Skip to content

v4.183.1

v4.183.1: fix: four detector-report false positives (#392–#395, #398)
@jessegall jessegall tagged this 22 Jul 21:52 
- NullableWireObjectDetector skips HOOKED (computed) properties: a get-hook is
  not a hydration slot, so 'T | Optional' cannot apply — 'T | null' is the honest
  computed wire shape (closes #394, closes #395).
- ContainerReachDetector's Container heuristic now sees SUBCLASS instantiation:
  an abstract base new'ed only via hand-built subclasses is not container-resolved
  (closes #392).
- MatchDefaultReturnsNullDetector no longer flags a match over an OPEN subject
  whose handled arms already call ?T-declared heuristics — null is the match's
  own answer vocabulary, not a swallowed case; an enum subject still flags
  (closes #393).
- ManufacturedFakeFillDetector no longer treats '?? []' as a fake: an empty
  collection is the type's identity, a real domain answer (closes #398).

New engine tools: AstNode::isEmptyArrayLiteral / isHookedProperty,
NodeMatch::matchHandledArmsAdmitNull, Codebase::methodReturnsNullable.
Assets 2
Loading