<cole-h>
gchristensen: Weren't you gonna look at this tomorrow? :P
<danderson>
yeah sorry didn't mean to pull you back in
<gchristensen>
no worries :P I'm trying to distract myself from something which isn't very healthy
<danderson>
would definitely ssh in to help out, but I doubt anyone will give me the keys :P
<cole-h>
Me too! Except my help would consist of `systemctl restart <service>` and hoping it works
<cole-h>
:D
<gchristensen>
hey me too
hmpffff has quit [Ping timeout: 256 seconds]
<danderson>
I might start with `journalctl -u <service>` to see what happened and how to make it not happen again
<danderson>
but... yeah, poke it with a stick and see if a million bees falls out!
<cole-h>
Pfft, who looks at logs?
turona has quit [Ping timeout: 272 seconds]
<cole-h>
Wait, that's almost exactly what I do for ofborg
<cole-h>
Uh
<danderson>
... I do, but I'm reliably informed that I'm a damaged curmudgeonly sysadmin
Fare has joined #nixos
turona has joined #nixos
EdLin has joined #nixos
* cole-h
's first job was to watch logs for internal errors and ping gchr*stensen to remove the label because it was GitHub's fault and there was nothing we could do
<danderson>
all problems eventually boiled down to "someone unplugged the machine" or "disk is full"
<cole-h>
Ironic, because ofborg was failing to build stuff earlier because of 0% inodes free :D
<gchristensen>
a scourge
<gchristensen>
nix's garbage collector treats a byte as a byte and ignores inodes, and drvs take few bytes but many more inodes per mb, compared to packages
<cole-h>
Oh really?
<bdju>
god I hope the new FF update fixes some of these horrific FF issues like the file picker crashing the browser or the file picker getting bigger each time it opens so you have to keep resizing it
<bdju>
https://chunk.io/bard/e7a5ef0b88534c759a1c0f3fe5b3f331 having trouble updating, can anyone give advice? between the two sections of errors I commented out the epkgs.perl6-mode line because it looked like it was a problem, but still can't update. it's not giving me line numbers but something more confusing.
hlisp has quit [Remote host closed the connection]
hlisp has joined #nixos
magnetophon has joined #nixos
orivej has joined #nixos
<lovesegfault>
est31: You're not around anymore, right?
drakonis has quit [Quit: WeeChat 2.8]
morgrimm has joined #nixos
bigvalen has quit [Ping timeout: 256 seconds]
cyris212 has quit [Ping timeout: 256 seconds]
cyris212 has joined #nixos
bigvalen has joined #nixos
signaryk has quit [Ping timeout: 256 seconds]
hlisp has quit [Remote host closed the connection]
chagra_ has joined #nixos
signaryk has joined #nixos
Nabzi_ has joined #nixos
chagra has quit [Ping timeout: 256 seconds]
morgrimm has quit [Ping timeout: 256 seconds]
hlisp has joined #nixos
Nabzi_ has quit [Quit: -a- IRC for Android 2.1.55]
hlisp has quit [Ping timeout: 246 seconds]
nabzi has joined #nixos
dngray has joined #nixos
hlisp has joined #nixos
<dngray>
I had some questions about NixOS and whether it is right for my usecase. I have a server that is going to be a hypervisor, running LXD/LXC and KVM. It will make extensive use of a ZFS pool
<dngray>
i wanted to run the system in RAM, ie like Alpine Linux's diskless install
<dngray>
I'm wondering whether NixOS is the right distribution for this purpose, i am thinking alpinelinux might be simpler as i don't need to run multiple versions of things etc
<dngray>
I did look at GUIX briefly, but they haven't packaged LXD so gave that a miss
<gchristensen>
nixos isn't really necessarily about multiple versions, but about being able to go back after you've decided you didn't like the update
