type_repetition_in_bounds should pay attention to msrv when ?Sized bound is involved

[dtolnay]

Lint Name

type_repetition_in_bounds

Reproducer

# .clippy.toml

msrv = "1.14.0"

// lib.rs

pub trait Trait<X, Y, Z> {}

pub fn f<T: ?Sized, U>(arg: usize)
where
    T: Trait<Option<usize>, Box<[String]>, bool> + 'static,
    U: Clone + Sync + 'static,
{}

error: this type has already been used as a bound predicate
 --> src/lib.rs:7:5
  |
7 |     T: Trait<Option<usize>, Box<[String]>, bool> + 'static,
  |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  |
  = note: `-D clippy::type-repetition-in-bounds` implied by `-D clippy::pedantic`
  = help: consider combining the bounds: `T: Sized + Trait<Option<usize>, Box<[String]>, bool>`
  = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#type_repetition_in_bounds

Clippy prefers I guess one of the following signatures instead:

pub fn f<T, U>(arg: usize)
where
    T: ?Sized + Trait<Option<usize>, Box<[String]>, bool> + 'static,
    U: Clone + Sync + 'static

// or:

pub fn f<T: ?Sized + Trait<Option<usize>, Box<[String]>, bool> + 'static, U>(arg: usize)
where
    U: Clone + Sync + 'static

The first one is great on modern compilers, but is a parse error on Rust 1.14 and older. Meanwhile, the second is not an improvement in my opinion. The bounds next to a type parameter (as opposed to in a where-clause) are only meant to be used for small/simple bounds.

I believe clippy should not trigger type_repetition_in_bounds if the msrv is declared as 1.14.0 or older and the type parameter has a single bound of ?Sized, with all the rest of the bounds being in the where-clause.

Version

rustc 1.62.0-nightly (7c4b47696 2022-04-30)
binary: rustc
commit-hash: 7c4b47696907d64eff5621a64eb3c6e795a9ec77
commit-date: 2022-04-30
host: x86_64-unknown-linux-gnu
release: 1.62.0-nightly
LLVM version: 14.0.1

Additional Labels

No response