[OpenJDK 2D-Dev] [14] Review Request: 8235547 Video memory leak in the OGL pipeline on macOS

[Philip Race]

OK, +1

-phil.

On 12/9/19, 11:15 AM, Sergey Bylokhov wrote:
> On 12/9/19 10:23 am, Phil Race wrote:
>> How did this get found ? Is it a regression caused by some other bug 
>> fix ?
>
> I had run the test for the old bug, to check that we do not have new 
> regressions.
> And found this one by running a simple system monitor, the usage 
> memory grows rapidly.
> Behavior is the same for all JDK I have tested 8, 11, 12, 13, 14.
>
> I think that one of the reasons why I have found it is that the driver
> starts to use systems memory on my machine instead of video memory, 
> probably this is a
> reason of performance degradation in OGL pipeline which we discussed 
> some day ago.
>
>> Do we have any reports of bugs that can be attributed to this ?
>>
>> How did you test that this does not cause any problems (crashes) ?
>
> I have run jtreg/jck on my system, no new issues were found.
>
>>
>> -phil.
>>
>> On 12/8/19 6:36 PM, Sergey Bylokhov wrote:
>>> Hello.
>>> Please review the fix for JDK 14.
>>>
>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8235547
>>> Fix: http://cr.openjdk.java.net/~serb/8235547/webrev.01
>>>
>>> Every window's peer on macOS has a native OGL surface which is used 
>>> as a back buffer
>>> for CAOpenGLLayer. This surface should be disposed in two cases, 
>>> when the layer
>>> is resized or if the layer is disposed. But the second case is not 
>>> implemented,
>>> the OGL surface is leaked when the layer is dispoed.
>>>
>>
>
>