<AWT Dev> <Swing Dev> [PATCH] 7168064: SwingUtilities.sharedOwnerFrame multiplies window close event
Jose Luis Martin
jlm at joseluismartin.info
Thu May 9 03:23:08 PDT 2013
I update the Test and include two new TestCases:
- Test if a Window that has been never shown receive the WINDOW_CLOSED
- Test if a child window that never has been shown receive the
WINDOW_CLOSED event on parent dispose().
With the patch applied the second test fails and the doDispose() method
in the child Window is never called so I suppose that the patch is
Looking at code I don't find any reason to call doDispose() on already
disposed window or a in a Window that never has been shown. Is there?
If not, we can simply test if window is displayable on doDispose() but
then the following code will not fire a WINDOW_CLOSED event.
If it's not acceptable, I think that the fix should:
- Fire the event and run the dispose action on never shown Windows.
- Run the dispose action but not fire the event when re-disposed.
boolean fireClosedEvent = isDisplayable() || hasBeenNeverShown;
As Anthony suggested.
-- Jose Luis Martin
El mié, 08-05-2013 a las 16:29 +0400, Anthony Petrov escribió:
> This indeed looks like a bug in AWT. I'm BCC'ing swing-dev@, and CC'ing
> awt-dev@ instead (please subscribe to this mailing list before posting
> to it).
> Here's a question: why do we consider this issue affects child windows
> only? If you call dispose() on a regular window several times in a row,
> you'll receive that many WINDOW_CLOSED events. Isn't it the real bug
> here actually?
> Note that simply guarding a call to dispose()/disposeImpl() with a check
> is not the best option because it changes the behavior. If user code
> overrides certain methods, they may be stopped being called after your
> fix, and this may break that code.
> Since the DisposeAction is executed synchronously anyway, perhaps it
> should set a flag based on which the doDispose() method would decide to
> send this event or not. What do you think? Also, have you run existing
> jtreg tests (in jdk/test/java/awt/) to verify that this fix doesn't
> introduce a regression?
> PS. I can't find the bug filed by you in our bug database. I'll file a
> new one once we review your fix on this mailing list.
> best regards,
> On 05/06/2013 12:28 PM, Jose Luis Martin wrote:
> > Hello,
> > I reported this issue one year ago but seems that it has been removed
> > from the public bug database and is still not fixed so I try now to
> > speak directly with you to see if this really is a bug or not.
> > The problem that I see is that a child window already disposed is
> > re-disposed when the owner frame is disposed, multiplying the
> > WINDOW_CLOSED events.
> > I guess that it could be fixed by testing if the child window is
> > displayable before disposing it in the Window dispose action. (May be
> > better in the dispose method itself).
> > I attach patch over last jdk8 repository and a test case for the issue.
> > Thanks,
> > Best Regards.
> > -- Jose Luis Martin
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 3090 bytes
Desc: not available
Url : http://mail.openjdk.java.net/pipermail/awt-dev/attachments/20130509/921a399f/DialogTest.java
More information about the awt-dev