saar.raz added inline comments.
================
Comment at: lib/Sema/SemaDecl.cpp:8377-8381
+ } else if (D.hasTrailingRequiresClause()) {
+ // C++2a [class.virtual]p6
+ // A virtual method shall not have a requires-clause.
+ Diag(NewFD->getTrailingRequiresClause()->getLocStart(),
+ diag::err_constrained_virtual_method);
----------------
rsmith wrote:
> This is the wrong place for this check. We don't yet know whether the
> function is virtual here in general. A function can become virtual due to
> template instantiation:
>
> ```
> template<typename T> struct A : T { void f() requires true; };
> struct B { virtual void f(); };
> template struct A<B>; // error, A<B>::f is constrained and virtual
> ```
>
> This is perhaps a wording defect: it's not clear that `A::f()` really should
> override `B::f()`, but that is the consequence of the current rules. I've
> posted a question to the core reflector.
I don't really see why A::f() should override B::f() indeed - since it is not
marked virtual nor override, shouldn't it just hide B::f()? or am I missing
something here?
Repository:
rC Clang
https://reviews.llvm.org/D43357
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
