Q: 8071326: ThreadPoolExecutor in endless thread creation loop if workQueue.take() throws RuntimeException

Doug Lea dl at cs.oswego.edu
Tue Jan 27 20:22:13 UTC 2015


[Oops. I dropped core-libs-dev CC on reply. Resending...)

On 01/27/2015 10:43 AM, Lev Priima wrote:
> Using TPE w/ custom BlockingQueue and if RuntimeException happens in blocking
> BlockingQueue.take() method then this code
>
> new ThreadPoolExecutor(1, 1, 0, TimeUnit.NANOSECONDS,
>      new ArrayBlockingQueue<Runnable>(1) {
>          public Runnable take() throws InterruptedException {
>              throw new RuntimeException();
>          }
>      }
> ).prestartAllCoreThreads();
>
> has an unbounded thread creation loop.
>
> As a result there are many created unbounded threads in RUNNING state after
> printing stack trace to stderr by default UncaughtExceptionHandler. And these
> thread will be cleaned only when whole TPE finished.
>
> Is this "Not an Issue"?
>

It is apparently an issue for someone! But I don't see any spec
violation: If a supplied BlockingQueue not only does not block but also
continuously throws exceptions, then threads using it will be
continuously replaced, given the other TPE construction arguments
here and the call to prestartAllCoreThreads.

Any attempt to change TPE to somehow guess that it should stop trying
to replace threads seems unlikely to solve any actual problem, since
the TPE remains unusable either way.

-Doug




More information about the core-libs-dev mailing list