Jan Uhlig
Jan Uhlig
@alex88 Well, yes and no =^^= That was why I submitted #124... One option is to implement your workers in a way so that the crash-prone part is not part...
Hm, yes, but I would argue that this is not poolboy's responsibility. It just gives you a worker, it doesn't say anything about that worker being ready. I don't know...
> Another solution to not crash the worker on init is to use {continue, Term} and handle_continue/2 in your worker. Or use `proc_lib:start_link` in in you worker's `start_link`, then `proc_lib:init_ack`...
Yeah, this is just what I'm wishing for right now. Maybe an (optional) callback in `poolboy_worker` would be the best thing to do, which would be called on `checkin`. The...
Or a function to call that could be given as a pool arg.
> Since we are in the middle of the vacation period now it will, however, take a while before we'll look further into it. Sure, no problem, I wish you...
I'll get back to this next week, right now I'm still on vacation 😉
> No, no timeouts will be lost even without this change. They will however be triggered late. A timer set to an absolute time that has already passed or is...
@rickard-green I removed the special handling for interval zero in the last commit. There is a strange failure in the doc build that seems unrelated to this PR.
> You know what, I think I have an idea ... and now I'm ~worried~ thrilled... :grin: > that may significantly reduce the congestion on the timer server that may...