lint against "silent elision" in structs/enums/etc

[nikomatsakis]

Now that support for '_ has landed, as part of #44524, it would be good to start linting for cases where it ought to be used. In particular, we wish to warn if there is a struct with lifetime parameters that are elided:

struct Foo<'a> { x: &'a u32 }
fn foo(x: &Foo) {
//         ^^^ warning: hidden lifetime parameters are deprecated, try `Foo<'_>`
}

This should be fairly straightforward to do during resolve_lifetimes.

[Dylan-DPC-zz]

Interested in taking this up 😄

[nikomatsakis]

@Dylan-DPC great =)

Sorry, I didn't get a chance to write up any instructions yesterday. I just took a look here and I think this should be fairly straightforward.

The first thing is that we need to add a lint. There are no perfect directions for adding a new lint that I can find. The best that I know of are in this forge document, under the section "Issuing future compatibility warnings". This section describes how to add a future compatibility lint. We don't want a future compatibility lint, just a regular lint, but most of the steps are the same. The one that does not apply is step 3 ,"Register the lint in src/librustc_lint/lib.rs". You can just skip that step.

Let's call the lint elided_lifetime_in_path. For now, let's set the default level to Allow, since we don't want to bothering people until the whole in-band lifetimes RFC is in place.

Once we have the lint, we want to modify the resolve_elided_lifetimes function in resolve_lifetimes.rs. We should add a deprecated: bool argument to it that indicates whether this elided lifetime occurs in a deprecated position.

Most of the callers can use false. The only one that is deprecated is this caller right here:

rust/src/librustc/middle/resolve_lifetime.rs

Line 948 in 88a28ff

self.resolve_elided_lifetimes(&params.lifetimes);

Then we just need to add some tests and we're done! To add tests, just make some ui tests that use lifetimes in the deprecated places (as well as places that are not deprecated) and write #![deny(elided_lifetime_in_path)] at the top.

[nikomatsakis]

We're also going to want to feature-gate this, though I don't think we have a good mechanism for that. We should figure that out. But let's do this first part first. =)

[nikomatsakis]

@Dylan-DPC how goes? Did you get a chance to take a look?

[Dylan-DPC-zz]

Hey Niko. Was busy. Will try it by this weekend.