JTreg test fail : java/lang/invoke/lambda/LambdaStackTrace.java

[ben-walsh]

Test https://github.com/ibmruntimes/openj9-openjdk-jdk11/blob/openj9/test/jdk/java/lang/invoke/lambda/LambdaStackTrace.java fails with ...

Actual trace did not match expected trace at frame 1
Expected frame patterns:
  0: LambdaStackTrace\\..*
  1: LambdaStackTrace.testBasic
Actual frames:
  0: LambdaStackTrace.lambda$testBasic$0(LambdaStackTrace.java:63)
  1: LambdaStackTrace$$Lambda$20.000000000000D7F0.run(Unknown Source)
java.lang.Exception: Incorrect stack frames found
	at LambdaStackTrace.verifyFrames(LambdaStackTrace.java:191)
	at LambdaStackTrace.testBasic(LambdaStackTrace.java:75)
	at LambdaStackTrace.main(LambdaStackTrace.java:53)
	at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
	at java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
	at java.base/jdk.internal.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
	at java.base/java.lang.reflect.Method.invoke(Method.java:566)
	at com.sun.javatest.regtest.agent.MainWrapper$MainThread.run(MainWrapper.java:115)
	at java.base/java.lang.Thread.run(Thread.java:825)

The synthetic frame (i.e. "LambdaStackTrace$$Lambda$20.000000000000D7F0.run(Unknown Source)") should not be present in the exception stack. See OpenJDK bug https://bugs.openjdk.java.net/browse/JDK-8025636.

Test passes against Hotspot.

Log archive : LambdaStackTraceProb.tar.gz

[tajila]

@fengxue-IS can you please take a look at this

[tajila]

The code responsible for creating stack traces is in /runtime/jcl/common/jclexception.c, specifically the getStackTraceIterator.

If the romClass has the J9AccClassAnonClass bit set in the extraModifiers it is an anonymousClass so you can remove that frame from the exception stack trace

---

Note: the steps above assume that they are removing the anonClass frames from the StackTraceElement[] (see Throwable.getInternalStackTrace()) and that they are not simply ignoring this frame when printing stack traces in Throwable.printStackTrace() . This should be verified first.

[fengxue-IS]

looking into this

[tajila]

Update: There is an annotation called java/lang/invoke/LambdaForm$Hidden which is used to hide frames. This is the mechanism that hotspot uses to decide which frames to hide. We will have to look at this instead of the J9AccClassAnonClass (as previously mentioned) bit to determine whether to hide the frame or not. We can you use the existing mechanism (ClassFileOracle::KnownAnnotation) to detect this.

Also, -XX:+UnlockDiagnosticVMOptions -XX:+ShowHiddenFrames are used to force the frames to show up in stack traces. We may want to see if these options map to some of our existing options and just update their behaviour. If not we can implement one of them.

[tajila]

we have an existing annotiation Ljava/lang/invoke/MethodHandle$FrameIteratorSkip; that does something similar.

[smlambert]

Test excluded: adoptium/aqa-tests#657

[babsingh]

The test was re-enabled via adoptium/aqa-tests#3302.