As of b99, TreePath.iterator() is implemented as follows, and as you see it's guaranteed to cause NPE because curr is not initialized to this. Also, it fails to iterate the leaf node correctly:

    public Iterator<Tree> iterator() {
        return new Iterator<Tree>() {
            public boolean hasNext() {
                return curr.parent != null;
            }
            
            public Tree next() {
                curr = curr.parent;
                return curr.leaf;
            }
            
            public void remove() {
                throw new UnsupportedOperationException();
            }
            
            private TreePath curr;
        };
    }

The correct implementation is:

    public Iterator<Tree> iterator() {
        return new Iterator<Tree>() {
            public boolean hasNext() {
                return curr != null;
            }
            
            public Tree next() {
                TreePath t = curr;
                curr = curr.parent;
                return t.leaf;
            }
            
            public void remove() {
                throw new UnsupportedOperationException();
            }
            
            private TreePath curr = TreePath.this;
        };
    }

The iterator method should also document in which order Trees are iterated. I find it more useful to be able to iterate from root to leaf, but no matter which way it goes, it should be documented.