gchristensen changed the topic of #nixos-chat to: NixOS but much less topical || https://logs.nix.samueldr.com/nixos-chat
drakonis has joined #nixos-chat
waleee-cl has quit [Quit: Connection closed for inactivity]
<colemickens> I'm connected to a network, but there's nothing in the /etc/NM/sys-conn/BSSID file, and nothing in `nmcli --show-secrets connection show id $example | fgrep 802-1x.password`.
<colemickens> no idea how I'm connected to it :S
<colemickens> Oh! I bet this is an IWD thing.
<colemickens> yup, /var/lib/iwd/{bssid}.psk contains it.
<colemickens> +1 for runtime data out of /etc :D
<colemickens> well, I guess thats arguable.
drakonis has quit [Ping timeout: 250 seconds]
ajs124 has joined #nixos-chat
endformationage has quit [Ping timeout: 265 seconds]
rardiol has quit [Ping timeout: 268 seconds]
cransom has quit [Ping timeout: 268 seconds]
cransom has joined #nixos-chat
Guanin has quit [Ping timeout: 240 seconds]
<NinjaTrappeur> I'm about to renew my desktop's graphics card, is anybody familiar with the current state of the nvidia vs amd drivers on NixOS? Should I avoid either of these brands? (I don't mind using a proprietary blob)
<NinjaTrappeur> r/either/one/g
<ar> i have a radeon vii, and it works. had a vega64, and rx480 before that, also worked (with nixos). had r9-280x before that, on gentoo, also worked
<ar> and no screwing around with blobs being incompatible with a particular X/kernel version
<etu> The AMD drivers are open source and built into the kernel/mesa. So with a modern card you may need modern versions of said things.
<etu> While the blobs as ar states can be incompatible and stuff.
<ar> haven't tried the new 5700/5700xt cards, and i'll probably end up skipping those as they're not really faster than vii
<ar> (though a lot more efficient, even if at half the vram)
Jackneill has joined #nixos-chat
<NinjaTrappeur> Right, did not know that. Thanks.
__Sander__ has joined #nixos-chat
__monty__ has joined #nixos-chat
rardiol has joined #nixos-chat
__Sander__ has quit [Quit: Konversation terminated!]
<gchristensen> hrm. I need to force some udev-executed programs to run sequentially
<gchristensen> to assign devices sequential numbers
<joepie91> NinjaTrappeur: another +1 here for AMD. the amdgpu drivers are open-source, and even the proprietary amdgpu-pro(?) drivers afaik are built against the same stable API so that *even there* you don't get aforementioned X/kernel version compat issues anymore
<joepie91> (you used to get those for the proprietary drivers)
<joepie91> (but that was resolved with the amdgpu switchover)
<joepie91> I haven't actively checked up on nvidia for the past 2 years or so, but last I checked they still built binary blobs against specific kernel/Xorg versions
<joepie91> and the open-source drivers for nvidia are... not great
<qyliss> Nvidia aren't good players in Linux, and it shows for users -- new graphics stuff (I'm especially thinking Wayland compositers) has either taken years to support Nvidia's proprietary drivers, or flat-out refused to ever implement them.
<qyliss> So even if you personally don't care about proprietary blobs, your experience is still going to suffer because of them.
<joepie91> ^
<NinjaTrappeur> Right, thanks for the heads up.
<joepie91> same story with things like WebGL and Vulkan actually
<ajs124> gchristensen: which kind of devices? And why does it need to be run sequentially to assign those numbers?
<gchristensen> block devices
<gchristensen> and, well, something needs to be sequential. I guess I could have some file I read and lock
<NinjaTrappeur> wow, yeah, seems like wlroots doesn't support the nvidia proprietary driver. (or the other way around)
<NinjaTrappeur> too bad, the nvidia GPU seemed more power efficient :/ (at least on paper)
<gchristensen> nvidia GPUs are really nice
<ajs124> What are you doing that needs sequential numbers assigned to block devices, if I might ask?
<joepie91> NinjaTrappeur: note that power consumption numbers originating from Windows may not necessarily hold up under Linux
<joepie91> due to power management differences in drivers etc.
<gchristensen> abstracting away physical devices to devices by category and number
<gchristensen> basically, I have N servers with 12 storage drives, 2 cache drives, 1 boot drive. they all enumerate slightly differently under by-id, by-label, by-uuid, by-path, and even worse of all /dev/sdXY
<gchristensen> I can examine them and determine if tehy're storage, cache, or boot based on if they're NVME, SSD, HDD and what their capacity is
<ajs124> And where is the number supposed to come from/why is it important?
<NinjaTrappeur> joepie91: ack. Well, looks like I'm good to go into yet another rabbit hole. Thanks for the infos.
