Grey

I've immediately realized a problem with my last proposal: a `Box` interface wouldn't be able to support generic methods (e.g. `fn publish`) which would make it extremely difficult to implement...

I'll need to update the unit tests for the parameter services to match the new behavior.

I just ran into this myself. It's disappointing to not be able to use `flatten` with RON. Has there been any consideration for allowing RON's map syntax `{ }` to...

Thanks for pointing that out, I didn't think of trying the latest `master`. That does exactly what I was proposing, so that's great. I did run into a separate bug...

I've been planning on doing something similar to this, but I think we can take it even further and make it so that rclrs can infer the service type without...

On a separate note, I strongly urge us to keep the signature of services as `Fn` and not `FnMut` because we will not be able to support `FnMut` for unconstrained...

After more thought, I've decided that my previous suggestion that involved introducing more traits wouldn't accomplish much. But I do think we should bundle everything into one argument instead of...

@luca-della-vedova That's a good point. Since Noble is an LTS we should expect that the build farm will be targeting it for quite a while now, and any build farm...

It turns out the FFI issue is thornier than we originally realized. The real origin of the problem is that [`rosidl_dynamic_typesupport`](https://github.com/ros2/rosidl_dynamic_typesupport/tree/rolling) introduces [float128](https://github.com/ros2/rosidl_dynamic_typesupport/blob/96d1cb662b1b69409d80383ed20a41cc92267aa1/include/rosidl_dynamic_typesupport/api/serialization_support_interface.h#L1006-L1010) support, even though float128 is not a...

After doing even more research on long double, I've reached the conclusion that bindgen **shouldn't** support interpreting long doubles from a C API, now or ever, because the binary representation...