https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117046
--- Comment #7 from Carlos Galvez <carlosgalvezp at gmail dot com> --- Actually in the patch that would address this issue for Clang (https://reviews.llvm.org/D53860), they mentioned 2277: https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2018/p0968r0.html#2227 With emphasis: > The destructor for each element of class type is potentially invoked (15.4 > [class.dtor]) from the context where the aggregate initialization occurs. Some more comments from the patch: > I've checked the minutes, and we did indeed discuss that when resolving C++ > DR 2227, and decided that the context for the access check should be the > context of the aggregate initialization, not the context of the class > definition. This makes sense with my very limited knowledge, but I'd like to hear your opinion on 2277. It does feel like a separate issue than 2244.
