<AWT Dev> [9] Review request for JDK-8046391: Hang displaying JFileChooser with Windows L&F

[Phil Race]

I didn't roll back .. I am leaving that to you guys.

-phil.

On 6/10/2014 11:20 AM, Alexey Ivanov wrote:
> Thank you Anthony and Sergey.
>
> Phil hasn't rolled back them yet.
> Now anyone with committer rights can do it.
>
> Regards,
> Alexey.
>
> On 10.06.2014 22:14, Sergey Bylokhov wrote:
>> Looks fine,
>> But double check, probably Phil already rolled back these changes.
>>
>> On 6/10/14 9:58 PM, Anthony Petrov wrote:
>>> +1
>>>
>>> -- 
>>> best regards,
>>> Anthony
>>>
>>> On 6/10/2014 9:52 PM, Alexey Ivanov wrote:
>>>> Hello AWT team,
>>>>
>>>> Could you please review the reverse changeset:
>>>> http://cr.openjdk.java.net/~aivanov/8046391/jdk9/webrev.02/
>>>>
>>>> This undoes the fixes for
>>>>      - JDK-8039383 "NPE when changing Windows Theme", and
>>>>      - JDK-8046239 "Build failure in 9-client on all non-Windows 
>>>> platforms"
>>>> which cause a regression where JFileChooser hangs in Windows 
>>>> Look-and-Feel.
>>>>
>>>> The proper fix for JDK-8039383 requires some more time.
>>>>
>>>>
>>>> Thank you in advance,
>>>> Alexey.
>>>>
>>>> On 10.06.2014 20:22, Sergey Bylokhov wrote:
>>>>> This fix is identical to JDK-8042590, which was fixed in cpu 
>>>>> workspace.
>>>>> I guess the push under the new bugid will lead to integration 
>>>>> problems
>>>>> in case of merge.
>>>>>
>>>>> On 6/10/14 7:21 PM, Alexey Ivanov wrote:
>>>>>> Hello AWT team,
>>>>>>
>>>>>> Could you please review the fix for bug:
>>>>>>     bug: https://bugs.openjdk.java.net/browse/JDK-8046391
>>>>>>     webrev: 
>>>>>> http://cr.openjdk.java.net/~aivanov/8046391/jdk9/webrev.01/
>>>>>>
>>>>>> Description:
>>>>>> The fix for JDK-8039383 "NPE when changing Windows Theme" caused a
>>>>>> regression where JFileChooser hung. The root cause is the deadlock
>>>>>> between Windows Toolkit thread and AWT Event Queue which is 
>>>>>> caused by
>>>>>> the fact that ThemeReader.flush() is called on the toolkit thread.
>>>>>>
>>>>>> The fix:
>>>>>> Modify ThemeReader so that flush() does not do the real work but
>>>>>> marks the current state as invalid. The old theme data are removed
>>>>>> when getTheme() is called which is done only from an Event Dispatch
>>>>>> thread.
>>>>>>
>>>>>> Regards,
>>>>>> Alexey.
>>>>>
>>>>>
>>>>
>>
>>
>