[Bug c++/9807] [3.3 regression] Derived class object used as default parameter for base class reference requires explicit cast to reference to base class

["reichelt at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=9807


reichelt at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|rejects-valid               |diagnostic
            Summary|Derived class object used as|[3.3 regression] Derived
                   |default parameter for base  |class object used as default
                   |class reference requires    |parameter for base class
                   |explicit cast to reference  |reference requires explicit
                   |to base class               |cast to reference to base
                   |                            |class


------- Additional Comments From reichelt at gcc dot gnu dot org  2003-08-05 14:55 -------
On mainline we now get

x.cc:9: error: invalid initialization of non-const reference of type 'A&' 
   from a temporary of type 'B'
x.cc:7: error: in passing argument 1 of `C::C(A&)'

for Wolfgang's small example. And I think that gcc is right here.

However, the error message on the 3.3 branch is still hosed:

x.cc:7: error: invalid type `B' for default argument to `A&'
x.cc:9: error: could not convert `B()' to `A&'
x.cc:7: error: in passing argument 1 of `C::C(A&)'

That's why I changed the keyword from "rejects-valid" to "diagnostic".

Since we already did better in 2.95.x (see below) I rate this as a
regression on the 3.3 branch

x.cc:7: invalid type `B' for default argument to `A &'
x.cc: In function `void __static_initialization_and_destruction_0(int, int)':
x.cc:9: initialization of non-const reference type `struct A &'
x.cc:9: from rvalue of type `B'

May others decide whether this will be fixed on the branch or not.
Btw, the error message got hosed with gcc 3.0.