Finalizability should be documented

[Pr0methean]

Objects created without running a constructor cannot be automatically finalized; see https://bugs.openjdk.java.net/browse/JDK-6590839. For ObjenesisStd's default strategy on OpenJDK, this isn't a concern, because it uses the munged Object constructor; but for other Objenesis strategies, including the defaults for other JVMs, it may be. It'd be nice if this was documented (and better yet if there was a way to check it at runtime). Is ReferenceQueue behavior also affected (which would mean users couldn't just implement their own finalizer either)?

[henri-tremblay]

Interesting. Do you have a use-case of an object created with Objenesis for which you would like the finalizer to be called? For me, it looks like it should never be called in fact.

Which strategies were you thinking about?

[Pr0methean]

I believe the old sun.misc.Unsafe strategy (which probably requires an --add-opens since Java 9) is one that would not finalize.