macros can close over their environment, “breaking” hygiene

playground

I’d like to give a link to documentation here, but I don’t think this is actually documented anywhere? neither reference nor TLBORM mention it as far as I can see

UPD: ferrocene specification

cyclic associated types are okay under lazy normalization / next solver
(you need ?Sized though, otherwise it tries to prove Assoc: Sized and fails)

godbolt
tracking issue

wrote down some thoughts about error handling, specifically a pattern that I call “progress pattern” (probably it has another name somewhere, but I’m not aware of it)
https://github.com/GoldsteinE/gh-blog/blob/master/error_progress/README.md
thanks to @meowaffle for reviewing this!

(btw, are y’all ok with breaking the format like this sometimes?)

my fix for this problem was merged and will be stable in rustc 1.81, unless something goes wrong
https://github.com/rust-lang/rust/pull/123043

GitHub

Disable dead variant removal for `#[repr(C)]` enums. by GoldsteinE · Pull Request #123043 · rust-lang/rust

This prevents removing dead branches from a #[repr(C)] enum (they now get discriminants allocated as if they were inhabited). Implementation notes: ABI of something like #[repr(C)] enum Foo { F...

found this one while thinking about size of uninhabited types. Ralf seems to agree it’s a bug, so the patch will probably be accepted
playground
UCG issue
proposed patch