5 Replies Latest reply on Jan 12, 2009 5:03 PM by alrubinger

    Revert of r82615 / EJBTHREE-1653

    alrubinger

      Due to regression:

      Edit:removed link to private Hudson

      I've reverted r82615 and reopened EJBTHREE-1653.

      S,
      ALR

        • 1. Re: Revert of r82615 / EJBTHREE-1653
          wolfc

          The service unit is failing, because it fails to deploy.

          • 2. Re: Revert of r82615 / EJBTHREE-1653
            pferraro

            I've run this unit test against both revisions of EJBContainer (using the latest ejb trunk and as Branch_5_0) and get the same results:
            [junit] Running org.jboss.ejb3.test.service.unit.ServiceUnitTestCase
            [junit] Tests run: 20, Failures: 0, Errors: 1, Time elapsed: 39.25 sec
            [junit] Test org.jboss.ejb3.test.service.unit.ServiceUnitTestCase FAILED

            Unless I'm missing something, from my perspective, it doesn't appear that the changes to EJBContainer for EJBTHREE-1653 caused any regressions. Can you email me that hudson link?

            • 3. Re: Revert of r82615 / EJBTHREE-1653
              alrubinger

              Today/yesterday I refactored and patched around ServiceContainer, getting to 1 remaining error. That work may be hiding the problems we saw earlier.

              I'll send you privately the Hudson link showing the regression.

              S,
              ALR

              • 4. Re: Revert of r82615 / EJBTHREE-1653
                pferraro

                Ah - I finally found the issue...
                Changes to JBossASKernel in ejb3 as module (r82684) dropped a re-entrancy requirement for EJBContainer.create() - which previous prevented the service mbean from deploying.
                In r82615, EJBContainer.create() is not re-entrant. The previous version (that utilized a ReentrantReadWriteLock) was at least thread re-entrant. I'll add re-entrancy for create() and destroy() (as it was for start() and stop() are re-entrant already) and get this tested and patched today.

                • 5. Re: Revert of r82615 / EJBTHREE-1653
                  alrubinger

                   

                  "pferraro" wrote:
                  Ah - I finally found the issue...
                  Changes to JBossASKernel in ejb3 as module (r82684) dropped a re-entrancy requirement for EJBContainer.create()...


                  Just cleared this up w/ Paul. The requirement is of the failing test, not of the Container contracts.

                  Also @see http://www.jboss.com/index.html?module=bb&op=viewtopic&t=148497. I'll be looking at a backwards-compatible solution until we can safely patch AS5 GA.

                  S,
                  ALR