On 16/09/2015 02:49, Caldarale, Charles R wrote: >> From: Mark Thomas [mailto:ma...@apache.org] >> Subject: Re: RV-Predict bugs > >> My expectation is that once T2 has seen the updated value (originated >> from another thread) all subsequent reads in T2 of the same field are >> going to see the same value rather than some of those reads seeing an >> older value (assuming there are no further changes in other threads). > > The memory model does permit out-of-order reads of the same location, but I > can't find any CPUs that would do this. (Out-of-order reads of different > locations occur frequently, of course.)
OK. Thanks. I think that wraps up this part of the discussion for me. I think we are all agreed that, purely from a re-ordering of reads point of view (deliberately ignoring transformations), there is a theoretical issue here based purely on the rules of the JMM but in practise the issue is never observed because no CPU takes advantage of what the JMM allows. Whether or not issues in the category above are worth fixing or not is a different debate and not one I really want to get into here. I'll just note that we did fix this bug in Tomcat. I'll follow up with the questions I have on the transformation aspects in that part of the thread. Mark --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
