On 30/09/2011 6:39 PM, Holger Hoffstaette wrote:
Addendum..
On Fri, 30 Sep 2011 18:27:19 +0200, Holger Hoffstaette wrote:
What I was trying to say is that the bug is much more likely in ZK and one of its racy/buggy codepaths that merely exposes this symptom. They
This is further backed by the fact that this happens during a (one of many?) test case during ZKServer shutdown. If they have multiple threads shutting down the server, zombie threads hanging around in the background and maybe shutting down logging (?) or doing other crazy things then it's easy to see how this could result in bungled lock acquisition.
On branch lbcore224, I modified AppenderAttachableImpl so that only addAppender and appendLoopOnAppenders could be invoked. See [1]. IllegalMonitorStateException was still being thrown. Given commit [1], how could you trigger a IllegalMonitorStateException ? It seems quite impossible although the code for ReadWriteLock is quite complicated so there is ample room for bugs. [1] https://github.com/ceki/logback/commit/379c40573e208b2c5bf52c7e
-h -- http://twitter.com/ceki