Deriving: Include bound generic params in type parameters for where clause

[audunhalland]

Fixes #89188.

The derive macro ignored the for<'s> needed with the Fn trait in that code example.

edit: I'm unsure if this might cause regressions. I'm not an experienced compiler developer so I'm not used to thinking about unwanted side effects code changes like this might have.

[rust-highfive]

r? @matthewjasper

(rust-highfive has picked a reviewer for you, use r? to override)

[jackh726]

r? @jackh726

[jackh726]

I think this looks correct. I was trying to come up with an example without GATs where we could run into this case, but I haven't been able to yet. We basically have to name a late-bound lifetime within the type but without also being required to name it outside. (So, <T as Trait<'a>::As wouldn't work, because we would need a T: Trait<'a> bound).

I think I'd like to see the following added to the test:

trait CallWithShim2: Sized {
    type Shim<T>;
}

struct S<'s>(&'s ());

#[derive(Clone)]
struct ShimMethod2<T: CallWithShim2 + 'static>(
    pub &'static dyn for<'s> Fn(&'s mut T::Shim<S<'s>>),
);

trait Trait<'s, 't> {}

#[derive(Clone)]
struct ShimMethod3<T: CallWithShim2 + 'static>(
    pub &'static dyn for<'s> Fn(&'s mut T::Shim<dyn for<'t> Trait<'s, 't>>),
);

trait Trait2 {
    type As;
}

#[derive(Clone)]
struct ShimMethod4<T: Trait2 + 'static>(
    pub &'static dyn for<'s> Fn(&'s mut T::As),
);

[audunhalland]

It would probably break by HRTB variable shadowing, e.g.

for<'a> (... for<'a> ( ... ))

by outputting for<'a, 'a>? Unsure if it's important enough to address?

[jackh726]

I was trying to come up with a test case where this might happen. But I'm unsure; I couldn't find one.

[jackh726]

Okay, can you instead actually change the test case to:

trait Trait<'s, 't, 'u> {}

#[derive(Clone)]
struct ShimMethod3<T: CallWithShim2 + 'static>(
    pub &'static dyn for<'s> Fn(&'s mut T::Shim<dyn for<'t> Fn(&'s mut T::Shim<dyn for<'u> Trait<'s, 't, 'u>>)>),
);

This better tests multiple levels of binding. Made me remember that we don't allow lifetime shadowing, so we should be good.

[jackh726]

After that change, this PR LGTM to merge.

[jackh726]

@bors r+ rollup=never

Given that this effects built-in derives, I want to make sure this is easy to bisect if there are any regressions.

[bors]

📌 Commit 87241e9 has been approved by jackh726

[bors]

⌛ Testing commit 87241e9 with merge f03eb6b...

[bors]

☀️ Test successful - checks-actions
Approved by: jackh726
Pushing f03eb6b to master...

[rust-timer]

Finished benchmarking commit (f03eb6b): comparison url.

Summary: This benchmark run did not return any relevant changes.

If you disagree with this performance assessment, please file an issue in rust-lang/rustc-perf.

@rustbot label: -perf-regression