Bug#477010: taking the address of a macro

["Christopher L Conway" <cconway@cs.nyu.edu>]

The problem is that the relevant lines are trying to create function
pointers, but the identifiers named are really macros.

The following example gives a similar error:

macroptr.c:
int f(int x, int y, int z) { return 0; }
#define FOO(x) f(x,0,0)

int (*p)() = &f;
int (*q)() = &FOO;

$ gcc macroptr.c
macroptr.c:5: error: 'FOO' undeclared here (not in a function)

It's strange this hasn't been a problem before, because the
declarations have been macros for several years in the Python trunk.