[jboss-dev-forums] [Design of Messaging on JBoss (Messaging/JBoss)] - Why failoverValve has the ThreadLocal routine

Basically this is the scenario... You have several threads trying to acquire the writeLock simultaneously... Different from a Tomcat Valve where from what I remember when I read the code, then don't have such requirement. You have several threads... each one with a readLock... then all of these threads will try to acquire a writeLock... Any thread will be able to assume the writeLock... as there are other threads with readLocks. so... before acquiring the writeLock, I would need to release the readLock... But... as the Valve is Reentrant.. I needed to store the number of reentrances before releasing readLocks executed before. The only solution to avoid the ThreadLocal would be to avoid reentrance on the Valve and aways release one readlock.. but I don't think that's completely possible. But if anyone can find a way to have this following test work (without releasing a readLock), we can apply the fix to the Valve: public void testOnReadlocks() throws Exception | { | int THREAD_COUNT = 10; | final ReadWriteLock readWriteLock = new WriterPreferenceReadWriteLock(); | final Object sync = new Object(); | | Thread threads[] = new Thread[THREAD_COUNT]; | | for (int i = 0; i < THREAD_COUNT; i++) | { | threads = new Thread() | { | public void run() | { | try | { | synchronized (sync) | { | sync.wait(); | } | readWriteLock.readLock().acquire(); | readWriteLock.writeLock().acquire(); | } | catch (Exception e) | { | e.printStackTrace(); | } | } | }; | threads.start(); | } | | Thread.sleep(2000); | | synchronized (sync) | { | sync.notify(); | } | | for (int i=0;i<threads.length;i++) | { | threads.join(); | } | } | This test would hang forever View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4021809#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...