-
Notifications
You must be signed in to change notification settings - Fork 724
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Re-enable openjdk test - jdk_bean #20531
Comments
Issue Number: 20531 |
@annaibm could you run the test on all platforms? Thanks |
|
The failure on jdk8 is already resolved via ibmruntimes/openj9-openjdk-jdk8#780 @babsingh pls take a look at the jdk11 failures. |
Thanks @pshipton With the latest JDK8, the
|
- Enable jck_beans on xlinux except for java version 8 and 11 related: eclipse-openj9/openj9#20531 Signed-off-by: Anna Babu Palathingal <[email protected]>
OpenJ9 JDK8 and JDK11 currently use OpenJ9 MethodHandles. Enabling jdk_bean in JDK8 and JDK11 triggers certain OpenJDK MethodHandles code paths. However, OpenJ9 MethodHandles are in use for JDK8 and JDK11, so OpenJDK MethodHandles paths shouldn't be activated. To address this, a MethodHandleImpl stub was added, used only with OpenJ9 MethodHandles, and not when OpenJDK MethodHandles are enabled. Related: eclipse-openj9#20531 Signed-off-by: Babneet Singh <[email protected]>
- Enable jck_beans on xlinux except for java version 8 and 11 related: eclipse-openj9/openj9#20531 Signed-off-by: Anna Babu Palathingal <[email protected]>
OpenJ9 JDK8 and JDK11 currently use OpenJ9 MethodHandles. Enabling jdk_bean in JDK8 and JDK11 triggers certain OpenJDK MethodHandles code paths. However, OpenJ9 MethodHandles are in use for JDK8 and JDK11, so OpenJDK MethodHandles paths shouldn't be activated. To address this, a MethodHandleImpl stub was added, used only with OpenJ9 MethodHandles, and not when OpenJDK MethodHandles are enabled. Related: eclipse-openj9#20531 Signed-off-by: Babneet Singh <[email protected]>
- Enable jdk_beans on xlinux except for java version 8 and 11 related: eclipse-openj9/openj9#20531 Signed-off-by: Anna Babu Palathingal <[email protected]>
- Enable jdk_beans on xlinux except for java version 8 and 11 related: eclipse-openj9/openj9#20531 Signed-off-by: Anna Babu Palathingal <[email protected]>
- Enable jdk_beans on xlinux except for java version 8 and 11 related: eclipse-openj9/openj9#20531 Signed-off-by: Anna Babu Palathingal <[email protected]>
- Enable jdk_beans on xlinux except for java version 8 and 11 related: eclipse-openj9/openj9#20531 Signed-off-by: Anna Babu Palathingal <[email protected]>
- Enable jdk_beans on xlinux for all java versions related: eclipse-openj9/openj9#20531 Signed-off-by: Anna Babu Palathingal <[email protected]>
- Enable jdk_beans on xlinux for all java versions related: eclipse-openj9/openj9#20531 Signed-off-by: Anna Babu Palathingal <[email protected]>
|
@dmitripivkine pls take a look at the jdk8 failure. |
@babsingh pls take a look at the jdk11 xmac failure:
|
There are some JIT asserts on xmac jdk21, which matches #20267
|
re #20531 (comment): JDK11 xmac failures have already been fixed via #20570, which was merged 2 days ago. The grinder uses a JDK with changes only till 7 Nov 24 and doesn't include the fix. Using a newer JDK, which includes the fix, should resolve the issue.
|
Regarding OOM case for Test5102804:
This allocation has failed because neither Tenure nor Nursery can take this object. OpenJ9 currently does not support expanding Tenure over Nursery (particularly for fully expanded heap |
@dmitripivkine I think the OOM is intentional. The test catches the OOM and checks if a Reference has been freed. |
Ok, thanks, I missed it. Do we know which Reference it is watching? |
I probably see it: there is Weak Reference stored in O-slot with referent pointed to class object. So I guess it expects |
re #20531 (comment), @annaibm amac runs are invalid. Please rerun with
|
Regarding OOM case for Test5102804: Test seems works with |
@annaibm please also list the failed subtests and the common failures. The Grinder links would expire after several days. |
Failures
2.
Common Failures
2. java/beans/PropertyEditor/TestColorClass.java.TestColorClass, java/beans/PropertyEditor/TestColorClassJava.java.TestColorClassJava, java/beans/PropertyEditor/TestColorClassNull.java.TestColorClassNull, java/beans/PropertyEditor/TestColorClassValue.java.TestColorClassValue,
3. java/beans/PropertyEditor/TestFontClass.java.TestFontClass, java/beans/PropertyEditor/TestFontClassJava.java.TestFontClassJava, java/beans/PropertyEditor/TestFontClassValue.java.TestFontClassValue,
|
re #20531 (comment), thanks @dmitripivkine . @annaibm please add |
- Updated the @run with -Xalwaysclassgc jvm option related: eclipse-openj9/openj9#20531 Signed-off-by: Anna Babu Palathingal <[email protected]>
- Updated the @run with -Xalwaysclassgc jvm option - This is a temporary workaround and should be reviewed once the root cause is identified. related: eclipse-openj9/openj9#20531 Signed-off-by: Anna Babu Palathingal <[email protected]>
Copied from ibmruntimes/openj9-openjdk-jdk11#849 (comment) Dmitri said:
Well, I have a plan to find out exact scenario. However I don't think there is something wrong with handling. This is combination class unloading - > clearing Weak Ref. OpenJ9 does not do class unloading every time. It does it for aggressive GC to prevent OOM but it is too late to process Weak Ref. If you need this test pass without pushing class unloading every single Global GC an alternative can be to have another run of aggressive Global GC. I am not sure it is necessary. |
It just seems wrong. The GC should do everything necessary to clear memory before throwing OOM. Throwing the OOM would likely stop the app from working, but if there are classes to unload or references to clear that would prevent the OOM, it seems it should be done. Also it would apparently match Hotspot behavior. |
- Exclude jdk_beans subtests as excluded in Adoptium related: eclipse-openj9/openj9#20531 Signed-off-by: Anna Babu Palathingal <[email protected]>
- Exclude jdk_beans subtests as excluded in Adoptium related: eclipse-openj9/openj9#20531 Signed-off-by: Anna Babu Palathingal <[email protected]>
- Exclude jdk_beans subtests as excluded in Adoptium related: eclipse-openj9/openj9#20531 Signed-off-by: Anna Babu Palathingal <[email protected]>
- Exclude jdk_beans subtests as excluded in Adoptium related: eclipse-openj9/openj9#20531 Signed-off-by: Anna Babu Palathingal <[email protected]>
Testing jdk_beans on xmac on openj9 Jenkins: https://openj9-jenkins.osuosl.org/view/Test/job/Grinder/4013/ |
- Exclude jdk_beans subtests for OpenJ9 related: eclipse-openj9/openj9#20531 Signed-off-by: Anna Babu Palathingal <[email protected]>
jdk_beans got excluded as part of the general exclude a while back via #10757.
https://github.com/adoptium/aqa-tests/blob/a498becc695f7bdeaa7c4b3752c827e65aa7e123/openjdk/playlist.xml#L755
My limited testing can pass on xlinux JDK17, 21 and 23. We still need to look into xlinux JDK8 and JDK11 failures. But I think we should look into re-enabling this test.
This would help us to discover issues like https://bugs.openjdk.org/browse/JDK-8336862 early.
Presumably, we will hit the same issue on Windows: adoptium/aqa-tests#5690 (comment)
The text was updated successfully, but these errors were encountered: