[Bug c++/13548] New: SFINAE fails to eliminate overload

[Bug c++/13548] New: SFINAE fails to eliminate overload

[dave at boost-consulting dot com]

It's not clear which operator== is invoked by the main() function in the 
enclosed example, but none of them should work.  Since neither of 
reverse_iterator<int**> or indirect_iterator<int**> is convertible to the 
other, SFINAE should eliminate the templated operator== in namespace boost.

-- 
           Summary: SFINAE fails to eliminate overload
           Product: gcc
           Version: 3.4.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: dave at boost-consulting dot com
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13548

[Bug c++/13548] SFINAE fails to eliminate overload

[dave at boost-consulting dot com]

------- Additional Comments From dave at boost-consulting dot com  2004-01-02 18:47 -------
Created an attachment (id=5400)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=5400&action=view)
preprocessed source


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13548

[Bug c++/13548] SFINAE fails to eliminate overload

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
            Version|3.4.0                       |3.3.1


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13548

[Bug c++/13548] SFINAE fails to eliminate overload

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-01-02 20:55 -------
The preprocessed source is for 3.3.1 which does not compile at all on 3.4.0 but does compile just 
fine on 3.3.3, Can you attach the one for 3.4?

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13548

[Bug c++/13548] SFINAE fails to eliminate overload

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-01-02 21:35 -------
You are wrong as it should match:
boost::detail::enable_if_interoperable<boost::indirect_iterator<int**, boost::use_default, 
boost::use_default, boost::use_default, boost::use_default>, boost::reverse_iterator<int**
>, bool>::type 
boost::operator==<boost::indirect_iterator<int**, boost::use_default, boost::use_default, 
boost::use_default, boost::use_default>, int, boost::random_access_traversal_tag, int&, 
int, boost::reverse_iterator<int**>, int*, boost::random_access_traversal_tag, int*&, int>
(boost::iterator_facade<boost::indirect_iterator<int**, boost::use_default, 
boost::use_default, boost::use_default, boost::use_default>, int, 
boost::random_access_traversal_tag, int&, int> const&, 
boost::iterator_facade<boost::reverse_iterator<int**>, int*, 
boost::random_access_traversal_tag, int*&, int> const&)

As iterator_adaptor's base is a iterator_facade.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
         Resolution|                            |INVALID


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13548

[Bug c++/13548] SFINAE fails to eliminate overload

[dave at boost-consulting dot com]

------- Additional Comments From dave at boost-consulting dot com  2004-01-02 22:45 -------
Bluntly starting your reply with "you are wrong", especially when you, in 
fact, are wrong, is impolite.  I'm just trying to help by reporting this one, 
and wish to be treated accordingly.

That base is not enough to make it match.  What is the return type of this 
function?

boost::detail::enable_if_interoperable<boost::indirect_iterator<int**, 
boost::use_default, 
boost::use_default, boost::use_default, boost::use_default>, 
boost::reverse_iterator<int**
>, bool>::type

Is not a valid type, therefore the overload should be removed.  Substitution 
failure is not an error

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13548

[Bug c++/13548] SFINAE fails to eliminate overload

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-01-02 22:57 -------
How is enable_if_interoperable<boost::indirect_iterator<..>, boost::reverse_iterator<..>, bool> an 
invalid type, there is only a definition of enable_if_interoperable:
    template <class Facade1, class Facade2, class Return>
    struct enable_if_interoperable
    {
        typedef Return type;
    };


template < class Derived1, class V1, class TC1, class R1, class D1 , class Derived2, class V2, class 
TC2, class R2, class D2 >

typename detail::enable_if_interoperable< Derived1, Derived2, bool >::type 
operator ==
( 
iterator_facade<Derived1, V1, TC1, R1, D1> const& lhs , 
iterator_facade<Derived2, V2, TC2, R2, D2> const& rhs) 

Derived1 = boost::indirect_iterator<int**, boost::use_default, boost::use_default, boost::
use_default, boost::use_default>

Derived2 = boost::reverse_iterator<int**>

V1 = int
V2 = int *
TC1 = boost::random_access_traversal_tag
TC2 = boost::random_access_traversal_tag
R1 = int&
R2 = int*&
D1 = int
D2 = int

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13548

[Bug c++/13548] SFINAE fails to eliminate overload

[dave at boost-consulting dot com]

------- Additional Comments From dave at boost-consulting dot com  2004-01-03 00:36 -------
You're right; enable_if_interoperable was given that definition as a 
workaround for earlier versions of GCC.  Let me look into this further and let 
you know by tomorrow.  Sorry...

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13548

[Bug c++/13548] SFINAE fails to eliminate overload

[dave at boost-consulting dot com]

------- Additional Comments From dave at boost-consulting dot com  2004-01-03 02:50 -------
OK, this one was my mistake.  It was a bug in previous versions of GCC, but it 
is fixed in the latest gcc-3.4 snapshot.

Sorry to waste time on this.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13548