Review request for 7124393: [macosx] JCheckBox in JTable: checkbox doesn't alaways respond to the first mouse click

[Alexander Potochkin]

Hello Anthony
> Hi Alex,
>
> I'm OK with realSync() calls, but unsure about the stopCellEditing() 
> call. Indeed, using a Swing method should produce correct results. 
> However, I think that this regression test is supposed to emulate 
> user's actions, not Swing calls, isn't it? If so, we should either 
> find a cross-platform events that finish editing and use Robot to 
> emulate them, or use the "os.name" property and synthesize different 
> user actions on different platforms.
>
> Please correct me if my assumption is wrong.

Your assumption is correct
it is indeed better to always emulate the user's action.

However in this case it is not as simple because Aqua LaF doesn't 
provide a key shortcut which stops the cell editing
see BasicLookAndFeel line 1545, F2 is bound to "startEditing" action (it 
calls stopEditing() method)
there is no such an action bound to any key in  
AquaKeyBinding.getTableInputMap()

So calling stopCellEditing inside the test is a good tradeoff
because it is simple,  makes the test compatible with every LaF
and doesn't interfere with what the *bug6711682 *actually tests

Thanks
alexp

>
> -- 
> best regards,
> Anthony
>
> On 1/23/2012 9:33 PM, Alexander Potochkin wrote:
>> Hello
>>
>> Please review the following fix:
>> http://cr.openjdk.java.net/~alexp/7124393/webrev.00/
>>
>> the bug's description is here:
>> http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=7124393
>>
>> The bug is in the test, F2 doesn't stop editing on MacOS,
>> we should not use the keyboard to commit the data
>>
>> the test is also updated
>> to access the Swing methods on the Event Dispatching Thread only
>>
>> Thanks
>> alexp