https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93210
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |jakub at gcc dot gnu.org
--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
>From C++ POV, Should when passed by reference to the operator== is odr-used and
therefore the static data member definition is required (unless C++17 or newer
where it is inline variable and the definition is already in the class
definition).
The rest is just a matter of optimization, where we for yet to be determined
reason punt in the anonymous struct in union case, as can be seen on C testcase
(but likewise with C++):
union U { int a[2]; long long b; };
static const union U u = { .a = { 1, 2 } };
union V { struct { int a, b; }; long long c; };
static const union V v = { { 1, 2 } };
void qux (const union U *, const union V *);
long long
foo (void)
{
if (sizeof (long long) == 2 * sizeof (int))
return u.b;
else
return -1;
}
long long
bar (void)
{
if (sizeof (long long) == 2 * sizeof (int))
return v.c;
else
return -1;
}
void
baz (void)
{
qux (&u, &v);
}
