fix: Goto definition for deref_mut

[ShoyuVanilla]

Fixes #16520

rust-analyzer/crates/hir/src/source_analyzer.rs

Lines 375 to 393 in a3236be

	pub(crate) fn resolve_prefix_expr(
	&self,
	db: &dyn HirDatabase,
	prefix_expr: &ast::PrefixExpr,
	) -> Option<FunctionId> {
	let (lang_item, fn_name) = match prefix_expr.op_kind()? {
	ast::UnaryOp::Deref => (LangItem::Deref, name![deref]),
	ast::UnaryOp::Not => (LangItem::Not, name![not]),
	ast::UnaryOp::Neg => (LangItem::Neg, name![neg]),
	};
	let ty = self.ty_of_expr(db, &prefix_expr.expr()?)?;
	let (op_trait, op_fn) = self.lang_trait_fn(db, lang_item, &fn_name)?;
	// HACK: subst for all methods coincides with that for their trait because the methods
	// don't have any generic parameters, so we skip building another subst for the methods.
	let substs = hir_ty::TyBuilder::subst_for_def(db, op_trait, None).push(ty.clone()).build();
	Some(self.resolve_impl_method_or_trait_def(db, op_fn, substs))
	}

As we can see from the above, current implementation routes all dereferencing prefix operations to Deref::deref implementation, not regarding mutabilities.

rust-analyzer/crates/hir-ty/src/infer/mutability.rs

Lines 134 to 151 in a3236be

	Expr::UnaryOp { expr, op: UnaryOp::Deref } => {
	if let Some((f, _)) = self.result.method_resolutions.get_mut(&tgt_expr) {
	if mutability == Mutability::Mut {
	if let Some(deref_trait) = self
	.db
	.lang_item(self.table.trait_env.krate, LangItem::DerefMut)
	.and_then(|l| l.as_trait())
	{
	if let Some(deref_fn) =
	self.db.trait_data(deref_trait).method_by_name(&name![deref_mut])
	{
	*f = deref_fn;
	}
	}
	}
	}
	self.infer_mut_expr(*expr, mutability);
	}

Since we are resolving them already in mutability inferences, we can use those results for proper deref / deref_mut routing.

[Veykril]

Thanks! I wonder if we record that for indexing expressions as well so that we can resolve IndexMut in a similar fashion.
@bors r+

[bors]

📌 Commit 6124431 has been approved by Veykril

It is now in the queue for this repository.

[bors]

⌛ Testing commit 6124431 with merge a41cec9...

[ShoyuVanilla]

I wonder if we record that for indexing expressions as well so that we can resolve IndexMut in a similar fashion.

Good point. I'll check for that case tomorrow

[bors]

☀️ Test successful - checks-actions
Approved by: Veykril
Pushing a41cec9 to master...