Adrian Taylor

Thanks. I wasn't quite sure. On the one hand, manipulating the returned pointer could break the contents of the `UniquePtr` (so it's unsafe in that sense) but on the other...

Thanks for working on this!

Yep, I'll take a look.

I added a design sketch to #682 - please take a look. It looks OK, but non-trivial, so unlikely I'll get to it especially soon. Maybe though.

There's a comment here with more thoughts about feature detection - https://github.com/dtolnay/cxx/pull/117#issuecomment-613699681

I've been playing around with this [here](https://github.com/adetaylor/nonconst-reference-wrapper/blob/main/src/main.rs) (implementation [here](https://github.com/adetaylor/nonconst-reference-wrapper/blob/main/src/reference_wrapper.rs)). The ergonomics seem OK - no worse than `Pin` which it would be replacing. Unless I'm missing something. More things to...

Should be fixed by #874.

Thanks, I agree! Maybe with a comment too. It would be great if you could raise a pull request to make this change :)

Hey @dtolnay @rodrigodesalvobraz sorry, I messed up here! I thought this was a concern about `autocxx` which is why I responded and said I thought this was a good idea....

There's been some work towards this in #80 and #410.