boost.python解释器实例化

Question

I need to use boost::python , but I want the interpreter context to be destroyed after the python interaction has finished.

How and when exactly is the python interpreter being instantiated, and how do I destroy all the resources allocated by it?

Update Apparently Py_Finalize is not designed to be called many times in the same process. It seems this approach won't work: http://docs.python.org/2/c-api/init.html#l2h-652

Answer 1

Py_Finalize is designed to only be called once per process. The way I handled it is to put it into the shutdown hook for our application, or run it at the end of the main function. You can have multiple instances of the interpreter context and those get cleaned up when their "__main__" object (the one created by import("__main__") ) is destroyed.

Answer 2

Boost.Python does not support safely calling Py_Finalize() .

Per the Embedding - Getting started section:

Note that at this time you must not call Py_Finalize() to stop the interpreter. This may be fixed in a future version of boost.python.

Additional details are noted in the TODO . In short, some internal Boost.Python objects will remain alive during Py_Finalize() , and only attempt to be deleted when Boost.Python unloads, causing the objects to attempt deletion with a non-existent interpreter.