I observed a code clone from the following files.

function : PyErr_Restore @ (file: "Python-3.3.0a4/Python/errors.c", line: 27)~49
function : PyErr_SetExcInfo @ (file: "Python-3.3.0a4/Python/errors.c", line: 341)~353

The function PyErr_Restore performs some checking on traceback while PyErr_SetExcInfo does not. I wonder if this is necessary in the function PyErr_SetExcInfo. The source code is included in the attachment. Hope it helps.

From what I can see the code is completely the same in default but does this matter or not?  As a matter of interest is there an official policy regarding code clones, whether in C or pure Python?

The code is not really cloned - note that PyErr_Restore is operating on tstate->curexc (the exception currently being handled) while PyErr_SetExcInfo is operating on tstate->exc_info (the earliest exception currently on the exception stack).

If there is a bug here, I think it is the silent conversion of traceback to NULL in the case where it is neither NULL nor Py_None. (The case in PyErr_Restore that the comment says should not happen).

Otherwise, the difference is that PyErr_Restore transforms Py_None to NULL for the traceback, and PyErr_SetExcInfo doesn't.  That conversion seems neither necessary nor wrong.

Adding extra checks to PyErr_SetExcInfo doesn't have a lot of value because is already out of of contract. If you pass an invalid object (let's say a list) to any of the specific APIs (like PyDict_GetItem) it will crash anyway. We could add some assert, though.