[issue24554] GC should happen when a subinterpreter is destroyed

[STINNER Victor]

STINNER Victor <vstinner at python.org> added the comment:

This issue is partially fixed in the master branch. Extract of the finalize_interp_clear() function, called by Py_EndInterpreter():

    /* Clear interpreter state and all thread states */
    PyInterpreterState_Clear(tstate->interp);

    /* Trigger a GC collection on subinterpreters*/
    if (!is_main_interp) {
        _PyGC_CollectNoFail();
    }

gc.collect() is now called.

It's only "partially" fixed because I would prefer to trigger a GC collection before or during PyInterpreterState_Clear(). IMHO trigger it after PyInterpreterState_Clear() creates a risk of crash in finalizers written in C which don't handle well before called very late during Python finalization. After PyInterpreterState_Clear(), Python is basically unusable. All modules are cleared.

----------

_______________________________________
Python tracker <report at bugs.python.org>
<https://bugs.python.org/issue24554>
_______________________________________