[Bug target/79671] [7 Regression] mapnik miscompilation on armv7hl since r235622

["bernd.edlinger at hotmail dot de" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79671

--- Comment #88 from Bernd Edlinger <bernd.edlinger at hotmail dot de> ---
(In reply to rguenther@suse.de from comment #87)
> On April 1, 2017 4:40:19 PM GMT+02:00, "bernd.edlinger at hotmail dot de"
> <gcc-bugzilla@gcc.gnu.org> wrote:
> >https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79671
> >
> >--- Comment #86 from Bernd Edlinger <bernd.edlinger at hotmail dot de>
> >---
> >Created attachment 41101 [details]
> >  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=41101&action=edit
> >another trial patch
> >
> >This is another approach for a patch that reflects what I understood
> >so far what the C++17 spec wants from us.
> >I think it should resolve the issue Richard raised.
> 
> That doesn't properly work with LTO

oops...
I think this does not work either,
at cp/decl.c (start_enum):

          /* std::byte aliases anything.  */
          if (enumtype != error_mark_node
              && TYPE_CONTEXT (enumtype) == std_node
              && !strcmp ("byte", TYPE_NAME_STRING (enumtype)))
            TYPE_ALIAS_SET (enumtype) = 0;

I looked at -flto -fdump-rtl-final
and see the alias set is 1:

#include <cstddef>
//namespace std { enum class byte: unsigned char {}; };
union  t  {
  int x;
  std::byte k[1];
};
t t1;
t t2;
void test()
{
  t1.k[0] = t2.k[0];
}
int main()
{
  test();
}

g++ -std=c++17 -flto -fdump-rtl-final t.cc

/tmp/cc*.final:
(insn 5 2 6 2 (set (reg:QI 0 ax [orig:87 _1 ] [87])
        (mem/j/c:QI (symbol_ref:DI ("t2") [flags 0x2]  <var_decl 0x7f2a492dfab0
t2>) [1 t2.k+0 S1 A32])) "t.cc":11 84 {*movqi_internal}
     (nil))
(insn 6 5 13 2 (set (mem/j/c:QI (symbol_ref:DI ("t1") [flags 0x2]  <var_decl
0x7f2a492dfb40 t1>) [1 t1.k+0 S1 A32])
        (reg:QI 0 ax [orig:87 _1 ] [87])) "t.cc":11 84 {*movqi_internal}
     (nil))


[1 = alias set 1 but should be 0