[Bug c++/56782] New: [C++11] Regression with empty pack expansions

[Bug c++/56782] New: [C++11] Regression with empty pack expansions

[daniel.kruegler at googlemail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56782

             Bug #: 56782
           Summary: [C++11] Regression with empty pack expansions
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: daniel.kruegler@googlemail.com


As of gcc 4.8.0 the following code is now rejected when compiled with flags

-std=c++11 -Wall -W -pedantic

//---------------------------
template<class T>
T&& declval();

struct is_convertible_impl {
  template<class T>
  static void sink(T);

  template<class T, class U, class = decltype(sink<U>(declval<T>()))>
  static auto test(int) -> char;

  template<class, class>
  static auto test(...) -> char(&)[2];
};

template<class T, class U>
struct is_convertible : is_convertible_impl
{
  static const bool value = sizeof(test<T, U>(0)) == 1;
};

template<bool, class>
struct enable_if {};

template<class T>
struct enable_if<true, T> { typedef T type; };

template<bool, class If, class Else>
struct conditional { typedef If type; };

template<class If, class Else>
struct conditional<false, If, Else> { typedef Else type; };

template<class...>
struct and_;

template<>
struct and_<>
{
  static const bool value = true;
};

template<class P>
struct and_<P> : P
{
};

template<class P1, class P2>
struct and_<P1, P2> : conditional<P1::value, P2, P1>::type
{
};

template<class... T>
struct Tuple {
  template<class... U,
    class = typename enable_if<and_<
      is_convertible<U, T>...
    >::value, int>::type
  >
  Tuple(U&&...){}
};

static_assert(is_convertible<Tuple<int>, Tuple<int>>::value, "Ouch"); // OK

static_assert(is_convertible<Tuple<>, Tuple<>>::value, "Ouch"); // Error
//---------------------------

The diagnostics being:

"
Compilation finished with errors:
source.cpp:18:48: error: template instantiation depth exceeds maximum of 900
(use -ftemplate-depth= to increase the maximum) substituting 'template<class,
class> static char (& is_convertible_impl::test(...))[2] [with
<template-parameter-1-1> = Tuple<>; <template-parameter-1-2> = Tuple<>]'
static const bool value = sizeof(test<T, U>(0)) == 1;
^
source.cpp:55:5: recursively required from 'const bool is_convertible<Tuple<>,
Tuple<> >::value'
source.cpp:55:5: required from 'const bool is_convertible<Tuple<>, Tuple<>
>::value'
source.cpp:64:49: required from here

source.cpp:18:48: error: no matching function for call to
'is_convertible<Tuple<>, Tuple<> >::test(int)'
source.cpp:18:48: note: candidates are:
source.cpp:9:15: note: template<class T, class U, class> static char
is_convertible_impl::test(int)
static auto test(int) -> char;
^
source.cpp:9:15: note: template argument deduction/substitution failed:
source.cpp:12:15: note: template<class, class> static char (&
is_convertible_impl::test(...))[2]
static auto test(...) -> char(&)[2];
^
source.cpp:12:15: note: substitution of deduced template arguments resulted in
errors seen above
source.cpp:64:1: error: non-constant condition for static assertion
static_assert(is_convertible<Tuple<>, Tuple<>>::value, "Ouch"); // Error
^
source.cpp:64:1: error: the value of 'is_convertible<Tuple<>, Tuple<> >::value'
is not usable in a constant expression
source.cpp:18:21: note: 'is_convertible<Tuple<>, Tuple<> >::value' was not
initialized with a constant expression
static const bool value = sizeof(test<T, U>(0)) == 1;
^
"

The code is accepted with gcc 4.7.2 and with Clang 3.2.

It seems that for empty expansions the compiler erroneously does enter into the
actually empty expansion:

is_convertible<U, T>...

[Bug c++/56782] [4.8/4.9 Regression] Regression with empty pack expansions

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56782

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-03-29
                 CC|                            |jason at gcc dot gnu.org
            Summary|[C++11] Regression with     |[4.8/4.9 Regression]
                   |empty pack expansions       |Regression with empty pack
                   |                            |expansions
     Ever Confirmed|0                           |1

--- Comment #1 from Paolo Carlini <paolo.carlini at oracle dot com> 2013-03-29 13:51:01 UTC ---
Confirmed.

[Bug c++/56782] [4.8/4.9 Regression] Regression with empty pack expansions

[jason at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56782

Jason Merrill <jason at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dodji at gcc dot gnu.org
         Depends on|                            |53609

--- Comment #2 from Jason Merrill <jason at gcc dot gnu.org> 2013-03-29 17:59:18 UTC ---
This seems to be an issue with Dodji's patch for bug 53609; we aren't using
PACK_EXPANSION_EXTRA_ARGS for the partial instantiation of the tuple
constructor, but we should be.

[Bug c++/56782] [4.8/4.9 Regression] Regression with empty pack expansions

[dodji at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56782

Dodji Seketeli <dodji at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |dodji at gcc dot gnu.org

[Bug c++/56782] [4.8/4.9 Regression] Regression with empty pack expansions

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56782

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.8.1

[Bug c++/56782] [4.8/4.9 Regression] Regression with empty pack expansions

[dodji at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56782

--- Comment #3 from Dodji Seketeli <dodji at gcc dot gnu.org> ---
A candidate patch was sent to
http://gcc.gnu.org/ml/gcc-patches/2013-05/msg00841.html for this.