Comment 1 Eike Stepper 2011-03-22 17:50:17 EDT

About to attach a patch that does all notifications asynchronously if doInvalidate() has already been called asynchronously...

Comment 2 Eike Stepper 2011-03-22 17:51:41 EDT

Created attachment 191713 [details]
Patch v1

With this patch it seems that only 2 test cases fail. Probably because the test logic has to be rewritten to expect the notifications a little later...

Comment 3 Eike Stepper 2011-04-04 00:46:26 EDT

Created attachment 192433 [details]
Patch v2

Issue with conflict resolvers remains!

The invalidation events that trigger the conflict resolution arrive too late in the game...

Comment 4 Martin Fluegge 2011-04-11 15:41:31 EDT

Created attachment 192968 [details]
Test v1

After a while I managed to create a test cases which reproduces the dead lock. It is only a starting point for the different synchronized places. I also haven't tried your patch yet to see whether the test passed with it applied.

Comment 5 Eike Stepper 2012-06-05 07:28:49 EDT

Moving all open bug reports to 4.1 because the release is very near and it's hghly unlikely that there will be spare time to address 4.0 problems.

Please make sure that your patches can be applied against the master branch and that your problem is not already fixed there!!!

Comment 6 Eike Stepper 2012-08-14 22:50:43 EDT

Moving all open issues to 4.2. Open bugs can be ported to 4.1 maintenance after they've been fixed in master.

Comment 7 Eike Stepper 2012-10-31 14:35:47 EDT

The following code at the beginning of the attached test case makes the test totally useless:

    // In case we discover a deadlock we will throw a TimeOutException
    new PollingTimeOuter()
    {
      @Override
      protected boolean successful()
      {
        return false;
      }
    }.assertNoTimeOut(3000);

Without it the test passes, as expected.

Comment 8 Eike Stepper 2012-10-31 14:36:28 EDT

commit 8e8791fe60f3b7db907c4ee0d9b74080ec5765dc

Comment 9 Eike Stepper 2013-06-27 04:00:42 EDT

Closing

Comment 10 Eike Stepper 2020-05-29 02:14:50 EDT

The test case that was added through this bugzilla is harmful:

package org.eclipse.emf.cdo.tests.bugzilla;

import org.eclipse.emf.cdo.eresource.CDOResource;
import org.eclipse.emf.cdo.session.CDOSession;
import org.eclipse.emf.cdo.tests.AbstractCDOTest;
import org.eclipse.emf.cdo.tests.model1.Category;
import org.eclipse.emf.cdo.transaction.CDOTransaction;
import org.eclipse.emf.cdo.view.CDOViewInvalidationEvent;

import org.eclipse.net4j.util.event.IEvent;
import org.eclipse.net4j.util.event.IListener;

import java.util.concurrent.CountDownLatch;

/**
 * @author Martin Fluegge
 */
public class Bugzilla_340709_Test extends AbstractCDOTest
{
  private transient CountDownLatch latch = new CountDownLatch(1);

  public void test() throws Exception
  {
    CDOSession session = openSession();

    {
      CDOTransaction transaction = session.openTransaction();
      addTransactionListener(transaction);

      CDOResource resource = transaction.createResource(getResourcePath("test"));

      Category category = getModel1Factory().createCategory();
      category.setName("name1");

      resource.getContents().add(category);
      transaction.commit();
    }

    {
      CDOTransaction transaction = session.openTransaction();
      CDOResource resource = transaction.getResource(getResourcePath("test"));

      Category category = (Category)resource.getContents().get(0);
      category.setName("name2");

      transaction.commit();
    }
  }

  private void addTransactionListener(final CDOTransaction transaction)
  {
    transaction.addListener(new IListener()
    {
      @Override
      public void notifyEvent(IEvent event)
      {
        System.err.println(event);
        if (event instanceof CDOViewInvalidationEvent)
        {
          handleNotify(transaction);

          await(latch);
        }
      }

      private void handleNotify(final CDOTransaction transaction)
      {
        Thread handleNotifyThread = new Thread(new Runnable()
        {
          @Override
          public void run()
          {
            msg("DEAD");
            transaction.getRootResource();

            msg("LOCK");
            latch.countDown();
          }
        });

        handleNotifyThread.start();
      }
    });
  }
}

The started thread times out after 15 seconds, long after the test case has finished executing. That produces stack traces that are bogus in the scope of later test cases.

As commit 8e8791fe60f3b7db907c4ee0d9b74080ec5765dc has not added an actual fix for the stated problem (and more generally the stated problem is a known/accepted limitation) I'm going to remove the dangerous test case...

Comment 11 Eike Stepper 2020-05-29 02:17:19 EDT

The removal is committed:
https://git.eclipse.org/c/cdo/cdo.git/commit/?id=b312ebf9f3c0dec37e246d77c80e6a7807490ede