kimgr added a comment. @aaron.ballman Thanks for the valuable feedback!
> Assuming that the `to_lvalue_ref(A{})` case should diagnose, then yes, I > agree, I think the `ret_a()` portion should as well. My primary focus is not to make sure Clang follows the standard closely for `consteval`, but rather to make sure it's self-consistent and doesn't provoke undefined behavior for `consteval` inputs. So another way to think about it is that `to_lvalue_ref` and `ret_a` might both diagnose, or not, but absence of diagnostics should not be a result of misinterpreting the AST the way we do now. I'm preparing a different patch to do the short-circuiting in `skipImplicitTemporary`, so we can more easily compare and contrast. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D117391/new/ https://reviews.llvm.org/D117391 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits