Assess/resolve Python 3.9 deprecation warnings

[nrnhines]

../src/nrnpython/nrnpy_nrn.cpp: In function ‘PyObject* NPySecObj_psection(NPySecObj*)’:
/home/hines/.pyenv/versions/3.9.0/include/python3.9/ceval.h:24:66: warning: ‘PyObject* PyEval_CallObjectWithKeywords(PyObject*, PyObject*, PyObject*)’ is deprecated [-Wdeprecated-declarations]
     PyEval_CallObjectWithKeywords(callable, arg, (PyObject *)NULL)
                                                                  ^
../src/nrnpython/nrnpy_nrn.cpp:831:24: note: in expansion of macro ‘PyEval_CallObject’
     PyObject* result = PyEval_CallObject(nrnpy_psection, arglist);
                        ^~~~~~~~~~~~~~~~~
/home/hines/.pyenv/versions/3.9.0/include/python3.9/ceval.h:17:43: note: declared here
 Py_DEPRECATED(3.9) PyAPI_FUNC(PyObject *) PyEval_CallObjectWithKeywords(
                                           ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~

I see that:

hines@hines-T7500:~/neuron/unicode-error/src/nrnpython$ grep -r PyEval_Call *
nrnpy_nrn.cpp:    PyObject* result = PyEval_CallObject(nrnpy_psection, arglist);
rxd.cpp:                result = PyEval_CallObject(source[i], NULL);

There are simple replacements with

PyObject* PyObject_CallNoArgs(PyObject *callable)
PyObject* PyObject_CallObject(PyObject *callable, PyObject *args)

[nrnhines]

@adamjhn
What is the meaning of this in geometry3d.cpp

    int geometry3d_test_call_function(PyObject* obj) {
        printf("inside\n");
        if (obj == NULL) printf("obj is NULL\n");
        Py_INCREF(obj);
        PyEval_CallObject(obj, obj);
        return 0;
    }

Shouldn't the return value of PyEval_CallObject be checked? It seems hardly correct for me merely to replace the line with

        PyObject_CallObject(obj, obj);

just to avoid the deprecation message for Python 3.9

[ramcdougal]

I assume that's test code that snuck into the repository that can be safely removed.

There shouldn't be any printfs in code used for simulation. (Also, note that Py_INCREF requires obj to be non-NULL.)

[nrnhines]

From ninja >& temp I see that there are 42 deprecation statements of three kinds.
So I guess this issue can be expanded to deal with them all.

hines@hines-T7500:~/neuron/pyeval_callobject/build$ grep deprecate temp | sed -n 's/.*warning:'//p | sort | uniq
 ‘PyObject* PyEval_CallObjectWithKeywords(PyObject*, PyObject*, PyObject*)’ is deprecated [-Wdeprecated-declarations]
 ‘Py_ssize_t _PyUnicode_get_wstr_length(PyObject*)’ is deprecated [-Wdeprecated-declarations]
 ‘Py_UNICODE* PyUnicode_AsUnicode(PyObject*)’ is deprecated [-Wdeprecated-declarations]
 #warning "Using deprecated NumPy API, disable it with " "#define NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION" [-Wcpp]

[olupton]

Looks like this was addressed in a combination #1334 and #1806.

[nrnhines]

Is it relevant that with python3.10 and

cmake .. -G Ninja -DCMAKE_INSTALL_PREFIX=install -DPYTHON_EXECUTABLE=`which python3` -DCMAKE_BUILD_TYPE=Debug -DNRN_ENABLE_RX3D=OFF -DNRN_ENABLE_INTERVIEWS=OFF
ninja

I see at the end:

/home/hines/neuron/temp/build/src/nrnpython/setup.py:2: DeprecationWarning: The distutils package is deprecated and slated for removal in Python 3.12. Use setuptools or check PEP 632 for potential alternatives
  from distutils.core import setup, Extension

[alexsavulescu]

Is it relevant that with python3.10 and

To be tackled with: #1605