On Fri, Mar 23, 2018 at 04:07:30PM +0100, Christopher Faulet wrote: > You're right, there is a bug. But it was not introduced by the commit you > pointed out, but this one: 4f0c64ca ("MINOR: session: release the listener > with the session, not the stream").
Interesting case. I'm pretty sure we already got this report, or maybe I confuse it with my memories of the problem before the patch above. At first I was worried that removing the sub(actconn) in case of LI_O_UNLIMITED could cause an imbalance between tracked and untracked connections, but now I'm seeing it's also present in listener_release() so this should be fine. > I attached the patch to fix this issue. Willy, could you merge it ? Sure, now merged, and just in time for 1.8.5 :-) Thank you! Willy