>It looks like a bug. I have filed an issue to look into further: > https://bugs.openjdk.java.net/browse/JDK-8134625 >Mandy Ok. I was starting to check my jdeps-like library(an upgrade of sourceforge.net/projects/jcyclesto compute dependencies in various ways and with more info(deps byte size, classes causing packages dependencies, etc.))against jdeps on all rt.jar class files, and I'm starting to see a lot moredifferences. The idea was both:- to check correctness of libraries against each other,- and, where dependency-or-not would be conventional (such as, should a nested class always depend on its surrounding class, or not, or on all of them up to top level one (since they are all in InnerClasses attributes), etc.) to use jdeps convention to keep things simple. I will send you details (off list?) when I got them (with and without -apionly) (test does Runtime.exec(jdeps) for each class, takes time :). Then you could tell me what looks like a bugand what looks like a convention.
-Jeff