<{^_^}>
[nixpkgs] @Ma27 opened pull request #75193 → nixos/networkd: only set gateway if it's explicitly specified in the module system → https://git.io/JeSel
<zeta_0>
evils: thanks
<zeta_0>
evils++
<{^_^}>
evils's karma got increased to 2
Church- has joined #nixos
mlen has joined #nixos
zeta_0 has quit [Quit: rcirc on GNU Emacs 26.3]
prusnak has quit [Quit: Connection closed for inactivity]
<eraserhd>
I've installed tracker and tracker-miners, which install some user .service files, but they don't install automatically... they _are_ linked into the /run/current-system/etc/systemd/user/ directory.
<eraserhd>
I can't figure out how to make Nix link them into /run/current-system/etc/systemd/user/default.wants/
qubasa_ has joined #nixos
neeasade has joined #nixos
<neeasade>
hello -- there wouldn't happen to anyone here with mattermost hosted on nixos
qubasa has quit [Ping timeout: 265 seconds]
<codygman>
Does anyone use mu4e or notmuch with emacs/home-manager? How do you handle setting the load path?
sondr3 has quit [Quit: WeeChat 2.6]
Enzime has joined #nixos
dingenskirchen has quit [Remote host closed the connection]
neeasade has quit [Remote host closed the connection]
jgt2 has joined #nixos
<slack1256>
Has anyone installed quodlibet? It doesn't install because it doesn't pass the test.
orivej has quit [Ping timeout: 250 seconds]
jgt2 has quit [Ping timeout: 276 seconds]
das_j has quit [Remote host closed the connection]
fragamus has joined #nixos
das_j has joined #nixos
<delroth>
didn't see it documented anywhere -- why are shell paths in /etc/passwd not directly pointing inside the nix store? instead they use system env symlinks
<butchery>
setting up a new install, can't get anything from nix cache, machine has network connectivity but timing out on nixos-rebuild, is something up with nix cache right now or is it on my end?
felixfoertsch has joined #nixos
felixfoertsch23 has quit [Ping timeout: 276 seconds]
fragamus has quit [Read error: Connection reset by peer]
<samueldr>
butchery: possible, but unlikely, the cache is served by a CDN network, so if there's issues it's possibly the node at your end in their network
<samueldr>
though, is it always timing out, or a random one, and others are fine?
aw has joined #nixos
<aleph->
Hmm, how's nix support on macOS?
spacefrogg has joined #nixos
<samueldr>
aleph-: try it :)
<samueldr>
it's there, but with catalina there may be teething issues I'm not much aware of
<aleph->
Haha planning to. Got a '11 MacBook pro off a mate.
<samueldr>
(there's an issue open on the nix project for that)
<aleph->
Figure it'll be a decent travel laptop
LLED2_2 has quit [Ping timeout: 265 seconds]
<aleph->
Given my main laptop is 5LB
<aleph->
And nevermind this is an old one and is about the same weight. Dang
ajs124 has quit [Quit: Gateway shutdown]
<butchery>
samueldr: its fastly, right? is fastly-debug.com still relevant? Its just spinning forever for me on any machine on the network, but their status page shows no problem...
<samueldr>
in my experience most content blockers break fastly-debug :)
<samueldr>
if it's an edge node, it might be "eventually consistent" in that it'll eventually come back, not sure exactly how it works, just know that's what serves the cache currently
<butchery>
ah yeah you're right, seems to be working fine on this machine with ub off
<aleph->
Ahhhh fastly.
<aleph->
I need to go do some automation with regard to them at work
<butchery>
any clues on nix timeout on the new machine? can ping out just fine, but nix is timing out, hitting a bit of a wall here
pamplemousse has joined #nixos
<dmj`>
so whats the latest these days with nixos
<dmj`>
or just nixpkgs
gm152 has quit [Quit: Lost terminal]
<dmj`>
what are the best derivations, the best modules added etc.
<dmj`>
what's the new jazz
<aleph->
Well you can install it in your brain now
<aleph->
Pretty cool stuff
<dmj`>
no way
<dmj`>
pics or it did not happen
iyzsong has quit [Read error: Connection reset by peer]
<kolbycrouch>
anyone know how to use something like nixpkgs.config.retroarch in a shell.nix? Declaring it in configuration.nix works but not when using let/in in a shell.nix
demfloro has left #nixos [#nixos]
<{^_^}>
[nixpkgs] @Chiiruno opened pull request #75218 → tinygo: init at 0.10.0 → https://git.io/JeSJK
mexisme has joined #nixos
jgt2 has joined #nixos
jgt2 has quit [Ping timeout: 252 seconds]
notgne2 is now known as notgne3
notgne3 is now known as notgne2
kolbyjcrouch has joined #nixos
<clever>
kolbyjcrouch: pkgs.config.retroarch ?
<clever>
kolbyjcrouch: if you want to read it
<clever>
to write to it, import <nixpkgs> { config.retroarch = something; }
<clever>
that would return a version of pkgs that obeys the config flag
kolbycrouch has quit [Ping timeout: 245 seconds]
<kolbyjcrouch>
clever: Thanks! that's exactly what I wanted.
<danderson>
hmm, how does nixos update Go packages? Coredns is a bunch of versions behind latest, and I see a 'deps.nix' full of git repos and hashes... Seems terrible to update by hand, I'm assuming there has to be tools for that
<danderson>
is there a way to stop activation from restarting a particular systemd unit?
<danderson>
I'm setting up a router, and when I screw up, the network drops. It's cold in my basement. I'd like my lifeline interface to be left alone :P
<dminuoso>
danderson: What activation are you referring to?
