On 09/04/2014 18:39, Seán Coffey wrote:
On re-read, I believe extending the sync block in read(..) to cover the reading and setting of the rem variable works to resolve this fix. It should preserve behaviour as well.This is probably okay for a test case that attempts concurrent reads but as it's not synchronized with skip then I don't think you can trust the value of rem or pos without taking copies and validating.http://cr.openjdk.java.net/~coffeys/webrev.8038491.v2/webrev/
-Alan.