Re: Ping: 4 C++ patches to review, all from serious regression list

[Jason Merrill <jason at redhat dot com>]

Nathan Sidwell wrote:
> sje@cup.hp.com wrote:
>
> > http://gcc.gnu.org/ml/gcc-patches/2008-12/msg00737.html (PR c++/37646)
>
> hm, I can't see where the C++ std allows deferring type resolution of ? operator
> operands (I can't see where it allows it for comma operators for that matter).

I agree.  I think the bug in this case is that real_value_p is returning 
true for "foo", which is incorrect; an expression that names a class 
member is an lvalue only if it names a data member or static member 
function.

The EDG front end gives an error to the effect that 'foo' is a 
nonstandard way of writing &A::foo, which seems sensible to me.  I'd 
guess that in VC++ compatibility mode it's not an error.

Jason