I can update UnixMultiResolutionSplashTest.java to re-use the new logic
of MultiResolutionSplashTest.java.
Setting GDK_SCALE=2 will still show the @2x splash screen. But this does
not correspond to a real-world scenario any more.
Therefore, I don't think it's worth doing.
When https://bugs.openjdk.org/browse/JDK-8375687 is fixed, the splash
screen code will need updating… It is only at that time in the future
that we'll know if it's possible to simulate High DPI env on Linux.
Would you recommend submitting a new bug to remove
UnixMultiResolutionSplashTest.java?
Or should I use JDK-8372546 with a justification?
--
Regards,
Alexey
On 2026-04-21 19:36, Philip Race wrote:
If there's no way to continue to simulate it, then I'm OK to delete it.
-phil.
On 4/20/26 12:48 PM, Alexey Ivanov wrote:
Hello colleagues,
The test
java/awt/SplashScreen/MultiResolutionSplash/unix/UnixMultiResolutionSplashTest.java
has been frequently failing in Oracle CI systems, the failure is
tracked in JDK-8372546 [1].
The test fails due to the same reason why
java/awt/SplashScreen/MultiResolutionSplash/MultiResolutionSplashTest.java
failed. In short, the splash screen may not be on the screen when a
pixel is captured. More details are available in the pull request
#29851 [2].
I updated the generic MultiResolutionSplashTest.java under
JDK-8374304 [3]. The updated test has never failed afterwards.
UnixMultiResolutionSplashTest.java differs from
MultiResolutionSplashTest.java in that it sets the GDK_SCALE
environment variable to 2 to simulate a High DPI screen and then
starts the java launcher.
But GDK_SCALE is no longer set in GNOME, so it doesn't simulate a
real environment any more. Additionally, both macOS and Windows rely
only on the DPI setting of the host.
I propose removing UnixMultiResolutionSplashTest.java.
Does anyone have any comments or objections?
