[Bug c++/54319] New: Assignment to rvalue

[Bug c++/54319] New: Assignment to rvalue

[aschepler at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54319

             Bug #: 54319
           Summary: Assignment to rvalue
    Classification: Unclassified
           Product: gcc
           Version: 4.7.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: aschepler@gmail.com


g++ seems to think "A = B" is a prvalue if expression A is a prvalue of a class
type with no members.  Example code:

-----

template<typename T, typename U>
struct is_same {
  static const bool value = false;
};

template<typename T>
struct is_same<T,T> {
  static const bool value = true;
};

struct X {};

struct Y { int m; };

void f() {
  static_assert(is_same<decltype((Y{} = Y{})), Y&>::value,
                "Y assignment does not result in lvalue");
  static_assert(is_same<decltype((X{} = X{})), X&>::value,
                "X assignment does not result in lvalue");
}

-----

g++ -std=c++0x -c rval_assign.cpp
rval_assign3.cpp: In function ‘void f()’:
rval_assign3.cpp:18:3: error: static assertion failed: X assignment does not
result in lvalue

-----

The difference in behavior between X and Y is surprising, since both are POD.

Although assigning to a prvalue is questionable practice, it's allowed in this
case and the assignment expression is an lvalue because of 12.8 paragraph 22:

  The implicitly-declared move assignment operator for a class X will have the
form
  X& X::operator=(X&&);

-----

g++ -v
Using built-in specs.
COLLECT_GCC=g++
COLLECT_LTO_WRAPPER=/usr/lib/gcc/i486-linux-gnu/4.7/lto-wrapper
Target: i486-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Debian 4.7.1-2'
--with-bugurl=file:///usr/share/doc/gcc-4.7/README.Bugs
--enable-languages=c,c++,go,fortran,objc,obj-c++ --prefix=/usr
--program-suffix=-4.7 --enable-shared --enable-linker-build-id
--with-system-zlib --libexecdir=/usr/lib --without-included-gettext
--enable-threads=posix --with-gxx-include-dir=/usr/include/c++/4.7
--libdir=/usr/lib --enable-nls --with-sysroot=/ --enable-clocale=gnu
--enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-gnu-unique-object
--enable-plugin --enable-objc-gc --enable-targets=all --with-arch-32=i586
--with-tune=generic --enable-checking=release --build=i486-linux-gnu
--host=i486-linux-gnu --target=i486-linux-gnu
Thread model: posix
gcc version 4.7.1 (Debian 4.7.1-2)

[Bug c++/54319] Assignment to rvalue

[daniel.kruegler at googlemail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54319

Daniel Krügler <daniel.kruegler at googlemail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |daniel.kruegler at
                   |                            |googlemail dot com

--- Comment #1 from Daniel Krügler <daniel.kruegler at googlemail dot com> 2012-08-19 07:30:30 UTC ---
Just for the record: This problem does no longer exist in gcc 4.8 (tested for 
4.8.0 20120729 (experimental))

[Bug c++/54319] Assignment to rvalue

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54319

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|aschepler at gmail dot com  |

--- Comment #2 from Paolo Carlini <paolo.carlini at oracle dot com> 2012-08-19 11:55:16 UTC ---
Daniel can you double check? Did we regress between the end of July and now or
what? Because with today's mainline the static_assert for X fires for me.

[Bug c++/54319] Assignment to rvalue

[daniel.kruegler at googlemail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54319

--- Comment #3 from Daniel Krügler <daniel.kruegler at googlemail dot com> 2012-08-19 12:13:32 UTC ---
(In reply to comment #2)
> Daniel can you double check? 

Good that you ask. There must by some problem with my gcc installation, because
I get different results from different contexts calling gcc. I can confirm that
gcc 4.8 invoked from my command line also fires the static assertion. I
apologize for any confusion.

[Bug c++/54319] Assignment to rvalue

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54319

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-08-19
     Ever Confirmed|0                           |1

--- Comment #4 from Paolo Carlini <paolo.carlini at oracle dot com> 2012-08-19 12:43:25 UTC ---
No problem, let's confirm these two bugs and, hey, privately maybe let me know
how it goes, maybe we can ask the help of a mingw maintainer, or something.

[Bug c++/54319] empty class causes error while non-empty does not

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54319

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |rejects-valid
            Summary|Assignment to rvalue        |empty class causes error
                   |                            |while non-empty does not

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Note even the following X causes an error:
struct X {int :1;};

Maybe that will give a hint of where the bug is located.

[Bug c++/54319] [9/10/11/12 Regression] empty class causes error while non-empty does not

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54319

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|empty class causes error    |[9/10/11/12 Regression]
                   |while non-empty does not    |empty class causes error
                   |                            |while non-empty does not
   Target Milestone|---                         |9.5

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #5)
> Note even the following X causes an error:
> struct X {int :1;};
> 
> Maybe that will give a hint of where the bug is located.

Which makes this a regression even :).
The above with just one anonymous bitfield field is accepted until GCC 6.

But this should give even more of what is the underlaying bug.

[Bug c++/54319] [9/10/11/12 Regression] empty class causes error while non-empty does not

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54319

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|                            |4.4.7

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #6)
> (In reply to Andrew Pinski from comment #5)
> > Note even the following X causes an error:
> > struct X {int :1;};
> > 
> > Maybe that will give a hint of where the bug is located.
> 
> Which makes this a regression even :).
> The above with just one anonymous bitfield field is accepted until GCC 6.
> 
> But this should give even more of what is the underlaying bug.

r6-7811 exposed the latent bug for the anonymous bitfield case (which is
expected).