<danderson>
`nixos-rebuild switch` stops, starts, and restarts a bunch of systemd units
<danderson>
that activation
<dminuoso>
danderson: Well it would only restart systemd units if it has to.
<danderson>
the network configuration is declarative, and seems to want to reconfigure eno1 no matter what I do.
polman has quit [Read error: Connection reset by peer]
polman has joined #nixos
<danderson>
scratch that, found the particular configuration that makes it leave eno1 alone.
<srhb>
betawaffle: Recently I had to take out the interfaces part locally until 20.03, but that's easily done with disabledModules and a custom import.
mexisme has joined #nixos
<betawaffle>
well, two things bother me a bit: 1) it isn't always a clear direct translation to/from the nix attribute sets and the config files. i wish there was a lower-level mode where i specify files, sections in those files, and config options in those sections
<srhb>
betawaffle: There mostly is. Are you using the system.networking interface?
<betawaffle>
and 2) the validation isn't always aligned with what networkd actually supports
<srhb>
betawaffle: Why not use systemd.networkd. inster?
<srhb>
betawaffle: sorry, systemd.network
<betawaffle>
srhb: i am using that
<betawaffle>
afaict, that's the only way to configure it at all
<srhb>
betawaffle: Ah, I thought you said you were using system.networking.
<srhb>
betawaffle: It's not. What are you actually trying to do?
<srhb>
We're talking about this in a very roundabout way currently :P
<betawaffle>
i'm using networkd for my nixos-based router's network config
mexisme has quit [Ping timeout: 276 seconds]
<MichaelRaskin>
Well, you can jhust write unit files directly …
<betawaffle>
i guess i hadn't considered that, but you're right
<betawaffle>
will nixos-rebuild know it needs to reload networkd?
<MichaelRaskin>
Using system.something to keep this in configuration.nx
<srhb>
What is it you're unable to do currently?
<srhb>
I'm not sure we've determined that systemd.network can't do what you need just yet.
<pistache>
using systemd.network, you have access to extraConfig and co.
jgeerds has joined #nixos
<pistache>
however, in my experience, there aren't much cases where this is required
gm152 has quit [Ping timeout: 268 seconds]
mexisme has joined #nixos
<betawaffle>
it's working fine, there are a few things i need extraConfig for, but yes, it mostly works just fine
gm152 has joined #nixos
* srhb
shrugs
<betawaffle>
but i would like a more direct translation, so that i can rely more on the networkd docs, rather than looking up what the nixos module wants me to do
<MichaelRaskin>
I just think that if the goal is «I want to specify networkd configuration in networkd terms», it makes sense to NixOS support for direct unit generation
chloekek has quit [Ping timeout: 268 seconds]
<srhb>
The networkd module is basically a low-level, direct translation with validation of the raw structure in networkd. If you're having trouble with some part of the translation it would help to talk about where you think the mapping doesn't match the systemd-networkd docs.
nicolasnicolas has joined #nixos
mexisme has quit [Ping timeout: 246 seconds]
<betawaffle>
ok, so one point of weirdness for me is that some options (like DHCP) get specified at the top-level of a network, rather than in the `networkConfig` set. similar story with other things
<foxmean>
Hello, I've some newbie question. It's doesn't cause so much problem on me, but I wonder how to make it work. The question is: how can I make Emacsclient use the same package as regular Emacs use (I've follow the Nix way of configure and build explained in https://nixos.org/nixos/manual/index.html#module-services-emacs). I would like to run Emacs as a service (but not necessary need it). I'm successful on build and run it but
<foxmean>
Emacsclient come with only pure emacs without my packages defined in emacs.nix. For clearity, I've made a change in code block 1 of "22.2.1. Enabling the Service" to https://paste.debian.net/1120084/ (my emacs.nix is in the same directory as configuration.nix).
mexisme has joined #nixos
<betawaffle>
then completely unrelated to nixos... i've got problems with networkd seeming to not behave the way the docs suggest it will, like with ipv6
<srhb>
betawaffle: You can definitely use networkConfig for that..
<betawaffle>
srhb: nixos-rebuild complained at me when i tried originally. maybe it has changed since 18.0x?
<srhb>
betawaffle: That's pretty likely.
<srhb>
betawaffle: Basically, you can always use networkConfig, linkConfig, dhcpConfig etc. to jump straight to the low level translation
<foxmean>
I've try to change the line "services.emacs.package = import /home/cassou/.emacs.d { pkgs = pkgs; };" to "services.emacs.package = import /home/[myuser]/.emacs.d { pkgs = pkgs; };" but it fail (the output said that it didn't found /home/[myuser]/.emacsd/default.nix). So I decied to change it to my emacs.nix.
<srhb>
Or at least that's the intent as I understand it.
<srhb>
The only issue is if our validation is incomplete somewhere, which does happen.
<pbb>
which service or module should create the nix per-user profiles?
<pbb>
on a nixos installation
<srhb>
betawaffle: As far as networks go, and I *think* every other type, the top level attributes are merely convenience for documentation. They end up setting networkConfig.foo in the end.
nicolasnicolas has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
mexisme has quit [Ping timeout: 245 seconds]
<betawaffle>
wow, ok
<shyim>
Have someone pantheon running without warnings with lightdm?
<srhb>
betawaffle: So as far as I know, it does exactly what you want.
<betawaffle>
looking at the code that definitely seems to be the case now. i can only hope it wasn't the case before, and not that i just missed it before
<foxmean>
Now, I can use Emacsclient but non-of-all-my-packages-list in it. I can delete the old service and start it with my Emacs then Emacsclient come with all my packages install and my init.el configure work. But how can I start it whitout open "Emacs"? Is there a way to do that?
<srhb>
pbb: There are multiple modules involved, ultimately pulled together by activation scripts.
