<AWT Dev> Review request for 4908075: Press shift and another key using robot does not trigger events properly

[Volker Simonis]

On Tue, Jul 26, 2016 at 1:28 AM, Sergey Bylokhov
<Sergey.Bylokhov at oracle.com> wrote:
> On 15.07.16 17:05, Volker Simonis wrote:
>>
>> However, I'm a little bit worried about the test. Even on Windows, the
>> various tested key-combination may get be partially intercepted by
>> other applications and this can leave the whole system in an
>> inconsistence state. For example on my Windows machine with Intel
>> Graphics the key combinations "Ctl-Alt-F11" and "Ctl-Alt-F12" trigger
>> some Intel Graphics utilities by default. If I run the test, some key
>> events of the test are intercepted by those utilities with the result
>> that the test fails and no more keyboard input will be possible at
>> all.
>
>
> Note that we have some configuration steps for test systems. And one of them
> is to disable default system shortcuts, otherwise it is not possible to test
> some key combinations. Since the fix was pushed already it is not a big
> problem, but note that the provided test will never be run on any systems,
> which means that this functionality can be easily broken(and is not
> necessary to work on linux/osx/solaris/ and others).
>

Hi Sergey,

thanks for your feedback. You're right that this test is now not
running on any system by default and I'm not happy with it. However,
after trying it on various Windows systems and after the test locked
my system several times in a way I even had to reboot, it was simply
to dangerous for me to enable it by default. It is good that you have
a special configuration for test systems, but how can we mark a test
to run only in such an environment? I just wanted to prevent an
ingenuous user who simply runs the regression test to hit these
problems. If there's a way to classify a test such that it only runs
in specific environments, please let me know.

Regards,
Volker


>
>>> ---------- Ursprüngliche Nachricht ----------
>>> Von: Alexandr Scherbatiy <alexandr.scherbatiy at oracle.com>
>>> An: Oliver Bendig <java at obendig.de>, Sergey Bylokhov
>>> <sergey.bylokhov at oracle.com>, awt-dev <awt-dev at openjdk.java.net>
>>> Datum: 11. Juli 2016 um 19:12
>>> Betreff: Re: <AWT Dev> Review request for 4908075: Press shift and
>>> another
>>> key using robot does not trigger events properly
>>>
>>>
>>> The fix looks good to me.
>>>
>>> Thanks,
>>> Alexandr.
>>>
>>> On 7/8/2016 3:09 PM, Oliver Bendig wrote:
>>>>
>>>>
>>>> Ping....
>>>>
>>>> It would be very nice if somebody could have a look at this webrev.
>>>>
>>>>
>>>> Thanks and kind regards, Oliver
>>>>
>>>>
>>>>> Oliver Bendig <java at obendig.de> hat am 29. Juni 2016 um 15:54
>>>>> geschrieben:
>>>>>
>>>>>
>>>>> Hi,
>>>>>
>>>>> here is a test for this. I updated the webrev.
>>>>> Webrev: http://cr.openjdk.java.net/~simonis/webrevs/2016/4908075.v2/
>>>>>
>>>>> This test is for windows only as other OS might catch some of the
>>>>> tested
>>>>> keyboard events before the are passed to the canvas.
>>>>> I hope this is ok as the problem itself occurs on windows only.
>>>>>
>>>>> Regards, Oliver
>>>>>
>>>>>
>>>>>> Sergey Bylokhov <Sergey.Bylokhov at oracle.com> hat am 21. Juni 2016
>>>>>
>>>>> um 21:53
>>>>>>
>>>>>> geschrieben:
>>>>>>
>>>>>> Hi, Oliver.
>>>>>> Is it possible to write a test for this fix?
>>>>>>
>>>>>> On 21.06.16 17:39, Oliver Bendig wrote:
>>>>>>>
>>>>>>> Hi,
>>>>>>>
>>>>>>> can you please review the following fix:
>>>>>>>
>>>>>>> Webrev: http://cr.openjdk.java.net/~simonis/webrevs/2016/4908075/
>>>>>>> BugID: https://bugs.openjdk.java.net/browse/JDK-4908075
>>>>>>>
>>>>>>> Following some mre details: we are facing issues when trying to send
>>>>>>> keyboard events via awt robot. When trying to send a keyboard
>>>>>
>>>>> event from
>>>>>>>
>>>>>>> the extended keys with the SHIFT-key pressed, this doesn't send a
>>>>>>> correct key combination. Instead, the Shift-Key is released
>>>>>
>>>>> before the
>>>>>>>
>>>>>>> second keycode is sent. This makes it impossible to send combinations
>>>>>>> like e.g. shift+delete.
>>>>>>>
>>>>>>> The bug id 4908075 for this issue is rather old. The suggested
>>>>>
>>>>> idea to
>>>>>>>
>>>>>>> switch to SendInput() instead of keybd_event() was delayed at
>>>>>
>>>>> that time
>>>>>>>
>>>>>>> because of missing support in Win98. For testing purposes, I
>>>>>
>>>>> implemented
>>>>>>>
>>>>>>> SendInput instead of keybd_event, but the issue stays the same. The
>>>>>>> problem seems to be caused by the missing KEYEVENTF_EXTENDEDKEY flag
>>>>>>> when calling keybd_event or SendInput. Bug id 8155742 introduces this
>>>>>>> flag for other reason for VK_ALT_GRAPH. If this is enhanced to
>>>>>
>>>>> cover the
>>>>>>>
>>>>>>> extended keys that were introduced as enhancement of the old 84
>>>>>
>>>>> key AT
>>>>>>>
>>>>>>> keyboard, the correct key events are sent.
>>>>>>>
>>>>>>> The bug 4908075 was closed as "Won't fix" but I think the fix
>>>>>
>>>>> would be
>>>>>>>
>>>>>>> rather simple. Should I reopen 4908075 or is it better to create
>>>>>
>>>>> a new
>>>>>>>
>>>>>>> bug for this issue.
>>>>>>>
>>>>>>> Thank you and best regards,
>>>>>>> Oliver
>>>>>>>
>>>>>>
>>>>>> --
>>>>>> Best regards, Sergey.
>>>>>
>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>
>>>>
>
>
> --
> Best regards, Sergey.