[zeromq-dev] ZeroMQ (2.1.7) subscriber "memory leak" (Ian Barber)

[Jess Morecroft]

> This unexpected requirement to service the queue to some degree at various
> points has been bought up before, and I think it is a fair concern. A
> periodic noblock recv would fix the issue, but I agree this is not a very
> clear way of handling the issue - though hopefully as 3.1 moves towards
> stable the issue will recede somewhat.
FYI - we've gone with the solution of passing the sub socket to the 
zmq_poll call regardless of subscriptions (or lack thereof). It wasn't 
as delicate an operation as I'd expected and has resulted in immediate 
elimination of the issue.

Looking forward to the 3.1 release. We at least have some big 
performance gains to expect from the upstream sub filtering.