[RFC][plugin]: Class loading problem.

[Denis Lila]

> If it's a warning and it bothers you, I think we should fix the cause
> (thought that should be a separate patch), not merely suppress the
> warning.

Ok. Although, I don't think we should "fix" it. Doing that would
involve inserting a serialVersionUID field, which we only need
if we wanted to serialize NetxPanel (which we don't). The
compiler generates one itself, so even if we were doing
serialization we might not need to introduce that field.
Not to mention that putting in that field even though we don't
serialize might confuse future readers.


> No, I think it's fine for now. AFAIK, knowing the name of the thread
> group is not sufficient to create other threads in it. As long as
> other applets cant access/use these other threadgroups, we are fine.

Hm... other applets can access these thread groups, but only if
they have the same uKey, and hence share a classloader. Is that ok?

Regards,
Denis.

----- Original Message -----
> On 04/29/2011 01:53 PM, Denis Lila wrote:
> >>> @@ -39,14 +40,51 @@
> >>>     *
> >>>     * @author Francis Kung<fkung at redhat.com>
> >>>     */
> >>> + at SuppressWarnings("serial")
> >>>    public class NetxPanel extends AppletViewerPanel {
> >>
> >> Is there a particular reason you are suppressing this warning?
> >>
> >
> > It's a pet peeve of mine ;)
> > Warning marks in eclipse bother me more than they reasonably
> > should. Should I change it back?
> >
> 

> 
> >>> - // create the frame.
> >>> - PluginAppletViewer.framePanel(identifier, handle, panel);
> >>> + // Framing the panel needs to happen in a thread whose thread
> >>> group
> >>> + // is the same as the threadgroup of the applet thread. If this
> >>> + // isn't the case, the awt eventqueue thread's context
> >>> classloader
> >>> + // won't be set to a JNLPClassLoader, and when an applet class
> >>> needs
> >>> + // to be loaded from the awt eventqueue, it won't be found.
> >>> + Thread panelInit = new Thread(panel.getFutureTG(), new
> >>> Runnable()
> >>> {
> >>> + @Override public void run() {
> >>> + panel.createNewAppContext();
> >>> + // create the frame.
> >>> + PluginAppletViewer.framePanel(identifier, handle, panel);
> >>> + panel.init();
> >>> + // Start the applet
> >>> + initEventQueue(panel);
> >>> + }
> >>> + }, "NetXPanel initializer");
> >>>
> >>> - panel.init();
> >>> -
> >>> - // Start the applet
> >>> - initEventQueue(panel);
> >>> + panelInit.start();
> >>> + while(panelInit.isAlive()) {
> >>> + try {
> >>> + panelInit.join();
> >>> + } catch (InterruptedException e) {
> >>> + }
> >>> + }
> >>>
> >>
> >> This part looks fine to me, but I am not too familiar with the
> >> concurrency requirements and communication between threads that
> >> goes
> >> on in PluginAppletViewer.
> >
> > This code is meant to be as close to the original as possible, with
> > the
> > exception that it should happen in the same TG as the one of the
> > thread
> > that will do the loading in panel.init() (and with a new AppContext
> > so
> > that we get new eventqueue threads for new applets).
> >
> > What do you think about the TODO?
> > I think you suggested something like that to me yesterday. That's
> > why
> > I put the TODO in there. However, what we have now is what we've
> > always
> > had (as far as the TG name goes), and I wasn't sure I should change
> > it.
> >
> 
> 
> Cheers,
> Omair
-------------- next part --------------
A non-text attachment was scrubbed...
Name: awt-classloading.patch
Type: text/x-patch
Size: 16858 bytes
Desc: not available
Url : http://mail.openjdk.java.net/pipermail/distro-pkg-dev/attachments/20110429/434cc0c4/awt-classloading.patch