Warn on unused lifetime parameters

[clarfonthey]

Somehow this made it into master without being warned:

#[stable(feature = "os_string_from_box", since = "1.17.0")]
impl<'a> From<Box<OsStr>> for OsString {
    fn from(boxed: Box<OsStr>) -> OsString {
        boxed.into_os_string()
    }
}

This should definitely trigger a warning.

[Thiez]

Yikes, no warnings at all.

fn main() {
    trait T {}
    impl<'a> T for u8 {}
}

[estebank]

Triage: no change.

[varkor]

Note that this is currently intentional behaviour:

rust/src/librustc_typeck/impl_wf_check.rs

Line 115 in e575773

// Disallow unconstrained lifetimes, but only if they appear in assoc types.

rust/src/librustc_typeck/impl_wf_check.rs

Lines 139 to 156 in e575773

	// (*) This is a horrible concession to reality. I think it'd be
	// better to just ban unconstrianed lifetimes outright, but in
	// practice people do non-hygenic macros like:
	//
	// ```
	// macro_rules! __impl_slice_eq1 {
	// ($Lhs: ty, $Rhs: ty, $Bound: ident) => {
	// impl<'a, 'b, A: $Bound, B> PartialEq<$Rhs> for $Lhs where A: PartialEq<B> {
	// ....
	// }
	// }
	// }
	// ```
	//
	// In a concession to backwards compatbility, we continue to
	// permit those, so long as the lifetimes aren't used in
	// associated types. I believe this is sound, because lifetimes
	// used elsewhere are not projected back out.

[clarfonthey]

@varkor they should be accepted, just linted.

[scottmcm]

I think this still needs a warn-by-default lint, because I just spotted this in liballoc:

rust/src/liballoc/collections/vec_deque.rs

Lines 2227 to 2236 in 9a3392e

	#[stable(feature = "collection_debug", since = "1.17.0")]
	impl<'a, T: fmt::Debug> fmt::Debug for IterMut<'_, T> {
	fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
	let (front, back) = RingSlices::ring_slices(&*self.ring, self.head, self.tail);
	f.debug_tuple("IterMut")
	.field(&front)
	.field(&back)
	.finish()
	}
	}

Edit: found another one:

rust/src/libstd/panic.rs

Lines 322 to 330 in 9a3392e

	#[unstable(feature = "futures_api", issue = "50547")]
	impl<'a, F: Future> Future for AssertUnwindSafe<F> {
	type Output = F::Output;
	fn poll(self: Pin<&mut Self>, waker: &Waker) -> Poll<Self::Output> {
	let pinned_field = unsafe { Pin::map_unchecked_mut(self, |x| &mut x.0) };
	F::poll(pinned_field, waker)
	}
	}

[JustAPerson]

pub type UnusedHrtb = for<'a> fn();

Found something like this in my codebase today. No warning from rustc or clippy.

[clarfonthey]

Any plans to add a warn-by-default lint?

[Mark-Simulacrum]

We have an unused_lifetimes lint now.

[estebank]

@Mark-Simulacrum it is not warn by default and I think it probably should be (gated may be on ecosystem prevalence).

[Mark-Simulacrum]

Hm, yeah, that makes sense. Reopening. I personally think we can probably "just" land this as a PR with FCP T-lang.

[varkor]

#64493 should be considered blocking for making this warn-by-default.

[tmandry]

#86615 should also be considered blocking.

[Aaron1011]

With #86615 now fixed on the latest nightly, is there anything blocking this?

[obi1kenobi]

I was assuming there aren't any blockers so I started looking into it a little bit, but I'm new to the codebase so it might take me a while especially given the holidays. If you'd like to take it over, be my guest!

There's no shortage of open issues, and I'm sure I can both find something else to work on and also learn from reading the PR that fixes this issue 😃

[Mark-Simulacrum]

Closing this in favor of #94038, which is tracking the enabling of this by default -- I opted to file a new issue so that we have a cleaner history given its tracking status.