erichkeane wrote:
> I think the interesting template cases are like:
>
> ```
> template<typename T> struct S {
> T foo();
> };
>
> S<int> a;
> S<SWIFT_SENDABLE int> b;
>
> void f() {
> a.foo();
> b.foo();
> }
> ```
>
> The question is, whether we actually have the type attributes where we expect
> to have them, or do we lose them? Hopefully @erichkeane can correct me if I'm
> wrong.
>
> That being said, one question is whether we actually need type attributes to
> be propagated in this case for Swift's interop.
>
> For the release notes, you'd want to mention that `swift_attr` can now be
> applied to types (maybe a small motivation why) in
> `clang/docs/ReleaseNotes.rst`. It has a section for attribute changes.
You're correct, the problem with attributed type is that it is going to be lost
A LOT every time we canonicalize the type, else they need to be different
types, and thus those would be different instantiations. This is SOMETIMES not
problematic (and sometimes even wanted, see vector types), but it IS
problematic if you want your attribute to not cause problems being used in
containers/etc, since `std::is_same<int, MY_INT_ATTRIBUTED_TYPE>` being 'false'
is problematic for many use cases.
https://github.com/llvm/llvm-project/pull/108631
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
