Sorry, that merge mistake is probably mine. It looks like we'll have to fix it on 1.3.x as well. I'll do it over there, and we'll leave this PR as-is to cover it on master.

Oops. Forget that. 1.3.x seems fine.

This seems like the right thing to do -- moderizing rather than putting band-aids that work around unicode_literals.

One thing, though -- I had to do the following, or I got a segfault at startup:

diff --git a/lib/matplotlib/backends/backend_qt4.py b/lib/matplotlib/backends/backend_qt4.py
index 6ae8083..991edc7 100644
--- a/lib/matplotlib/backends/backend_qt4.py
+++ b/lib/matplotlib/backends/backend_qt4.py
@@ -82,7 +82,7 @@ def _create_qApp():
                 if display is None or not re.search(':\d', display):
                     raise RuntimeError('Invalid DISPLAY variable')

-            qApp = QtGui.QApplication([" "])  # probably fine, not used by QT to resolve anything
+            qApp = QtGui.QApplication([str(" ")])  # probably fine, not used by QT to resolve anything
             qApp.lastWindowClosed.connect(qApp.quit)
         else:
             qApp = app

There are many failures in travis from errors coming out of locale.py.

@mdboom What OS are you testing on? I find it a bit worrying that I was not getting segfaults without that last change.

I'm on Fedora 19, which has pyside 1.1.0 installed from the standard package (and this is Python 2.7).

Does this need anything else to get merged?

Nope. Merging.

It doesn't look like this actually got merged, just closed.

Sorry about that. I think it's good now.