[hvitved]

As suggested by @asgerf, I want to use UnqualifiedMemberAccess.getAccessingClass in #22866 for determining implicit nested type qualifiers. However, it cannot always be used directly in cases like

class Node {
    class Leaf: Node {
        let x: Leaf = Leaf()
    }
}

where the unqualified Leaf access gets Node.Leaf as accessing class instead of Node. I think this is what we want for getAccessingClass, though, so this PR is only about documenting this intent with a test case.

[Copilot]

🟡 Changes recommended

The lookup guard leaves valid Swift references such as Node.Leaf.Leaf unbound and causes subclasses to lose inherited type bindings.

1 open issue

• Preserve qualified and inherited lookup for nested types · New

What changed in this PR

Updates unified static name binding to address incorrect results from UnqualifiedMemberAccess.getAccessingClass() for nested Swift subclasses.

Changes:

• Add self-inheritance filtering to member lookup.
• Remove binding pragmas from unqualified-access candidate resolution.
• Add a Swift regression fixture and query output for accessing classes.

File	Description
unified/​ql/​test/​library-tests/​static-name-binding/​unqualified-access.swift	Adds the nested Node.Leaf regression case.
unified/​ql/​test/​library-tests/​static-name-binding/​test.ql	Queries unqualified accesses and their accessing classes.
unified/​ql/​test/​library-tests/​static-name-binding/​test.expected	Records expected accessing-class query results.
unified/​ql/​lib/​codeql/​unified/​internal/​StaticNameBinding.qll	Filters self-inheritance and updates candidate resolution.

_{Review effort: Max}

[Copilot]

🟡 Changes recommended

The target-specific guard can discard valid overloads in nested-subclass lookups, and downstream database effects remain unverified without final DCA results.

2 open issues

• Preserve outer overloads when computing unqualified member access depth · New
• Validate production name-binding changes with DCA results · New

1 resolved since last review

• Preserve qualified and inherited lookup for nested types

_{Review effort: Max}

[asgerf]

Weird situation, but yes I agree this is technically correct for what getAccessingClass does.