JDK17-24: zLinux: Segmentation fault seen while changing focus between awt elements

[adamfarley]

Summary A Segmentation fault can be seen on zLinux machines when the user switches focus between multiple awt elements over 20 times.

Details

Here are some useful facts:

1. This issue is seen across several zLinux machines, and I've yet to find a machine that this issue doesn't occur on.

2. On JDK17, this issue is first seen on jdk-17.0.13+1, and is not present in jdk-17.0.12+7.

• Note: There are 881 file changes during this period. List here.

3. While this issue is seen clearly in builds that far back, I see no sign of this issue in the March release of JDK24, so it's possible that this is some form of machine configuration issue.

4. I don't think it's a memory problem, as adding the following code to the testcase's iterative loop shows no lack of free memory within the jvm.

Runtime.getRuntime().freeMemory();

5. The test runs to completion 100% of the time if the JIT is disabled.

6. The test passes on JDK11, but fails on JDK 17, 21, and 24.

Minimal testcase

Prep: Make sure you have a display set up, and that the DISPLAY environment variable is set to the appropriate value.

Command: java FocusTest.java

FocusTest.java.txt

Success: Test passed at 100 iterations.

Failure: Test fails with Segmentation fault.

[adamfarley]

I've requested access to an aqa-tests test machine here so I can see if the issue occurs on aqa-test machines as well.

If it doesn't, it could help us identify a potential machine config issue.

Update: It does happen on the aqa-tests machine. I'm raising an issue upstream and deprioritising this issue in favor of other work.

[sxa]

Update: It does happen on the aqa-tests machine. I'm raising an issue upstream and deprioritising this issue in favor of other work.

@adamfarley Do you add a link to the upstream issue?

[adamfarley]

Update: It does happen on the aqa-tests machine. I'm raising an issue upstream and deprioritising this issue in favor of other work.

@adamfarley Do you add a link to the upstream issue?

Ah yes, here you go: https://bugs.openjdk.org/browse/JDK-8356643

Thanks for the reminder.