Re: seperate keyword and seperate compilation with Gnat?

[dewar@cs.nyu.edu (Robert Dewar)]

Norman asked:

"I presume you mean that it is invalid to compute a+b this way because of
the potential overflow, and not that it is invalid to raise an exception
because of the overflow."

Well not quite, it is not invalid to compute a+b that way, it is invalid
to compute a+b that way if it will indeed raise an exception. If you
compute a+b that way (4 * a + b - 3 * a), and you do not raise an exception
during the computation (as is allowed by the optimization rules), then this
is a stupid and inefficient thing to do, but is not invalid.