[Bug c++/50864] New: SFINAE bug

[Bug c++/50864] New: SFINAE bug

[fate66260 at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

             Bug #: 50864
           Summary: SFINAE bug
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: fate66260@gmail.com


[C++0x] 
compiler option: -Wall -std=c++0x -O3

error message:
internal compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

source code:
namespace concept_impl {

template < class lhs, class rhs >
struct is_arrow_operable_impl
{
private:
  template < class T = lhs, class U = rhs
           , class = decltype(std::declval<T>() -> std::declval<U>()) // <-
here
           >
  static mpl::true_
  test (int);

  template < class... >
  static mpl::false_
  test (...);

public:
  typedef decltype(test (0)) type;
};


} // namespace concept_impl

template < class lhs, class rhs >
struct is_arrow_operable
  : public concept_impl::is_arrow_operable_impl<lhs, rhs>::type
{ };

// test type
struct foo
{
  int* p;
};

// call this meta function
static_assert(is_arrow_dereferencable<foo*, int* foo::*>::value ,"error");

[Bug c++/50864] SFINAE bug

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2011-10-25
     Ever Confirmed|0                           |1

--- Comment #1 from Paolo Carlini <paolo.carlini at oracle dot com> 2011-10-25 14:25:23 UTC ---
Apparently this has essentially to do with namespaces, TREE_CODE (code) is
NAMESPACE_DECL when tsubst_copy ICEs.

The testcase is a bit garbled (ie, cannot possibly compile anyway) - missing
std::declval, mpl::true_, mpl::false_, is_arrow_dereferencable - but
interestingly if all the missing bits are provided in the global namespace, no
ICE.

[Bug c++/50864] SFINAE bug

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |WAITING

--- Comment #2 from Paolo Carlini <paolo.carlini at oracle dot com> 2011-10-25 14:30:00 UTC ---
As is, this is a low priority ICE on *invalid*: do you have a version which
just triggers the static_assert but is otherwise well formed? (ie, a *proper*
sfinae bug)

[Bug c++/50864] SFINAE bug

[fate66260 at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

--- Comment #3 from fate66260 at gmail dot com 2011-10-25 15:00:06 UTC ---
Thank you for your response.

wrong: allow_dereferencable
right: allow_operable

mpl::true_, mpl::false_ are included in <boost/mpl/bool.hpp>
std::declval ... <utility>

The same result in the all cases.

[Bug c++/50864] SFINAE bug

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |ice-on-invalid-code

--- Comment #4 from Paolo Carlini <paolo.carlini at oracle dot com> 2011-10-25 15:01:21 UTC ---
Yep, please provide a self-contained testcase, thanks.

[Bug c++/50864] SFINAE bug

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

--- Comment #5 from Paolo Carlini <paolo.carlini at oracle dot com> 2011-10-25 15:45:10 UTC ---
To clarify: the reason why I'm asking a self-contained testcase from you is
that the way I quickly hacked what you filed myself, I got the static_assert
firing, but that's not a bug, it's supposed to do that for the involved types
(+ of course this is the normal policy..)

[Bug c++/50864] [4.6/4.7 Regression] ICE with decltype and "declval" from another namespace

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|ice-on-invalid-code         |ice-on-valid-code
             Status|WAITING                     |NEW
                 CC|                            |jason at gcc dot gnu.org
            Summary|SFINAE bug                  |[4.6/4.7 Regression] ICE
                   |                            |with decltype and "declval"
                   |                            |from another namespace

--- Comment #6 from Paolo Carlini <paolo.carlini at oracle dot com> 2011-10-25 17:04:41 UTC ---
Ok, this way at least part of the issue is more clear: the below doesn't parse,
remove the impl namespace and it works:

/////////////////

namespace impl
{
  template <class T> T create();
}

template < class lhs, class rhs >
struct is_arrow_operable_impl
{
  template <class T, class U,
        class = decltype(impl::create<T>() -> impl::create<U>())>
  void test();
};

[Bug c++/50864] [4.6/4.7 Regression] ICE with decltype and "declval" from another namespace

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |paolo.carlini at oracle dot
                   |                            |com

--- Comment #7 from Paolo Carlini <paolo.carlini at oracle dot com> 2011-10-25 17:14:51 UTC ---
The arrow operator (vs, eg, +) seems also essential.

[Bug c++/50864] [4.6/4.7 Regression] ICE with decltype and "declval" from another namespace

[daniel.kruegler at googlemail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

--- Comment #8 from Daniel Krügler <daniel.kruegler at googlemail dot com> 2011-10-25 18:08:01 UTC ---
(In reply to comment #7)
> The arrow operator (vs, eg, +) seems also essential.

That makes sense to me, because the code could never be valid, so I would
suggest that the keyword is changed to ice-on-invalid-code. The proper code

namespace impl {
  template <class T> T create();
}

template < class lhs, class rhs >
struct is_arrow_operable_impl
{
  template <class T, class U,
        class = decltype(impl::create<T>()->*impl::create<U>())>
  void test();
};


works as expected. I guess the reason for the ICE of the original code is that
the compiler runs into a branch where the second impl::create is interpreted as
a member access, which could never be valid, because there is no (base) class
impl. 

But this is probably only part of the whole story, because if I change the
erroneous example to

struct impl {
  template <class T> static T create();
};

template < class lhs, class rhs >
struct is_arrow_operable_impl
{
  template <class T, class U,
        class = decltype(impl::create<T>()->impl::create<U>())
  >
  void test();
};


I'm getting "internal compiler error: Segmentation fault" without further
context information. Note that *this* example *can* lead to valid code when
attempting to instantiate is_arrow_operable_impl with impl* and int,
respectively, for example.

