`wasm-bindgen` 0.2.92 triggers `clippy::mem_forget`

[Conaclos]

Summary

wasm-bindgen 0.2.92 triggers the Clippy rule mem_forget. Is it expected?

error: usage of `mem::forget` on `Drop` type
  --> crates/biome_wasm/src/lib.rs:24:1
   |
24 | #[wasm_bindgen]
   | ^^^^^^^^^^^^^^^
   |
   = note: argument has type `wasm_bindgen::JsValue`
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#mem_forget
   = note: this error originates in the attribute macro `wasm_bindgen` (in Nightly builds, run with -Z macro-backtrace for more info)

[Liamolucko]

I think this is the mem::forget it's picking up on: https://github.com/rustwasm/wasm-bindgen/blob/e78db23553e70a48767a9bafaac3a4ffb4cba08d/crates/backend/src/codegen.rs#L369

There's an #[allow(clippy::all)] right above it that's supposed to prevent issues like this, but I'm now realising that I've misunderstood how that works a bit: clippy::all isn't actually all of clippy's lints, it's clippy's default set of lints, which doesn't include mem_forget.

What we should really be using is #[automatically_derived], which does disable all lints - I'll make a PR switching it to use that.

[Liamolucko]

What we should really be using is #[automatically_derived], which does disable all lints - I'll make a PR switching it to use that.

Hm, seems like I was wrong: Clippy still throws a warning on this code even though it uses #[automatically_derived]:

#![warn(clippy::mem_forget)]
struct Foo;
#[automatically_derived]
impl Foo {
    fn test() {
        std::mem::forget(String::new());
    }
}

Searching for automatically_derived in rustc/clippy's source code, and looking back at the original PR where we started using it (#2719), it seems like specific lints manually disable themselves if they see it, wherever it makes sense.

So I'm not really sure what the best solution is then, other than just spamming #[allow(clippy::all, clippy::nursery, clippy::pedantic, clippy::restriction)] everywhere.