<Swing Dev> [14] RFR JDK-8231084:Large performance regression in SwingMark TextArea in 14-b13

[Prasanta Sadhukhan]

http://cr.openjdk.java.net/~psadhukhan/8231084/webrev.1/ problemlisting 
the original testcase to be fixed in new CR JDK-8233177 
<https://bugs.openjdk.java.net/browse/JDK-8233177> (clone of 8001470).

Regards
Prasanta
On 30-Oct-19 6:03 AM, Sergey Bylokhov wrote:
> I suggest to attach these tests to the new CR just as
> an additional verification step for the future fix.
>
> On 10/29/19 11:03 am, Phil Race wrote:
>> Ok then, keep just the one on the problem list.
>>
>> -phil.
>>
>> On 10/28/19 10:22 PM, Prasanta Sadhukhan wrote:
>>> We probably could keep bug8001470.java testcase but would like to 
>>> take a relook at the test during fix of the original issue. The 
>>> original issue was JDK-8001470was identified as generic but the 
>>> testcase fails only in linux even after the removal of fix. Other 
>>> testcases was for the regressions and since we are rolling those 
>>> back, my take is not to keep them.
>>>
>>> Regards
>>> Prasanta
>>> On 28-Oct-19 11:19 PM, Phil Race wrote:
>>>> Can we just problem list the tests rather than deleting them ?
>>>> Won't they be uiseful for the new version of the fix ?
>>>> Perhaps the JEditorLayoutPaneTest.java won't be needed if
>>>> it was only to confirm the first regression was resolved, but I am 
>>>> not sure.
>>>>
>>>> -phil.
>>>>
>>>>
>>>> On 10/23/19 4:07 PM, Prasanta Sadhukhan wrote:
>>>>> Hi All,
>>>>>
>>>>> Please review a rollback of the bug fix which caused large 
>>>>> performance regression for TextArea.
>>>>>
>>>>> JDK-8226513 removes the zero component size check to layout the 
>>>>> root view when text component is updated by resetting the 
>>>>> rootViewNeedsLayout flag, but it caused textarea performance 
>>>>> regression as rootview initialization is triggered multiple times.
>>>>>
>>>>> JDK-8226513 was a regression of JDK-8160246, which was again 
>>>>> regression of JDK-8130892, which in turn was a regression of 
>>>>> JDK-8001470
>>>>> where view's preferred size is set only when text component's 
>>>>> height or width is 0. This check is modified in subsequent fixes 
>>>>> which caused more regressions.
>>>>>
>>>>> Proposed fix for performance regression is to roll back to 
>>>>> original issue and figure out why negative insets are obtained for 
>>>>> JEditorPane.getPreferredSize()
>>>>>
>>>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8231084
>>>>>
>>>>> webrev: http://cr.openjdk.java.net/~psadhukhan/8231084/webrev.0/
>>>>>
>>>>> Changeset of 8001470 : 
>>>>> http://hg.openjdk.java.net/jdk9/jdk9/jdk/rev/ff8622450b29
>>>>>
>>>>> A new bug JDK-8232917 has been created to investigate a fix for 
>>>>> the original issue JDK-8001470.
>>>>>
>>>>> Regards
>>>>> Prasanta
>>>>>
>>>>>
>>>>
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.java.net/pipermail/swing-dev/attachments/20191030/95bc9e6d/attachment-0001.html>