<Swing Dev> [9] Review Request: 8144166 [macosx] Test java/awt/Component/CompEventOnHiddenComponent/CompEventOnHiddenComponent.java fails

[Semyon Sadetsky]

On 3/11/2016 12:40 PM, Sergey Bylokhov wrote:
> On 11.03.16 9:30, Semyon Sadetsky wrote:
>> Hi Sergey,
>>
>> Could you explain why do you suppose that the icon cannot be requested
>> more then once between the frame minimize and restore?
>
> Every time when the frame is minimized the new 
> AquaInternalFrameDockIconUI is created. Every 
> AquaInternalFrameDockIconUI have its own ScaledImageLabel(which is 
> JLabel). At the beginning this label is initialized by the icon from 
> the cache, and this label is used until the frame is restored(cache is 
> not used).
What if user requests the dock icon to paint it somewhere else while the 
frame is hidden? Before the fix the user got the same icon image from 
the cache, but after the fix he gets another image because it is being 
re-created on each paint. So it becomes impossible to paint the same 
icon that is shown in the dock currently.
>
>>
>>  > The bug is in the last step. When the frame will be minimized in the
>> second time, it can have another content. But minimized icon
>>  > will show the miniature from the previous minimization...
>> Doesn't it contradict with :
>>  >    * If the frame will be resized/show the cache will be reset.
>> ?
>>
>> --Semyon
>>
>> On 3/10/2016 11:35 PM, Sergey Bylokhov wrote:
>>> Hello.
>>> Please review the fix for jdk9.
>>>
>>> Problem description:
>>>  - CompEventOnHiddenComponent test was written not according the
>>> specification but according an assumption(performance related) - "the
>>> component should not post move/resize events if no listeners were
>>> registered". The test creates a JInternalFrame and adds
>>> ComponentListener to it. No events should come to the listeners,
>>> because the frame is not resized or moved by the test(except initial
>>> layout). Initial move/resize events are skipped usually, because
>>> during initialization there are no any listeners. But this is not the
>>> case in Aqua. AquaInternalFrameDockIconUI adds a listeners to the
>>> JInternalFrame during initialization -> move/resize events are posted
>>> -> the test adds own listeners -> events dispatched -> the test
>>> listeners called -> boom.
>>>
>>> According above description the bug could be closed as not a defect.
>>> But I found another related problem in the AquaInternalFrameDockIconUI.
>>>
>>>  - These component listeners in AquaInternalFrameDockIconUI are used
>>> to maintain the cache of image for the dock(the icon which is used
>>> when the internal frame is minimized). The logic is next:
>>>    * Before the frame will be minimized it draws to the special
>>> ImageIcon which will be used as an icon for the dock.
>>>    * After the frame is minimized and dock is showing the saved image
>>> will be used(and placed to the cache)
>>>    * If the frame will be resized/show the cache will be reset.
>>>    * When in the next time the same frame will be minimized the cached
>>> image will be used.
>>>
>>> The bug is in the last step. When the frame will be minimized in the
>>> second time, it can have another content. But minimized icon will show
>>> the miniature from the previous minimization. Note this is the only
>>> step where the cache is used.
>>>
>>> Solution:
>>> As a solution all cache related stuff were dropped. The new test was
>>> added.
>>>
>>>
>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8144166
>>> Webrev can be found at:
>>> http://cr.openjdk.java.net/~serb/8144166/webrev.01
>>>
>>
>
>