<srhb>
pbb: eg shaow and user-groups.
<srhb>
shadow*
gm152 has quit [Quit: leaving]
mexisme has joined #nixos
<pbb>
ok hm, it seems to not work correctly in a nixos vm built with "nixos-rebuild build-vm". /etc/profiles/per-user is set up correctly but /nix/var/nix/profiles/per-user isn't
<pbb>
and all the tools get confused, for example nix-env is looking in /nix/var/nix/profiles/per-user
<srhb>
pbb: iirc the per user profiles are created on the fly by nix.
<pbb>
they aren't
<srhb>
Oh well. :-)
smatting has quit [Ping timeout: 265 seconds]
<betawaffle>
ok, my next question is what helper function should i use to convert a list of things into dynamically-named attributes?
<betawaffle>
like i want to generate network unit names from something higher level
<chrisaw>
Folks - I'm trying to get an icon to show up for an app I package (should be simple!) but I just can't figure it out: https://termbin.com/bqqs
<chrisaw>
I've created an overlay to test before I update my real package with the fix but no matter what I do - GNOME seems to just ignore it completely and I never end up with the app icon showing! It's driving me nuts! D:
knupfer has quit [Remote host closed the connection]
knupfer has joined #nixos
captn3m0 has quit [Ping timeout: 246 seconds]
<srhb>
foxmean: I'm having trouble understanding exactly what you mean. You're having trouble giving services.emacs.package the emacs you've created with emacsWithPackages?
<srhb>
foxmean: Can you check whether your user emacs service contains the correct path to the emacs with packages you created?
mexisme has joined #nixos
ddellacosta has quit [Ping timeout: 240 seconds]
<foxmean>
srhb: Yes, I'm having troble giving services.emacs.package the emacs I've created with emacsWithPackages. :D But I don't know how to check my urser emacs service contains the correct path to the emacs with packages I crated. T-T
<foxmean>
How can I check it?
brightone has joined #nixos
<manveru>
foxmean: `systemctl --user cat emacs`
<{^_^}>
[nixpkgs] @orivej-nixos pushed 2 commits to master: https://git.io/JeStC
polman has quit [Read error: Connection reset by peer]
polman has joined #nixos
<chrisaw>
Is there a simple way to use makeDesktopItem without it actually creating the folder structure? I'm starting to think that may be the problem here. :|
<{^_^}>
[nixpkgs] @teto pushed commit from @r-ryantm to master « grip: 4.0.0 -> 4.0.1 »: https://git.io/JeSmf
Boko_cecdc has joined #nixos
mexisme has joined #nixos
<bdju>
Is the deluge PR still being worked on? It looks almost ready, but no activity since October.
ng0_ has joined #nixos
<das_j>
umm dbus currently fetches configs from services.dbus.packages and automagically includes system.path there. Is there any reason not to take the impure approach and add the paths to environment.pathsToLink and just point dbus to /run/current-system/sw/whatever?
ng0 has quit [Ping timeout: 260 seconds]
Boko_cecdc has quit [Read error: Connection reset by peer]
<das_j>
I'm asking because that causes the dbus unit to be changed when the system path is changed
kleisli has joined #nixos
jco has joined #nixos
mexisme has joined #nixos
dansho has quit [Ping timeout: 240 seconds]
<betawaffle>
what's the deal with this: "Dec 08 10:12:11 router systemd-networkd[22445]: Could not set hostname: Changing system settings via systemd is not supported on NixOS."
mexisme has quit [Ping timeout: 245 seconds]
<srhb>
betawaffle: What do you mean?
<betawaffle>
i get that error in the journal when i do nixos-rebuild switch
<betawaffle>
my hostname is getting set properly, though...
<srhb>
betawaffle: It's (probably) not systemd doing that though -- our systemd fork stubs all that out entirely.
<betawaffle>
the message is coming from systemd-networkd (which i'm using for network config)
<srhb>
betawaffle: That functionality is entirely disabled in systemd.
<betawaffle>
should i be disabling the hostnamed unit or something?
<srhb>
No need.
<betawaffle>
(i didn't enable it explicitly)
<srhb>
As long as it's being set through other means you're fine.
<srhb>
Just don't rely on systemd-networkd setting it (eg. via dhcp)
<betawaffle>
i don't like error messages
pointfourone has joined #nixos
mexisme has joined #nixos
pointfourone has left #nixos [#nixos]
<betawaffle>
what does nixos use to set the hostname in whatever way it's happy with?
Boko_cebde has quit [Read error: Connection reset by peer]
<srhb>
betawaffle: system activation scripts.
<srhb>
(ie. the hostname command)
<jonge[m]>
hi there! i have an nginx service running and that proxies further to hydra. sometimes i get http error 500 when accessing that page. should nginx restart once hydra is restarted, e.g. by saying `systemd.services.hydra.wantedBy = [ "nginx.service" ];` or is that not the right thing to do?
<{^_^}>
[nixpkgs] @Lassulus pushed commit from @dsg22 to master « transmission: Configurable download directory permissions »: https://git.io/JeSmj
captn3m0 has joined #nixos
<srhb>
jonge[m]: Well, without better logging all we know is that it can't talk to the backend. You might want to fiddle with the logs to know more. But I guess you could also just take a guess that the resource nginx is proxying to has gone missing and try to remedy that so it won't disappear.
<scasc>
If having a single-user setup other users will clearly not be able to install new packages etc. Is it still possible to have user dependent "profiles", different set of _activated_ apps per user, or does a single-user install mean that all users will have exactly the same environment?
<srhb>
jonge[m]: (eg. this could happen if you're proxying to some hostname and the ip of the backend changed, since nginx won't look that up again) -- but really, the best thing to do here is probably to log more :)
<srhb>
scasc: You can make as many profiles as you like, but they should then probably be read-only.
