<AWT Dev> [8] Review request for 7177173: [macosx] JFrame.setExtendedState(JFrame.MAXIMIZED_BOTH) not working as expected in JDK 7

[Artem Ananiev]

That's not the clearest code I've ever seen... Is it possible to 
refactor it to make easier to review and support? For example, replace 
zoom() with two separate methods, maximize() and restore(), or zoom() 
and unzoom(), or whatever. Each of those two methods would be no-op, if 
the window is already in the required state.

isZoomed() method is not that trivial as well. In each of the two 
methods, I would have isUndecorated() condition, and based on the 
result, check for "isZoomed" or "normalBounds" field... Something like that.

I would also enhance the test to include both decorated and undecorated 
frames.

Thanks,

Artem

On 7/4/2012 7:30 PM, Anthony Petrov wrote:
> Hi Sergey,
>
> Please find an updated webrev at:
>
> http://cr.openjdk.java.net/~anthony/8-36-MaximizedBoth-7177173.2/
>
> I renamed the local variable from isZoomed to doZoom.
>
> --
> best regards,
> Anthony
>
> On 7/4/2012 7:26 PM, Sergey Bylokhov wrote:
>> Hi,Anthony.
>> I just point to the same name "isZoomed", which contains different state:
>> first is "true" for this.normalBounds == null and the second "true"
>> for this.normalBounds != null.
>> This is strange.
>> I believe that correct code should be isZoomed = isZoomed();?
>>
>> 04.07.2012 19:15, Anthony Petrov wrote:
>>> Hi Sergey,
>>>
>>> Thanks for the review.
>>>
>>> On 7/4/2012 7:07 PM, Sergey Bylokhov wrote:
>>>> Why there are two different methods for isZoomed?
>>>>
>>>> 481 final boolean isZoomed = *this.normalBounds == null*;
>>>
>>> Here we determine a new state that needs to be set to an undecorated
>>> window. This is what isZoomed() will return after the zoom() method
>>> returns.
>>>
>>>> But in the method we use another code.
>>>>
>>>> 471 private boolean isZoomed() {
>>>> 472 return zoomed || *this.normalBounds != null*;
>>>> 473 }
>>>
>>> This is a method that returns the current zoomed state for any window
>>> regardless whether it is decorated or not.
>>>
>>>> Probably it can be unified?
>>>
>>> No. These are completely different operations. We might change the
>>> first line to "final boolean isZoomed = !isZoomed();", but the
>>> current code looks a lot clearer to me.
>>>
>>> --
>>> best regards,
>>> Anthony
>>>
>>>
>>>>
>>>>
>>>> 04.07.2012 18:54, Anthony Petrov wrote:
>>>>> Hello,
>>>>>
>>>>> Please review a fix for
>>>>> http://bugs.sun.com/view_bug.do?bug_id=7177173 at:
>>>>>
>>>>> http://cr.openjdk.java.net/~anthony/8-36-MaximizedBoth-7177173.1/
>>>>>
>>>>> The bug synopsis may sound misleading: the original problem with
>>>>> applying the extended state after showing a frame has been resolved
>>>>> a while back with another fix. The only issue currently left is
>>>>> that the state is reset when a frame is disposed/hidden, which
>>>>> prevents apps from tracking the last known state of a window.
>>>>>
>>>>> With this fix we avoid canceling/reapplying the maximized state to
>>>>> windows on each showing/hiding operation. Instead we track the
>>>>> state in the CPlatformWindow instance, and apply it on showing of
>>>>> the window if the state differs from what shared code expects. The
>>>>> behavior now matches that of Apple JDK (including handling of the
>>>>> ICONIFIED state).
>>>>>
>>>>> --
>>>>> best regards,
>>>>> Anthony
>>>>
>>>>
>>>> --
>>>> Best regards, Sergey.
>>
>>