[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Re: [Libevent-users] Should my code be able to deal with EAGAIN?



Hi!

Thanks for the clarifications. 

I will try to make a small POC to reproduce the issue and get an strace from that. The app I'm working on (Envoy) is quite big and makes it a bit hard to quickly test assumptions. Especially if you suspect the eventports issue might be related to the EAGAIN issue.

Kind regards,

Jasper 

> On 26 juni 2019 at 14:55 Azat Khuzhin <azat@xxxxxxxxxxxx> wrote:
> 
> 
> On Wed, Jun 26, 2019 at 3:07 PM Jasper Siepkes <siepkes@xxxxxxxxxxxxxxxx> wrote:
> >
> > Hi!
> >
> > Thanks for your answer! Sorry about the mix-up with example. That definitely didn't help getting my question across ;-).
> >
> > What actually my problem is is that libevent gives me an EAGAIN error in a connection listener (callback) when running on SmartOS (ie. Solaris). And I was wondering if that was normal? As far as I can tell the documentation for suchs listeners is specified "A callback that we invoke when a listener encounters a non-retriable error.". Hence the inquiry whether I as user of libevent should be able to handle EGAIN? The documentation seems to imply I should only be getting fatal errors, not something like EGAIN?
> 
> Can you provide strace-like output for such an issue?
> Also where did you get EAGAIN in the callback or an error callback?
> (actually you should not get it in any of them).
> 
> > I'm actually running libevent with 'EVENT_NOEVPORT=yes' because I encountered an situation in which libevent seems to be getting in an unbound loop and starts consuming 100% CPU. But thats not related I think (and something I will delve in later) to the more conceptual question on how and where to deal with EAGAIN.
> 
> Hm, could be related, if accept() always returns EAGAIN and always got
> an event like there is some activity on this socket, again strace-like
> output (for minimal example of course) will help to figure out what is
> going wrong here.
> 
> Regards,
> Azat.
> ***********************************************************************
> To unsubscribe, send an e-mail to majordomo@xxxxxxxxxxxxx with
> unsubscribe libevent-users    in the body.
***********************************************************************
To unsubscribe, send an e-mail to majordomo@xxxxxxxxxxxxx with
unsubscribe libevent-users    in the body.