The issue is a difference between how PhiAnalysis and TypeAnalysis
handle OSR compilations: type analysis is osr-aware but phi analysis
isn't.  I think the right solution to this is to make phi and
definedness analysis also osr aware, but I want to get these other changes
in.

Set correct site-package directory, fix 'str * -1', add sys.dont_write_bytecode

Attach BoxedFile::gcHandler.

cmake explicitly list all c++ source files

Beginnings of a testsuite, starting with a simple django test.
Increases testing time by about 45 seconds which isn't great, but
tests a lot of stuff which is good.

We don't support arbitrary dict-like objects as the globals, but
in the special case that that dict-like object is just an attrwrapper
around the current module, we can fall back on our module-as-globals
support.

(Django 1.7.5)

We had a few callers but they weren't actually using the fact
that the strings were interned.

Implement misc stuff

This both lets us reuse the frame object and verify
that the BoxedFrame is still valid.

Conflicts:
	src/codegen/unwinding.cpp
	src/codegen/unwinding.h

The first change is to move to a PIMPL strategy so that
we can provide a more extensive unwinding API to other modules.
In particular, you can get access to a python frame iterator
and make repeated calls on it.  This is in preparation of the
sys._getframe implementation.

This also exposed an issue with our usage of libunwind -- basically
you have to do all of your unwinding in one function, but we were
doing it in several different functions and happening to get away
with it.  So, switch from a C++11 range-for loop which is nicer, to
a callback approach which is less nice in my opinion, but safer.

fix recursive searching problem in Makefile

-target_link_libraries already adds the dependencie

Update README.md

strip U out of the mode passed to fileNew before calling fopen

-fix link to Makefile
-update -x argument to current usage