We've had a few large-scale cleanup PRs before, but it doesn't always stick.
This is a quick example of what I'm referring to (as the agreed-upon style):
use rustc::ty::{Ty, TyCtxt};
fn foo(tcx: TyCtxt<'_, '_, 'tcx>, ty: Ty<'tcx>) {...}
TyCtxt and Ty are referred to unqualified (without ty::) and passed by value.
(Note that types can be behind & due to a trait, but that should be the only exception to the rule)
Currently both &TyCtxt and &Ty<'tcx> show up in the codebase (although in small quantities), but I only accidentally noticed and I won't be replacing them any time soon (someone else can, if they want to, but I'm more interested in how we keep this from happening - maybe with clippy?).
cc @rust-lang/compiler
cc https://github.com/rust-lang/rust/issues/49509 (Add compiler-internal lints)
cc @flip1995
I'll write internal lints for these two cases. I try to open a PR by the end of this week.
Quick update on this:
The lint writing is done: https://github.com/flip1995/rust/commits/internal_lints
Now I have to fix the findings of the lints. This may take a little bit of time (I had to touch 18 files for librustc alone)
Most helpful comment
I'll write internal lints for these two cases. I try to open a PR by the end of this week.