<dngray>
i also require reproducable builds, which both alpine linux and NixOS do
<dngray>
i like the way Alpine Linux works, for example with the diskless install it boots from USB stick, and just installs the cached packages (into RAM)
nabzi is now known as YourNi
YourNi is now known as nabzi
<dngray>
as it's a hypervisor, i require keeping the system ultra thin, and easy to maintain
<dngray>
one of the things that attracts me to Nix, style system is the idea everything can be comitted asa recipie to a version control repo
<dngray>
gchristensen: that's why i am thinking it might not really be good for this
<dngray>
because as a hypervisor, realistically KVM, LXD/LXC/ZFS aren't updated regularly those would be fairly conservative anyway
butcher has quit [Ping timeout: 246 seconds]
<dngray>
as far as encryption goes, i plan to leave that to ZFS
<dngray>
you can update the system and that will update the cache
<{^_^}>
[nixpkgs] @bhipple pushed commit from @r-ryantm to master « intel-media-sdk: 20.1.0 -> 20.1.1 »: https://git.io/JfZ8p
<gchristensen>
I think NixOS could easily be great for what you're asking of it, but since your requirements of the host are SO thin, it is maybe not worth it
waleee-cl has quit [Quit: Connection closed for inactivity]
marcusr has joined #nixos
<bqv>
That's exactly how nixos works too :)
<dngray>
yeah that's kinda what i am thinking
<{^_^}>
[nixpkgs] @rardiol opened pull request #87050 → Hedgewars server fix haskell → https://git.io/JfZ8h
<bqv>
I would try it and see, but up to you.
<dngray>
its not even really devops or anything like that
<dngray>
what i might do is use alpinelinux, and them run nixos as a guest of some kind
<dngray>
and possibly change over in the future, which should be easy enough
<dngray>
the only disks this machine has is a nvme (for l2arc) and 8x 10TB disks for the zpool
<gchristensen>
nice
<dngray>
also alpinelinux packages by default use musl, so the binaries are even smaller than say glibc based distributions
<dngray>
i don't really want to be compiling from source on my server
<dngray>
so i think i would be hard pressed to get anything leaner than the leanest alpinelinux 😁
<dngray>
and as i said, as its a hypervisor i'm not really "using" it much anyway
<dngray>
everything i do would either be in a container or KVM
<dngray>
so i can rock out with some nixos containers!
<gchristensen>
:D
<dngray>
and i can already think of a few use cases there where nixos would be awesome
<gchristensen>
nice! what are you thinking?
hlisp has quit [Ping timeout: 260 seconds]
<dngray>
I very much appreciate the honesty too guys 🙂
<dngray>
because some people get very fanatical about a distribution and try to make it fit in every hole
<dngray>
well one of the containers runs a bunch of webapps
<dngray>
and it would be nice to be able to manage that with nixos, and fall back to good versions if something breaks
<gchristensen>
nice
<cole-h>
dngray: I dunno, squares fit perfectly fine in round holes.
<dngray>
oh nice, so i guess you know what i am saying then 🙂
<dngray>
but i am thinking if i like nixos, i might start using it instead of archlinux
<gchristensen>
=)
<dngray>
on my workstation/laptop
<hyper_ch>
hmmm, not sure what to do... for a PR I did include the patch file... however reviewers said to use the upstream patch files. However using upstream patch files does not work because of offsets I think....
<lopsided98>
Is is possible to get GTK applications that use wrapGAppsHook to match the system on non-NixOS?
