Hi Sergey, I had to update the test, since on Mac, it is always selecting the second element, and the test fails.
Thanks, Krishna -----Original Message----- From: Sergey Bylokhov Sent: Monday, October 1, 2018 7:57 AM To: Krishna Addepalli <krishna.addepa...@oracle.com> Cc: Ajit Ghaisas <ajit.ghai...@oracle.com>; Shashidhara Veerabhadraiah <shashidhara.veerabhadra...@oracle.com>; awt-dev@openjdk.java.net Subject: Re: <AWT Dev> [12]RFR : JDK-8014503: AWT Choice implementation should be made consistent across platforms. Hi, Krishna. Why did you update the test? Before the fix the robot used the middle of the first element, but after the fix it will click on the first pixel of the first element(it can missclick on the choice itself), not sure it will be stable enough. On 21/09/2018 15:30, Krishna Addepalli wrote: > Hi Sergey, > > I have not been able to find any example of different behaviour > between index based and object based logic, so I have made the changes > on Mac as you suggested. Also, on Linux, I found that the Choice.java > class holds a selectedIndex already, so I removed it in XChoicePeer.java, and > tested it. > Here is the new webrev: > http://cr.openjdk.java.net/~kaddepalli/8014503/webrev03/ > <http://cr.openjdk.java.net/%7Ekaddepalli/8014503/webrev03/> > > Thanks, > Krishna > >> On 23-Aug-2018, at 4:34 PM, Sergey Bylokhov >> <sergey.bylok...@oracle.com <mailto:sergey.bylok...@oracle.com>> wrote: >> >> Hi, Krishna. >> On 20/08/2018 04:30, Krishna Addepalli wrote: >>> Yes, the behavior is not because of the fix I made. I was trying to >>> understand the implications of the index based selection approach, >>> and now it looks to be consistent on all the platforms. >>> On Mac, as per your suggestion, we can get the same behavior by >>> removing the "setSelectedItem", but my question is, should I make >>> this change or keep the index based change, since that would keep >>> the implementation also consistent across platforms. >> >> Can you please provide an example which will show the difference >> between these two solutions. >> >> btw initially the code in setSelectedItem(..) was added to fix the >> opposite bug. And current fix will skips some events which are >> generated to fix the opposite bug, it looks strange. >> >>> Thanks, >>> Krishna >>> -----Original Message----- >>> From: Sergey Bylokhov >>> Sent: Tuesday, August 14, 2018 7:52 AM >>> To: Krishna Addepalli <krishna.addepa...@oracle.com >>> <mailto:krishna.addepa...@oracle.com>>; Ajit Ghaisas >>> <ajit.ghai...@oracle.com <mailto:ajit.ghai...@oracle.com>>; >>> Shashidhara Veerabhadraiah <shashidhara.veerabhadra...@oracle.com >>> <mailto:shashidhara.veerabhadra...@oracle.com>>; >>> awt-dev@openjdk.java.net <mailto:awt-dev@openjdk.java.net> >>> Subject: Re: <AWT Dev> [12]RFR : JDK-8014503: AWT Choice >>> implementation should be made consistent across platforms. >>> On 13/08/2018 11:06, Krishna Addepalli wrote: >>>> The fix I propose is modeled after the code in Windows, which >>>> tracks items by their selected index. The code in Mac is storing >>>> currently selected object, which could raise the question about >>>> which object to select next, when the selected object is removed/changed. >>>> With index based approach, I see that, the index remains unchanged, >>>> even if the object at that location is removed/changed, unless it >>>> causes the index to be out of bounds. >>>> This behavior is the same across Windows, Linux and Mac. >>> I guess behavior is the same, because of the shared code in >>> Choice->remove->removeNoInvalidate >>> which selects the new element if the old selected element was removed. >>> And it is not affected by the code changed in the fix. >> >> >> -- >> Best regards, Sergey. > -- Best regards, Sergey.
