The article is a bit confusing because they also write:

> Regardless of which process had the bug—the compiler or compiled program—in both cases the processor only did the bad thing because the compiler told it to. And in both cases the fix is the same: the compiler's code must change, since that code was what caused the memory corruption.

But yeah, I wonder what those 1,200 unsafe uses actually did?