On 25 December 2011 11:58, Peter <j...@zeus.net.au> wrote: > I've noticed with the event bases tests, the more logging, security debug > enabled, less events are dropped, or the test passes. >
That's symptomatic of a race condition - as you slow things down with logging etc the effect is to space things out and reduce the amount of interleaving. > On one test, I've found after teardown commences, that permissions are > denied, the number denied coincided (at least on one test, will try more > later) with the number of missing events. > > The new DynamicPolicyProvider implementation, responds to interruption, by > returning a false result early, if interrupted, rather than continue, this > makes sense for ProtectionDomain checks occuring in parallel, where the first > to fail causes PD checks belonging to the same context to be cancelled using > Future.cancel. This may be behind the failures. > > Cheers, > > Peter.
