Bump elided_lifetimes_in_associated_constant to deny

[compiler-errors]

It's been 5 versions since this was last bumped. Let's bump it up again.

[rustbot]

r? @cjgillot

rustbot has assigned @cjgillot.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[tmandry]

@rfcbot fcp merge

[rfcbot]

Team member @tmandry has proposed to merge this. The next step is review by the rest of the tagged team members:

• @joshtriplett
• @nikomatsakis
• @pnkfelix
• @scottmcm
• @tmandry

Concerns:

• let's just use 'static resolved by Bump elided_lifetimes_in_associated_constant to deny #124211 (comment)
• wait for FCP to default to 'static (Bump elided_lifetimes_in_associated_constant to deny #124211 (comment))

Once a majority of reviewers approve (and at most 2 approvals are outstanding), this will enter its final comment period. If you spot a major issue that hasn't been raised at any point in this process, please speak up!

cc @rust-lang/lang-advisors: FCP proposed for lang, please feel free to register concerns.
See this document for info about what commands tagged team members can give me.

[tmandry]

Actually, I don't believe we should have this lint at all and should just use 'static everywhere. I'm about ready to file a different issue and FCP to remove the lint, but we wanted to hear from @nikomatsakis if he has any rationale for landing the lint.

@rfcbot concern let's just use 'static

[compiler-errors]

relevant:

rust-lang/rfcs#1623 (comment)

#38831 (comment)

[scottmcm]

I'd be happy continuing with our previous direction here, so
@rfcbot reviewed

But I'm also sympathetic to tyler's idea that it might be reasonable to decide to add an elision rule for this. I don't like an elision rule for fn foo() -> &T, but given that even associated it's still a const, it might be fine to have elision force unspecified to 'static for it.

[compiler-errors]

'static is a more meaningful value than what's currently happening (recall from the tracking issue #115010 that elided lifetimes in associated consts turn into new generic lifetimes, which is very certainly not correct), but i still think we should still first unstabilize this behavior before deciding on what actually to do here.

[tmandry]

@compiler-errors Does going straight to 'static only allow strictly more code to compile than before?

I still think we should still first unstabilize this behavior before deciding on what actually to do here.

Generally sympathetic, and from the notes we decided to do this before. At the same time, it is churn that feels unnecessary if we are going to turn around and allow all of the code again. Speaking of which: Did we ever look at how much churn there is in crater?

While the decision of using 'static looks trivial to me, I see references in the tracking issue to it needing an RFC and being more subtle than I expect. Modulo the "level of churn" question, I'm comfortable letting this FCP + stabilization cycle be a forcing function for us to find a better way, rather than block an already-agreed path (undoing the stabilization) on somebody doing more work.

@rfcbot resolve let's just use 'static

[rfcbot]

🔔 This is now entering its final comment period, as per the review above. 🔔

[tmandry]

Looking at the references to #115010 at the bottom of the issue, this lint has created a lot of churn already. This makes me think we should leave this PR nominated and discuss again at the next lang team meeting: Do we see a feasible alternative to going down this route?

[nikomatsakis]

@tmandry I feel like I'm missing context here. I'm not sure, for example, what the current behavior is, etc.

I'm sympathetic to defaulting to 'static for consistency with const elsewhere and the reasons that I (apparently) outlined before.

I can certainly imagine reasons to not want 'static, but they are unusual, e.g., const Foo<'a> = None.

[traviscross]

@rustbot labels -I-lang-nominated

We discussed this in the triage meeting today. We were in agreement that 'static is in fact the right default when there are no lifetimes in scope lexically and syntactically. @tmandry is going to write that up as a separate issue on which we can propose FCP.

[tmandry]

@rfcbot concern wait for FCP to default to 'static

[tmandry]

The FCP in question has been proposed here: #125190