https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88572
--- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> ---
The relevant code is:
/* It is invalid to initialize a non-aggregate type with a
brace-enclosed initializer before C++0x.
We need to check for BRACE_ENCLOSED_INITIALIZER_P here because
of g++.old-deja/g++.mike/p7626.C: a pointer-to-member constant is
a CONSTRUCTOR (with a record type). */
if (TREE_CODE (stripped_init) == CONSTRUCTOR
/* Don't complain about a capture-init. */
&& !CONSTRUCTOR_IS_DIRECT_INIT (stripped_init)
&& BRACE_ENCLOSED_INITIALIZER_P (stripped_init)) /* p7626.C */
{
if (SCALAR_TYPE_P (type))
{
if (cxx_dialect < cxx11
/* Isn't value-initialization. */
|| CONSTRUCTOR_NELTS (stripped_init) > 0)
{
if (complain & tf_error)
error ("braces around scalar initializer for type %qT",
type);
init = error_mark_node;
}
}
The condition means we give an error for any braced-init-list in C++98, and for
a non-empty braced-init-list in later dialects . But the latter condition is
wrong, it should allow a single element (and presumably more than one element
will be rejected as an invalid initializer anyway).
I'm not even sure we should warn about this.
