round two preliminaries for bumping nightly to 2023-08-25 #33064
Conversation
seems to be something related to `const ref`s in generic contexts...
| // | ||
| // allow lint false positive trait bound requirement (`CryptoRng` only | ||
| // implemented on `&'a mut T` | ||
| #[rustversion::attr(since(1.73), allow(clippy::needless_pass_by_ref_mut))] |
There was a problem hiding this comment.
This doesn't seems like a false positive to me. The rng param seems wrong, or at least very weird, here.
It doesn't need to be mut rng: &'a mut R, it should be rng: &'a mut R and just passed as rng to Ping::new_rand(rng,...)`
Currently we're passing a &mut &mut R to Ping::new_rand which is probably immediately deref'd, but because we have &mut rng it's forcing us to have mut rng?
There was a problem hiding this comment.
Nevermind, I pulled the branch and even with those changes it still flags it falsely.
There was a problem hiding this comment.
It doesn't need to be
mut rng: &'a mut R,it should berng: &'a mut R
haha, yeah i tried that too 🙃
yeah i'm about to work on minimal repro to report upstream |
Problem
i left these off of #33047 since they might stir contention. the remainder of compatibility changes for bumping toolchain(s) need to go in with the bump as they reference lints that the current toolchain(s) don't know about
Summary of Changes
const refin generic contextclippy::needless_pass_by_ref_mutto avoid false-positive ignoring trait bounds