This Bugzilla instance is a read-only archive of historic NetBeans bug reports. To report a bug in NetBeans please follow the project's instructions for reporting issues.
1. Add connection to MySQL DB, but uncheck "Save Password" checkbox. 2. Disconnect, then connect, specify valid password and click "OK". Result: "Unable to add connection. Connection already exists." message appears and then it starts to appears all the time and only one way to stop it is to restart NetBeans.
Created attachment 70672 [details] IDE's log file.
David, you beat me to it. You're fast so early in the morning.
Don't know why my comment changed the TM. I'm putting it back.
The repeating error message is because the logic to add a connection is fired again and again, each time causing the same exception. If I have it right, the loop is caused by this: - We run the DatabaseConnection.doConnect() method on a request processor thread - This fires the "connected" property change event - This calls the property change listener in ConnectUsingDriverAction.java, in the PropertyChangeListener. - Here, when the event is "connected" we first call "fireConnectionFinished" and then attempt to add the connection to the RootNodeInfo - the attempt to add the connection fails with a "connection already exists" - The "connected" property change event then invokes the next listener, which is ConnectionNodeInfo. This listener tries to attach the CNI to the new active JDBC connection. This includes calling getSpecification - getSpecification() notices that there is no connection (it never got set up) and calls createJDBCConnection() to open the connection - This fires the "connected" property change, which sends us around the circle again - meanwhile fireConnectionFinished causes ConnectionNodeInfo.finishConnect() to be called, which calls createSpecification() which then notices the connection is not valid and thus calls createJDBCConnection() , which create a *whole new connection" to the database
Sorry, some typos in the last comment. I'll rewrite here, fixing the typos so it's more clear: The repeating error message is because the logic to add a connection is fired again and again, each time causing the same exception. If I have it right, the loop is caused by this: - We run the DatabaseConnection.doConnect() method on a request processor thread - This fires the "connected" property change event - This calls the property change listener in ConnectUsingDriverAction.java. - Here, when the event is "connected" we first call "fireConnectionFinished" and then attempt to add the connection to the RootNodeInfo - the attempt to add the connection fails with a "connection already exists" - The "connected" property change event then invokes the next listener, which is ConnectionNodeInfo. This listener tries to attach the CNI to the new active JDBC connection. This includes calling getSpecification - getSpecification() notices that there is no connection (it never got set up) and calls createJDBCConnection() to open the connection - This fires the "connected" property change, which sends us around the circle again
Sigh, can't reproduce on my machine. Why is it all these things are happening for Roman and not for me?
Probably because I have magic machine. :)
411fdef62a9c The reason this reared its ugly head I am pretty sure is because we are now holding on to existing infos across connect/disconnect. The NewConnectionFromDriver action added a connection listener but never removed it. This was fine before when we deleted the CNI on a disconnect, but now that the CNI is retained, the listener attached to it is also retained. This meant it fired when the user reconnects and tries to add the connection, but the connection is already added. Then the connection listener for the connect action fires, and the DriverSpecification detects the connection is null, and tries to create it, and we get into the infinite loop scenario; this won't happen now that I remove the old connection listener from the NewConnectionFromDriver dialog. It makes me think we have the risk of some other memory leaks across multiple connect/disconnects; I'll ask Roman to see if there is some memory testing he can do.
Integrated into 'main-golden', will be available in build *200809300201* on http://bits.netbeans.org/dev/nightly/ (upload may still be in progress) Changeset: http://hg.netbeans.org/main/rev/411fdef62a9c User: David Van Couvering <davidvc@netbeans.org> Log: #148489: Message appears all the time and prevents from work Also a little cleanup of logging and message formatting
Verified with build 080930. Good job, David.