<dngray>
oh i should also mention this server is in my local network
kleisli has quit [Remote host closed the connection]
charukiewicz has joined #nixos
hlisp has quit [Ping timeout: 256 seconds]
<charukiewicz>
I'm trying to bump one of my AWS EC2 instances from nixos-19.09 to nixos-20.03 by updating my nixos channel and running 'nixos-rebuild switch' but I'm getting the following error: sbin/grub-install: error: cannot find a GRUB drive for /dev/xvda. Check your device.map.
kleisli has joined #nixos
<charukiewicz>
Is there anything I can do to fix this problem?
proofofkeags has quit [Remote host closed the connection]
proofofkeags has joined #nixos
<hyper_ch>
(1) I would not use switch on major updates but boot (2) what's your configuration.nix / hardware-configuration.nix and output of ls -al /dev/disk/by-id
proofofkeags has quit [Ping timeout: 256 seconds]
marcusr has quit [Remote host closed the connection]
<hyper_ch>
charukiewicz: boot.loader.grub.device = if (cfg.hvm && !cfg.efi) then "/dev/xvda" else "nodev";
<hyper_ch>
my guess would be to replace /dev/xvda with /dev/nvme0n1 but that's just my guess
<charukiewicz>
so just a question: in order to find that file I just ran 'locate amazon-image.nix' and found one of the ones in nix-store. Where would the location <nixpkgs/nixos/modules/virtualisation/amazon-image.nix> actually resolve to on the machine?
<hyper_ch>
no idea
<hyper_ch>
you set that up
<hyper_ch>
that's amazon stuff
proofofkeags has joined #nixos
<charukiewicz>
That was one of the two default lines in /etc/configuration.nix when I deployed the AMI, I avoided touching it, just added my own stuff
<charukiewicz>
that is, /etc/nixos/configuration.nix
<Ashy>
hmm, i've noticed that all my nixos machines seem to wait for NetworkManager during boot recently
<Ashy>
slows down booting quite a lot
iqubic has joined #nixos
<{^_^}>
[nixpkgs] @peterhoeg opened pull request #87063 → mu: allow configuring the BatchSize → https://git.io/JfZ0X
zupo has joined #nixos
<bqv>
> Allows to `from nixpkgs import` stuff in interactive Python sessions
<{^_^}>
error: syntax error, unexpected $undefined, expecting ')', at (string):309:11
<bqv>
woah that's awesome
<fps>
hmm, i setup virtualbox as described in the wiki https://nixos.wiki/wiki/Virtualbox. but there's no virtualbox executable after rebuilding the system
<{^_^}>
[nixpkgs] @kampka opened pull request #87068 → kops: 1.15.2 -> 1.15.3 and 1.16.1 -> 1.16.2 → https://git.io/JfZue
<DamienCassou>
there is a proprietary app with a linux version that I would like to use on nixos. It downloads and updates .so files (e.g., libusb4java.so). How would you recommend to package it?
rauno has joined #nixos
mallox has quit [Quit: WeeChat 2.8]
charukiewicz has quit [Quit: leaving]
choward has joined #nixos
<fps>
DamienCassou: isn't there a LSB wrapper in nixos? let's see
<fps>
DamienCassou: or take a look at one of the many unfree packages that do patchelf trickery, etc..
timmw has joined #nixos
<DamienCassou>
the problem is that the *.so files are downloaded when the application executes
<DamienCassou>
and the application might update them a few days later
sigmundv__ has joined #nixos
choward has quit [Ping timeout: 272 seconds]
__monty__ has joined #nixos
choward has joined #nixos
thc202 has joined #nixos
<monsieurp>
DamienCassou: you're going to have to work against the app internal package manager if you want to package it properly for nix and that's going to be a PITA
<DamienCassou>
I also thought about putting all of it in a (docker?) container and let it do its stuff
<Mic92>
Izorkin: if you change the test to just test the php version that is enabled by default and instead of rebuilding unit with all php versions, I am merging this.
<turion>
evax: Not sure what channels have to do with it
<bdju>
ugh firefox crashes when I paste some lines of text from emacs as well. not just the file picker. I wonder if this is wayland-specific.
nabzi has joined #nixos
<evax>
turion: right, sorry, I misread
<turion>
No problem :)
mxco86 has quit [Ping timeout: 256 seconds]
<turion>
I think I'm pinning nixpkgs to a certain commit, and a commit that has the version 1.10.3. I don't understand why it evaluates to 1.10.2.
