------- Comment #4 from joseph at codesourcery dot com 2008-07-25 22:59 -------
Subject: Re: gcc does not reject invalid cast
On Fri, 25 Jul 2008, sabre at nondot dot org wrote:
> Though that does raise a question. Does GCC normally emit errors for
> undefined
> behavior? I thought the policy was to insert runtime traps? If so, doesn't
> the { x; } case qualify, or does it violate some other constraint?
What's valid for expressions of incomplete types - including whether the
cases of undefined behavior here are undefined only at execution or
undefined as a property of the program (the former being where traps are
inserted, the latter being where diagnostics are OK) is exceedingly
obscure. We do have a bit of guidance in the form of DR 106 allowing the
dereference of pointers to qualified void in certain cases (note that
qualified void is an incomplete type other than void). It doesn't help
that neither C90 nor C99 got the definition of "lvalue" right.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36941
