[Bug c++/83778] [8 regression] g++.dg/ext/altivec-cell-2.C fails starting with r256448

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83778

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

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

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Minimal reproducer:
int f4 (__attribute__((altivec(vector__))) int a)
{
  return __builtin_vec_extract (a, 4);
}
./cc1plus -quiet -std=gnu++14 -maltivec pr83778.C -isystem include/
-mcpu=power8

The difference is that before r256448 rs6000_resolve_overloaded_builtin was
called with INTEGER_CST in the arg2, but now with NON_LVALUE_EXPR wrapper
around it.