Description
Quoted from the mailing list:
2008-02-20 (수), 20:36 +0900, 이희승 (Trustin Lee) 쓰시길:
> 2008-02-20 (수), 08:23 +0100, Niklas Therning 쓰시길:
> > Wilson Yeung wrote:
> > > After sending a bunch of UDP datagrams, sleep for 30 seconds, then
> > > this gets printed to console:
> > >
> > > [java] Exception in thread "pool-1-thread-1" java.lang.NullPointerException
> > > [java] at
> > > org.apache.mina.filter.executor.OrderedThreadPoolExecutor.getSessionBuffer(OrderedThreadPoolExecutor.java:447)
> > > [java] at
> > > org.apache.mina.filter.executor.OrderedThreadPoolExecutor.access$900(OrderedThreadPoolExecutor.java:52)
> > > [java] at
> > > org.apache.mina.filter.executor.OrderedThreadPoolExecutor$Worker.run(OrderedThreadPoolExecutor.java:492)
> > > [java] at java.lang.Thread.run(Thread.java:613)
> > >
> > > And then no more messages are ever sent despite how I might try.
> > >
> > > If I modify OrderedThreadPoolExecutor like so:
> > >
> > > clara:executor wilson$ svn diff OrderedThreadPoolExecutor.java
> > > Index: OrderedThreadPoolExecutor.java
> > > ===================================================================
> > > — OrderedThreadPoolExecutor.java (revision 627903)
> > > +++ OrderedThreadPoolExecutor.java (working copy)
> > > @@ -489,7 +489,9 @@
> > > }
> > >
> > > try {
> > > - runTasks(getSessionBuffer(session));
> > > + if (session != null)
> > > } finally
{ > > > idleWorkers.incrementAndGet(); > > > }> > >
> > > Then all is well. Any ideas? Am I abusing Mina, or is it an honest
> > > to goodness Mina bug?
> > >
> > > Wilson
> > >
> >
> > Do you have a test case which triggers this problem? You should open up
> > a JIRA issue and attach your test case and we will have a look at it.
>
> The stack trace Wilson provided is very weird. According to the trace,
> it seems like OrderedThreadPoolExecutor.getSessionBuffer() throws a NPE
> because the specified session is null. The specified session is get
> from IoEvent object. Now if you look into the IoEvent class, its
> constructor throws NPE if the specified session is null. This means
> that the session parameter that is passed to
> OrderedThreadPoolExecutor.getSessionBuffer() cannot be null at all.
>
> So.. any clue?
Ugh... my reasoning was wrong. It's
OrderedThreadPoolExecutor.fetchSession() that returns null, and it's
possible. Wilson's patch looks reasonable. Let me fix it soon...