[2026-10-23T17:37:06Z] ended up ditching the s6 suite on basix [2026-10-23T17:37:16Z] it's just way too fragile [2026-10-23T17:37:37Z] replaced with nitro + s6-log + chpst from runit + still execline for services [2026-10-23T17:37:43Z] luckily i get to keep the bits that work from s6 [2026-10-23T18:13:44Z] kris_: what was fragile about s6 specifically? [2026-10-23T18:14:12Z] also re "keeping the bits that work" that's really cool [2026-10-23T18:24:43Z] dery: well, admittedly im still not very sure what kept breaking, but tldr my package manager has post install hooks and i had one trigger on every package installation to recompile the s6 database as a tldr [2026-10-23T18:24:58Z] this hook was just interfacing with s6-frontend, initially i was driving s6 without its frontend via that [2026-10-23T18:25:25Z] regardless, no matter what, without fail, it would always get itself into a state where it couldn't recompile the system db or know what services are even running [2026-10-23T18:25:34Z] probably some sort of race condition in s6 [2026-10-23T18:25:42Z] ...with literally like its primary function [2026-10-23T18:25:54Z] weird. maybe you had to lock something before updating the db? [2026-10-23T18:25:54Z] the bpm hook for that is kinda unrelated it happened when running by hand too [2026-10-23T18:26:07Z] well i had figured i was just being stupid initially [2026-10-23T18:26:19Z] which is why i switched to s6-frontend which should be handling everything but that was also broken [2026-10-23T18:26:20Z] that sounds like either a very severe bug or some badly documented quirk [2026-10-23T18:26:27Z] there shouldn't be any locking required [2026-10-23T18:26:36Z] the database on the disk only changes when a new one is compiled and put in its palce [2026-10-23T18:26:38Z] place* [2026-10-23T18:26:57Z] either way i ended up giving up after a few months of like [2026-10-23T18:27:00Z] >s6 breaks itself [2026-10-23T18:27:09Z] >reboot with nitro, eventually fix s6 and not have any clue how i did [2026-10-23T18:27:11Z] >reboot with s6 [2026-10-23T18:27:49Z] idk either way i wasn't *really* super onboard with it anyway just because it's super unbelievably complex even if it is little to no code [2026-10-23T18:28:12Z] nitro gives me the benefits of runit and s6 both [2026-10-23T18:29:16Z] interesting [2026-10-23T18:29:21Z] either way ive kinda had it and at a certain point it's sunken cost fallacy [2026-10-23T18:29:25Z] ive been trying to figure this out for months [2026-10-23T18:29:37Z] and i was spending more time using nitro because i need my computer to work [2026-10-23T18:29:44Z] yeye if it doesn't work, it doesn't work [2026-10-23T18:30:06Z] still as you said you can pick and pair whatever is comfiest [2026-10-23T18:30:31Z] i may also end up going back to busybox tbh [2026-10-23T18:30:38Z] busybox is still the default on my distro but ive been using chimerautils for a while [2026-10-23T18:30:48Z] the mv/cp behavior on freebsd utils is about to drive me actually batshit [2026-10-23T18:31:13Z] what changes [2026-10-23T18:31:27Z] isn't this stuff supposed to be at least somewhat posix compliant [2026-10-23T18:31:31Z] running cp or mv on /path/to/dir/ copies the contents of dir/ [2026-10-23T18:31:33Z] not dir itself [2026-10-23T18:31:40Z] it has to be /path/to/dir [2026-10-23T18:31:45Z] bleh, like rsync [2026-10-23T18:32:04Z] did rsync take this from freebsd [2026-10-23T18:32:08Z] plus busybox has literally like my favorite shell anyway [2026-10-23T18:32:15Z] openbsds ksh is, also, driving me fucking nuts [2026-10-23T18:32:31Z] also i'm in the weird camp that dirs must always have a slash, to avoid surprises [2026-10-23T18:32:44Z] even if string concatenation causes double slashes [2026-10-23T18:33:04Z] yeah i pass root with its / to bpm hooks [2026-10-23T18:33:06Z] that's kinda exactly why path handling allows as many slashes as you want, imo [2026-10-23T18:33:10Z] i may like, stop doing that [2026-10-23T18:33:17Z] because right now im just stripping / from all of them [2026-10-23T18:33:21Z] like, from the hook itself [2026-10-23T18:33:37Z] https://github.com/kkrruumm/basix-packages/blob/main/sys/tinyramfs/files/hooks/post-install [2026-10-23T18:33:40Z] see: the top of this script [2026-10-23T18:34:11Z] yea [2026-10-23T18:34:21Z] sounds unnecessary?? [2026-10-23T18:34:33Z] it is [2026-10-23T18:34:52Z] bothers me nonetheless [2026-10-23T18:37:11Z] kris_ dumb question about the s6 db thing [2026-10-23T18:37:23Z] cause i was interested in trying that [2026-10-23T18:37:49Z] i'm reading about how the db docs, and it says that you can only replace the db with a specific comman [2026-10-23T18:38:17Z] you've been using it or have you been moving the file directly? i assume the former but i'm curious [2026-10-23T18:38:49Z] i dind't even know thing had a db. sounds unnecessary but i have no idea what's the rationale [2026-10-23T18:38:53Z] https://github.com/kkrruumm/basix-packages/blob/7599ec94bd33f32989031e85140c78e4ee76b7c7/sys/baseinit/files/hooks/20-s6-rc [2026-10-23T18:39:04Z] this requires you use s6-frontend [2026-10-23T18:39:14Z] which you should be anyway given s6 ships like fucking 60 binaries [2026-10-23T18:39:18Z] not exaggerating [2026-10-23T18:40:07Z] also the rationale behind having a database that your services basically huge tldr compile to is having everything more or less hardcoded [2026-10-23T18:40:13Z] there's a few benefits like zero fs traversal or anything like that [2026-10-23T18:40:53Z] its also like [2026-10-23T18:41:05Z] if you fuck up the state of your system service wise all you gotta do is reboot and you're back to a guaranteed working state [2026-10-23T18:41:10Z] unless you fucked up your db somehow [2026-10-23T18:41:35Z] sbin/init on s6 systems isnt a static thing [2026-10-23T18:41:44Z] generally what you do is symlink that to the init that s6 has compiled [2026-10-23T18:41:55Z] you do this every time you modify your default service state or add a new service to the system or whatever [2026-10-23T18:42:27Z] there's a lot of moving parts in pursuit of zero moving parts [2026-10-23T18:43:16Z] i see [2026-10-23T18:43:32Z] by all means try it [2026-10-23T18:43:38Z] i absolutely love the idea of s6 [2026-10-23T18:43:42Z] i just failed to get it to be reliable [2026-10-23T18:43:54Z] https://github.com/leahneukirchen/nitro [2026-10-23T18:43:58Z] instead i'm just using this [2026-10-23T18:44:03Z] better UX anyway imho given its basically just runit [2026-10-23T18:44:13Z] you know a lot more than me so i feel sorry if i sound like pretentious or smth [2026-10-23T18:44:28Z] i'm reading the docs and like, it says one thing you aren't doing [2026-10-23T18:44:31Z] you dont [2026-10-23T18:44:36Z] also, ? [2026-10-23T18:44:42Z] the commit command says [2026-10-23T18:44:45Z] >This command must be run once all desired modifications have been done to the set and that its consistency has been ensured, for instance by a s6 set check -F command. [2026-10-23T18:45:18Z] kinda weird that you're supposed to fix inconsistencies all the time but maybe that's why you push broken dbs? [2026-10-23T18:46:01Z] also re nitro: the runit model is really comfy yea [2026-10-23T18:46:01Z] i really wish you hadn't mentioned that [2026-10-23T18:46:11Z] oops xD [2026-10-23T18:46:47Z] you mean it like in a "i'm gonna mess with my system again" right [2026-10-23T18:46:58Z] can you link that specifically [2026-10-23T18:47:01Z] i dont know how i missed that [2026-10-23T18:47:12Z] https://skarnet.org/software/s6-frontend/s6_set.html [2026-10-23T18:47:17Z] at the end [2026-10-23T18:47:20Z] in the commit section [2026-10-23T18:48:40Z] for fucks sake [2026-10-23T18:48:46Z] i think it may be best to just ignore that [2026-10-23T18:48:50Z] zero clue how i didn't see that [2026-10-23T18:49:07Z] it's fine! [2026-10-23T18:49:25Z] and yea you wasted enough time with s6 already xD [2026-10-23T18:49:26Z] it also seems kind of odd to me to have it, on literally every invocation, repair itself [2026-10-23T18:49:34Z] it is [2026-10-23T18:49:43Z] it gets itself into an inconsistent state constantly [2026-10-23T18:49:52Z] i would argue that correctly designed software doesn't do that [2026-10-23T18:49:52Z] there must be some reason [2026-10-23T18:50:04Z] i mean we could ask to the author [2026-10-23T18:50:06Z] im sure there is but there's basically zero real error handling [2026-10-23T18:50:10Z] yeah i could just go talk to laurent he doesnt bite [2026-10-23T18:50:16Z] but im probably not going to [2026-10-23T18:50:40Z] also, again hope i didn't sound too pedant the curiosity got the better of me [2026-10-23T18:50:49Z] you didnt sound pedantic at all [2026-10-23T18:51:12Z] yay [2026-10-23T18:51:44Z] this stuff just tingles my brain the right way but yea it is really hard to do without wasting a lot of time lol [2026-10-23T18:51:50Z] well [2026-10-23T18:51:58Z] the issue is that ive already updated the basix repo and website to use nitro [2026-10-23T18:52:09Z] had i not already "committed to the bit" i'd have tried that [2026-10-23T18:52:25Z] you don't have to do it right now [2026-10-23T18:52:39Z] maybe in the future if you get curious again [2026-10-23T18:52:44Z] probably won't, even [2026-10-23T18:52:54Z] 30 minutes into using nitro i got the point of it and had no questions [2026-10-23T18:52:57Z] it is also the software's fault in this case imo [2026-10-23T18:53:20Z] kris_: even better [2026-10-23T18:53:26Z] hate to admit that even claude couldnt figure out what was going on here [2026-10-23T18:53:42Z] and before anyone dogpiles me yes i pay for claude to ask it questions and for information on stuff when i can't figure shit out [2026-10-23T18:53:45Z] it is what it is [2026-10-23T18:53:49Z] that's what i love about obscure software [2026-10-23T18:53:52Z] no training data [2026-10-23T18:53:57Z] well lol [2026-10-23T18:53:58Z] no answers :P [2026-10-23T18:54:01Z] claude just fetched the websites and read everything [2026-10-23T18:54:06Z] hell even wayland is very obscure actually [2026-10-23T18:54:22Z] i mean, how it works [2026-10-23T18:54:46Z] that's why most slop prs are almost always unfixably wrong, cause it's all so hidden and undocumented [2026-10-23T18:54:48Z] fair enough [2026-10-23T18:55:07Z] and yeah the only two people who have tried to help me with basix a bit have just been having their LLMs do it [2026-10-23T18:55:09Z] which like [2026-10-23T18:55:12Z] i mean, the wire and stuff. client bindings are a lot more regular and predictable [2026-10-23T18:55:18Z] i'm not even upset by that honestly i love AI [2026-10-23T18:55:24Z] but it's unbelievably fucking wrong 100% of the time [2026-10-23T18:55:34Z] with theirs anyway [2026-10-23T18:55:37Z] some mistral AI garbage [2026-10-23T18:55:49Z] i won't comment on the ai thing, sorry [2026-10-23T18:55:52Z] no worries [2026-10-23T18:56:02Z] at some point i may try to write a logger [2026-10-23T18:56:10Z] s6-log is good but it doesn't seem like it's able to cycle logs on boot [2026-10-23T18:56:19Z] which i figure may require some form of integration with pid1 anyway [2026-10-23T18:56:29Z] what's the current scene with loggers? [2026-10-23T18:56:32Z] cycling logs just when it comes up doesnt make sense [2026-10-23T18:56:41Z] there's oughta be more options before having to roll your own [2026-10-23T18:56:46Z] altho rolling your own is cool too [2026-10-23T18:56:52Z] idk s6-log and syslogd are the only things ive looked into really [2026-10-23T18:56:56Z] syslogd from runit i mean [2026-10-23T18:57:09Z] id argue systemd has the only good logging stuff [2026-10-23T18:57:13Z] from what ive found [2026-10-23T18:57:23Z] it's also like unbearably complex tho isn't it [2026-10-23T18:57:30Z] like a binary db for logs is way overkill imo [2026-10-23T18:57:34Z] usage wise no [2026-10-23T18:57:36Z] otherwise yes [2026-10-23T18:57:39Z] binary logging is incredibly dumb [2026-10-23T18:57:42Z] at least i heard it works like that [2026-10-23T18:57:47Z] but it's really nice to be able to like [2026-10-23T18:57:51Z] >i want logs from the last boot my machine did [2026-10-23T18:57:56Z] when your system crashes the log stops at the crash [2026-10-23T18:58:01Z] yea won't lie that's comfy [2026-10-23T19:00:45Z] OH [2026-10-23T19:00:51Z] new libreboot release came out, edk2 is now the default uefi impl [2026-10-23T19:00:53Z] i need to test it [2026-10-23T19:01:06Z] i was working on getting that going myself but leah finished it in like a week [2026-10-23T19:01:08Z] seemingly [2026-10-23T19:05:32Z] dery: https://github.com/finit-project/finit [2026-10-23T19:05:34Z] there's also this [2026-10-23T19:05:38Z] i haven't tried it and don't know muhc about it but it seems neat [2026-10-23T19:11:25Z] yea sounds neat! [2026-10-23T19:50:48Z] I'm happy to have banned AI out from my own kiss hard fork [2026-10-23T19:51:40Z] it's really costly to let it go around and I kept seeing dumb and/or mean people come up in a few😑 [2026-10-23T19:51:48Z] *few places [2026-10-23T19:52:33Z] the cost to deal with this crap is too invasive