scottmcm

> We don't use the byval attribute, and this seems to be by design. I'm curious about why; any chance you have a link handy to a discussion about it?

This seems like a policy decision to me, so I'm going to flip it over to r? @joshtriplett to have a team discussion.

I'm pretty sure this'll be fine, but just in case since it's updating sip128: @bors try @rust-timer queue I'd like to get an official team opinion on direction here before...

(Dup of https://github.com/belluzj/fantasque-sans/issues/122)

Pointers have `cast_mut` and `cast_const` to go back and forth between `*const T` and `*mut T` without the possibility of accidentally changing the pointee. Having `cast_signed` and `cast_unsigned` for integers...

I think part of this got approved in https://github.com/rust-lang/libs-team/issues/359#issuecomment-2033209931 as `cast_(un)signed`? Still worth continuing the discussion here for the other parts.

Linking another conversation: - https://github.com/rust-lang/rust/pull/110702#issuecomment-1518887370

I think we should do this regardless of it can can be done with casts, like how we have .

> `use impl Trait` is also more greppable than `impl use Trait`. Hmm, do we have `impl for` already?

> For instance, I think this PR would annotate every single method of `String` with `1.0.0`, right? I wonder if that gives another alternative -- rather than "omitted is same...