On Fri, 20 Feb 2026 19:24:34 GMT, Alexey Ivanov <[email protected]> wrote:

> I added capturing a screenshot of the splash screen into the test, and I 
> found that the splash screen is already closed when the screenshot is taken. 
> I realised that I had changed the behaviour of the test with my screenshot 
> code. The original test gets the color of the pixel on the screen before it 
> calls `getScaleFactor` that displays a dialog, which results closing the 
> splash screen.
> 
> https://github.com/openjdk/jdk/blob/e0b040a6c6713827033e9ba51c9ded920dd0203b/test/jdk/java/awt/SplashScreen/MultiResolutionSplash/MultiResolutionSplashTest.java#L110-L111
> 
> I analysed the code of the splash screen, and it never opens the images with 
> decorations for 100% scale: `[email protected]` or `[email protected]`.
> 
> To avoid any confusion, I modified the test code to ensure it never creates 
> decorated files for 100%. If the scale factor of the main screen is 1.00, the 
> second image remains with the default `@2x` decoration.
> 
> I also simplified the test. Now, `getScaleFactor` doesn't display a dialog, 
> but reads the scale from `GraphicsEnvironment`. It did so anyway because the 
> `Graphics` object passed to `paint` is an instance of `SunGraphics2D` in 
> nearly all the cases.
> 
> The test always creates a screenshot of the splash screen and reads the color 
> from the screenshot to determine whether the test fails or not.
> 
> The updated test is stable on all the platforms, I ran it many times, and the 
> test never failed.
> 
> If the test ever fails again, we'll have the screenshot to analyse why.

Marked as reviewed by prr (Reviewer).

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

PR Review: https://git.openjdk.org/jdk/pull/29851#pullrequestreview-3884406258

Reply via email to