Found a bug in replier clean-up.
While refactoring code, Charlie found a bug cleaning up resources to a replier task. Rather than use the request ID, it was using the reply ID. This would cause it to remove information to other reply tasks. Possible symptom would be not receiving replies.
This problem wasn't more widespread because during
acnetd startup we randomize the upper bits of request IDs and replyIDs. The IDs from different nodes would have to have the same, random upper bits in order for their request to be lost.
The fix: 388b886a
Charlie will make the fix generally available on Monday. He's also been doing tests running Java engines on
acnetd and is getting great results with this patch (no more "out-of-data" errors from
- Related to Bug #13703: Inconsistent behavior with D105 added
acnetd over the weekend on
DCE16 with no problems. Before this commit, engines running on
acnetd couldn't run for more than 1/2 hour. He's going to push out
acnetd so, little by little, our nodes will run the update..
- Status changed from Assigned to Closed
Using this version of
acnetd with some changes he made to his Java interface resulted in zero "out-of-data" messages over the weekend. This fix is good.
- Related to Bug #20222: Assigned DPM that has an Issue added
- Tracker changed from Support to Bug
This was a "bug", not "support".
Also available in: Atom