New lint: "unnecessary `return foo()` because foo returns ()"

[alco]

As suggested by @huonw in IRC.

From sha1 implementation here:

fn mk_result(st: &mut Sha1, rs: &mut [u8]) { ... }
fn result(&mut self, out: &mut [u8]) { return mk_result(self, out); }

This last line should emit the warning.

[msullivan]

I don't really buy it. In this particular case, the return is silly, but there are plenty of places where it isn't.

if something {
   return foo();
}
// more stuff

We could have something more specific, but I don't really see the point.

[nikomatsakis]

I sometimes intentionally write return foo() where foo() returns (), because I want there to be a compile error should foo() not return unit at some point in the future (i.e., if it should change to return a Result)

[alexcrichton]

Closing, this is an aspect of the language which I think is quite useful, and it makes sense in a typesystem sort-of-way because if a function returns unit you can then continue to return the unit.

[nikomatsakis]

Just to agree with @alexcrichton -- I often purposefully do return foo() when I know foo() has unit return type, because I want to get a compilation error if the return type of foo() should change.

[nikomatsakis]

Oh, I see I wrote that already. Old age.