<AWT Dev> Regression test jawa/awt/Window/AlwaysOnTopEvenOfWindow failures on OpenJDK6 & Gnome

[Anthony Petrov]

Hi Pavel,

Thank you for your testing efforts!

It looks like the test does not actually belong in JDK 6, and should 
fail there - please refer to the comment section in the very beginning 
of the test source. [1] shows that the fix for which the test is 
written, has been available since 7b03, and isn't present in 6 at all. 
Since 7b20 included all the tests present in 7b03, the test got to 
OpenJDK 6 as well [2]. Weird, I agree.

I doubt if we want to back-port [3] though (which is the actual fix for 
[1]). Therefore, I think the test had better be removed from OpenJDK 6.


[1] http://bugs.sun.com/view_bug.do?bug_id=5028924

[2] http://openjdk.java.net/projects/jdk6/

[3] http://bugs.sun.com/view_bug.do?bug_id=6380835

--
best regards,
Anthony

On 10/19/2009 6:05 PM Pavel Tisnovsky wrote:
> Hi,
> 
> regression test jawa/awt/Window/AlwaysOnTopEvenOfWindow regularly fails 
> on these platforms:
> 
> Fedora 10/11, RHEL 5.3
> i386, x86_64
> OpenJDK6
> * Gnome as desktop environment *
> 
> It looks like that AWT Window constructed by calling new Window(null) is 
> always shown on top (at least on Gnome), regardless of other settings. 
> For example frame.setAlwaysOnTop(true) does not change order of top 
> level containers.
> 
> I'm not sure about Window constructed by new Window(null) correct 
> behaviour. That probably means that this Window does not have parent 
> container, but JavaDoc (only) says: "A window must have either a frame, 
> dialog, or another window defined as its owner when it's constructed."
> 
> When I change line 49 to Frame win = new Frame(); (Frame is constructed 
> instead of Window), test passes.
> 
> My question is: is this regression test written incorrectly or has 
> OpenJDK6 & Gnome bug in top level containers handling?
> 
> Thank you for your answers,
> Pavel Tisnovsky