<joepie91> #nixos-chat, handing people rabbit-holes since 20XX :P
<NinjaTrappeur> :P
<gchristensen> so at the time of installation I want to create some udev rules which make links at /dev/disk/by-intent/{storage,cache,boot}{n} so then my install instructions can be: zpool create tank raidz1 /dev/disk/by-intent/storage{1,2,3} raidz1 /dev/disk/by-intent/storage{4,5,6} raidz1 /dev/disk/by-intent/storage{7,8,9} raidz1 /dev/disk/by-intent/storage{10,11,12} cache /dev/disk/by-intent/cache1 log
<gchristensen> /dev/disk/by-intent/cache2 and then every device gets the exact same tank configuration with all the correct drives, without having to customize it per chassis
<gchristensen> or more precisely, the installation instructions will convert a /dev/disk/by-intent/{type}{n} in to a /dev/sdXY and then in to /dev/disk/by-id/{id} path, and pass those paths to the system configuration
<ajs124> How about type%e. Or is that racy?
Guanin has joined #nixos-chat
<gchristensen> ooh I'll take a look
<gchristensen> what is %e, ajs124?
<gchristensen> dudle is so slow :|
<ajs124> man 8 udev says it basically does what you want. if there is a device with that name already, it adds a number or if it ends a number, increments by 1
<gchristensen> oh cool
waleee-cl has joined #nixos-chat
endformationage has joined #nixos-chat
<adisbladis> ar: I'm running a 5700XT on Nixos :)
<adisbladis> There are still some bugs, like the occasional non-recoverable freeze
<adisbladis> NinjaTrappeur: I'd stay away from Nvidia (or anything not in mainline mesa really)
<NinjaTrappeur> adisbladis: ack, seems like there's a consensus around that here. I'll probably go with a rx590 in the end.
<adisbladis> NinjaTrappeur: Not a bad choice :)
<adisbladis> I almost went for one of those too
<NinjaTrappeur> :)
rardiol has quit [Ping timeout: 250 seconds]
rardiol has joined #nixos-chat
noonien has joined #nixos-chat
lopsided98 has quit [Quit: Disconnected]
lopsided98 has joined #nixos-chat
Shoubit has quit [Remote host closed the connection]
Shoubit has joined #nixos-chat
lopsided98 has quit [Quit: Disconnected]
lopsided98 has joined #nixos-chat
avn has quit [Remote host closed the connection]
drakonis has joined #nixos-chat
<drakonis> i forgot now, but one of our dudes here's running weechat from half a decade ago
<drakonis> profpatsch whenever he shows up in here
ravndal has quit [Quit: WeeChat 2.6]
ravndal has joined #nixos-chat
Jackneill has quit [Read error: Connection reset by peer]
jD91mZM2 has quit [Quit: ZNC 1.7.4 - https://znc.in]
jD91mZM2 has joined #nixos-chat
<samueldr> at the beginning of the month, I'm aiming for the first tuesday, I'll be writing up an update on the Mobile NixOS project
<samueldr> oops
<samueldr> and there's an RSS feed for news items https://mobile.nixos.org/index.xml
<srhb> Very nice!
<srhb> Regular reports are great.
<samueldr> yes, and focusing on a date rather than goals makes it easier to act upon
<srhb> Very reasonable. :D
<gchristensen> wooo!
<infinisil> Nice
<samueldr> it's amusing to see chromeos techs being used outside chromeos, again :)
<samueldr> and nice job
<qyliss> Hoping to get all the stuff needed to do this into Nixpkgs
<samueldr> I figure this needs virtualization, right?
<qyliss> Yeah, it's running inside crosvm, the Chrome OS hypervisor
<qyliss> (which we have an open PR for by hyperfekt)
<samueldr> too bad for qualcomm "protecting" their hypervisor from use :(
<qyliss> Would be pretty sweet if you could run NixOS VMs that supported this
<qyliss> The hypervisor needs special virtio-wayland support fwiw
<qyliss> So it has to be crosvm, for now at least.
<qyliss> crosvm, btw, is the base for Amazon Firecracker, which more people seem to have heard of
<samueldr> the chromeos side of the big Alphabet soups seems to be quite... apt... at doing their things
<qyliss> Their packaging/build system is terrible but the software is very nice
<samueldr> imo, both software and hardware specificities to chromeos are interesting
<qyliss> Yep. A lot we can learn from it.
<samueldr> more to the point: the other "operating system" teams and "hardware teams" at the big ol' soup should learn from them
<qyliss> lol
<qyliss> They do share some software stuff
<qyliss> I'm building a couple of libraries that have "aosp" in their repo paths
<samueldr> I'm curious if that's coming from their need of having "an android" for android app support, rather than full sharing
* samueldr checks
<samueldr> inconclusive
<qyliss> nah, they have utility libraries and stuff that they share
<qyliss> See the ones starting with aosp here: https://chromium.googlesource.com/
<ajs124> In hindsight, I think fetching an aosp tree with fetchGit might not have been the best design decision I ever made.
<ajs124> I'm not sure I've made significantly worse, if I think about it.
<qyliss> what are you referring to?
<gchristensen> oh good grief ajs124 :P
<gchristensen> how many GB? :)
<ajs124> gchristensen: I kind of don't want to check, tbh.
<samueldr> ajs124: seen the approach in this fork? https://github.com/danielfullmer/NixDroid
<qyliss> fwiw Gitiles (Google's Git web software) will give you tarballs
<samueldr> gchristensen: I tried (nix-less) to build a GSI using "supposedly reproducible" docker builds, and it was ~40GIB for the .git metadata without git-lfs files
<gchristensen> lmao
<ajs124> samueldr: I looked though some of it, but not that much.
