Failing JTReg: test/jdk/java/lang/ref/NullQueue.java

[M-Davies]

Failure link

https://github.com/ibmruntimes/openj9-openjdk-jdk11/blob/502ab02dbda2e8b021cffedc3f52e25246cdeec1/test/jdk/java/lang/ref/NullQueue.java#L1 fails for JDK8/11-openj9 (test can't be found for JDK13 and hotspot is unaffected)

• test category, openjdk
• OS/architecture, openjdk_x86-64_linux, openjdk_x86-64_windows

16:30:13  openjdk version "11.0.5" 2019-10-15
16:30:13  OpenJDK Runtime Environment AdoptOpenJDK (build 11.0.5+8-201909260313)
16:30:13  Eclipse OpenJ9 VM AdoptOpenJDK (build master-61255304e, JRE 11 Linux amd64-64-Bit Compressed References 20190926_340 (JIT enabled, AOT enabled)
16:30:13  OpenJ9   - 61255304e
16:30:13  OMR      - c30a4776
16:30:13  JCL      - 502ab02dbd based on jdk-11.0.5+8)

Optional info

• intermittent failure (no):
• Part of the excluded Investigate JTreg tests failing against Openj9 Java 11 builds adoptium/aqa-tests#585
• ThreadGroup output:

Current Thread: GC Slave
Current Thread: Concurrent Mark Helper
Current Thread: IProfiler
Current Thread: JIT-SamplerThread
Current Thread: JIT Diagnostic Compilation Thread-7 Suspended
Current Thread: JIT Compilation Thread-6 Suspended
Current Thread: JIT Compilation Thread-5 Suspended
Current Thread: JIT Compilation Thread-4 Suspended
Current Thread: JIT Compilation Thread-3 Suspended
Current Thread: JIT Compilation Thread-2 Suspended
Current Thread: JIT Compilation Thread-1 Suspended
Current Thread: JIT Compilation Thread-0
Current Thread: Attach API wait loop
Current Thread: main
Current Thread: MainThread
Current Thread: Common-Cleaner

Failure output (captured from console output)

16:33:44  STDERR:
16:33:44  java.lang.Exception: Couldn't find Reference-handler thread
16:33:44  	at NullQueue.main(NullQueue.java:64)
16:33:44  	at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
16:33:44  	at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
16:33:44  	at java.base/jdk.internal.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
16:33:44  	at java.base/java.lang.reflect.Method.invoke(Method.java:566)
16:33:44  	at com.sun.javatest.regtest.agent.MainActionHelper$AgentVMRunnable.run(MainActionHelper.java:298)
16:33:44  	at java.base/java.lang.Thread.run(Thread.java:831)

Rerun test: https://ci.adoptopenjdk.net/job/Grinder/parambuild/?JDK_VERSION=11&JDK_IMPL=openj9&BUILD_LIST=openjdk&JenkinsFile=openjdk_x86-64_linux&CUSTOM_TARGET=test/jdk/java/lang/ref/NullQueue.java&TARGET=jdk_custom_0

openjdk_test_output (1).tar.gz

[pshipton]

Likely Hotspot specific.

[M-Davies]

Thanks for replying to this @pshipton! Just to dot the I's and cross the T's, how did you come to this conclusion?

[M-Davies]

FYI, /test/jdk/java/lang/ref/OOMEInReferenceHandler.java fails in the same way.

09:47:34  STDERR:
09:47:34  java.lang.IllegalStateException: Couldn't find Reference Handler thread.
09:47:34  	at OOMEInReferenceHandler.main(OOMEInReferenceHandler.java:78)
09:47:34  	at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
09:47:34  	at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
09:47:34  	at java.base/jdk.internal.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
09:47:34  	at java.base/java.lang.reflect.Method.invoke(Method.java:566)
09:47:34  	at com.sun.javatest.regtest.agent.MainWrapper$MainThread.run(MainWrapper.java:127)
09:47:34  	at java.base/java.lang.Thread.run(Thread.java:831)

[pshipton]

OpenJ9 is an independent implementation. It seems the "Reference Handler" thread is related to the Hotspot implementation of java.lang.ref.Reference, and OpenJ9 has it's own implementation that doesn't use a thread with this name.

[M-Davies]

Since the test is looking for a specific thread and to avoid making the code specific to one implementation, I will perm exclude this and OOMEInReferenceHandler.java from openj9 testing.

[M-Davies]

https://github.com/ibmruntimes/openj9-openjdk-jdk11/blob/b44d19c4cbca42541aef5132f21870c1b924a68c/test/jdk/java/lang/ref/ReachabilityFenceTest.java#L1 is also using the hotspot implementation. Will perm exclude this too

10:00:25  STDERR:
10:00:25  java.lang.IllegalStateException: The object had been finalized without a fence, even though we don't expect it.
10:00:25  	at ReachabilityFenceTest.main(ReachabilityFenceTest.java:83)
10:00:25  	at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
10:00:25  	at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
10:00:25  	at java.base/jdk.internal.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
10:00:25  	at java.base/java.lang.reflect.Method.invoke(Method.java:567)
10:00:25  	at com.sun.javatest.regtest.agent.MainWrapper$MainThread.run(MainWrapper.java:127)
10:00:25  	at java.base/java.lang.Thread.run(Thread.java:831)

[pshipton]

@M-Davies how do we know ReachabilityFenceTest is also Hotspot specific? I can't tell from the stack trace.

[M-Davies]

It's using a Hotspot implementation of getBoolean() to find a specific variable. It returns false when it cannot find it and throws an error later in the test when an if statement checks if it exists