Vadim Petrochenkov

I think my preferred solution would be to find some non-keyword identifier (e.g. `unsafe_attr`) to avoid the syntactic can of worms. As for expansion, my preference would be an active...

@RalfJung >that's a whole separate language anyway, isn't it? Why is the keyword a problem? Not at all, the syntax is `#[PATH(TOKENS)]`/`#[PATH[TOKENS]]`/`#[PATH{TOKENS}]`/`#[PATH]`, and only the `TOKENS` part is the whole...

>I don't think `unsafe` expands do anything? It's simply decorating the attribute inside it, e.g. `unsafe(no_mangle)`. `no_mangle` doesn't expand to anything either, does it? I am very confused by this...

@rfcbot concern field names are unnecessary (I guess it's a trade off between symbol length and readability of demangled symbols.) @rfcbot concern unit structs are unnecessary

@rfcbot resolve field names are unnecessary @rfcbot resolve unit structs are unnecessary

Dunno. You can do this in C++, but it's usually frowned upon ([example](https://google.github.io/styleguide/cppguide.html#auto)) Maybe type inference should be allowed, but only on private constants/statics. (I had one weaker suggestion in...

I suspect that in a typical case you do want to pass the created or inherited jobserver to children processes. In all cases caught by https://github.com/rust-lang/rust/pull/113730 the right solution was...

https://github.com/rust-lang/jobserver-rs/pull/65 is a draft PR that reverses the current behavior - jobserver is inherited by default, but the inheritance can be disabled by explicitly calling a method.

> I'm not sure how this compares with #119412. This PR changes representation of `NtIdent`, #119412 removes it entirely.

https://github.com/rust-lang/rust/pull/119412 is viable, I just didn't have time to address the regressions (I more or less know how to do that). >It changes a lot of test outputs in ways...