c++/9865: [3.2/3.3/3.4 regression] Template matching for reference types

[nathan@gcc.gnu.org]

Synopsis: [3.2/3.3/3.4 regression] Template matching for reference types

State-Changed-From-To: analyzed->closed
State-Changed-By: nathan
State-Changed-When: Thu Mar 27 14:01:43 2003
State-Changed-Why:
    Not a bug. It is doing exactly what the standard says.
    
    template <typename T> struct Foo; //#1
    template <typename T> struct Foo<const T>; //#2
    Foo<short &> ;;// #1 or #2?
    14.5.4.1/2 says a partial specialization matches if the template arguments of the partial specialization can be deduced from the actual template argument list. So, can 'const T' deduce a T from 'short &'? One deduction is that T is 'short &', which leads to a const qualified reference type. [8.3.2]/1 says that that's ill-formed *except* when
    the cv qualifiers are introduced through the use of a typedef or of a template type argument, in which case they are ignored. So the partial specialization #2 is selected, rather than the primary template #1.
    
    That is PR 2645.
    DR295 was the similar case for function types, and I think it is bad for exactly the same reasons PR 9865 is giving a surprise.
    
    To avoid the surprise, you need another partial specialization
    template <typename T> struct foo<T &>;

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=9865