arrays of size 0 are no problem for gcc both in C and C++, so you should use something like

#define ct_assert(x) typedef char _ct_assert__[(x)? 1 : -1];

tamlin@algonet.se schrieb:
Actually, I'd say the stupidity is in using a runtime assert, when a 
compile-time assert will do the trick (not using CPU at runtime, while 
enforcing the condition).

Unless memory fails me, in C++ you can implement ct_assert(x) like:

typedef char[x!=0] foo;

as an array of size zero is an error.

How does current C compilers react to typedefs of zero-sized arrays? 
Sure, one can't use it exactly as in C++, but it could still be useful 
to at least know about this idiom.

Problems should be trapped as early as possible.