[Bug c++/23227] New: SFINAE bug

[Bug c++/23227] New: SFINAE bug

[sylvain dot pion at sophia dot inria dot fr]

The following code refuses to compile :
----------------------------
struct C;

template < typename T >
struct B;

template < typename T >
struct A;

void f(const C &c);    // this one is fine
void f(const B<C> &a); // this one is fine
void f(const A<C> &a); // this one triggers the bug
void f(double) {}

template < typename T >
struct A
{
  B<T> b;
};


int main()
{
  f(1.0); // => instantiates A<C> => instantiates B<C> => fails.
}
-------------------------------

The error message is :
instance.C: In instantiation of 'A<C>':
instance.C:23:   instantiated from here
instance.C:17: error: 'A<T>::b' has incomplete type
instance.C:4: error: declaration of 'struct B<C>'

I am not 100% sure it is a SFINAE bug, but it looks like one to me.

Note that if you comment the declaration of f(A<C>), then it works.
Similarly if you comment the definition of A.

So the problem here is probably that if the definition of A is
available, then the compiler instantiates it, which triggers the
instantiation of B<T>, which fails, but the compiler does not
recover from this instantiation as it should (following SFINAE).

There is the same problem with g++ 3.3, 3.4 and 4.0.1.

-- 
           Summary: SFINAE bug
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: sylvain dot pion at sophia dot inria dot fr
                CC: gcc-bugs at gcc dot gnu dot org,sylvain dot pion at
                    sophia dot inria dot fr
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23227

[Bug c++/23227] SFINAE bug

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-04 13:07 -------
IIRC SFINAE does not mean not instantiating the template class.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23227

[Bug c++/23227] SFINAE bug

[gdr at integrable-solutions dot net]

------- Additional Comments From gdr at integrable-solutions dot net  2005-08-04 13:26 -------
Subject: Re:  SFINAE bug

"pinskia at gcc dot gnu dot org" <gcc-bugzilla@gcc.gnu.org> writes:

| IIRC SFINAE does not mean not instantiating the template class.

That is true.  However, the real issue has nothing to do with SFINAE.  
The compiler is just plain buggy.

Sylvain -- don't describe a plain compiler bug as SFINAE, otherwise
people might be sidetracked :-)

-- Gaby


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23227

[Bug c++/23227] SFINAE bug

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-05 04:02 -------
I don't think this is a bug as A<C> will be tried to be instantiating while trying to convert it from float.

ICC gives the same result as GCC, an error.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23227

[Bug c++/23227] SFINAE bug

[bangerth at dealii dot org]

------- Additional Comments From bangerth at dealii dot org  2005-09-14 15:35 -------
I believe the compiler is correct. In order to check whether there 
is a conversion sequence from float to A<C>, it needs to instantiate the 
type, parts of which are declared by incomplete. This should be an error. 
icc says the same, whatever this means. 
 
W. 

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23227

[Bug c++/23227] SFINAE bug

[gdr at integrable-solutions dot net]

------- Additional Comments From gdr at integrable-solutions dot net  2005-09-14 21:03 -------
Subject: Re:  SFINAE bug

"bangerth at dealii dot org" <gcc-bugzilla@gcc.gnu.org> writes:

| I believe the compiler is correct. In order to check whether there 
| is a conversion sequence from float to A<C>, it needs to instantiate the 
| type, parts of which are declared by incomplete. This should be an error. 

I don't think I agree.

Consider

    stuct A;
    void f(const A&) { }
    void f(int) { }

    void g()
    {
       f(2.0);
    }

We don't reject the call in g() just because A happens to be incomplete
-- therefore we could not "look into" it for determining whether it
needs an implicit conversion or not.  In another words,  It is not
clear whether the completeness of A<C> is required in the testcase
submitted. 