<pbogdan>
turion: have you perhaps changed the rev in the url without updating the hash?
<pbogdan>
I'm guessing you could be getting a different archive if you already fetched something with that checksum and only changed the url
yourfate has joined #nixos
<turion>
pbogdan: Arrrrgh yes that must be it!
<turion>
I was wondering already why it wouldn't download the tarball
<nschoe>
Hi everyone, what should I install / wha option should I enable in my configuration.nix to be able to ping / reach other devies on same network with hostname?
<nschoe>
I haven't paid much attention to this since I switched to NixOS, but now I need it: I'm on my mobile phone, connection sharing. There is another device to which I am trying to ssh. I kno wits hostname on the network (the mobile gives me that) but 'ssh' or 'ping' it doesn't work (system error).
<nschoe>
Is that related to mDNS? nss? I've looked a bit and found these, I'm not much in the know-how.
<b42>
nschoe: perhaps services.avahi.enable and services.avahi.nssmdns? just guessing
<nschoe>
hum so that might be it.. I'll try that. Thank you
chagra_ has joined #nixos
Maxdamantus has quit [Read error: Connection reset by peer]
Alexey63 has joined #nixos
<{^_^}>
[mobile-nixos] @thefloweringash opened pull request #141 → initrd-usb: provide modules required for gadgets in initrd → https://git.io/JfZwR
Maxdamantus has joined #nixos
o1lo01ol1o has joined #nixos
<turion>
How do I use nix with elixir/mix properly? For example, assume I have a mix project packaged. How can I open a shell with that package and run a mix task?
<turion>
Or create a derivation that contains the output of running a particular mix task?
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
woffs has quit [Quit: Gateway shutdown]
knupfer has quit [Remote host closed the connection]
<Mic92>
Izorkin: could you also link those tests in the mariadb package?
nschoe_ has quit [Client Quit]
zupo has quit [Ping timeout: 240 seconds]
nschoe has joined #nixos
lsix has quit [Ping timeout: 260 seconds]
disasm has joined #nixos
fendor__ has joined #nixos
<{^_^}>
[hydra] @edolstra pushed to remove-releases « Remove the "releases" feature »: https://git.io/JfZrD
<{^_^}>
[hydra] @edolstra opened pull request #752 → Remove the "releases" feature → https://git.io/JfZrS
vidbina has quit [Ping timeout: 260 seconds]
<betaboon>
any wayland/sway users having problems with gtk3-font-rendering as well? for me it seems like antialiasing doesnt work: https://i.imgur.com/50UtaDa.png
iyzsong has joined #nixos
o1lo01ol1o has quit [Remote host closed the connection]
<Alexey63>
I have in config https://pastebin.com/2gDhVAS9 php-fpm configured with user and group `nginx`, but it creates `/run/phpfpm/rainloop.sock` with owner root:root, so nginx cannot connect (writes permission denied in logs and return 502). How can it be fixed?
<b1000101>
c00w: are you here? I've tried your blockbook patch but the build is failing with go/src/blockbook/bchain/coins/dcr/decredrpc.go:367:3: cannot use 0 (type untyped int) as type string in field value
<{^_^}>
[nixpkgs] @FRidh pushed commit from @marcus7070 to master « python3Packages.spyder_3: remove broken application »: https://git.io/JfZiy
hexa- has quit [Quit: WeeChat 2.7.1]
lukash_|away is now known as lukash_
hexa- has joined #nixos
o1lo01ol1o has quit [Remote host closed the connection]
<unclechu>
hey. can i somehow patch file list of config dir of `nixpkgs.neovim` or `nixpkgs.vim_configurable`? for instance for neovim apart form `init.vim` there's also `ginit.vim` file which is an additional config for GUIs.
