False alarms on inner class?

This is a multipart message in MIME format.
For the following class, findbugs generates
two warnings:
---------------------------
import java.util.*;

public final class Foo {

   private final static Comparator COMPARATOR
= new Comparator() {
      public int compare(Object o1,
Object o2) {
         int result = o1.hashCode()
- o2.hashCode();
         assert(result >
0);
         return result;
      }
   };

   TreeMap getMap() {
      return new TreeMap(COMPARATOR);
   }
}
-----------------------

The first warnings is: H C UMAC: Uncallable
method Foo$1.<static initializer>() defined in anonymous class  At
Foo.java:[line 6]

If I'm not missing something, this is
a plainly false alarm: the static initializer is not only callable, but
will be called if the anonymous class is ever used. Perhaps the
rule used to generate this warning should make an exception of the static
initializer method?

The second warning is: M D IC: Initialization
circularity between Foo and Foo$1  At Foo.java:[lines 3-14]

Again if I'm not missing anything, this
warning, too, results from a perfectly innocent code. But I realize this
is a much harder case for findbugs to distinguish from a really buggy code.

I peeked at the bytecode of the two
classes to find out where static initializers call each other. It's obvious
why Foo's static initializer needs Foo$1. The opposite direction is less
obvious - the reason is the assertion in Foo$1, which inherits the assertion
enablement status from Foo. Naturally (it's all compiler-generated) there
is no actual bug - the static initializer of Foo$1 contains code that explicitly
takes care of the case that Foo is not completely initialized. I realize
this is probably a challenge for findbugs to identify this pattern - or
is it?

Assuming the warning is inevitable,
it seems to me a good idea to document it in the bug description page.
Also, what's the best way to suppress this warning? I use an exclude rule
that matches this type of warning and (the entire) Foo class, but this
seems more general than I'd want. Would have been nice if findbugs would
offer a way to suppress this warning for a particular pair of classes.

Best regards,
Yarden Nir-Buchbinder
Software Verification and Testing
IBM Haifa Research Lab
Tel. 972-4-8296010