Sorry if this is ignorant, but timeouts don't have to be precise.
Everybody uses them in a rough way anyway (I think). So it doesn't
need to be checked on every call. Ignore me if this makes no sense/is
totally irrelevant, you're the experts (and are awesome).

Thanks,
N

On 3 September 2010 13:03, Martin Sustrik <[email protected]> wrote:
> On 09/03/2010 01:50 PM, Martin Lucina wrote:
>
>> I'd still rather get "Martin Sustrik level of imperfection" APIs than most
>> of what's out there ... :-)
>
> Oh, you are being so good to me.. sniff, sniff!
>
> :)
>
> Anyway, this is once again a problem that cannot be solved using POSIX
> API cleanly (looks like we hitting those quite often recently). The
> problem is that POSIX doesn't allow user-space application to create a
> file descriptor to be signaled/unsignaled at will. Linux' event_fd comes
> close but it's not perfect still (it doesn't allow to signal !POLLIN and
> !POLLOUT at the same time).
>
> So, there has to be zmq_poll instead of standard poll/select/etc.
>
> To honour timeouts in the poll we have to measure time. Which can be
> inefficient in tight loops using zmq_poll.
>
> This can be worked around using internal 0MQ's file descriptors (ZMQ_FD)
> and stadard polling (poll) but in such case the application is going to
> be signaled even on 0MQ's internal signals.
>
> Martin
> _______________________________________________
> zeromq-dev mailing list
> [email protected]
> http://lists.zeromq.org/mailman/listinfo/zeromq-dev
>
_______________________________________________
zeromq-dev mailing list
[email protected]
http://lists.zeromq.org/mailman/listinfo/zeromq-dev

Reply via email to