> On 20 Nov 2018, at 17:55, Seán Coffey <sean.cof...@oracle.com> wrote: > > Access to the handlers Hashtable and the factory should be made while holding > the streamHandlerLock lock
I'm not sure about that. I wouldn't go with coarsening this lock's region. I would rather go with an additional check for handler being null. It doesn't look like this lock has ever had that big of a scope and bear in mind this place might be very startup-time sensitive. So, in order to not steer clear of any potential issues with that or deadlocks I would simply introduce an unconditional null check. -Pavel