The original testcase worked in 4.4.7 also.

[Bug c++/54319] [9/10/11/12 Regression] empty class causes error while non-empty does not

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54319

--- Comment #8 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I think the problem is here in init.c:
      /* If the type has data but no user-provided default ctor, we need to
zero
         out the object.  */
      if (type_has_non_user_provided_default_constructor (type)
          && !is_really_empty_class (type, /*ignore_vptr*/true))
        {
          tree field_size = NULL_TREE;
          if (exp != true_exp && CLASSTYPE_AS_BASE (type) != type)
            /* Don't clobber already initialized virtual bases.  */
            field_size = TYPE_SIZE (CLASSTYPE_AS_BASE (type));
          init = build_zero_init_1 (type, NULL_TREE,
/*static_storage_p=*/false,
                                    field_size);
          init = build2 (INIT_EXPR, type, exp, init);
          finish_expr_stmt (init);
        }

[Bug c++/54319] [9/10/11/12 Regression] empty class causes error while non-empty does not

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54319

--- Comment #9 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #8)
> I think the problem is here in init.c:
>       /* If the type has data but no user-provided default ctor, we need to
> zero
>          out the object.  */
>       if (type_has_non_user_provided_default_constructor (type)
>           && !is_really_empty_class (type, /*ignore_vptr*/true))
>         {
>           tree field_size = NULL_TREE;
>           if (exp != true_exp && CLASSTYPE_AS_BASE (type) != type)
>             /* Don't clobber already initialized virtual bases.  */
>             field_size = TYPE_SIZE (CLASSTYPE_AS_BASE (type));
>           init = build_zero_init_1 (type, NULL_TREE,
> /*static_storage_p=*/false,
>                                     field_size);
>           init = build2 (INIT_EXPR, type, exp, init);
>           finish_expr_stmt (init);
>         }

Nope.

[Bug c++/54319] [9/10/11/12 Regression] empty class causes error while non-empty does not

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54319

--- Comment #10 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
It is the code in call.c that makes a difference:
      if (is_really_empty_class (type, /*ignore_vptr*/true))
        {
          /* Avoid copying empty classes.  */
          val = build2 (COMPOUND_EXPR, type, arg, to);
          suppress_warning (val, OPT_Wunused);
        }
      else if (tree_int_cst_equal (TYPE_SIZE (type), TYPE_SIZE (as_base)))
        {
          if (is_std_init_list (type)
              && conv_binds_ref_to_prvalue (convs[1]))
            warning_at (loc, OPT_Winit_list_lifetime,
                        "assignment from temporary %<initializer_list%> does "
                        "not extend the lifetime of the underlying array");
          arg = cp_build_fold_indirect_ref (arg);
          val = build2 (MODIFY_EXPR, TREE_TYPE (to), to, arg);
        }
      else
        {
          /* We must only copy the non-tail padding parts.  */
          tree arg0, arg2, t;
          tree array_type, alias_set;

          arg2 = TYPE_SIZE_UNIT (as_base);
          to = cp_stabilize_reference (to);
          arg0 = cp_build_addr_expr (to, complain);

          array_type = build_array_type (unsigned_char_type_node,
                                         build_index_type
                                           (size_binop (MINUS_EXPR,
                                                        arg2, size_int (1))));
          alias_set = build_int_cst (build_pointer_type (type), 0);
          t = build2 (MODIFY_EXPR, void_type_node,
                      build2 (MEM_REF, array_type, arg0, alias_set),
                      build2 (MEM_REF, array_type, arg, alias_set));
          val = build2 (COMPOUND_EXPR, TREE_TYPE (to), t, to);
          suppress_warning (val, OPT_Wunused);
        }

if we do the middle one only, we get the correct type.
I won't be able to debug this any further. But I suspect because we are
lowering the AST too early, we don't do the correct type dealing here due to
the COMPOUND_EXPR.

[Bug c++/54319] [9/10/11/12 Regression] empty class causes error while non-empty does not

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54319

--- Comment #11 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
For the original testcase, I think r0-98505 introduce the failure.

[Bug c++/54319] [9/10/11/12 Regression] empty class causes error while non-empty does not

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54319

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ihorelo at mail dot ru

--- Comment #12 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
*** Bug 86032 has been marked as a duplicate of this bug. ***

[Bug c++/54319] [9/10/11/12 Regression] empty class causes error while non-empty does not

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54319

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |accepts-invalid, wrong-code

--- Comment #13 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Here is a run time testcase of the same issue from PR 86032:
struct Empty
{
        Empty& operator=(const Empty&) = default;
};

template<typename T>
class ShowType;

template<>
struct ShowType<Empty&>{int t = 0;};
template<>
struct ShowType<Empty>{int t = 1;};

int main()
{
  ShowType<decltype(Empty() = Empty())> example;
  return example.t;
}

[Bug c++/54319] [9/10/11/12 Regression] empty class causes error while non-empty does not

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54319

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

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

[Bug c++/54319] [9/10/11/12 Regression] empty class causes error while non-empty does not

[ppalka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54319

Patrick Palka <ppalka at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ppalka at gcc dot gnu.org
             Status|NEW                         |RESOLVED
         Resolution|---                         |DUPLICATE

--- Comment #14 from Patrick Palka <ppalka at gcc dot gnu.org> ---
Looks like this is essentially a dup of the recently fixed on trunk PR59950.

*** This bug has been marked as a duplicate of bug 59950 ***