Petr,
according to the simplified way of finding the top-level, have you
tried this code with applets?
I'm not sure this will work there, applet's top level component is
instance of Window.
With best regards,
Alexander Zuev
On 1/28/13 17:18, Petr Pchelko wrote:
Hello, Leonid, Alexander.
Artem told me that I could ask you for the second review on this fix.
Could you please have a look?
For your convenience:
http://bugs.sun.com/view_bug.do?bug_id=8005405
The fix is available at:
http://cr.openjdk.java.net/~pchelko/8005405/webrev.02/
<http://cr.openjdk.java.net/%7Epchelko/8005405/webrev.02/>
Thank you.
With best regards. Petr.
On Jan 25, 2013, at 1:56 PM, Sergey Bylokhov wrote:
Hi, Petr.
Fix looks good.
25.01.2013 12:39, Petr Pchelko wrote:
Hello, AWT Team.
Sorry for this mess with updates, please review the updated fix for
the issue:
http://bugs.sun.com/view_bug.do?bug_id=8005405
The new version is available at:
http://cr.openjdk.java.net/~pchelko/8005405/webrev.02/
<http://cr.openjdk.java.net/%7Epchelko/8005405/webrev.02/>
As Sergey noticed there was a bug in the previous version: the
getParent method could return an owner of the window an coordinates
would be computed incorrectly. In the updated version this issue is
solved and the code is further simplified.
Thank you.
With best regards. Petr.
On Jan 24, 2013, at 7:35 PM, Petr Pchelko wrote:
Hello, AWT Team.
Please, review an updated fix for:
http://bugs.sun.com/view_bug.do?bug_id=8005405
at
http://cr.openjdk.java.net/~pchelko/8005405/webrev.01/
<http://cr.openjdk.java.net/%7Epchelko/8005405/webrev.01/>
Sergey suggested to simplify the loop which computes the component
offset. I simplified it and deleted the check that a peer is an
instance of LWComponentPeer because as I understand it is always
true for a root component in the hierarchy on the mac.
The updated fix is tested on toy apps and on netbeans.
With best regards, Petr.
On Jan 23, 2013, at 6:35 PM, Petr Pchelko wrote:
Hello, AWT team.
Please, review the fix for the issue
http://bugs.sun.com/view_bug.do?bug_id=8005405
The fix is available at:
http://cr.openjdk.java.net/~pchelko/8005405/webrev.00/
<http://cr.openjdk.java.net/%7Epchelko/8005405/webrev.00/>
2 problem existed:
1. Calculation of the dragOrigin and componentOffset relied on the
component.isLightweight() method, which considers all the AWT
components heavyweight on Mac, however we really wanted there to
find the component which has a real NSView or NSWindow under it.
Replacing it with instanceof Window solves the problem.
2. On the native level the dragOrigin and location of the
dragEvent were calculated without respect to the fact than Cocoa
coordinate system is flipped.
The fix is tested on toy apps with both AWT and Swing components.
Also I have run netbeans on the JDK with this fix and all drag
images look good.
With best regards. Petr.
--
Best regards, Sergey.
