugwarrior/lib/python2.7/site-packages/future-0.16.0-py2.7.egg/future/standard_library/__init__.py", line 483, in install_aliases
    import test
  File "test.py", line 5, in <module>
    with open('/tmp/npm.txt', 'rb') as npm_file:
IOError: [Errno 2] No such file or directory: '/tmp/npm.txt'

Anybody else seeing this?

[ralphbean]

Looks like I only get that failure with future==0.16.0, the 0.15.2 release is a little better?

[ralphbean]

@irl, have you looked at packaging future for debian? It is non-trivial: https://bugzilla.redhat.com/show_bug.cgi?id=1393502

[ryneeverett]

Thanks for fixing the build @ralphbean! I had no idea what the problem was but I learned the hard way that doing a github rebase-merge breaks the ability to auto-revert a PR.