[Bug c++/3478] [3.4 regression] Accepts invalid enum declarations from template 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=3478


reichelt at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |REOPENED
           Keywords|rejects-valid               |accepts-invalid
           Priority|P3                          |P1
         Resolution|FIXED                       |
            Summary|Rejects code when using enum|[3.4 regression] Accepts
                   |from template class         |invalid enum declarations
                   |                            |from template class
   Target Milestone|---                         |3.4
            Version|2.95.2                      |3.4


------- Additional Comments From reichelt at gcc dot gnu dot org  2003-09-01 21:46 -------
I reopened the PR for two reasons:
Wolfgang was not quite right and the new parser changed its behavior.

To get it really right this time, here's an updated example (with Mark's help):

-------------------------------------------------------------------------------
template <typename> struct A
{
    enum E {};
};

template <typename T> void foo()
{
    enum          A<void>::E e1; // OK
    typename      A<T>::E    e2; // OK
    enum          A<T>::E    e3; // invalid, A<T>::E is dependent and hence
                                 // [7.1.5.3]/3 cannot be enforced
    enum typename A<T>::E    e4; // invalid, violates [7.1.5.3]
}
-------------------------------------------------------------------------------

Alas, mainline now accepts all four definitions,
which leaves us with a double accepts-invalid bug.

E4 was correctly rejected before the new parser merge.
E3 was correctly rejected by the new parser until 20030812.
Thus, we have a 3.4 regression in both cases.