In ceval.c there is such code:

PyObject *
PyEval_CallObject(PyObject *func, PyObject *arg)
{
	return PyEval_CallObjectWithKeywords(func, arg, (PyObject *)NULL);
}
#define PyEval_CallObject(func,arg) \
        PyEval_CallObjectWithKeywords(func, arg, (PyObject *)NULL)

Is this needed any longer? (both #define and function have the same name)

Well, #defines are not exported in shared libraries. I suppose PyEval_CallObject() was part of an old definition of the Python C-API and it was kept for compatibility.

The #define dates back from 2007, this changeset:

branch:      trunk
user:        guido
date:        Sat Aug 30 17:02:50 1997 +0200
files:       Include/ceval.h Python/ceval.c
description:
[svn r8683] Inline PyObject_CallObject (Marc-Andre Lemburg).

Antoine Pitrou wrote:
> 
> Antoine Pitrou <pitrou@free.fr> added the comment:
> 
> The #define dates back from 2007, this changeset:
> 
> branch:      trunk
> user:        guido
> date:        Sat Aug 30 17:02:50 1997 +0200
> files:       Include/ceval.h Python/ceval.c
> description:
> [svn r8683] Inline PyObject_CallObject (Marc-Andre Lemburg).

That's not useless: it's a typical backwards compatibility macro
to auto-upgrade existing software via recompile.

It's also still used a lot in the Python source code and elsewhere:

http://www.google.de/search?q=PyEval_CallObject

The function PyEval_CallObject() can probably be removed by
now, though.

Yeah, I meant the function. Sorry for not being specific enough. I suppose the function is there to preserve ABI, but definitely code needed recompilation since 2007.

Right, we can certainly suppress the function definition now and just keep the macro.
(especially given it's 1997, not 2007, there was a typo in my message)

Fixed in r79555 (trunk), r79556 (py3k).