<scasc>
"In Nix, different users can have different “views” on the set of installed applications. That is, there might be lots of applications present on the system (possibly in many different versions), but users can have a specific selection of those active — where “active” just means that it appears in a directory in the user’s PATH. Such a view on the set of installed applications is called a user environment, which is just a director
<scasc>
y tree consisting of symlinks to the files of the active applications."
<srhb>
scasc: See the nix-env manual and the --profile parameter
<srhb>
Or you can generate the profiles by hand, they're not very magical.
mexisme has joined #nixos
Boko_ededa has joined #nixos
<scasc>
here the manual says: "such a view ... is called a >>user environment<<." -- that's really just what's called "profile" otherwise in Nix, right?
<srhb>
scasc: Yes.
polman has quit [Read error: Connection reset by peer]
<scasc>
Somewhat confusing :-) If "profile" is the terminology, maybe the manual should use it ^^.
<srhb>
Mind that the users won't be able to generate these profiles themselves, because the symlink tree is built as just another nix package. So you'll have to provide them up front.
<srhb>
scasc: I would tend to agree.
polman has joined #nixos
Boko_ededa has quit [Write error: Connection reset by peer]
<scasc>
@srhb: with (pregenerated) profiles, how would I make the correct profile to be loaded on graphical login?
Boko_deaec has joined #nixos
lsyoyom has quit [Quit: WeeChat 2.6]
<srhb>
scasc: I think all the loading mechanisms are still in place in /etc/profile, so the user just needs $HOME/.nix-profile to be a symlink to /nix/var/nix/profiles/$someprofile -- usually /nix/var/nix/profiles/per-user/$username
mexisme has quit [Ping timeout: 246 seconds]
<srhb>
scasc: Mind, why are you trying to do this? It sounds very cumbersome compared to a multi user setup.
mexisme has joined #nixos
endformationage has joined #nixos
Boko_deaec has quit [Write error: Connection reset by peer]
mexisme has quit [Ping timeout: 246 seconds]
polman has quit [Read error: Connection reset by peer]
<schmittlauch[m]>
Hi, I'm currently running the installer with a manually configured network. Any ideas on how to set up DNS resolution? Looks like systemd-resolvd is used.
<schmittlauch[m]>
just modifying the /etc/resolv.conf doesn't do the trick.
Boko_ebdea has quit [Read error: Connection reset by peer]
<scasc>
@srhb: multi-user nix (on an other OS than NixOS) doesn't have a "common" system profile, so every user -- non tech-savvy ones too -- have to manage their environment/profile themselves?
<scasc>
that was the guiding idea, but I bet there's a good solution
mexisme has joined #nixos
<srhb>
scasc: I think the default profile (that is, the profile used by root when nix-env'ing) is global by default.
<srhb>
scasc: It's the one called /nix/var/nix/profiles/default
<clever>
nixos also has that profile
<{^_^}>
[nixpkgs] @kampka opened pull request #75263 → coturn: Support secrets file for configuring the static-auth-secret → https://git.io/JeSOt
<clever>
but that default profile cant do system level things
bvdw has quit [Read error: Connection reset by peer]
simpson has joined #nixos
bvdw has joined #nixos
polman has quit [Ping timeout: 250 seconds]
<simpson>
(This might be the wrong channel, lemme know.) I'm using Home Assistant on NixOS. Things are working well, generally. However, some services don't seem to get their Python dependencies installed, even with autoExtraComponents = true;
<simpson>
Is there some additional configuration that I need to do? Perhaps I need to override the hass package itself?
nicolasnicolas has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<keithy[m]>
Hi guys, I want to use nix with mac os 10.11, any chance?
<keithy[m]>
there I was thinking that nix would be my salvation
<keithy[m]>
hmm perhaps I should upgrade my mac pro
<simpson>
I'm not sure whether there's limitation on supported versions of OSX. The manual doesn't seem to have any minimum or maximum versions listed.
polman has quit [Read error: Connection reset by peer]
<{^_^}>
nix#2925 (by mroi, 26 weeks ago, open): /nix will not be writable on macOS Catalina
polman has joined #nixos
<{^_^}>
[nixpkgs] @lsix opened pull request #75274 → vmTools.fillDiskWithRPMs: fix cpio use → https://git.io/JeSss
[Leary] has joined #nixos
ixxie has joined #nixos
Lears has quit [Remote host closed the connection]
racoonslug has joined #nixos
<Thra11>
/sys/class/backlight/edp-backlight is owned by root:root. Which system component should be responsible for e.g. setting a different group so non-root users can control the brightness?
polman has quit [Read error: Connection reset by peer]
<aterius>
Nice! Sorry, new user, but since a nix system configuration is really a nix project, this should work for home-manager/my nixos configuration.nix?
<clever>
aterius: yeah, niv should work nicely there
<aterius>
Perfect. Thank you! I'll give this a shot
polman has joined #nixos
knupfer has joined #nixos
slack1256 has quit [Remote host closed the connection]
<betawaffle>
is there a helper function to re-indent a string to a deeper level?
<{^_^}>
[nixpkgs] @c0bw3b pushed commit from @r-ryantm to master « libebml: 1.3.9 -> 1.3.10 (#75246) »: https://git.io/JeSsS
<betawaffle>
clever: so instead i should?
<clever>
betawaffle: does it need to be indented, or is it just to make the config pretty?
<betawaffle>
the latter
<clever>
id just ignore it then
<betawaffle>
:'(
<clever>
my OCD isnt that bad, lol
<betawaffle>
what if it did need to be indented?