<ajs124> How does he clone the repo?
<samueldr> the final with-build (that failed to run correctly) size of the folder was ~300GiB
<samueldr> ajs124: haven't looked closely about how it *builds*, but how it speeds up build is documented here https://github.com/danielfullmer/NixDroid#speeding-up-the-build
<ajs124> I know he uses bindfs, because his filesystem doesn't support reflink copies (through mount namespaces, with the copy_file_range syscall)
<ajs124> Ah, yeah. I have my own solution for that part. I just patched coreutils.
<qyliss> I'm trying to add fetchFromGitiles to Nixpkgs, which will just download a tarball like fetchFromGitHub: https://github.com/NixOS/nixpkgs/pull/74862
<{^_^}> #74862 (by alyssais, 20 hours ago, open): fetchFromGitiles: init
<ajs124> Wait, googlesource supports downloading zips?
<samueldr> qyliss++
<{^_^}> qyliss's karma got increased to 14
<samueldr> yeah, but it's extremely undocumented
<ajs124> fml
<qyliss> It links you to them!
<ajs124> Doesn't matter, I won't touch this code. Unless it breaks. Which it seems to have done. Seems like I'll just have a vulnerable phone from now one.
<ajs124> It does? I think I never looked at the website much >.<
<qyliss> Yeah
<ajs124> I just assumed it wouldn't?
<qyliss> This page has a tgz link
<samueldr> ah, the undocumented bit I had in mind is raw file
<qyliss> samueldr: HOW
* samueldr tries to remember
<qyliss> I've only managed to find the base64-ed version
<qyliss> (why??)
<ajs124> randomly applying base64? What is this? M$ Exchange?
<qyliss> Same with Gerrit. It won't let you download a patch. Only a base64-ed version.
<qyliss> So you can't use fetchpatch.
<ajs124> inb4 fetchpatchbase64?
<qyliss> fetchPatchFromGerrit
<qyliss> The patch cleaning stuff should just be factored out of fetchpatch, I guess
<ajs124> or just debase64. so you can patch = debase64 fetchpatch {}.
<qyliss> That wouldn't work
<samueldr> might have been another forge I have in mind, still :(
<qyliss> Because fetchpatch modifies the patch.
* samueldr continues looking
<gchristensen> and no debasing nixpkgs!
<gchristensen> (that is a joke)
<ajs124> I assumed that word had some kind of meaning
<ajs124> qyliss: what does it modify? and why?
<samueldr> qyliss: yeah, it must have been another forge, OH, it was sourceforge's
<samueldr> though asking for plain/text for any blob seems to, indeed, give a base64 stream
<qyliss> It tries to normalize the patches and strip out anything that doesn't change the actual effect of the patch
<qyliss> So patches with different comments or whatever hash to the same thing
<qyliss> (after being run through this function)
<ajs124> Interesting. I didn't know it did that. Makes sense, I guess.
<qyliss> Otherwise you could just fetchurl
<ajs124> True. Idk, I just don't tend to think too much about these things, apparently.
<tilpner> qyliss: Can you .override fetchpatch with a fetchurl with a custom postFetch?
<ajs124> On the other hand, doing these things without putting much thought into them leads to the best error messages https://gist.github.com/ajs124/9080f16b3380fbbb7e72b0a07cfa1cc0
<tilpner> qyliss: https://tx0.co/EBw
rardiol has quit [Ping timeout: 268 seconds]
rardiol has joined #nixos-chat
<qyliss> tilpner: oh that's neat
pie_ has quit [Ping timeout: 265 seconds]
pie_ has joined #nixos-chat
vesper11 has quit [Ping timeout: 246 seconds]
vesper11 has joined #nixos-chat
noneucat has joined #nixos-chat
Jackneill has joined #nixos-chat
drakonis has quit [Ping timeout: 240 seconds]
drakonis has joined #nixos-chat
noneucat has quit [Quit: WeeChat 2.6]
noneucat has joined #nixos-chat
<worldofpeace> uhhn, I really didn't come off the way I wanted here https://github.com/NixOS/nixpkgs/pull/74943#issuecomment-561381719 :(
Jackneill has quit [Remote host closed the connection]
waleee-cl has quit [Quit: Connection closed for inactivity]
<infinisil> Ah that's unfortunate, duplicated work can be a bit annoying
<infinisil> A github feature that shows PR's that change certain files in a tree would be nice
<samueldr> blargh, I really want to work some on that github porcelain idea I have started, which would trivially enough enable such a feature
<samueldr> could I get a transfusion of time?
<worldofpeace> lol, throw in the weekly iv drip too
<worldofpeace> ok so the nixos-python-test-driver channel was helpful https://github.com/NixOS/nixpkgs/pull/74943#issuecomment-561387063
drakonis has quit [Ping timeout: 250 seconds]
drakonis has joined #nixos-chat
rardiol has quit [Ping timeout: 250 seconds]