Resource Leak Checker: false negative related to owning fields

[kelloggm]

Based on my review of #6271, I wrote the following test case, which demonstrates an unsoundness in the current implementation of the Resource Leak Checker. The same problem exists in the code in #6271, but @Calvin-L has helpfully added a "TODO: this is very wrong" comment to the code that's responsible (thanks to @Calvin-L for that - I would not have noticed the issue otherwise):

// Test case for https://github.com/typetools/checker-framework/issues/6276

import org.checkerframework.checker.mustcall.qual.*;
import org.checkerframework.checker.calledmethods.qual.*;
import java.net.Socket;

@InheritableMustCall("a")
public class OwningFieldStringComparison {

  // :: error: required.method.not.called
  @Owning Socket s;

  // important to the bug: the name of this field must contain
  // the name of the owning socket
  /* @NotOwning */ Socket s2;

  // Note this "destructor" closes the wrong socket
  @EnsuresCalledMethods(value="this.s2", methods="close")
  public void a() {
    try {
      this.s2.close();
    } catch(Exception e) {

    } finally {
      try {
        this.s2.close();
      } catch (Exception e) {

      }
    }
  }
}

This problem is that String#contains is used by the RLC internally to check whether the value field of a destructor's @EnsuresCalledMethods annotation contains the simple name of an owning field. Instead, something like program element equality should be checked, instead.