-- Gaby


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23227

[Bug c++/23227] SFINAE bug

[bangerth at dealii dot org]

------- Additional Comments From bangerth at dealii dot org  2005-09-14 21:10 -------
Fair enough. And to get more to the point of only using user-defined 
conversion sequences (instead of the standard conversion from double 
to int): 
-------------------- 
struct A; 
struct B { B(const double &); }; 
 
void f(const A&) { } 
void f(const B&) { } 
 
void g() 
{ 
  f(2.0); 
} 
---------------------- 
This is accepted by all compilers I have as well. 
 
I retract my point of view. This doesn't mean, however, that I'm convinced 
that the opposite would be true. 
 
W. 

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23227

[Bug c++/23227] SFINAE bug

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-09-14 21:23 -------
(In reply to comment #6)
> I retract my point of view. This doesn't mean, however, that I'm convinced 
> that the opposite would be true. 
But both of those testcases represent "void f(const B<C> &a); ".
Here is how I see it:
A<T> is instantiated because it is defined as a way too instantiate it and then we reject it as having 
referencing an incomplete type.

The question here is should we be instantiating A<C>?  I saw yes because otherwise we would reject 
other valid code where A<C> defines a constructor taking double.

And the other question is if we instantiate A<C> should we be erroring out, I say yes as there is no part 
of the standard as far as I can see says we should not error out here.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23227

[Bug c++/23227] SFINAE bug

[gdr at integrable-solutions dot net]

------- Additional Comments From gdr at integrable-solutions dot net  2005-09-14 22:11 -------
Subject: Re:  SFINAE bug

"bangerth at dealii dot org" <gcc-bugzilla@gcc.gnu.org> writes:

| Fair enough. And to get more to the point of only using user-defined 
| conversion sequences (instead of the standard conversion from double 
| to int): 
| -------------------- 
| struct A; 
| struct B { B(const double &); }; 
|  
| void f(const A&) { } 
| void f(const B&) { } 
|  
| void g() 
| { 
|   f(2.0); 
| } 
| ---------------------- 
| This is accepted by all compilers I have as well. 
|  
| I retract my point of view. This doesn't mean, however, that I'm convinced 
| that the opposite would be true. 

In fact, I've come to be on the same position: I don't know.  This is
probably something we need to forward to CWG.

-- Gaby


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23227

[Bug c++/23227] SFINAE bug

[fang at csl dot cornell dot edu]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |fang at csl dot cornell dot
                   |                            |edu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23227

[Bug c++/23227] SFINAE bug

[jason at gcc dot gnu dot org]

------- Comment #13 from jason at gcc dot gnu dot org  2009-03-04 01:29 -------
Invalid.


-- 

jason at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23227

[Bug c++/23227] SFINAE bug

[jason at gcc dot gnu dot org]

------- Comment #12 from jason at gcc dot gnu dot org  2009-03-04 01:28 -------
oops, wrong resolution.


-- 

jason at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |UNCONFIRMED
         Resolution|FIXED                       |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23227

[Bug c++/23227] SFINAE bug

[jason at gcc dot gnu dot org]

------- Comment #11 from jason at gcc dot gnu dot org  2009-03-04 01:28 -------
This can't be a SFINAE issue, as the f's in question aren't templates.  The
compiler instantiates A<C> in order to determine whether or not there's a
conversion from double to A<C>, and that instantiation fails.  This isn't a
bug.


-- 

jason at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23227

[Bug c++/23227] SFINAE bug

[paolo dot carlini at oracle dot com]

------- Comment #10 from paolo dot carlini at oracle dot com  2009-02-22 16:38 -------
CC-ing Jason seems a good idea...


-- 

paolo dot carlini at oracle dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jason at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23227

[Bug c++/23227] SFINAE bug

[steven at gcc dot gnu dot org]

------- Comment #9 from steven at gcc dot gnu dot org  2009-02-22 16:29 -------
>3 years of inaction.  What is going to be done about this?


-- 

steven at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23227