Bugreport: View looses current row if user parameters are passivated!

Question

Hi
I think I've discovered a bug in JDeveloper:
I realized that a ViewObject is loosing its current row under the following circumstances:
- a user parameter is passivated in method "prepareForPassivation" AND
- the ViewObject is passivated, removed, recreated and reactivated (pooling)
See the following example:
  protected void prepareForPassivation(Document out,
                                       Element parent)
    super.prepareForPassivation(out, parent);
    Element userdata = out.createElement("SamUserData");
    parent.appendChild(userdata);
    // create tag named "myData"
    Element elem = out.createElement("mSecurityLevel");
    userdata.appendChild(elem);
    elem.setAttribute("Value", "4);
  } This code makes the current row to be lost if passivation occures. For testing you should set the following parameters:jbo.ampool.resetnontransactionalstate = false
jbo.ampool.dynamicjdbccredentials = false
jbo.dofailover = false
jbo.maxpoolcookieage = -1
jbo.ampool.timetolive = 30000
jbo.ampool.initpoolsize=0
jbo.ampool.maxpoolsize=5000
jbo.recyclethreshold=0
jbo.ampool.monitorsleepinterval=500
jbo.ampool.maxavailablesize=0
jbo.ampool.minavailablesize=0
jbo.ampool.maxinactiveage=1000 This parameters will cause heavy pooling even if you access by two browser sessions only.
If anyone from oracle can have a look at this I'd be happy. Unfortunately I do not know how to workaround this problem, so any hints are appreciated.
Thanks
Frank Brandstetter

Answer 1

And here comes a workaround:
First: All user data will be passivated below a common node ("SamUserData") in "prepareForActivation". See code sample above.
Second: "prepareForActivation" will be overridden to look like this:  protected void prepareForActivation(Element parent) 
    // apply a bugfix to prevent current row to be lost if userdata is passivated
    fixVoActivationBug(parent);
    super.prepareForActivation(parent);
    // find common node "SamUserData"
    Element userdata = (Element) findChildByTagname(parent, "SamUserData");
    // read user value
    Element elem = (Element) findChildByTagname(userdata, "mSecurityLevel");
    mSecurityLevel = elem.getAttribute("Value");
  } And here comes the bugfix code:  public static void fixVoActivationBug(Element parent)
    // find user data
    Element userdata = (Element) parent.getFirstChild();
    if (userdata==null || !userdata.getNodeName().equals("SamUserData"))
      return;  // no userdata found, nothing to do
    // move userdata element to end of parent tag (this is the bugfix)
    parent.removeChild(userdata);
    parent.appendChild(userdata);
  }I'm quite sure that this is a serious bug in the BC4J passivation and activation logic!
Thanks
Frank Brandstetter