🚀 go-pugleaf

> This point has been debated before.  Some people think the standard
> says `f()+f()' has undefined behavior when f has side effects
> (e.g. if f is defined by `int f() { static int i; return i++; }').
> Other people think the standard defines behavior for `f()+f()'.
> ... 
> It is clearly an unnecessarily messy area in the C Standard.

On the contrary, the problem is that some people (i.e. the first group
above) insist on reading the completly un-messy sentence (in 6.6/3.6,
semantics):

#  Except as indicated, statements are executed in sequence.

as if there was something in the standard$ that indicated that statements
could be interleaved.
 
> I don't know if the ISO C committee has issued an official ruling on this.

We can expect one soon, as part of the package of Record of Response 2
and Technical Corrigendum 2.  Balloting on that should have been finished
a while ago now, but I haven't heard the results of it.

> As a practical matter, if the implementation does function inlining,
> I wouldn't rely on the behavior being well-defined here.

As a practical matter, any implementer that *I* caught not producing
the behavior I expect here could expect a bug report from me.

$ - Okay, the "as if" rule allows it, but only if you can't tell it's
    happening, so that doesn't affect the "debate".
-- 
Mark Brader, msb@sq.com | "Ah, determinism (likewise, forgetfulness) reigns."
SoftQuad Inc., Toronto  |                                      --Steve Summit

My text in this article is in the public domain.