Originally reported by: Anselm Kruis (Bitbucket: akruis, GitHub: akruis)

Issue #62 discusses problems caused by the non recursive pickling of frames. In the end I closed #62 as "won't fix", because the proposed solution (recursive frame pickling) had compatibility problems.

But there is still room for improvements:

1. Linkage of frames in unpickled tracebacks. It is fairly simple to reconstruct the linkage for the inner frames - as returned by inspect.getinnerframes(tb) - of a traceback. I already coded a test case and a patch to static PyObject * tb_setstate(PyObject *self, PyObject *args) in prickelpit.c. This could go into 2.7-slp, because it shouldn't cause any compatibility problems.

2. In issue pickling of frames without a tasklet #62 we discussed the idea of a FrameReducer, that would avoid the requirement to add a reduce function for frames to copy_reg. See https://bitbucket.org/stackless-dev/stackless/issues/62/pickling-of-frames-without-a-tasklet#comment-10422733. I'll try to implement this proposal. I'm undecided, if this is a feature for 2.7-slp, but if it works well, it should definitely go into 3.4-slp. And it will help to fix 3.5-slp.

• Bitbucket: https://bitbucket.org/stackless-dev/stackless/issue/107