hi all, im not sure if this is a core.async issue or a clojurescript/nodejs issue, but I am struggling to figure out the best way to shut down an app that has at least one (timeout) pending

in a nutshell, I have an (alt!) that waits on two channels: one of which is a core.async/timeout

but I find that my program hangs at shutdown until the timeout eventually fires...which is the common scenario because the timeout just an exceptional condition

so, i am wondering if there are some best practices here, or other advice...for instance, canceling the timeout, telling core.async to shutdown, a way to wait on a channel with a timeout, etc

the way to wait on a channel with a timeout is to use alt! or alts!

if your channel has a value before the timeout does, that's the value you will get

right, understood...i think i am explaining poorly

my (alt!) works as expected, but then my program cannot exit until the timer also fires

oh, add a third channel - called a poison-pill

close that channel when you want to exit

but never write to it

a closed channel returns nil immediately, so if it is one of your alts it will fire

hmm, the problem is not with the (alt!) firing...the logic is part of a command line app....so start -> (alt!) -> exit

the (alt!) is written something like (alt! normal-ch timeout-ch)

normally, the normal-ch fires first

but the program hangs until an interval approximately equal to the timeout

what is it waiting for then?

its not hanging in the (alt!), but in some seeming system shutdown

in clojure I would say you need to run (shutdown-agents) but I don't think that applies in node

yeah, it feels like that kind of problem

but I am not sure what the node solution is

if the app is waiting on a timeout that you are not reading, that sounds like a bug to me

but, what if you call close! on the timeout channel?

i tried that (even though the doc comments said "dont") but it had no effect

using close! on a timeout works in my repl - let me double check in a node repl

it "feels" like node is still seeing something with active callbacks in the event loop, or whatever (not a node-internals expert)

ala it needs a (shutdown-agents) whack

I can probably write this a half dozen other ways, its just seemed like (alt!)+(timeout) was the idiomatic way to do it

just curious, where did you see the comment not to call close! on a timeout?

https://clojuredocs.org/clojure.core.async/timeout

see commands at bottom

comments, not commands

granted, its a user comment, not the official doc

but it seemed legit

but if you want your whole app to exit, that caveat doesn't apply

too bad that it does that though

yeah, thats why I went for it anyway

i wasnt 100% sure it wouldnt have a reentrancy problem, but I think for my use case, that would have been ok

but like I said, it seemed to have no impact

I suspect there aren't a whole bunch of people writing cljs+node command line apps

thats probably part of the problem

just based on my experience getting them working