[Bug c++/55436] New: g++ compiles invalid code with child class of nested class in template class

[Bug c++/55436] New: g++ compiles invalid code with child class of nested class in template class

[conradsand.arma at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55436

             Bug #: 55436
           Summary: g++ compiles invalid code with child class of nested
                    class in template class
    Classification: Unclassified
           Product: gcc
           Version: 4.7.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: conradsand.arma@gmail.com


Created attachment 28760
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28760
test.cpp: source for invalid code that compiles under g++

The code below compiles under g++ 4.7.2 and 4.4.6, but is apparently invalid,
ie. it is not rejected by g++.

We initially thought this was a bug in the clang compiler, but the clang folks
are saying that this is a bug in g++ (ie. gcc shouldn't be accepting this
code).

Related clang bugzilla bug entry:
http://llvm.org/bugs/show_bug.cgi?id=14402

The clang folks are quoting this part of the standard:

See [temp.dep]p3: "In the definition of a class or class template, if a base
class depends on a template-parameter, the base class scope is not examined
during unqualified name lookup either at the point of definition of the class
template or member or during an instantiation of the class template or member."

#include <iostream>

template<typename T>
class A
{
 public:
  void m() { std::cout << "A::m()" << std::endl; }
  void z() { m(); }

  class B
  {
   public:
    void m() { std::cout << "B::m()" << std::endl; }
    void y() { m(); }
  };

  class C : public B
  {
   public:
    void x() { m(); }
  };
};

int main()
{
  A<double> a;
  a.z();

  A<double>::B b;
  b.y();

  A<double>::C c;
  c.x();
}


Clang 3.1 reports the following error:
test.cpp:20:16: error: call to non-static member function 'm' of 'A' from
nested
      type 'C'
    void x() { m(); }


Shouldn't g++ also report a similar error?

[Bug c++/55436] g++ compiles invalid code with child class of nested class in template class

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55436

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-11-22 04:13:58 UTC ---
Comeau online tester (which is EDG based) says:
"ComeauTest.c", line 20: error: a nonstatic member reference must be relative
to a
          specific object
      void x() { m(); }
                 ^
          detected during instantiation of "void A<T>::C::x() [with T=double]"
                    at line 33

[Bug c++/55436] g++ compiles invalid code with child class of nested class in template class

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55436

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-11-22
     Ever Confirmed|0                           |1

--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-11-22 12:53:54 UTC ---
Yes, A<T>::B is a dependent base of A<T>::C, so lookup should find A::m not
B::m

[Bug c++/55436] g++ compiles invalid code with child class of nested class in template class

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55436

Jonathan Wakely <redi at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |rejects-valid, wrong-code

--- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-11-22 12:58:19 UTC ---
This is also a wrong-code and rejects-valid bug.

This correct program shows the wrong m() being called (and so results in a
linker error)


template<typename T>
struct A
{
  static void m() { }

  struct B
  {
    void m();
  };

  struct C : B
  {
    void x() { m(); }
  };
};

int main()
{
  A<double>::C c;
  c.x();
}

and if we make B::m private we get rejects-valid

.cc: In member function ‘void A<T>::C::x() [with T = double]’:
x.cc:21:7:   instantiated from here
x.cc:9:10: error: ‘void A<T>::B::m() [with T = double]’ is private
x.cc:14:16: error: within this context

[Bug c++/55436] g++ compiles invalid code with child class of nested class in template class

[conradsand.arma at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55436

Conrad <conradsand.arma at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |redi at gcc dot gnu.org

--- Comment #4 from Conrad <conradsand.arma at gmail dot com> 2012-12-18 08:55:23 UTC ---
Compiling the invalid test.cpp (original attachment) under 4.8.0 20121209
(x86_64) also works, when it shouldn't.

The code presented in Comment #3 by Jonathan Wakely also results in a linked
error when using 4.8.0 20121209.

Should this bug be upgraded to P2, given its wide ranging effects ?

[Bug c++/55436] g++ compiles invalid code with child class of nested class in template class

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55436

--- Comment #5 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-12-18 10:46:57 UTC ---
(In reply to comment #4)
> Should this bug be upgraded to P2, given its wide ranging effects ?

If it was a serious problem it would have been reported before by more people.

[Bug c++/55436] g++ compiles invalid code with child class of nested class in template class

[conradsand.arma at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55436

Conrad <conradsand.arma at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Version|4.7.2                       |4.8.0

--- Comment #6 from Conrad <conradsand.arma at gmail dot com> 2012-12-18 14:54:50 UTC ---
> If it was a serious problem it would have been reported before by more people.

I'm not sure if the degree of popularity of a bug should diminish its
seriousness.  This is an "accepts-invalid, rejects-valid, wrong-code" bug,
which is pretty nasty as far as bugs go. The use of C++ is expanding (including
GCC's own code base), meaning this bug is going to bite more people as time
goes on.

However, I do understand that bugs which affect more people right now should
get looked at first. Furthermore, given its wide ranging impact, perhaps fixing
this bug in 4.7.x would not be doable (ie. too invasive).  As such, I'll
reassign this bug to 4.8.0.

[Bug c++/55436] g++ compiles invalid code with child class of nested class in template class

[redi at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55436

--- Comment #7 from Jonathan Wakely <redi at gcc dot gnu.org> 2012-12-18 15:29:59 UTC ---
(In reply to comment #6)
> > If it was a serious problem it would have been reported before by more people.
> 
> I'm not sure if the degree of popularity of a bug should diminish its
> seriousness.

If noone else has ever reported it in all the time it's been in GCC it's not
causing a lot of problems, then it's not causing serious problems to most GCC
users.  It's not a regression and there's a workaround (qualify the names) so
it's not P2.

>  This is an "accepts-invalid, rejects-valid, wrong-code" bug,
> which is pretty nasty as far as bugs go.

Almost any name lookup bug can be turned into all three categories, it doesn't
alter its seriousness.

> The use of C++ is expanding (including
> GCC's own code base), meaning this bug is going to bite more people as time
> goes on.

It will get fixed, it doesn't need to be P2 for that.

[Bug c++/55436] g++ compiles invalid code with child class of nested class in template class

[conradsand.arma at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55436

--- Comment #8 from Conrad <conradsand.arma at gmail dot com> 2013-01-09 06:03:07 UTC ---
Can we change this bug status to _confirmed_, and the target milestone to 4.8.0
?

[Bug c++/55436] g++ compiles invalid code with child class of nested class in template class

[conradsand.arma at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55436

Conrad <conradsand.arma at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |conradsand.arma at gmail dot com

--- Comment #9 from Conrad <conradsand.arma at gmail dot com> ---
Paolo, any reason for twice removing my address from the CC list for this bug?

[Bug c++/55436] g++ compiles invalid code with child class of nested class in template class

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55436

Paolo Carlini <paolo.carlini at oracle dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|conradsand.arma at gmail dot com   |

--- Comment #10 from Paolo Carlini <paolo.carlini at oracle dot com> ---
Yes, it's redundant, you get all the messages by default being the bug
submitter

[Bug c++/55436] g++ compiles invalid code with child class of nested class in template class

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55436

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|NEW                         |RESOLVED
   Target Milestone|---                         |7.0
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=10200,
                   |                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=69753

--- Comment #12 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Fixed in GCC 7 by r7-755.