«

"Douglas A. Gwyn" <> wrote in message news:...
> Jun Woong wrote:
> > But, is there behavior which doesn't affect the program's output but
> > which still affects the program's conformance?
>
> I still don't understand what the problem is supposed to be.

The real problem is the meaning of the definition for "behavior,"
especially the purpose of the phrase "external appearance" in it.

> A program might not *have* any output if it is rejected by the
> implementation for containing undefined behavior etc.

This was what I said in my previous posting as quoted above.

> There are essentially three kinds of program behavior:
> (1) specified.
> (2) unspecified but "stable".
> (3) undefined.
> A s.c. program contains no instances of (3).
> A s.c. program can contain instances of (2), but its output
> shall not depend on which of the allowed choices the
> implementation has made for each such instance (except for
> locale-specific behavior, which we intentionally allow to
> affect the output as a special kludge to encourage use of
> locales).

Thanks for the answer.


--
Jun, Woong ()
Dept. of Physics, Univ. of Seoul