[concurrency-interest] Memory Leak in LinkedBlockingQueue

Doug Lea dl@cs.oswego.edu
Thu, 27 Jan 2005 06:55:32 -0500


Thanks for reporting this!

Sorry for the problem. The leak is real, although your diagnosis is a
little off.  Space from aborted timed condition waits was designed to
be cleaned up if the condition is EVER signalled. I now see that this
isn't good enough for some applications.  (We should know better too!
Java.util.Timer has long had essentialy the same problem.)  We'll work
on improving this to self-limit in the case of an unbounded series of
aborted waits without a signal.  In the mean time, you might be able
to work around this by occasionally artificially inserting a dummy
element.

-Doug