opengl and borland C++ cause floating point exceptions