new C++ parser bug

[Matt Austern]

Anyone tried building QT with top of mainline?  I just tried this 
morning
(using top of tree as of yesterday morning), and it failed because of a
compiler bug. Here's a small test case:

     struct Str
     {
       Str( const char *str );
       Str frob() const;
       operator const char *() const;
     };

     void foo(int x, const Str& str)
     {
       x ? str : str.frob();
     }

The error message I get is:
     foo.ii: In function `void foo(int, const Str&)':
     foo.ii:10: error: non-lvalue in unary `&'

This code compiles cleanly with 3.3, and the error message is of
course nonsense.  And I could be missing something, but I believe
this code ought to compile.  Yes, it's bad style to have a class with
a const char* constructor and also an operator const char*, but this
class isn't being used in a way that should choke on any ambiguities.

So, before I spend time trying to understand how to fix this...
Is this  a known bug?

			--Matt