“That is either genius, or a seriously diseased mind.” – Linus Torvalds

In that post Linus says something I don't understand in that post:

"... So with a constant, we have

  sizeof( 1 ? NULL : (int *) 1)
  

and the rule is that if one of the sides of a ternary operation with pointers is NULL, the end result is the other type (int *)."

Is Linus saying that the condition of the ternary operation doesn't matter if one of the sides is NULL? That it then always evaluates to the other side? That's not what happens when I try it. Or are there special rules when sizeof is applied?

I really need this for libc optimizations. So far it only worked in C++ mode or newer clang's, but not in gcc.

Isn't this what GCC's `__builtin_constant_p()` does?