<jluttine>
DamienCassou: no. it asks for my user password for authentication
<DamienCassou>
I don't know then. I confirm it shouldn't ask for any password in the normal case
zupo has joined #nixos
terrorjack has joined #nixos
o1lo01ol1o has quit [Remote host closed the connection]
srid has joined #nixos
o1lo01ol1o has joined #nixos
fusion809 has quit [Remote host closed the connection]
<emilsp>
hrm, it just so happens that the stable kernel version doesn't support my GPU all too well, and the unstable (5.6) isn't much better. The sweet sport seemed to be 5.5.17, what could I do to ensure that I'm using that golden kernel version whilst still upgrading all other packages as normal? Should I import a specific commit of nixpkgs with the kernel I want?
AluisioASG has quit [Read error: Connection reset by peer]
<emily>
emilsp: importing the kernel package from a specific nixpkgs version sounds like your best bet, yes
<emily>
you could also vendor the linux-X.Y.nix file from that version in your config
<emily>
srid: `@srid:matrix.org` :p
pingiun has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<jluttine>
udisksctl asks authentication with this kind of header: ==== AUTHENTICATING FOR org.freedesktop.udisks2.filesystem-mount-other-seat ====
AluisioASG has joined #nixos
virus_dave has joined #nixos
<Alexey63>
Seems to be trivial, but cannot find: how to ensure directory exists (create if missing)?
jakobrs has joined #nixos
pingiun has joined #nixos
<jakobrs>
What's the difference between fileSystemOpts and coreFileSystemOpts?
<emily>
jluttine: it seems like your polkit configuration may be wrong
opticnerve has joined #nixos
<emily>
Alexey63: in a derivation build phase? `mkdir -p path/to/dir`
<jluttine>
emily: not sure if relevant but i'm trying to mount it via SSH while there's another user logged in to the computer in a normal desktop environment. but the disk wasn't mounted when i ran that command.
<{^_^}>
[nixpkgs] @7c6f434c merged pull request #86837 → quicklisp-to-nix: fix missing '&&' after test command → https://git.io/JfGCt
lorimer has quit [Read error: Connection reset by peer]
<emily>
jluttine: I think that is intended, then; only users who have "physical access" to the machiine are granted rights to mount stuff by default, iirc
<emily>
so you'd need a local session, unless you tweak your polkit config to allow it
lorimer has joined #nixos
<emily>
Alexey63: systemd.tmpfiles or making it the StateDirectory of a systemd.service are probably what you want to look at
<jluttine>
emily: ok. well i can mount it, i just need to provide my password to authenticate. so not a problem. i just thought it should mount without asking it
<emily>
I think that's probably because your user is in wheel or whatever, so it lets you do anything as long as you elevate to administrator with your password
<jluttine>
emily: maybe. that's what i thought too first. but in those cases it asks whether to authenticate as root or jluttine. now it only says that authenticate as jluttine, give password. and this documentation seems to say that indeed authentication is needed in this case: http://storaged.org/doc/udisks2-api/latest/udisks-polkit-actions.html
<jluttine>
emily: so, if i understand correctly, it's not that it needs root permissions but just that it wants to authenticate that i'm myself. a bit weird to me though.
<kayg04>
when I try to do: `borg create` manually, the backup seems to work fine but when it's triggered through a systemd service, I get a: Errno 30: Read-only filesystem.
oborot has quit [Ping timeout: 272 seconds]
<kayg04>
azazel, have you seen this happen before?
sridIrcCloud has joined #nixos
<azazel>
kayg04: that's because the borgbackup module configures the service to see all the FS as readonly, If I remember corretly it expect for the destination to be remote
<kayg04>
but the repos can be local, no?
<{^_^}>
[nix] @gilligan opened pull request #3571 → WIP: Add unit tests → https://git.io/JfZMj
thc202 has quit [Ping timeout: 256 seconds]
<kayg04>
how do I make it backup to a local repo then?
