c++/9341: [3.4 regression] Specialization after use gives error

[rguenth@tat.physik.uni-tuebingen.de]

>Number:         9341
>Category:       c++
>Synopsis:       [3.4 regression] Specialization after use gives error
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          rejects-legal
>Submitter-Id:   net
>Arrival-Date:   Thu Jan 16 02:46:02 PST 2003
>Closed-Date:
>Last-Modified:
>Originator:     Richard Guenther
>Release:        g++ (GCC) 3.4 20030115 (experimental)
>Organization:
>Environment:
ix86-linux
>Description:
template<class T, class U>
struct UnaryReturn {
        typedef T Type_t;
};
struct foo
{
        template <class T>
        typename UnaryReturn<T, int>::Type_t foo();
};
template<class T>
struct UnaryReturn<T, int> {
        typedef bool Type_t;
};

gives

bellatrix:~/src/tests$ ~/ix86/gcc3.4/bin/g++ -c -Wall insbefdecl.cpp 
insbefdecl.cpp:8: error: return type specification for constructor invalid
insbefdecl.cpp:11: error: specialization of `UnaryReturn<T, int>' after instantiation

>From Mark Mitchell:

This is a bug -- but not in the new parser.  This is lingering fallout from the now-almost-gone implicit typename extension.  I will use this as the excuse to kill the last vestiges of that ugliness.
>How-To-Repeat:
compile the testcase with the new parser
>Fix:

>Release-Note:
>Audit-Trail:
>Unformatted: