Aleksey Shipilev wrote: > The change in monitor_enter, even the safest one, touches every > application that runs. Just because it's touching threading and > synchronization. So I'm very pessimistic that ordinary reasoning on > "change locality" is enough here. Like we can't miss this NPE bug in > M8, we can't miss the occasional regression while fixing this bug :)
Maybe. Just because the monitor enter is used everywhere doesn't mean that the fix has to be tested in every usage -- if it is a well contained change. In this case the null check should be a simple enough change with little chance of side effects. Can you think of a situation when the null check will introduce some instability or regression? I could be persuaded otherwise, but right now I think limited retesting is sufficient for this fix. BTW I'm -1 on releasing r711744 as-is, so we will need to rebuild and restart the vote anyway. Regards, Tim