<azazel>
kayg04: you have to give to it a repo url that's non local... i.e manually it should work from the same host (where the sshd is running) with borgbackup create borg@localhost:.::your_archive_name
pingiun has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<ircuser5678>
hi. somehow google is dumb or I am. It's probably me... I'd like to change the configuration.nix file before I invoke the installation and I want it to create some folders automatically. E.g. /mnt/share1, /mnt/share2. So that I don't need to do that manually after I've booted up the newly installed system. Is this possible?
<{^_^}>
[nixpkgs] @FRidh pushed to master « python3Packages.mortgage: init at 1.0.5 »: https://git.io/JfZyy
<{^_^}>
[nix] @Mic92 opened pull request #3572 → nix/search: no error for empty search results if json is enabled → https://git.io/JfZy7
<kayg04>
azazel, btw what's the reasoning behind making local files read-only to borg?
<azazel>
keyg04: ensure that for whatever reason your backup process doesn't modify the fs it's expected to snapshot
<kayg04>
I see, I see
proofofkeags has quit [Remote host closed the connection]
<kayg04>
but borg itself doesn't do that, right?
proofofkeags has joined #nixos
rardiol has joined #nixos
<azazel>
kayg04: sorry, wrong nick... it was for you. Anyway I admit it's a bit... hairy... anyway you can disable it by directly configuring your backup job
<kayg04>
I mean it makes sense but it's kinda accounting for the user's stupidity
<azazel>
kayg04: have fun with "borgfs" to restore the files
<kayg04>
azazel, what's borgfs?
<azazel>
kayg04: the fuse fs interface to the borg repos...
<kayg04>
azazel, the manpage says: local or remote repo so it might be misleading to new users
<kayg04>
oh lol, my bad
<atemu12[m]>
ircuser5678: I don't know if there's a proper way of doing that but a qnd hack would be to declare a systemd service which runs a script that creates them for you on boot
<ircuser5678>
atemu12[m] That sounds... Rather complicated for such a simple thing :)
<atemu12[m]>
ircuser5678: what do you need those dirs for? Mounting things?
<ircuser5678>
atemu12[m]: Yeah, e.g. for predefined samba shares
<atemu12[m]>
ircuser5678: Could those things be mounted by Nix?
<{^_^}>
[nix] @edolstra pushed to flakes « Remove the nixpkgs.<attr> compatibility hack »: https://git.io/JfZHk
erasmas has joined #nixos
<Alexey63>
`services.opendkim.socket` is created without group write access, so how can it be used by another user, postfix? It seems the only way is to run opendkim as postfix:postfix (and manually change owner of private keys).
<betaboon>
Alexey63: add postfix to opendkim group ?
<Alexey63>
betaboon: But socket will not be writeable anyway, will it?
<{^_^}>
#82379 (by Infinisil, 7 weeks ago, open): nixos/opendkim: Deprecate configFile in favor of settings option
<turion>
How do I build a derivation step by step in nix shell if it's defined somewhere in my fork of nixpkgs
Alexey8 has joined #nixos
Alexey63 has quit [Remote host closed the connection]
<turion>
I have a default.nix there, but it needs certain arguments, because it would usually be called as callPackage (import ./pkgs/foo/bar) {}
<cole-h>
You can probably build it with `nix-build -E '(import <nixpkgs> {}).callPackage ./path/stuff/here {}'
<clever>
,callPackage
<{^_^}>
If you're updating a package file in nixpkgs that starts with something like `{ stdenv, cmake }:`, use `nix-build -A the-package-attribute` in the nixpkgs root with the corresponding package attribute to build it. The mapping from package attributes to package files is in pkgs/top-level/all-packages.nix
<clever>
turion: but use nix-shell instead of nix-build
<clever>
weird, callPackage is missing from the ,callPackage help text!
<cole-h>
lol
<turion>
clever++
<{^_^}>
clever's karma got increased to 406
<turion>
clever: I can't quite get it to work... I thought I can open a nix-shell and then I say configurePhase and it does the configurePhase, but here it says "no configure script"
<clever>
turion: did you override configurePhase?
