<AWT Dev> [9] Review request for 8163101: dual-screen issue with JMenu, JPopupMenu

Semyon Sadetsky semyon.sadetsky at oracle.com
Mon Nov 7 15:37:04 UTC 2016

On 11/7/2016 6:13 PM, Sergey Bylokhov wrote:
> On 28.10.16 13:12, Semyon Sadetsky wrote:
>> On 10/25/2016 7:44 PM, Sergey Bylokhov wrote:
>>> On 20.10.16 9:36, Semyon Sadetsky wrote:
>>>>> The code which you tried to revert if not correct. The previous 
>>>>> change
>>>>> was done as a fix for HiDPI support. Lots of our code depends from 
>>>>> the
>>>>> XWindow.scaleUp/scaleDown/getScale which ware implemented on top of
>>>>> graphicsConfig. If it was not updated then this is a serious bug.
>>>> XWindow.scaleUp/scaleDown/getScale do not depend on screen on Linux.
>>> How it is possible since scaleXX use the graphics config of the window?
>> Because on linux we do not support different scales only the same scale.
> This is the question why it is unsupported, since Unity can set 
> different dpi per screen. 
It doesn't work. It only scales the Unity top panel and the launcher. 
Apps windows have always the same scale equals to the main screen scale.
> But this is unrelated. We support a different GraphicsDevice per 
> screen, each device should have its own scale(even if it is the same 
> value). The code which operates on screen bounds should use its own 
> graphics device.
Nope. Java does not support different scales for monitors. Attempt to 
set them different when Xinerama is used will cause unpredictable behavior.
>>>> I proposed to do same to investigate the problem. Does it mean that 
>>>> you
>>>> approve the current fix?
>>> No, this is one is a separate bug. and I already described why.
>> But do you have any other concerns to the current fix?
> The code in question should use the bounds of the screen where the 
> component is located, not a bounds of the main screen.
>>>> But severity of this bug is rather low because it is internal problem
>>>> that does not affect user experience. It may be deferred to the next
>>>> major JDK release.
>>>> Also note, that the same bug was already closed as "won't fix" before:
>>>> https://bugs.openjdk.java.net/browse/JDK-5100801
>>>> --Semyon

More information about the awt-dev mailing list