Do not suggest unresolvable builder methods

[gurry]

Fixes #125303

The issue was that when a builder method cannot be resolved we are suggesting alternatives that themselves cannot be resolved. This PR adds a check that filters them from the list of suggestions.

[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

[gurry]

This is the extra check I have added. Hopefully it isn't a performance concern as it will run only during diagnostics and only for builder methods not all the assoc items of an impl.

[compiler-errors]

Now that we're just passing in the expr id, I feel like rcvr_opt: Option<_>, SelfSource: _, args: Option<_> are all redundant, since they can all be derived from the expression pointed to by the expr_id.

[compiler-errors]

And the span too

[gurry]

I looked into it. Not much is derivable from expr_id.

These are the two calls to report_method_error:

rust/compiler/rustc_hir_typeck/src/fn_ctxt/_impl.rs

Lines 835 to 845 in 9cdfe28

	if let Some(e) = self.report_method_error(
	span,
	None,
	ty.normalized,
	item_name,
	SelfSource::QPath(qself),
	error,
	args,
	Expectation::NoExpectation,
	trait_missing_method && span.edition().at_least_rust_2021(), // emits missing method for trait only after edition 2021
	) {

and

rust/compiler/rustc_hir_typeck/src/expr.rs

Lines 1347 to 1357 in 9cdfe28

	if let Some(err) = self.report_method_error(
	span,
	Some(rcvr),
	rcvr_t,
	segment.ident,
	SelfSource::MethodCall(rcvr),
	error,
	Some(args),
	expected,
	false,
	) {

As you can see rcvr_opt (second arg) is being explicitly passed as None in one case and Some(rcvr) in the other. Similarly rcvr_ty (third arg) is ty.normalized in one case and rcvr_ty in the other and both are obtained in different ways. source (fifth arg) too has different values with somewhat convoluted provenance and the same can be said of args (seventh arg).

So I don't feel we should change any of this.

[compiler-errors]

SelfSource, rcvr_opt, and args are all a property of whether the expr_id (which really should be passed in as a whole hir::Expr) is ExprKind::MethodCall or ExprKind::Path.

[compiler-errors]

Similarly, segment.ident should be derivable from that expression (it's directly stored in the MethodCall and is also as the final segment in ExprKind::Path, or something like that).

I don't think I said to touch rcvr_ty, which can continue to be passed in.

[compiler-errors]

Similarly, span can be derived from the expression without much trouble as well.

[gurry]

Okay, let me have another look at it

[gurry]

SelfSource, rcvr_opt, and args are all a property of whether the expr_id (which really should be passed in as a whole hir::Expr) is ExprKind::MethodCall or ExprKind::Path.

@compiler-errors Turns out the expr_id isn't necessarily pointing to an Expr always. it could also be a Pat because check_pat also calls report_method_error (check_pat->resolve_ty_and_res_fully_qualified_call->report_method_error) . So if I were to derive all these bits from expr_id I'll have to take Pat into account in addition to Expr, which means I might have to repeat code like this from resolve_ty_and_res_fully_qualified_call:

rust/compiler/rustc_hir_typeck/src/fn_ctxt/_impl.rs

Lines 749 to 772 in 78dd504

	let (ty, qself, item_segment) = match *qpath {
	QPath::Resolved(ref opt_qself, path) => {
	return (
	path.res,
	opt_qself.as_ref().map(|qself| self.lower_ty(qself)),
	path.segments,
	);
	}
	QPath::TypeRelative(ref qself, ref segment) => {
	// Don't use `self.lower_ty`, since this will register a WF obligation.
	// If we're trying to call a nonexistent method on a trait
	// (e.g. `MyTrait::missing_method`), then resolution will
	// give us a `QPath::TypeRelative` with a trait object as
	// `qself`. In that case, we want to avoid registering a WF obligation
	// for `dyn MyTrait`, since we don't actually need the trait
	// to be object-safe.
	// We manually call `register_wf_obligation` in the success path
	// below.
	let ty = self.lowerer().lower_ty_in_path(qself);
	(LoweredTy::from_raw(self, span, ty), qself, segment)
	}
	QPath::LangItem(..) => {
	bug!("`resolve_ty_and_res_fully_qualified_call` called on `LangItem`")
	}

within report_method_error. I have a feeling that complicates things instead of making them simpler 😫

[compiler-errors]

That didn't end up being too complicated, for the record: 2227c1f

[compiler-errors]

This needs IsSuggestion set to true (or needs to use lookup_method_for_diagnostic or something diagnostic-flavored instead), and I encourage you to also look into cleaning up redundant args (it's nice to leave code better than you found it)

[compiler-errors]

@rustbot author

[gurry]

@rustbot ready

[gurry]

@rustbot author

[gurry]

@rustbot ready

The suggested cleanup seems a bit convoluted. Please see my comment here: #125397 (comment)

[compiler-errors]

I'll do the cleanup myself I guess

@bors r+ rollup

[bors]

📌 Commit 273a78b has been approved by compiler-errors

It is now in the queue for this repository.

[bors]

⌛ Testing commit 273a78b with merge 865eaf9...

[bors]

☀️ Test successful - checks-actions
Approved by: compiler-errors
Pushing 865eaf9 to master...

[rust-timer]

Finished benchmarking commit (865eaf9): comparison URL.

Overall result: ❌ regressions - no action needed

@rustbot label: -perf-regression

Instruction count

This is a highly reliable metric that was used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	1.1%	[1.1%, 1.1%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Max RSS (memory usage)

Results (primary -3.4%)

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)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-3.4%	[-4.0%, -2.9%]	2
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-3.4%	[-4.0%, -2.9%]	2

Cycles

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

Binary size

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

Bootstrap: 667.872s -> 667.71s (-0.02%)
Artifact size: 318.92 MiB -> 318.93 MiB (0.00%)