On Sun, 25 Apr 2021 16:27:46 GMT, Pankaj Bansal <pban...@openjdk.org> wrote:
> The tests a javax/swing/JComboBox/8033069/bug8033069ScrollBar.java and
> javax/swing/JComboBox/8033069/bug8033069NoScrollBar.java fails
> intermittently. The tests have a few problems which could cause this
> behaviour and the are being fixed here.
>
> 1. Test access the Swing components without EDT thread at couple of places.
> 2. Test is not calling waitForIdle or delay after Robot.mouseMove operations.
>
> There is some additional cleanup done in the fix and moved the frame in
> centre. The tests pass with multiple iterations on CI. Link in the JBS.
test/jdk/javax/swing/JComboBox/8033069/bug8033069NoScrollBar.java line 64:
> 62:
> 63: private Point p;
> 64: private Dimension d;
Both should be `volatile` as they're accessed from two different threads.
test/jdk/javax/swing/JComboBox/8033069/bug8033069NoScrollBar.java line 120:
> 118: d = cb1.getSize();
> 119: });
> 120: robot.waitForIdle();
Is it needed here?
The block above does not modify the state of the UI.
test/jdk/javax/swing/JComboBox/8033069/bug8033069NoScrollBar.java line 123:
> 121:
> 122: robot.mouseMove(p.x + d.width / 2, p.y + d.height / 2);
> 123: robot.waitForIdle();
I believe the delay between mouseMove and mousePress/Release is unneeded, does
it really make a difference?
test/jdk/javax/swing/JComboBox/8033069/bug8033069NoScrollBar.java line 175:
> 173: if (frame != null) {
> 174: SwingUtilities.invokeAndWait(frame::dispose);
> 175: }
The frame is still accessed from two different threads.
-------------
PR: https://git.openjdk.java.net/jdk/pull/3678
