2 Replies Latest reply on Dec 13, 2008 3:01 PM by Ronald van Kuijk

    Wierd Problem in Hibernate session

    Shai Bentin Newbie

      I have to following code:

      protected TaskInstance locateTask(ContextInstance ctx, JbpmContext jbpmContext) throws ApplicationException {
       String searchValue = (String) ctx.getVariable(contextKey);
       if (searchValue == null || searchValue.trim().equals(""))
       return null;
       String query = String.format("select t.ID_ TASK_ID from JBPM_TASKINSTANCE t , JBPM_VARIABLEINSTANCE vi "
       + "where vi.STRINGVALUE_ = '%s' and vi.TOKEN_ = t.TOKEN_ and (t.ISOPEN_ = 1 or t.ISSUSPENDED_ = 1)"
       + " and t.END_ is null and t.NAME_ ", new Object[] {searchValue});
      
       String inString;
       if (!notInTaskType) {
       inString = String.format("in ( %s )", new Object[] {taskName});
       }
       else {
       // we're looking for tasks which are not of types given
       inString = String.format("not in ( %s )", new Object[] {taskName});
       }
      
       query += inString;
      
       SQLQuery sqlQuery = jbpmContext.getSession().createSQLQuery(query).addScalar("TASK_ID", Hibernate.LONG);
       List result = sqlQuery.list();
       // Collection<DbaseFieldList> result = JDBCHelper.getInstance().runSelectStatement(query,
       // (Connection) ctx.getTransientVariable(FlowVariables.CONNECTION));
       if (result == null || result.isEmpty()) {
       return null;
       }
       // always take the first one found! usually there should be only one task in existence
       // long taskId = Long.parseLong(result.iterator().next().get("ID_").getValueAsString());
       Long taskId = (Long) result.get(0);
       if (logger.isLoggable(Level.FINEST))
       logger.finest("Found a Task with ID: " + taskId);
       TaskInstance task;
       try {
       task = TaskManager.loadInstance(taskId, jbpmContext);
       if (logger.isLoggable(Level.FINEST))
       logger.finest(task.toString());
       }
       catch (ApplicationException ae) {
       // the task is either being opened by a user of is being finished by a user.
       Long userId = (Long) ctx.getTransientVariable(FlowVariables.USER_ID);
       if (userId != null) {
       String folderId = (String) ctx.getVariable(FlowVariables.FOLDER_ID);
       String text = ServerSideResources.translate("Msg38") + " " + folderId;
       Set<String> users = new HashSet<String>();
       users.add(userId.toString());
       MessageManager.getInstance().sendMessageToUsers(users, text);
       }
       return null;
       }
       return task;
       }
      
      public static EmiTaskInstance loadInstance(long taskId, JbpmContext ctx) throws ApplicationException {
       EmiTaskMgmntSession mgmtSession = (EmiTaskMgmntSession) ctx.getTaskMgmtSession();
       try {
       return (EmiTaskInstance) mgmtSession.loadTaskInstance(taskId);
       }
       catch (JbpmException e) {
       throw new ApplicationException("ALREADY_ASSIGNED", e.getMessage());
       }
       }
      
      


      In one process I call a subprocess. Before the subprocess is called I run an action which closes a Task (Hence Task A). At the beginning of the subprocess I run an action which runs the locateTask from above to locate open tasks that have a certain value. The select I run using the open hibernate session returns finding Task A although Task A is closed i.e has an end date, and is not marked as open or suspended. More over, on the next couple of lines of code I load the task with the id the select returned and try to resume it, this fails on exception that says I can't resume a task that has been closed. How can the select find this task if it closed?

      I'm running a Jbpm 3.2 in a clustered environment, which means I'm using an XA DataSource and XA transactions. A Transaction is opened before I begin and the commit is only after The subprocess has reached a blocking state.

      The only thing I can think of is that for some reason the connection used for the query is not the connection used for loading and closing the task. I also want to mention that the same exact code does not fail in a non clusterd environment.

      I'll be happy to read any thoughts on this matter that can point me to a direction....