Idea: Allow `#[allow(clippy::mutable_key_type)]` on type definitions

[anko]

Description

Problem

I am getting false positives from mutable_key_type on a type I own. (Illustrative Playground.) The current options to silence them are:

• Label each site at which it is used, with #[allow(clippy::mutable_key_type)].

  • Downside: The annotations add lots of noise if there are many use-sites. And since the actual type definition has no such annotations, it is possible someone might modify it to be mutable without realising that it's important not to.

• Add the type to clippy.toml or other config, under the key ignore-interior-mutability.

  • Downside: Nothing in the code warns about the expected invariant. You have to realise to look in the clippy configuration file.

My solution so far has been to do the latter, and write a comment on the key type's definition, explaining the situation and referring to the clippy config file.

Idea

What if type definitions could also be annotated with #[allow(clippy::mutable_key_type)]?

It feels a lot more intuitive than the current options, and would simultaneously

• make the lint ignore that type,
• document the expected invariant for other contributors at the right location,
• cause no noise at the call site where the code is already correct, and
• fit into the lint tooling ecosystem, allowing the reasoning to be listed attached to the allow as a lint reason, instead of an impromptu comment.

Version

rustc 1.85.0-nightly (6d9f6ae36 2024-12-16)
binary: rustc
commit-hash: 6d9f6ae36ae1299d6126ba40c15191f7aa3b79d8
commit-date: 2024-12-16
host: x86_64-unknown-linux-gnu
release: 1.85.0-nightly
LLVM version: 19.1.5

Additional Labels

@rustbot label +C-enhancement

[llogiq]

The problem with that approach is that this would make the #[allow] annotation have a double meaning, either allowing the lint in the annotated item or allowing the lint in all uses of the annotated type. Also if this would just make the mutable_key_type lint ignore the interior mutability, other lints might not benefit from that.

So how about we add a different annotation, like #[clippy::ignore_interior_mutability] instead? That would better encapsulate the intent of implicitly adding the type to the list of types for which all lints should ignore interior mutability.

[anko]

Oh! That sounds even better and I agree with all of that. I withdraw my idea in favour of that. Makes me feel silly for having written this issue in an RFC-ish way.