Hello,
yeah i found that swing bug as well. Commenting out the code in ClientInfo is not an option for us since we do not have the sources. so the only alternative would be to use openjdk7, but there's no official release yet. At least i couldn't find anything on the openjdk page.
I guess there's no way to configure the ClientInfo to exclude the JColorChooser?
Cheers,
Matthias
-----Peter Arrenbrecht <[email protected]> wrote: -----yeah i found that swing bug as well. Commenting out the code in ClientInfo is not an option for us since we do not have the sources. so the only alternative would be to use openjdk7, but there's no official release yet. At least i couldn't find anything on the openjdk page.
I guess there's no way to configure the ClientInfo to exclude the JColorChooser?
Cheers,
Matthias
To: [email protected], [email protected]
From: Peter Arrenbrecht <[email protected]>
Date: 06/10/2010 08:00AM
Cc: Daniel Grob <[email protected]>
Subject: Re: [ULC-developer] Problem with ULC Remote Tests
On Wed, Jun 9, 2010 at 5:26 PM, <[email protected]> wrote:
> Hello,
> we are running our ULC UI Tests on a Hudson build server and didn't use to
> have much problems. The build server was then migrated to a faster 64Bit
> Ubuntu Machine and since then the tests randomly freeze in almost every
> build. The build environment is only different in the hardware and uses a
> 64bit java version instead of a 32bit one. The tests always freeze in the
> setup method of the AbstractTestCase class. But it's never the same test
> that freezes. It looks like the instantiation of the JColorChooser causes a
> problem. Did anybody experience similar problems? I attached the build
> server info and the thread dump below.
Yes, indeed. Here's a copy of some information I sent to Daniel Grob
(in German):
<cite>
Es scheint im Zusammenhang mit diesen Bugs zu sein:
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4480705
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4177735
Ein Kommentar im zweiten Bug hat mich auf den Workaround gebracht:
"It seems we should force the thread to stop at some moment, and be
able to restart it later, should this be needed. E.g. we could stop
when HSB chooser panel becomes invisible, and restart when it's
made visible again."
Ich hatte es, meine ich, unter jre1.5u18, jre6u13 und jre6u14. Unter
openjdk7 ist es korrigiert. Es scheint, dass dort die ganze
SyntheticImageLoader-Geschichte umgeschrieben worden ist.
Der Fehler wird nur gravierend, wenn man in schneller Abfolge
UISessions erzeugt. Das geschieht natürlich, wenn viele kurze Tests
hintereinander laufen. Die blieben mir dann gelegentlich hängen, da
der Event-Thread wie im ersten Bug beschrieben in
java.awt.MediaTracker.waitForID wartete.
</cite>
I suggested a fix to Canoo, but alas it only fixed the thread leak,
not the freeze, so it was not incorporated. You might want to try
running your tests on OpenJDK 7.
@Dani: For the time being, we have resorted to commenting out
JColorChooser in ClientInfo to run our tests. You might want to make
this configurable.
Cheers,
-peter
_______________________________________________ ULC-developer mailing list [email protected] http://lists.canoo.com/mailman/listinfo/ulc-developer