<clever>
there are string split map, and concat functions, and that can let you add something
<keithy[m]>
The first install does this: T/nix-binary-tarball-unpack.JylwkKwT5m/unpack/nix-2.3.1-x86_64-darwin/install: macOS 10.11.6 is not supported, upgrade to 10.12.6 or higher
<das_j>
zimbatm: I have successfully created 19.09 boxes for i686 and x64. Do you want me to upload them into my own namespace or the nixos one?
selfsymmetric-pa has joined #nixos
<zimbatm>
das_j: are they for libvirt of virtualbox?
<das_j>
libvirt, but virtualbox is currently in the making
<zimbatm>
do you want to take over the maintainership of nixbox?
scasc has quit [Ping timeout: 276 seconds]
<betawaffle>
i'm trying to define some custom types for my module in a separate file. is there something i should be using instead of this: "types = types // (import ./types.nix { inherit lib; });"
<das_j>
zimbatm: Well I could partly maintain it (qemu and vbox), but not the vmware stuff
<betawaffle>
i want `types` to have the types that come with nixpkgs, and my custom ones
<zimbatm>
yeah the vmware stuff is unsupported
<zimbatm>
it requires a license to build it
<aterius>
yorick: Should I just import sources with a relative path in the specific nix file that needs the sources from niv? Or is there a better way in nix to have sources globally available to all submodules?
<aterius>
This may be more a nix language question than a niv question per se
selfsymmetric-pa has quit [Remote host closed the connection]
<scasc>
given nix store can grow somewhat unpredictably with transient dependencies, how much space would I provision for the NixOS partition?
polman has quit [Read error: Connection reset by peer]
<srhb>
scasc: It's really hard to gauge without knowing your use case, honestly
polman has joined #nixos
<srhb>
Developing in some languages can really take up a ton of space.
<scasc>
Or can nix on the classic somehow share the nix store with the NixOS?
<srhb>
You could, but I wouldn't.
<srhb>
compression and store optimisation helps a lot with store space too.
<scasc>
srhb: yeah, and I am going to play with Haskell, which is notorious for blowing up nixstore
<srhb>
scasc: Yeah, optimization and compression helps a ton there.
zupo has joined #nixos
<srhb>
scasc: fwiw I haven't felt comfortable doing haskell dev with less than, say 40 gigs of nix store.
<srhb>
And I hate gc'ing so I go much larger.
<scasc>
That's why I asked about sharing nixstore somehow, because 40G is really something I'd reuse, I guess.
<srhb>
scasc: Yes. The problem is that you have to be careful with indirect roots if you share nix store
polman has quit [Read error: Connection reset by peer]
<srhb>
scasc: eg. /nix/var/nix/profiles/.../someroot might point to a /home/user/someroot on one OS but not the other
polman has joined #nixos
<srhb>
scasc: So depending on the active system it's now free to be garbage collected.
<srhb>
scasc: A safer way would be to ensure that all indirect roots _also_ live on a shared filesystem
<srhb>
scasc: But this requires discipline.
<srhb>
scasc: The store itself really doesn't care.
<MichaelRaskin>
Well, not much discipline is needed to only have roots inside /home
<scasc>
I guess that's all over my head yet. I mean, the reason I want to parallel install is because I'm unsure I can climb the NixOS mountain (given my temporal resources). So screwing with nixstore is probably also something I don't want to do then..
<srhb>
scasc: If you share /home and /nix and you never ever put gc roots outside of /home, you're good.
<srhb>
scasc: And fwiw I felt the same, but my arch partition ended up living less than one day. :-P
amanjeev_ has joined #nixos
amanjeev_ has quit [Client Quit]
polman has quit [Read error: Connection reset by peer]
polman has joined #nixos
<scasc>
@srhb: but if /nix (and /home) are their own partitions, for the "root" partition for nixos itself probably < 1G suffices? I mean, there is really nothing outside of /nix, but some symlinks into the store in e.g. /bin etc, right?
<srhb>
scasc: You may want about that much space for /boot itself, but otherwise yes.
<{^_^}>
[nixpkgs] @risicle opened pull request #75284 → aiger, lingeling: enable for darwin → https://git.io/JeSZv
<srhb>
scasc: My root directories, excluding /home and /nix, currently take up about 600M, and most of that is /var
<MichaelRaskin>
Well, depending on your needs there might be /var
<MichaelRaskin>
There is always a question of /tmp
<srhb>
Ah, yes.. I have tmpontmpfs, but I have oodles of RAM.
<srhb>
In case my boss is reading, I will change that to "I have just about enough RAM"
<gchristensen>
srhb: I heard you had just barely enough RAM actually and definitely needed more
<srhb>
gchristensen: What do you know, I've heard this too... :-)
<MichaelRaskin>
The correct claim, I guess is «I have the necessary RAM for some rare but important operations, which allows me not to save RAM in normal operation»
<srhb>
There we go. ;-) #nixos is always able to help.
<scasc>
tmpontmpfs could be an option for me, 16G RAM on the Dell Latitude I'm setting up.
<srhb>
scasc: But you're going to be developing Haskell :P
<srhb>
(it's fine! I did that for quite a while)
<{^_^}>
[nixpkgs] @wucke13 opened pull request #75285 → Add autochanger support to bacula-sd → https://git.io/JeSZU
<MichaelRaskin>
Some builds are both /tmp hungry _and_ RAM hungry!
<gchristensen>
for Haskell you need a minimum of 1T of (virtual)mem
<scasc>
Good to know!
