[zeromq-dev] [PATCH] Improved response to socket violations
mato at kotelna.sk
Fri May 20 18:17:20 CEST 2011
ph at imatix.com said:
> On Fri, May 20, 2011 at 6:00 PM, Martin Sustrik <sustrik at 250bpm.com> wrote:
> > The problem is that using single socket from mutliple threads can cause
> > basically any error, whether assertion, segfault or whatever. Singling
> > this one out doesn't make much sense. Moreover, as Mato points out, the
> > assertion can happen because of limited OS resources, in which case the
> > message would be misleading.
> So the repeated instances of this particular assertion failure are
> just coincidence?
> Fair enough, but it reflects badly on 0MQ that it can't more
> accurately detect and report such a common application error.
There's always a tradeoff. I recall the discussion from last year about
whether or not ZeroMQ is a library for "Real men" or "Little old ladies".
Personally, I think you can have "Fastest messaging ever" or "safest
messaging ever", but not both. Of course, feel free to submit a patch which
accurately detects such conditions while keeping the lock-free and high
performance nature of the system.
More information about the zeromq-dev