[cfe-dev] C++ default arguments patch, rework Parse-Sema interaction for function parameters

[Eli Friedman]

On Sun, Apr 6, 2008 at 10:24 AM, Doug Gregor <doug.gregor at gmail.com> wrote:
>  You'll see a few new FIXMEs in this code that I haven't attacked yet.
>  The most annoying one (for me) is that
>
>   void f(int x, int y = x);
>
>  doesn't actually produce an error.

Watch out here: there are other similar restrictions.  For example,
the following is illegal:
void f()
{
    int i;
    extern void g(int x = sizeof(i));
}

Probably better to deal with this in Sema; any other behavior would
involve completely butchering scopes.

+      // Check for C99 6.7.5.3p10 - foo(void) is a non-varargs
+      // function that takes no arguments, not a function that takes a
+      // single void argument.  FIXME: Is this really the right place
+      // to check for this? C++ says that the parameter list (void) is
+      // the same as an empty parameter list, whereas the parameter
+      // list (T) (with T typedef'd to void) is not. For C++, this
+      // should be handled in the parser. Check C89 and C99 standards
+      // to see what the correct behavior is.

AFAIK, clang currently handles C89/99 correctly for this case.  Not
sure how this change affects that, though.

-Eli