On Tue, 23 Aug 2022 14:31:26 GMT, Сергей Цыпанов <d...@openjdk.org> wrote:
> > I agree that caution is warranted > > But assuming I've reverted the changes that are dubious, how could there be > bugs caused by replacement of multiple reads with a single one? I was sure > that such replacement improves thread safety as soon as we rid racy reads. Yes - that's what I was trying to say. I believe reading the volatile only once per method is much better for consistency. And I'm not sure I see why the other changes were dubious since these methods weren't checking for `null` either? My point is that if some method in the subclass can set the value to `null` asynchronously when some other method in the super class is running then that's a NPE time-bomb (and IMO there's a bug lurking if that can happen). ------------- PR: https://git.openjdk.org/jdk/pull/9956
