Artem Ananiev wrote:
On 1/26/2010 2:57 PM, Pavel Tisnovsky wrote:
Artem Ananiev wrote:
I've added synchronization of scrolling using method
SunToolkit.realSync(), it seems to works perfectly on RHEL x86_64 +
Gnome and Fedora 10 i386 + Gnome. Thanks Anthony Petrov for his tip!
the problem looks really, really strange... Util.waitForIdle() is
implemented exactly the same way: as a call to realSync() - while you
wrote that calling waitForIdle() didn't help.
Please, postpone your fix (don't integrate into AWT gate repository)
and check if Util.waitForIdle() does the job. realSync() is a
low-level and private Sun's API, so I'd prefer having it in a single
place (in Util) so we can change it easily, if necessary (e.g. when
realSync() becomes public).
Yes, you are right, Util.waitForIdle() simply calls realSync(). It is
interesting - without this sync (waifForIdle/realSync, does not matter
which I use) this test always fails on my machine (i.e. it does not
selects all lines of TextArea), with sync included it fails "only" in
circa 1/10 of its run. At this time, I don't plan to push this patch as
it isn't 100% correct.
realSync() only deals with events - native and Java - but it doesn't
cover things like native animations (e.g. on Vista/Seven with Aero
enabled, a window is not minimized instantly, but can be animated, and
realSync() couldn't be used to track this animation is over or in
progress). That's why sometimes, in rare cases, a simple Thread.sleep()
is a way to go.
Hmm, I see, the same problem can occur on Gnome+Compiz, for example. So
do I have to just add Thread.sleep() after each mouse move +
waitForIdle(), at least in this case? I tried that solution - whole text
in TextArea are selected during only 2-3 mouse moves due to autoscroll
feature (which is correct for this test case, IMHO).
Thanks,
Artem
Thanks,
Artem
This patch can be applied to OpenJDK7 too.
Thanks in advance
Pavel Tisnovsky
Red Hat QA
