> My impression was that for any object ({u}char, {u}int, pointer, double) X
> in the current stack frame (while we're not in the runtime), then
> (&X - stackTop) mod (sizeof X) == 0. The real problem is that
> (&X - stackBottom) mod (sizeof X) == 0 is not always true.
Right. And we can get the latter by for ints forcing frames to be of size = 0
mod 4.