From: Omar Polo Subject: Re: gotd: run authentication in separate process To: Stefan Sperling Cc: gameoftrees@openbsd.org Date: Thu, 29 Dec 2022 18:50:36 +0100 On 2022/12/29 17:51:48 +0100, Stefan Sperling wrote: > On Thu, Dec 29, 2022 at 04:48:31PM +0100, Omar Polo wrote: > > On 2022/12/29 16:11:57 +0100, Stefan Sperling wrote: > > > On Thu, Dec 29, 2022 at 12:55:03PM +0100, Omar Polo wrote: > > > > > + switch (sig) { > > > > > + case SIGHUP: > > > > > + break; > > > > > + case SIGUSR1: > > > > > + break; > > > > > > > > just for curiosity, what's the point of catching SIGUSR1 too? the > > > > listener process and gotwebd do that too (but gotwebd due to proc.c I > > > > guess), but otherwise they don't do anything with it. > > > > > > No idea. This was just copied along from other daemons. > > > The HUP and USR signals are often used to trigger actions such as > > > reloading config files. I suspect that is why they are present in > > > all these signal handlers. > > > > I see. diff below drops it for gotd but I wouldn't mind to keep it > > since it's there even in other daemons that don't do anything with it > > (e.g. see relayd.) > > The default action for USR1 is to terminate the process. > Do we really want gotd to terminate when someone sends this signal? wrote the patch in a hurry, i haven't thought it throughfully, sorry. don't have a strong opinion either way to honest. SIGUSR2 also terminates the process by default, and it's not handled; on the other way various daemons catch USR1 even if they don't do anything with it. ignoring SIGUSR1 is not a burden anyway, so let's drop this. Sorry for the noise.