Hi all,

I'm working with the BTree package (avi.67) and have noticed that
BTree>>testStringKeys fails consistently on 3.8/3.9, while it succeeds
consistently on 3.7. The exact failure point varies randomly, but the
symptom is consistent: keys are sorted properly by content, but
reverse by length (e.g., 'aa' comes before 'a', but both would come
before 'ab'). I'm pretty confident that the problem's due to the
String changes, but beyond that I'm stuck. Any ideas on what might be
happening?

Thanks.

--Benjamin