[Bug c++/50864] [4.6/4.7 Regression] ICE with decltype and "declval" from another namespace

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|ice-on-valid-code           |ice-on-invalid-code

--- Comment #9 from Paolo Carlini <paolo.carlini at oracle dot com> 2011-10-25 18:25:38 UTC ---
Thanks Daniel. Indeed, I was noticing strange things having to do with members
;) Let's move this back to ice-on-invalid. 

Now, however, in the light of the second half of your message, I guess we
should have another PR for the ICE on valid issue. Are you willing to open it?

[Bug c++/50864] [4.6/4.7 Regression] ICE with decltype and "declval" from another namespace

[daniel.kruegler at googlemail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

--- Comment #10 from Daniel Krügler <daniel.kruegler at googlemail dot com> 2011-10-25 18:28:56 UTC ---
(In reply to comment #9)
> Now, however, in the light of the second half of your message, I guess we
> should have another PR for the ICE on valid issue. Are you willing to open it?

I will do that

[Bug c++/50864] [4.6/4.7 Regression] ICE with decltype and "declval" from another namespace

[daniel.kruegler at googlemail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

--- Comment #11 from Daniel Krügler <daniel.kruegler at googlemail dot com> 2011-10-25 19:17:51 UTC ---
(In reply to comment #10)
See bug 50870

[Bug c++/50864] [4.6/4.7 Regression] ICE with decltype and "declval" from another namespace

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |paolo.carlini at oracle dot
                   |gnu.org                     |com

[Bug c++/50864] [4.6/4.7 Regression] ICE with decltype and "declval" from another namespace

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.6.3

[Bug c++/50864] [4.6/4.7 Regression] ICE with decltype and "declval" from another namespace

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2

[Bug c++/50864] [4.6/4.7 Regression] ICE with decltype and "declval" from another namespace

[paolo at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

--- Comment #12 from paolo at gcc dot gnu.org <paolo at gcc dot gnu.org> 2011-10-28 18:40:29 UTC ---
Author: paolo
Date: Fri Oct 28 18:40:22 2011
New Revision: 180623

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=180623
Log:
/cp
2011-10-28  Paolo Carlini  <paolo.carlini@oracle.com>

    PR c++/50864
    * pt.c (tsubst_copy_and_build): Fix qualified_name_lookup_error
    call in case COMPONENT_REF.

/testsuite
2011-10-28  Paolo Carlini  <paolo.carlini@oracle.com>

    PR c++/50864
    * g++.dg/template/crash109.C: New.

Added:
    trunk/gcc/testsuite/g++.dg/template/crash109.C
Modified:
    trunk/gcc/cp/ChangeLog
    trunk/gcc/cp/pt.c
    trunk/gcc/testsuite/ChangeLog

[Bug c++/50864] [4.6 Regression] ICE with decltype and "declval" from another namespace

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[4.6/4.7 Regression] ICE    |[4.6 Regression] ICE with
                   |with decltype and "declval" |decltype and "declval" from
                   |from another namespace      |another namespace

--- Comment #13 from Paolo Carlini <paolo.carlini at oracle dot com> 2011-10-28 18:41:49 UTC ---
Fixed in mainline so far.

[Bug c++/50864] [4.6/4.7 Regression] ICE with decltype and "declval" from another namespace

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[4.6 Regression] ICE with   |[4.6/4.7 Regression] ICE
                   |decltype and "declval" from |with decltype and "declval"
                   |another namespace           |from another namespace

--- Comment #14 from Paolo Carlini <paolo.carlini at oracle dot com> 2011-10-29 00:08:38 UTC ---
Fix reverted, unfortunately.

[Bug c++/50864] [4.6/4.7 Regression] ICE with decltype and "declval" from another namespace

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |NEW
         AssignedTo|paolo.carlini at oracle dot |unassigned at gcc dot
                   |com                         |gnu.org

--- Comment #15 from Paolo Carlini <paolo.carlini at oracle dot com> 2011-10-29 00:31:58 UTC ---
Unassigning: a lot is going on in case COMPONENT_REF of tsubst_copy_and_build
which I don't really understand at the moment.

[Bug c++/50864] [4.6/4.7 Regression] ICE with decltype and "declval" from another namespace

[paolo at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

--- Comment #16 from paolo at gcc dot gnu.org <paolo at gcc dot gnu.org> 2011-11-08 10:23:57 UTC ---
Author: paolo
Date: Tue Nov  8 10:23:53 2011
New Revision: 181151

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=181151
Log:
/cp
2011-11-08  Paolo Carlini  <paolo.carlini@oracle.com>

    PR c++/50864
    * parser.c (cp_parser_postfix_dot_deref_expression): Reject invalid
    uses of '->' and '.' as postfix-expression in namespace scope.

/testsuite
2011-11-08  Paolo Carlini  <paolo.carlini@oracle.com>

    PR c++/50864
    * g++.dg/parse/template26.C: New.
    * g++.dg/template/crash45.C: Adjust dg-error string.

Added:
    trunk/gcc/testsuite/g++.dg/parse/template26.C
Modified:
    trunk/gcc/cp/ChangeLog
    trunk/gcc/cp/parser.c
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/testsuite/g++.dg/template/crash45.C

[Bug c++/50864] [4.6/4.7 Regression] ICE with decltype and "declval" from another namespace

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50864

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

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

--- Comment #17 from Paolo Carlini <paolo.carlini at oracle dot com> 2011-11-08 10:33:13 UTC ---
Now in mainline a clear error message is produced by the parser. In the branch,
the fix for 50870 fixed the problems at template instantiation time.