<srhb>
gchristensen: #troll :P
<gchristensen>
=)
<srhb>
But yeah, those processes always look fun :P
amanjeev has joined #nixos
<MichaelRaskin>
In the meantime, in Common Lisp land, the default is having 1GiB address space and not a byte more.
mounty has joined #nixos
<MichaelRaskin>
(I am not saying I am never overriding this default)
polman has quit [Ping timeout: 250 seconds]
selfsymmetric-mu has joined #nixos
mexisme has joined #nixos
<wucke13>
That reminds me back when I was using Arch linux (btw. I used Arch). I wanted to build telegram-destkop from the aur, and the build was a real build (e.g. not a binary download).
<amanjeev>
:)
polman has joined #nixos
<das_j>
oh yes that sounds like my arch back then
<wucke13>
I was quite baffled, that my workstation with 64GB of RAM went OOM. Turned out, that I build a full Qt.
<das_j>
there was also telegram-desktop-bin btw
<wucke13>
Jep, that was the way that I went from that on :D
<MichaelRaskin>
I think my laptop with 32GiB RAM builds Qt fine from Nixpkgs. But I do have a real /tmp
<wucke13>
I didn't, since someone in the community told me to ramdisk/tmpfs everywhere possible
<wucke13>
(Which was nice for mostly everything but these kind of bigger builds)
<keithy[m]>
ok, I tried nix 2.2.2 and its got past the download stage!
<MichaelRaskin>
I think gchristensen can advise you to have / on tmpfs
<wucke13>
Why is that, gchristensen ?
<gchristensen>
I use a real filesystem for / actually, but I erase it on every boot to save on ram :)
<keithy[m]>
Might I suggest some docs, 2.2.2 is the last version that support mac os el capitain
<das_j>
s/erase/rollback/
<{^_^}>
[nixpkgs] @kalbasit merged pull request #75013 → go: do not replace path to zoneinfo.zip and mime.types → https://git.io/JeDPP
<das_j>
gchristensen: Btw, what was your goal with LDAP? SSH keys and users there?
<MichaelRaskin>
gchristensen: is your /etc/ a symlink to store already?
<das_j>
management of the users via gui or nix?
<MichaelRaskin>
Although I hope you won't read my branch until you read RFC#46·
<gchristensen>
das_j: management of users with nix ideally. SSH keys, users/passwords
<MichaelRaskin>
(because my branch is a bit of a horror summoned by a non-NixOS user to just boot the test VM with fully static /etc)
polman has quit [Read error: Connection reset by peer]
<das_j>
alright
polman has joined #nixos
mounty has quit [Quit: Konversation terminated!]
sudoforge has joined #nixos
polman has quit [Read error: Connection reset by peer]
<scasc>
*ponders* Hm..., shall I be overconfident and use btrfs or stick with ext4? (I know, this is not strictly a NixOS question :v)
<wucke13>
Vote for BTRFS
<clever>
scasc: ive been sticking to zfs
<wucke13>
Also an excellent choice, dempending on the required flexibility
<emilsp>
I vote against btrfs due to past experience
<wucke13>
scasc: Just a few notes; When using either zfs or btrfs, I would create at least three subvolumes/zfs': /home /nix /var
<scasc>
Thanks, @wucke13, good point
polman has joined #nixos
<ajs124>
xfs is also nice, if you want one more option
* gchristensen
moved another machine to zfs today
<emilsp>
What would be the best way to package up a binary artifact that should be derived from my config and then put somewhere in the /var/run/? Speaking specifically about dropbears host keys.
<wucke13>
And maybe four. For my btrfs installs I use /@ as root, /@home, /@var and /@var for the rest
<wucke13>
xfs isn't COW. scasc: Do you explicitly want a COW fs?
<ajs124>
wucke13: xfs has COW support :P
<MichaelRaskin>
I have a lot of !!FUN!! experience with BtrFS years ago. I run /nix on btrfs now, and it is boringly predictable
<ajs124>
since kernel 4.9
<gchristensen>
heck, xfs supports cp --reflink=always and zfs doesn't even
<MichaelRaskin>
(I use ext4 for my /home anyway)
<{^_^}>
[nixpkgs] @c0bw3b merged pull request #74214 → rdf4store: remove pkg and service modules → https://git.io/JeP6Y
<{^_^}>
[nixpkgs] @c0bw3b pushed to master « rdf4store: remove pkg and service modules »: https://git.io/JeSZc
<wucke13>
AFAIK that is more of an afterthought which doesn't offer the real-deal CoW experience, but please educate me/point me to a recent document stating otherwise
polman has quit [Excess Flood]
<MichaelRaskin>
scasc: if you go with ext4 for /nix, request more inodes
<clever>
gchristensen: ive looked at how --reflink=always, and i believe it was an xfs specific ioctl, lol
<gchristensen>
hah
<MichaelRaskin>
Nix loves symlinks, symlinks eat inodes and do not eat data sectors, so…
<das_j>
clever: nixdroid actually uses that btw
<clever>
das_j: reflink?
<ajs124>
It uses copy_file_range. A very good syscall.
<{^_^}>
[nixpkgs] @bjornfor pushed to staging « wrap-qt-apps-hook.sh: wrap binaries in ./sbin/ »: https://git.io/JeSZE
<dongcarl>
clever, Sorry I'm a n00b, but wouldn't systemd.tmpfiles.rules = config.systemd.tmpfiles.rules ++ [ <my new rule here ]; cause infinite recursion?
<wucke13>
A nice. I'm also quite satisfied with ZFS. The only thing that I'm in desperate desire towards ZFS is the dynamic adding/removing of devices and live migration of RAID level, that BTRFS offers.
<clever>
dongcarl: yes, nix already merges the value from other files for you
<clever>
dongcarl: if you just want to add rules, just systemd.tmpfiles.rules = [ <my new rule here ];
<{^_^}>
[nixpkgs] @bjornfor pushed to master « makeDesktopItem: use runCommandLocal »: https://git.io/JeSZg
<__monty__>
das_j: If your question was about stratis. It's at 2.0 so I think RH is already deploying it? Might've been wrong about its relation to COW though.
