Notices by q66 (q66@blahaj.social), page 5

a perfect day for a ride

oddly enough did not even get soaked (much)

@xerz relevant one i saw elsewhere today

@david_chisnall until you realize that everybody has a different idea of what's an idiot and what's a 0.1x developer

@domi @pj well it's less about testing it (though that part is also important) and more about core guarantees wrt. things like having user services or a session tracking framework

it's notoriously hard to do all this correctly, and if you allow multiple inits you are multiplying that work and increasing complexity for the common glue

i think stuff like application api etc should be generic (which is why i'm working on turnstile etc.) but specific oses should pick something and do it well

@domi @pj @lanodan i guess my main issue with alpine is that it has some neat components (like there is a reason i adopted apk-tools and stuff) but in the end it does not end up going the extra mile to give things real polish

by making the base system bare, you miss out on guaranteeing things (so e.g. session bus is still legacy stuff via dbus-launch, you have no usersrvs so no supervised pipewire ootb, etc.)

and the maintainer tooling could be a whole lot better even as a pile of shellscripts

@domi @pj @lanodan i guess i'm really used to all the fancy cbuild stuff these days (one tool that can do everything, is fast - parses all the templates in the collection under a second, does a ton of lints for all sorts of stuff, is always containerized, has helpful error messages, does staging so you get to know about forgotten rebuilds, integrates with git, has a bulk builder with deep topological sort, and is *the exact same tooling* that runs on the builders, on your local computer)

@lanodan @domi @pj i consider gentoo more of a metadistro than a particular system, so if they wanna go down that rabbit hole it's like the one place where it makes sense, but most distros really should not

@domi @pj i don't believe in init system freedom

@domi @pj (really, imo anybody who seriously talks about supporting multiple service managers is either unaware of what it takes to properly integrate one, or does not consider proper integration important, so i actually totally understand why somebody would go with just systemd, because i have no plans to allow more than one either)

@domi @pj is it an arc if it's been a consistent level since june 2021

@xerz tbh i ask myself that like all the time and i don't even need bot followers for it

@lanodan i ended up locating an old unused tv in the office and that supports 30hz

@lanodan yeah have fun getting that into yocto

@lanodan ensuring timings of a particular benchmark on a crappy device without vsync getting in the way

apparently it's surprisingly rare for a 1920x1080 display to support actual 30hz mode

@chris @mia hm, but /usr/sbin should have become a symlink? so i don't see how such breakage could have happened (unless arch did not make a symlink, but then it's just an arch thing and not a usrmerge thing)

@mia nixos approaches things differently than we do, and the result is pretty much the opposite

where we focus on doing as much possible with as little infrastructure as possible (and *really* care about being able to have *all* the tooling locally and being able to do everything the builders do, with very slim barrier of entry) nixos instead has a disgusting amount of infra for the sake of achieving subjective perfection and the result is rather intimidating

@xerz @4censord the actual fun part is that it still won't do what you expect at times

and then if you try bash compiled with sanitizers it will blow up regardless of how clean your shell code is, because the underlying code is bad

@xerz bash supports a lot of things but... ever tried to run it with sanitizers

messy code + feature kitchen sink == fun for days

btw i did add busctl :3

it works (and was less work than i expected but more than i hoped)

next up is manpages and tests, then i can cut a release... probably