On 12/18/25 1:34 AM, Jakub Jelinek wrote:
On Tue, Dec 16, 2025 at 10:48:40PM +0700, Jason Merrill wrote:
+ else if (VAR_P (r) && DECL_ANON_UNION_VAR_P (r))
+ return boolean_true_node;
This surprises me, I'd think that an anon-union variable is equivalent to a
member-access expression, but is not a class member itself, but perhaps it
makes sense for reflection to blur that distinction. It would be nice if
the standard were clear either way.
The artificial variable with the union type can be reflected through
members_of on the containing namespace. DECL_ANON_UNION_VAR_P is not that
through, it is what one gets from say ^^a after
static union { int a; long b; };
and that isn't a variable. Member-access expressions aren't one of the
https://eel.is/c++draft/basic.fundamental#17 categories, except for object
but I believe the only way to get an object is reflect_object.
And GCC agrees with clang++ on that, haven't tried EDG.
https://godbolt.org/z/Y36418a7x
Standard clarification would be appreciated.
Ah, this seems to be addressed by https://eel.is/c++draft/expr#reflect-5.8
"if lookup finds a class member of an anonymous union
([class.union.anon]), R represents that class member."
Which sounds like we should have looked through the variable to the
member in cp_parser_reflection_name.
Jason
