I'm surprised and confused about this working in Python 2!

I'm sort of curious about the failing tests in Python 2.7, I glanced at them and I didn't understand what was broken. No pressure for this not-going-to-be-merged PR, but if you look at them maybe leave a comment about what broke.

I was just looking into the tests, the problem is that some expectations fail because some tests are expecting some names to not be defined.

For test_traceback I could see what's going on: it expects to see a NameError when calling g() in a new interpreter, but the actual exception is TypeError: 'int' object is not callable -- which I figure means some module available in the path during test run has a global g constant. :)

For the other errors, they're trying to get the line with error by causing a NameError using abc. But with crazy locals, abc gives you <module 'abc' from '/usr/lib/python2.7/abc.pyc'>. 🤘

I'm pushing a fix changing the names abcd and gfunc. :)

Ah cool, that all makes sense. Wow, thanks for changing all those tests. I wonder if we should merge the changes to tests anyway since this is a conceivable way that things could fail?

It probably goes without saying that this isn't getting merged. What about as a module? The changes you make in bpython/curtsiesfrontend/interpreter:Interpreter could be monkeypatched in, so this could be a bpython-autoimport plugin of sorts.

Hm, I don't know, doesn't it seem like a maintenance headache? :)

I'll send a second PR with the test changes.

@eliasdorneles sure, but only a maintenance headache for you, not bpython! :)

Haha, right!
I'll see how to make a module, just for fun! :)

Closing to reduce confusion