I recall considering whether to fix this "bug" myself many years ago. I decided that the value was so low (throw the "required" NPE when this collection is empty and the argument collection is null) that users who where not ultra-pedantic (or tck testers) would prefer the legacy behavior. And I would make the same decision today. But it's hard to really object to something that appears to make the implementation a bit more compliant with the spec.
On Mon, May 6, 2013 at 3:03 PM, Mike Duigou <mike.dui...@oracle.com> wrote: > Hello all; > > Long, long ago, Brandon Passanisi proposed a fix to correct non-conformant > implementation of retainAll/removeAll(). It was reviewed but was never > committed to TL repo (it got lost amidst Christmas vacation and other > matters). Rather than surprise everyone with a rogue changeset they may > have forgotten I'm reposting the review webrev and giving everyone a chance > to change their mind. > > http://cr.openjdk.java.net/~mduigou/JDK-4802647/0/webrev/ > > Hearing no objections I will push this in about 24 hours. > > Mike
