needless_lifetimes ignores trait bounds

[Manishearth]

fn foo<'a, T: Foo<'a>>(&'a u8, T) will hit the lint. It shouldn't, whenever we have lifetimes in trait bounds we should bail.

[Manishearth]

https://github.com/Manishearth/rust-clippy/issues/417

[fhartwig]

I think this should be re-opened. It's not really the same problem as #417, even though they're both false positives. #417 is basically about mis-counting input and output lifetimes, while the problem here is that you're not allowed to elide lifetimes in trait bounds.

[llogiq]

I'd like to add that we still have false positives even after #599 – The followig (from rust-bio) is still linted:

pub fn find_all<'a, I: Iterator<Item=&'a u8>>(&'a self, text: I) -> Matches<I> {
    Matches { shiftand: self, active: 0, text: text.enumerate() }
}

[dtolnay]

Users of Serde are going to start hitting this really badly in the next release as we are adding a lifetime to the Deserializer trait.

fn f<'de, D: Deserializer<'de>>(deserializer: D) -> Result<F, D::Error> { /* ... */ }

It would be great to fix this before Serde 1.0.

[mcarton]

@dtolnay I believe a simple case like this is already not linted against?
https://github.com/Manishearth/rust-clippy/blob/0432f1efd31f8cd58faea180d8e44a786bc6fc4c/tests/ui/lifetimes.rs#L128-L129
https://github.com/Manishearth/rust-clippy/blob/0432f1efd31f8cd58faea180d8e44a786bc6fc4c/tests/ui/lifetimes.stderr#L88-L92