On Thu, 12 Nov 2026 08:33:39 GMT, Prasanta Sadhukhan <[email protected]> 
wrote:

> Test intermittently fails in linux citing cursor is not RESIZE cursor at 
> tableheader cell boundary when the mouse is moved from one table header cell 
> to another.
>  Although I am not able to get the test to fail locally or in CI standalone 
> but it seems the test can potentially fail as it moves the mouse across the 
> header and checks the cursor whenever the mouse enters another column but the 
> problem is that entering another column doesn’t always mean the cursor should 
> be a resize cursor. The resize cursor appears only very close to the 
> boundary. On Linux, some mouse movements may be combined or skipped, so the 
> test can receive a position already past the boundary. The normal cursor is 
> correct there, but the test reports a failure.
> 
> The test is made more robust:
> - Stops exactly at each column boundary and waits for mouse events to be 
> processed.
> - Checks that the resize cursor is already set when the boundary’s rollover 
> callback rolloverColumnUpdated() runs.
> - Checks that the normal cursor returns inside a column.
> - ui.prepare() ensures at which point mouse is expected, whether resize 
> cursor is expected, clear the event flags and records failure
> - Logs the actual mouse coordinates to help explain any future failure.
> - Also magic number 9 is removed
> 
> This makes the test more robust because it checks the cursor at known 
> positions, instead of assuming where the mouse was when a column change was 
> reported. It still detects the original cursor-update problem
> 
> 100 iterations in linux and other platforms were run and seen to be ok..
> 
> ---------
> - [x] I confirm that I make this contribution in accordance with the [OpenJDK 
> Interim AI Policy](https://openjdk.org/legal/ai).

Can we please get an evaluation in the JBS issue. Not just here.
I already added the testbug label but it needs an evaluation too.

test/jdk/javax/swing/JTableHeader/6889007/bug6889007.java line 48:

> 46:     static JTableHeader header;
> 47:     static MyTableHeaderUI ui;
> 48:     static Point screenPoint;

You removed volatile from all of these but (for example, it may not be the only 
case), I notice that screenPoint is used on both the EDT (invokeAndWait) and in 
moveAndCheck() on the main thread.

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

PR Review: https://git.openjdk.org/jdk/pull/33866#pullrequestreview-5899372584
PR Review Comment: https://git.openjdk.org/jdk/pull/33866#discussion_r4569726020

Reply via email to