Currently we have test failures like below:

Error processing opcode invokestatic @ 531 in java.lang.invoke.MethodHandles.loop : ([[Ljava.lang.invoke.MethodHandle;)Ljava.lang.invoke.MethodHandle;
    java.lang.IllegalArgumentException: Unknown signature for number: Ljava/lang/Boolean;
      At edu.umd.cs.findbugs.OpcodeStack$Item.<init>(OpcodeStack.java:659)
      At edu.umd.cs.findbugs.OpcodeStack.processMethodCall(OpcodeStack.java:2504)
      At edu.umd.cs.findbugs.OpcodeStack.sawOpcode(OpcodeStack.java:2152)

Error processing opcode invokestatic @ 2 in gcUnrelatedTypes.ListTests.test2Bugs : (Ljava.util.LinkedList;)V
    java.lang.IllegalArgumentException: Unknown signature for number: Ljava/lang/Integer;

How is that possible when we've checked that constValue is a Number?

Edit: Hang on, where are we even getting constValue from in this case? It's in a constructor and that's not one of the arguments...

It happens in MethodHandle class updated by Java9. I guess that Java9 javac changed some behavior, but cannot find related document.

→Update: it is not related with javac behavior, it it caused just by new implementation in MethodHandles.loop() method. i could reproduce it with Java8, see unit test in this PR.

@ThrawnCA please have a review again, I've added test case to reproduce problem with java.lang.Boolean constant value.

SonarQube analysis reported 4 issues

Note: The following issues were found on lines that were not modified in the pull request. Because these issues can't be reported as line comments, they are summarized here:

1. OpcodeStack.java#L2808: Merge this if statement with the enclosing one.
2. OpcodeStack.java#L3024: Merge this if statement with the enclosing one.
3. OpcodeStack.java#L3722: Remove these unused method parameters.
4. OpcodeStack.java#L3023: Use isEmpty() to check whether the collection is empty or not.

Not array, just a boxed types. I'll merge this PR.

Wait, I remember this. @mebigfatguy Didn't you discover that the compiler saves array length as a constValue?

Edit: Found it: mebigfatguy/fb-contrib#40 (comment)
Looks like this is something SpotBugs is doing to itself.

@mebigfatguy Bump. I know this is merged, but do you have more insight on what the compiler is doing here?