<gchristensen>
when 2 files define it, the code which imports the two files has logic which knows how to merge hte lists. Nix by default doesn't know how to do that
<dongcarl>
gchristensen, Oh that's very good to know
<dongcarl>
Lemme digest a bit, will come back, thanks for patient answers
<das_j>
__monty__: Maybe worth trying it once it supports some other RAID levels than 0
<das_j>
or does 2.0 do that?
<gchristensen>
dongcarl: glad that helped!
<__monty__>
das_j: No idea. A lot of talk about the DBus API in the release post, no mention of RAID.
<das_j>
yeah, that's what I saw as well
Church- has quit [Quit: WeeChat info:version]
<MichaelRaskin>
I thought Stratis is a unifying management level including, for example, Gluster?
<{^_^}>
[nixpkgs] @bjornfor pushed commit from @lsix to master « vmTools.fillDiskWithRPMs: fix cpio use »: https://git.io/JeSZy
<o1lo01ol1o>
I have some source that instructs me to make a $src/build directory and do some stuff there, what's the prudent way to do this in a derivation? I know there's a $tmp location, but I don't understand the permissions around it.
<schmittlauch[m]>
Is there a way to specify additional kernel cmdline arguments for booting? I only found a general `extraConfig` option for grub.
<drozdziak1>
Hey, I'm using i3 under NixOS but it won't read my config file under $HOME/.i3/config
chloekek has quit [Ping timeout: 250 seconds]
<drozdziak1>
That path is described as the default
<drozdziak1>
I did not alter `configFile` either
<schmittlauch[m]>
found it, `boot.kernelParams`
<tilpner>
drozdziak1: Does ~/.config/i3/config exist?
<chrisaw>
Guys, I package an Electron app up and that works great but the problem I've found is - that app defaults to autostarting on login and sadly it doesn't start the wrapped version of the app - it instead just starts electron itself. Any ideas?
<{^_^}>
[nixpkgs] @bjornfor merged pull request #75041 → kernel generate-config.pl: Show all errors, not just the first → https://git.io/JeDSh
<{^_^}>
[nixpkgs] @bjornfor pushed commit from @andersk to staging « kernel generate-config.pl: Show all errors, not just the first »: https://git.io/JeSnG
<drozdziak1>
tilpner: Oh crap it might, one sec
<drozdziak1>
I think I didn't import my own dotfiles before running i3
domogled has quit [Ping timeout: 265 seconds]
<schmittlauch[m]>
chrisaw: Try finding the part in the code setting up the autostart, looks like you need to patch it. But maybe `substituteInPlace` will do the job
<schmittlauch[m]>
if it's a binary, try grepping for the unwrapped file in the binary
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
talvdav has joined #nixos
zupo has joined #nixos
<{^_^}>
[nixpkgs] @marsam opened pull request #75290 → awsweeper: init at 0.4.1 → https://git.io/JeSno
<talvdav>
hi, i forgot to set my hostname during installation and i wanted to change that afterwards but it dose not work. is that normal or do i miss something?
<chrisaw>
schmittlauch[m]: Not a bad call - it's minified js though being called via Electron and it seems like an npm module controls the auto-launch but I don't see any params for it being defined so I assume it's using some defaults. D:
<tilpner>
talvdav: Set networking.hostName
<talvdav>
tilpner: it dose not work
alfborge has joined #nixos
<tilpner>
talvdav: You ran a nixos-rebuild switch, and it stays the same?
<talvdav>
tilpner: yes
<tilpner>
I don't know if it needs a reboot or works on a running system
<talvdav>
i tried everything, reboot setting networking.hosts aswell
<talvdav>
deleting the file befor rebuild switch
<talvdav>
editing /etc/hostname
<tilpner>
Which file?
<talvdav>
nothing works
<chrisaw>
schmittlauch[m]: Disregard, I've found it. So if I can deminify it then I can patch it on-the-fly. Hmm.
<talvdav>
tilpner: /etc/hostname
<talvdav>
might this be a bug?
<tilpner>
talvdav: Run nixos-option networking.hostName
<alfborge>
Where do I find best practice for managing custom packages? Currently I have a folder called ~/nix where I put stuff which I then install using nix-env -if ~/nix/name-of-package.nix
<schmittlauch[m]>
alfborge: You might want to try overlays
<tilpner>
Test with runCommand "foo" {} ''echo $out > $out''
<tilpner>
Also see builtins.placeholder "out"
ixxie has quit [Ping timeout: 250 seconds]
alfborge has quit [Remote host closed the connection]
<chrisaw>
I always thought that was just a placeholder dir until the package was actually "merged" (installed? whatever. :))
talvdav has joined #nixos
<schmittlauch[m]>
chrisaw: It will be the final store path. Just checked it in a nix-shell
<talvdav>
i think my installation is broken, i am dualbooting windows10 and i notice that if i select nixos it dose not offer me a selection of generations
<{^_^}>
[nixpkgs] @kampka opened pull request #75293 → postgresql: Quote role names when creating database users → https://git.io/JeSc3
<talvdav>
if run nix-env --list-generations as root it dose not print anything thats not normal right?
<tilpner>
Hold on
<tilpner>
Add -p /nix/var/nix/profiles/system to that nix-env invocation
<chrisaw>
schmittlauch[m] & tilpner: Thanks folks - really appreciate it! Managed to add a patch to my package (rambox-pro) which fixes the launch path now. Finally - no more blank Electron windows firing up whenever I login!
