Event-polling after a fork

Felipe Gasper felipe at felipegasper.com
Thu Feb 27 13:03:16 CET 2020


Follow-up/postscript:

Once I took a step back, I realized that because AE’s internal references to the watchers are weakened, AnyEvent->reset() would only remove those references and not fire any DESTROY handlers. So yeah, it would serve no purpose.

(With Mojo the reset() method works because Mojo::IOLoop stores normal references.)

Thank you for your responses.

-FG


> On Feb 25, 2020, at 10:55 AM, Marc Lehmann <schmorp at schmorp.de> wrote:
> 
> On Tue, Feb 25, 2020 at 07:27:59AM -0500, Felipe Gasper <felipe at felipegasper.com> wrote:
>> This appears to be Mojo::IOLoop’s solution to this problem.
> 
> It might appear so to you, but it clearly doesn't. For one thing, it's
> only one event loop, and AnyEvent of course has to work for all of
> them. But it also doesn't handle any of the actual issues you run into
> - sure, you can reset the loop, but how does that handle the resulting
> breakage when watchers silently stop working for example? In code that
> you have never seen because it was added after you wrote your program for
> example? And lastly, it doesn't solve any of the problems that fork brings
> with it.
> 
> And note that it relies on EV to do the limited work it does, and as far
> as I know, EV is (I think) the only event loop that even attempts to do
> soemthing sensible (e.g. allow you to remove watchers after a fork - with
> other event loops this might result in watchers in both parent and child
> becoming disabled for example. or worse).
> 
>> Would it be feasible to do something similar in AnyEvent? Most of AE’s event loop backends seem to be part of the distribution anyway, so whichever loops don’t add reset() could simply throw.
> 
> Well, first of all, it doesn't actually seem to solve an actual
> problem. And most importantly, if you rely on something like that in a
> specific event loop, you can just use that specific event loop - the point
> of AnyEvent is to make event loops generically usable by programmers,
> which is a much bigger group than group that controls a specific program
> (think CPAN modules). An option that breaks this while pretendinfg to be
> generic makes little sense - just use the event loop you think solves your
> problems directly then, and deal with the issues cleanly.
> 
> -- 
>                The choice of a       Deliantra, the free code+content MORPG
>      -----==-     _GNU_              http://www.deliantra.net
>      ----==-- _       generation
>      ---==---(_)__  __ ____  __      Marc Lehmann
>      --==---/ / _ \/ // /\ \/ /      schmorp at schmorp.de
>      -=====/_/_//_/\_,_/ /_/\_\




More information about the anyevent mailing list