output.log

@arifogel thanks for the report. The rule is effectively not using type resolution yet, so these non-straightforward scenarios will be missed for the time being.

I'd be willing to look into this with a little shepherding. Is there some documentation/code I could review to help me get started? Or is this blocked on something that is currently churning internally?

@jsotuyod any progress on this? I'm still willing to look into this with a bit of initial guidance.

Hi @arifogel,
Thanks for the offer!

First step would be (if not done already): https://github.com/pmd/pmd/wiki/Setup-IDE
Then you can add your case as a new test case into https://github.com/pmd/pmd/blob/master/pmd-java/src/test/resources/net/sourceforge/pmd/lang/java/rule/bestpractices/xml/UnusedPrivateMethod.xml (you would add here only your OuterClass). Since you are dealing with type resolution, you'll need additionally the two classes as java source, so that it is compiled and available during test runtime. You can add your two classes OuterClass and Referrer to the package net.sourceforge.pmd.lang.java.rule.bestpractices under src/test/java.
You can start the unit test then through BestPracticesRulesTest. You should now see a failure. You can also run the test in the debugger and step through the class https://github.com/pmd/pmd/blob/master/pmd-java/src/main/java/net/sourceforge/pmd/lang/java/rule/bestpractices/UnusedPrivateMethodRule.java .

Then the hard part comes: Fixing the issue. We currently don't resolve the return type of a method call - which would be needed here in order to know, that we call the method privateMethod().

I guess, changing the sample code to this one:

public final class OuterClass {
  private static class InnerClass {
    private void callPrivateMethod() {
      OuterClass outer = Referrer.getOuterClass();
      outer.privateMethod();
    }
  }

  private void privateMethod() {}
}

won't trigger this false positive, since we know the type of the variable outer. If this also triggers the rule, then this might be a first easier step to fix.

Let me know, if you run into any problems - either here or on https://gitter.im/pmd/pmd

I finally got my IDE set up. Your version of the sample code passes.
I was able to condense the false positive test code to a single file:

public final class OuterClass {
    public static class InnerClass {
        public void callPrivateMethod() {
            Referrer.getOuterClass().privateMethod();
        }
    }

    public static class Referrer {
        public static OuterClass getOuterClass() {
            return new OuterClass();
        }
    }

    private void privateMethod() {
    }
}

Now comes the hard part of processing the return type..

It seems like the failure occurs even earlier than I expected.
I changed net.sourceforge.pmd.lang.java.symboltable.TRACE to true and ran using pmd master for the concise snippet above.
The trace is unexpected:

new search for variable Referrer:4:class net.sourceforge.pmd.lang.java.ast.ASTName
 checking scope MethodScope: for name occurrence Referrer:4:class net.sourceforge.pmd.lang.java.ast.ASTName
 moving up from MethodScope: to ClassScope (InnerClass): Inner Classes ; Method callPrivateMethod, line 3, params = 0(begins at line 3, 0 usages), 
 checking scope ClassScope (InnerClass): Inner Classes ; Method callPrivateMethod, line 3, params = 0(begins at line 3, 0 usages),  for name occurrence Referrer:4:class net.sourceforge.pmd.lang.java.ast.ASTName
 moving up from ClassScope (InnerClass): Inner Classes ; Method callPrivateMethod, line 3, params = 0(begins at line 3, 0 usages),  to ClassScope (OuterClass): Method privateMethod, line 14, params = 0(begins at line 14, 0 usages), 
 checking scope ClassScope (OuterClass): Method privateMethod, line 14, params = 0(begins at line 14, 0 usages),  for name occurrence Referrer:4:class net.sourceforge.pmd.lang.java.ast.ASTName
 moving up from ClassScope (OuterClass): Method privateMethod, line 14, params = 0(begins at line 14, 0 usages),  to SourceFileScope: Class OuterClass
 checking scope SourceFileScope: Class OuterClass for name occurrence Referrer:4:class net.sourceforge.pmd.lang.java.ast.ASTName
found []

AFAICT, even Referrer by itself cannot be found. Is this the same problem of not using type resolution, or is this a separate issue?

It seems like we also encountered this issue.

This has been fixed with PMD 7.0.0-rc1.