<tilpner>
talvdav: That looks fine
<schmittlauch[m]>
yay
<chrisaw>
For my next trick... to port VMWare Workstation to run on NixOS.... this should be fun. D:
<schmittlauch[m]>
bah, proprietary software…
<{^_^}>
[nixpkgs] @kampka opened pull request #75294 → Postgres service refactor → https://git.io/JeSc0
<schmittlauch[m]>
chrisaw: You should already introduce yourself to `patchelf`, you gonna need it.
<chrisaw>
Yeah, sorry. :( Filling up the nixpkgs repo with proprietary junk. D:
<chrisaw>
schmittlauch[m]: Yeah, sadly I'm all too familiar with it. :)
<{^_^}>
[nixpkgs] @Lassulus pushed commit from @r-ryantm to master « duktape: 2.4.0 -> 2.5.0 »: https://git.io/JeScu
<aterius>
Are there any good examples of how to import/use niv in nix expressions? Having trouble going from just the example in the readme
<chrisaw>
First annoyance first though - I can't even extract the damn package yet because it's one of those "self-extracting shell script with a huge tar.gz file embedded" kind of larks and whilst it does have an extract option (woo) it doesn't work on NixOS for some reason (boo.) :(
alfborge has joined #nixos
<tilpner>
talvdav: You never replied to my suggestion about nixos-option
cosimone has joined #nixos
<tilpner>
Oops, in that gist
<talvdav>
tilpner: i includet my prompt so you can see that it sais nixos
<tilpner>
You need to figure out which generation is booted by default
<talvdav>
ok
<MichaelRaskin>
chrisaw: maybe shar would help
<talvdav>
tilpner:how do i do that?
<tilpner>
talvdav: /run/booted-system and /run/current-system are symlinks to store paths. Do those differ for you currently?
<tilpner>
That will tell us if you activated a new system after booting into an old one
alfborge has quit [Remote host closed the connection]
<tilpner>
And you can compare them to the links in /nix/var/nix/profiles
<tilpner>
This could happen if e.g. you aren't mounting your boot partition, so no persistent changes can be made to your bootloader config
<tilpner>
That would explain a lack of generations in the menu during boot
<talvdav>
ok so i ran ls -l /run/booted-system /run/current-system they are the same
<chrisaw>
MichaelRaskin: Hah, I'd never even heard of that - I'll have a play around and see if that can indeed help to extract this thing! Thanks!
patrl has quit [Quit: WeeChat 2.6]
<MichaelRaskin>
Also, you can just use csplit and look for the separator script uses to unpack itself
<talvdav>
ok but they differ from system-7-link
magneticduck has quit [Remote host closed the connection]
<wucke13>
From within the machine (nixos-enter) or from the iso's shell?
<lassulus>
shouldn't matter I guess
<gchristensen>
19.09 doesn't have 2.0
<gchristensen>
19.09 doesn't have 2.04
talvdav has quit [Ping timeout: 276 seconds]
<wucke13>
background: I ran in the issue of having a zstd compressed btrfs as root, and grub doesn't know how to mount that.
jgt2 has joined #nixos
<schmittlauch[m]>
wucke13: damn, good to know. I'm just installing a new 19.09 an planned to enable zstd compression
philr has joined #nixos
<lassulus>
if you have a separate boot partition your kernel/initrd must be able to mount the root fs
* schmittlauch[m]
reformats ALL the things
<gchristensen>
I think you'll need to use nix-channel to change from nixos-19.09 to nixos-unstable, update the channel, then run `nix-shell -p grub` to get a recent grub
<lassulus>
normally adding it to boot.supportedFilesystems should suffice, I think it also should get added automatically?
<wucke13>
Yup, have done that so far.
<wucke13>
Grub 2.02 cannot open zstd compressed btrfs
<gchristensen>
it is a shame that grub 2.04 broke things
<schmittlauch[m]>
wucke13: Does grub deal with lzo compression?
<wucke13>
I guess yes
<wucke13>
Unfortunately grub-install from within the nixos iso throws an error.
<lassulus>
If your boot partition is a separate one (ex. vfat) grub has nothing to do with your root fs
<lassulus>
whats the error?
<wucke13>
It was missing the --boot-directory
<schmittlauch[m]>
Btw, any ideas why the manual installation instructions still recommend ext4? I read that with all the small files in the store there's the danger of running out of inodes.
<schmittlauch[m]>
btrfs on the other hand doesn't have a fixed number of inodes
<gchristensen>
ext4 works great for nixos for all but the most strenuous deployments like big build machines
<wucke13>
Will anything but explicitly running either `grub-install` or `nixos-rebuild --install-bootloader` overwrite my new grub-2.04?
mumuluxi has joined #nixos
urban_murmeli has joined #nixos
<gchristensen>
I don't think I've heard of a user having problems with ext4 inodes, outside of that context, anyway
<tobim[m]>
patrl: The problem seems to have been introduced with https://github.com/NixOS/nixpkgs/pull/71287. `fusermount` from the nix store does not have suid. It worked before because `/run/wrappers/bin/fusermount` is created by default.
cosimone has quit [Remote host closed the connection]
cosimone has joined #nixos
chloekek has joined #nixos
<keithy[m]>
so no help here
polman has quit [Read error: Connection reset by peer]
brightone has quit [Quit: WeeChat 2.6]
Church- has joined #nixos
Church- has quit [Client Quit]
Church- has joined #nixos
noudle has quit []
cosimone_ has joined #nixos
cosimone has quit [Read error: Connection reset by peer]
NoctisLabs has joined #nixos
troydm has quit [Quit: What is Hope? That all of your wishes and all of your dreams come true? To turn back time because things were not supposed to happen like that (C) Rau Le Creuset]