Here's a possible fix based on call/reply cooperative link/unlink (resp.) via an additional call flag in RequestMessage.
The build was failing on JDK7 only on Travis (tried with the most recent JDK7 on 2 machines and with the same JDK7 version as Travis on the slower machine without being able to reproduce it) and it turns out one supervisor test had become too time-sensitive there, so increased slightly the timeout.
Not sure how to write a reproducible test for this race condition internal to RequestReplyHelper.
Here's a possible fix based on
call
/reply
cooperativelink
/unlink
(resp.) via an additionalcall
flag inRequestMessage
.The build was failing on JDK7 only on Travis (tried with the most recent JDK7 on 2 machines and with the same JDK7 version as Travis on the slower machine without being able to reproduce it) and it turns out one supervisor test had become too time-sensitive there, so increased slightly the timeout.
Not sure how to write a reproducible test for this race condition internal to
RequestReplyHelper
.