Thanks for this PR. I'm gonna review it.

@sebdah Any chance you could take a look at this? I promise the diff isn't as scary as it looks; most of it is just indenting code 4 spaces to put it under the context manager. I haven't fixed the merge conflicts yet since if you decide you don't like the overall approach I'm taking there may be no point. Let me know if you'd prefer me to fix those first.

@evanunderscore I'm fine with this PR, so if you get the conflicts resolved, I'll merge it.

@sebdah Thanks, done. Build failure seems to be due to something dropping support for 3.2 (see travis-ci/travis-ci#5517).

Just want to chime in and say that I run my own stash and pop before running git-pylint-commit-hook (not all my tools are nice enough to do this for me), using the same strategy outlined here. The two don't seem to conflict in the slightest and my hooks work, no extra stashes are made, no stashes are popped at the wrong time, no stashes are dropped, etc.

Didn't notice a test for this very situation, so wanted to provide some anecdotal evidence. Thanks guys! Looks like a solid implementation!