This looks good, in that it fixes the flakiness. I don't think we have tests yet that ensure Either completion when only one task completes. Consider writing one normal async supplier and one that waits on a latch; ensure that the Either future completes with the normal value, then trip the latch to allow the second one to finish.
On Mon, Apr 15, 2013 at 7:09 AM, Chris Hegarty <chris.hega...@oracle.com>wrote: > I missed three cases in the previous change [1]. That will teach me for > working on the weekend ;-) > > A full audit of the use of the XxxEitherXxx methods in the test has been > done, and there are still three particular checks that are possibly > incorrect. The failure is not always seen as this is racy code. > > http://cr.openjdk.java.net/~**chegar/8012237/webrev.00/** > webrev/test/java/util/**concurrent/CompletableFuture/** > Basic.java.udiff.html<http://cr.openjdk.java.net/~chegar/8012237/webrev.00/webrev/test/java/util/concurrent/CompletableFuture/Basic.java.udiff.html> > > Thanks, > -Chris. > > [1] http://mail.openjdk.java.net/**pipermail/core-libs-dev/2013-** > April/016013.html<http://mail.openjdk.java.net/pipermail/core-libs-dev/2013-April/016013.html> >
