RFR: 8346581: JRadioButton/ButtonGroupFocusTest.java fails in CI on Linux

[Alexey Ivanov]

On Thu, 9 Jan 2025 18:52:58 GMT, Damon Nguyen <dnguyen at openjdk.org> wrote:

>> **Problem:**
>> 
>> The `javax/swing/JRadioButton/ButtonGroupFocus/ButtonGroupFocusTest.java` test fails in CI on Linux.
>> 
>> The focus is on _Button 2_ instead of _Button 4_.
>> 
>> **Root cause:**
>> 
>> The additional logging revealed, an expected `KEY_PRESS` event. The test uses `Robot` API to press the <kbd>Tab</kbd> key and release it. When the test fails, there are two `KEY_PRESS` events followed by a single `KEY_RELEASE` event. Because the <kbd>Tab</kbd> key is pressed twice, the focus moves twice: from _Button 2_ (the initial state) to _Button 4_ and then back to _Button 2_.
>> 
>> **Fix:**
>> 
>> Use `CountDownLatch`es to track whether a radio button received focus. Detect the case where two `KEY_PRESS` events moved focus to _Button 2_ and report the failure.
>> 
>> Log focus movements and dispatched key events to facilitate failure analysis.
>> 
>> Take a screenshot of the test frame in case of failure.
>> 
>> These CI hosts seem to be quite slow, removing the delay added by `robot.setAutoDelay(100)` has reduced the number of unexpected `KEY_PRESS` events. This didn't affect Windows or macOS.
>
> test/jdk/javax/swing/JRadioButton/ButtonGroupFocus/ButtonGroupFocusTest.java line 172:
> 
>> 170:             BufferedImage image = robot.createScreenCapture(getFrameBounds());
>> 171:             ImageIO.write(image, "png",
>> 172:                           new File("image.png"));
> 
> Does this overflow the 80 character line limit? Or is this a standard?

Yes, it does… by two characters. I find this way easier to parse anyway.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/22977#discussion_r1909308082