debuginfo: Recursive types and some trait object pointer support

[michaelwoerister]

This pull request finally adds support for recursive type definitions and provides a stub implementation for object pointers.

[jdm]

This looks good to me. Let's just see if we can avoid using @fn, since their future is in doubt.

[michaelwoerister]

Let's just see if we can avoid using @fn, since their future is in doubt.

I've tried that but we can't use an &fn here because the struct containing the closure outlives the allocating stack frame. I tried using a ~fn but those can't capture span values (which contain managed pointers internally). Do you know of a way to the tell compiler that the ~fn does not need to be sendable?

All that being said, I think this will need a refactoring within the next few weeks anyway---so it might be OK to leave the @fn as is for now and get rid of it later as part of a bigger change.

[huonw]

It's possible to specific built-in trait bounds on a fn's environment with a colon: i.e. ~fn: (no bounds), ~fn:Send (the default), etc.

However, it still won't work, since ~fn: capturing @ makes things unhappy: #8983.

[jdm]

Very well. Fix up the nits and this is good to go!

[michaelwoerister]

OK, I fixed the style issues and did a rebase.

@huonw Thanks for hint :)

[michaelwoerister]

Sorry about that. Retry please...

[jdm]

Heh, every time. We should write a script that generates that list from the header file or something.

[michaelwoerister]

Yeah :) I had a post-it on my monitor once about the xfail-win32 stuff, which did actually help...