10:17 <kh1b> Is changing the init system in CRUX common, and is it worth it

10:32 <zorz> its up to you. personally no need to change it

10:36 <cruxbot> [opt/3.7]: linux-firmware: updated to version 20240811

10:36 <cruxbot> [opt/3.7]: gvim: updated to version 9.1.0672

10:36 <cruxbot> [opt/3.7]: graphviz: updated to version 12.1.0

10:36 <cruxbot> [core/3.7]: vim: updated to version 9.1.0672

10:36 <cruxbot> [core/3.7]: libedit: updated to version 20240808-3.1

11:37 <remiliascarlet> kh1b: By default, CRUX doesn't really have an init system, which is similar to the BSD's. Something I personally prefer over having 20 different init systems, and every one of them sucking.

12:04 <uwumeowmeow> How does it work then?

12:13 <remiliascarlet> BSD style init scripts.

12:13 <remiliascarlet> Kind of similar to how Linux used to work until the 2010s, back when there was just SysV Init and nothing else.

12:19 <cruxbridge> <jloc0> Basically crux is a O.G. init system and it’s better for it

12:50 <uwumeowmeow> Why is it better if other init systems let a user to set up dependencies, for example?

13:38 <remiliascarlet> I have never heard of init scripts being dependent on anything other than the user, binary, maybe config file, and shell.

13:39 <remiliascarlet> Sounds very soydevvery to me to have a depencency system thrown in there as well.

13:42 <farkuhar> remiliascarlet: I think what uwumeowmeow meant was the dependencies among the services. For example, you would want the network interface to come up before launching the service that syncs your hardware clock via NTP.

13:42 <uwumeowmeow> Yes, I meant this.

13:42 <remiliascarlet> Oh, like that.

13:43 <remiliascarlet> I think it should be done by ordering the services in rc.conf, but not sure.

13:43 <remiliascarlet> Never needed to deal with that.

13:43 <uwumeowmeow> They are indeed run in order.

13:45 <uwumeowmeow> But what if I want to run many services in parallel? Such thing requires init system to understand dependencies

13:45 <uwumeowmeow> By running I mean starting.

13:45 <farkuhar> zorz adapted the OpenBSD rcctl tool for CRUX, but currently it does not have any logic for determining where a new service should be inserted (with the `rcctl enable` command). I think it just appends the given argument at the end of the SERVICES array.

13:46 <zorz> farkuhar: i am open to improvements

13:46 <zorz> i thought of something

13:47 <zorz> https://codeberg.org/zorz/crux/src/branch/main/rcctl

13:49 <remiliascarlet> I don't think there's any parallelism in CRUX. Not like there's any need for it, because with the exception of wpa_supplicant and sometimes tor, everything just starts within a snap of a finger, even on hardware as old as a ThinkPad X200.

13:50 <remiliascarlet> In comparison, I found Artix + runit to be a lot slower on this same machine.

13:50 <farkuhar> interestingly, sysklogd is not an element of the SERVICES array, but is started earlier by rc.single. On CRUX-MUSL this service has a noticeable delay, in contrast to glibc-based CRUX where sysklogd does "start within a snap of a finger".

13:51 <farkuhar> So on CRUX-MUSL I am often tempted to edit rc.single and change the line that starts sysklogd, to avoid the noticeable delay at startup.

13:57 <jaeger> Like anything else an init system is a subjective choice. If you want to replace sysvinit with something else that has more features, you're of course free to do that on your own systems

13:57 <zorz> syvinit of crux runs good. no need to change for myself.

14:04 <zorz> as a matter of fact the best part of CRUX is that gives you the freedom to do whateveer you like with your system. Hassle-free