for someone like me [2026-10-23T19:52:51Z] and I won't fight fire with fire ofc, waste of time [2026-10-23T19:53:40Z] I want to focus on the community and sharing aspect, not only a result [2026-10-23T19:54:39Z] on that topic, there's some cool articles I could recommend taking a look at, specifically https://cyrneko.eu/posts/ai-philosophy.html and https://lumi.codeberg.page/blog/software-community.html [2026-10-23T20:22:53Z] i think it's invaluable for information and the resulting utter death of intellectual property kicks ass [2026-10-23T20:23:01Z] vibecoding is incredibly cringe though [2026-10-23T20:23:22Z] speaking of vibecoding, i need to package neovim to replace vim [2026-10-23T20:45:19Z] kris_: attribution won't die though [2026-10-23T20:46:22Z] LLMs don't care about it — that's the real issue [2026-10-23T20:47:10Z] and it's unlikely to get solved, no matter how many open-weight models are made [2026-10-23T20:47:54Z] and it happens to be a somewhat "good" search engine, maybe, but that's a backup plan [2026-10-23T20:48:34Z] it's a lot better as a search engine compared to anything else i've found unfortunately, i do remember a few years ago search engines seemed to be largely decent [2026-10-23T20:48:42Z] if we, humans, can't live without an llm as our search engine, that must mean something really bad happened [2026-10-23T20:48:53Z] kris_: fair [2026-10-23T20:48:56Z] i'm not really willing to cope with duckduckgo [2026-10-23T20:48:57Z] nor google [2026-10-23T20:49:08Z] always missing *some* bit of crucial information that way [2026-10-23T20:49:18Z] google without AI overview is good through [2026-10-23T20:49:28Z] at least in my experience [2026-10-23T20:50:03Z] https://codeberg.org/gabrielf/omnisearch [2026-10-23T20:50:05Z] i need to try this out [2026-10-23T20:50:09Z] searxng is unusably dogshit [2026-10-23T20:50:16Z] metasearch engines are a great idea but searxng is just awful [2026-10-23T20:50:59Z] and honestly I'm not really sure it's a good idea to take the info that comes from an agent's searching skill as-is, especially when you need it for more serious work where you need references [2026-10-23T20:51:23Z] people keep saying that and i never have any idea what they're talking about [2026-10-23T20:51:27Z] "claude, cite your sources" [2026-10-23T20:51:37Z] >sure, here's a list of literally everything and each thing that came from these individually [2026-10-23T20:52:13Z] of course, i've not used other llms [2026-10-23T20:52:28Z] kris_: honestly? avoid using the resume as is, cross reference it with the actual sources themselves [2026-10-23T20:52:44Z] well yeah i never just take claude at its word [2026-10-23T20:52:58Z] anyway [2026-10-23T20:53:05Z] the "Claude, find me sources" thing is probably your safest bet [2026-10-23T20:53:07Z] it seems like liberas palladium has gotten substantially more performant [2026-10-23T20:53:16Z] my messages are no longer taking like 3 seconds to send [2026-10-23T20:53:40Z] palladium being their tor entry [2026-10-23T20:57:26Z] on another note: neovim's website has a bit from a certain... dhh in it https://github.com/neovim/neovim.github.io/issues/501 [2026-10-23T20:57:33Z] sigh [2026-10-23T20:57:37Z] im aware [2026-10-23T20:57:41Z] why can't we have nice things [2026-10-23T20:57:50Z] i agree more with the neovim guy here than i do any of the people trashing him [2026-10-23T20:57:57Z] ? [2026-10-23T20:57:58Z] obviously i don't like DHH but people really are acting like a garbage patch [2026-10-23T20:58:13Z] strictly speaking on what i saw him say in the issues [2026-10-23T20:58:18Z] i havent seen any comments elsewhere [2026-10-23T20:58:48Z] I don't see why one should accept any attention from this guy, like what's the point in keeping that quote [2026-10-23T21:00:18Z] it's just stupid and proves the neovim maintainer doesn't really care about giving credit back to a... very disliked character in the FOSS community [2026-10-23T21:00:53Z] whatever keeps the business going I suppose.. [2026-10-23T21:02:56Z] i think there's definitely an aspect of the guy just not wanting to entertain people randomly making demands and threatening him [2026-10-23T21:03:08Z] and he did comply anyway [2026-10-23T21:03:23Z] https://github.com/neovim/neovim.github.io/issues/505 [2026-10-23T21:03:29Z] thought i had known you from somewhere