Only emit one error per unsized binding, instead of one per usage

[estebank]

Fix#56607.

[rustbot]

r? @BoxyUwU

(rustbot has picked a reviewer for you, use r? to override)

[compiler-errors]

I think this is too fragile of a heuristic to be worth landing, and it introduces potential future headaches when queries interleave and for the parallel compiler. I'm not convinced this PR is worth landing 😓

clarifying edit: Specifically the dependency on self.tcx.sess.has_errors().is_some(), which is global state that I think we want to eventually remove, or at least track a lot more delicately.

[compiler-errors]

I think we want to move forward with depending on less global state, rather than more? UI test instability due to global diagnostic state is one of the blockers for the parallel compiler, I think?

[estebank]

Note that a few lines above (689, hidden by the patch view) we already have a has_errors check. When we come up with an alternative for that one, this one can also go away. In this case this is to ensure that we don't unnecessarily silence obligations that happen to be in reported_trait_errors without actually having been emitted (I found a couple of examples in the test suite).

[estebank]

@compiler-errors let me know if my rationale above is sufficient to land the PR as is, otherwise we can close this PR.

[davidtwco]

Is tcx.sess.has_errors().is_some() required when we also check self.reported_trait_errors.borrow().get(&pat.span)?

[estebank]

The following tests ICE without that check

[ui] tests/ui/derives/derives-span-Clone-enum-struct-variant.rs
[ui] tests/ui/derives/derives-span-Clone-enum.rs
[ui] tests/ui/derives/derives-span-Hash-enum-struct-variant.rs
[ui] tests/ui/derives/derives-span-PartialOrd-enum-struct-variant.rs
[ui] tests/ui/derives/derives-span-PartialOrd-enum.rs
[ui] tests/ui/derives/derives-span-Hash-enum.rs

I think because report_fulfillment_errors inserts all the error predicates found in the global reported_trait_errors, but it later has suppression logic that avoids emitting some diagnostics. I'm sure we could change that logic instead, but will make the PR larger.

Edit: actually the change ended up being much smaller (and the slight perf loss due to tiny amount of extra work only happens in the error path).

[BoxyUwU]

r? @compiler-errors

[compiler-errors]

I am still a bit skeptical whether this is worthwhile given the limitations (only fn calls) and introducing more dependency on global state, but I'd rather not unilaterally make that call.

r? diagnostics

[estebank]

given the limitations (only fn calls)

I focused on dealing with the reported case, but would hope to extend this to more ObligationCauseCode cases, within reason.

[apiraino]

r? diagnostics

[davidtwco]

I think I'd like to see the stuff inside the if let here put into a function, just to keep report_selection_error a bit simpler.

[apiraino]

@estebank does this comment needs to be addressed?

Apart from this I think this PR can be r+'ed by @davidtwco (as per comment)?

[estebank]

This was already done

[davidtwco]

I think this is a useful improvement, particularly if extended to other reasonable obligation causes as @estebank has said is his intention.

I agree with the concerns about has_errors, but given that the function already uses has_errors only a few lines above the diff here, I don't think this addition makes removing has_errors or changing it materially more difficult - however the usage above is replaced, this usage can replaced the same way.

I've left a comment/reply or two, but I think I'm happy to r=me this after those are resolved.

[apiraino]

@estebank by reading the last comment, I think the review switch was left unchanged, when I think this is actually waiting on you.

@rustbot author

[estebank]

I can confirm that handling other ObligationCauseCodes isn't necessary because Sized requirements on bindings is the only case where this particular issue crops up. I found a case where a type is defined as dyn Trait, which gets a predicate for dyn Trait + 'static in the definition, but on the use the predicate is for dyn Trait + '_, so the deduplication doesn't happen, but I feel like adding logic to check for it might be too convoluted (will try it out anyways).

[estebank]

@bors r=davidtwco

[bors]

📌 Commit 568b316 has been approved by davidtwco

It is now in the queue for this repository.

[bors]

⌛ Testing commit 568b316 with merge aa91057...

[bors]

☀️ Test successful - checks-actions
Approved by: davidtwco
Pushing aa91057 to master...

[rust-timer]

Finished benchmarking commit (aa91057): comparison URL.

Overall result: no relevant changes - no action needed

@rustbot label: -perf-regression

Instruction count

This benchmark run did not return any relevant results for this metric.

Max RSS (memory usage)

This benchmark run did not return any relevant results for this metric.

Cycles

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	1.5%	[1.5%, 1.5%]	1
Regressions ❌ (secondary)	2.4%	[2.3%, 2.4%]	3
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	1.5%	[1.5%, 1.5%]	1

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 635.024s -> 636.976s (0.31%)
Artifact size: 304.50 MiB -> 304.49 MiB (-0.00%)