Alias analysis is one of the most used techniques that aim to optimize languages with pointers. It is no surprise that this topic has received much attention in many fields including compilation and program verification. However, despite all this attention, a deep study of its state-of-the art shows that it is not handled satisfactorily, by far. In this paper we demonstrate that, although many approaches have been proposed, we still cannot rely on alias analysis inside compilers. We illustrate our statement with an overview of the capabilities of state-of-the-art compilers.