On Wed, 15 Jul 2026 02:03:04 GMT, David Holmes <[email protected]> wrote:
>> I tend to agree with @ExE-Boss here. Vagaries of finalizers aside, if the >> subclass can be an identity class then the superclass's finalizer should be >> registered for instances of the subclass. Otherwise the JEP needs to state >> clearly that this is the case. > > FTR the behaviour was ratified by a spec update through JDK-8387417 For future reference, https://github.com/openjdk/valhalla/pull/2588/ updated the specification of Object.finalize() to specify: * <div class="preview-block"> * <div class="preview-comment"> * The {@code finalize} method of a value class is never directly * invoked by the garbage collector. This includes the case where an * abstract value class declares a {@code finalize} method and the * class is extended by an identity class; the garbage collector never * directly invokes the {@code finalize} method declared by the * abstract value class. * </div> * </div> ------------- PR Review Comment: https://git.openjdk.org/valhalla/pull/2589#discussion_r3583977425