<{^_^}>
[nixpkgs] @markuskowa merged pull request #85839 → dnnl: rename to oneDNN to track upstream project rename → https://git.io/JfIsG
<cole-h>
clever++ That's cool. I think that should even be a ,command here, tbh...
<{^_^}>
clever's karma got increased to 407
<turion>
At least I can't see it doing anything, it doesn't even say "... doing nothing"
<clever>
turion: what does `echo "configurePhase"` say?
<infinisil>
clever: Using callPackage from a nix-build -E is almost always bad practice
<turion>
$ echo "configurePhase"
<turion>
configurePhase
<infinisil>
Which is why I changed it to reflect what most people should want to do instead
<clever>
turion: with the $
<clever>
oops, managed to miss that
<turion>
clever: Ah yes, that says the right script
dramaturg has quit [Quit: dramaturg]
<cole-h>
infinisil: Why is it bad practice? If I have a `{ cmake, stdenv }:` nix file out of tree, that's my only choice, isn't it?
alp has quit [Ping timeout: 265 seconds]
fendor__ has quit [Ping timeout: 260 seconds]
<clever>
cole-h: i tend to add a default.nix that loads program.nix in that case
<infinisil>
^^
<clever>
cole-h: and i try to fetchTarball nixpkgs, so it is reproducable
<infinisil>
Otherwise people have to guess where the arguments should come from
<cole-h>
And how does it load default.nix? Just `import` or something?
dramaturg has joined #nixos
<turion>
infinisil, but aren't thousands of files in nixpkgs in that wrong way then?
<clever>
cole-h: default.nix should just contain: with import (builtins.fetchTarball ...) {}; callPackage ./app.nix {}
<codez>
Just spent 3hrs struggling to create a bootable windows usb in ubuntu, must of changed bios settings 100000 times, checked filesystem types, different usb ports, you name it..
<clever>
turion: what does the buildPhase actually do? what is the 1st line of it?
<turion>
clever: runHook preBuild
<turion>
clever: So should I do phases="configurePhase buildPhase installPhase" genericBuild?
<clever>
turion: installPhase will never work right in nix-shell, since you lack access to /nix/store
<clever>
turion: the other 2 should work fine though
<turion>
Ah right
rogue_koder has quit [Remote host closed the connection]
<turion>
Ok, that gives funny errors :) I'm a step further I guess
rogue_koder has joined #nixos
hlisp_ has quit [Remote host closed the connection]
<Alexey8>
infinisil: thank, UMask helps.
<aanderse>
dhess: still on for 6 minutes from now?
<aanderse>
oh... looks like i got that wrong, 1 hour and 6 minutes from now... ignore me :)
m0rphism has quit [Quit: WeeChat 2.7.1]
<emily>
turion: no, there is no phase-level caching (not sure what clever means)
<emily>
configurePhase will be run on each build
justanotheruser has joined #nixos
<cole-h>
(I certainly wish there was)
<clever>
but if your reusing the build dir, with the state from before, you dont have to re-run it
<cole-h>
(Not sure how it'd work, but I wish)
<clever>
nix-build wipes that dir every time
<emily>
oh, in a nix-shell? there's no guarding of configurePhase to only run once afaik
<emily>
right
morgrimm has quit [Ping timeout: 240 seconds]
<turion>
emily: Ok, thanks. Why isn't each phase its own derivation?
<turion>
emily++
<{^_^}>
emily's karma got increased to 20
<emily>
shrug, historical reasons + overhead I guess
<emily>
I'd prefer it wasn't the case, especially for installPhase
<emily>
go too far trying to fix it and you end up wanting to do ccache-style incremental compilation which is a bigger task though
<emily>
the recursive Nix stuf is relevant to that
<clever>
turion: it would mess with permissions and also require the entire build dir (potentially several gig) to be put into the store and kept long-term
<clever>
turion: it also moves that build dir, breaking more things
<tobi_>
hello, I'm trying to setup a samba server which authenticates against an ldap server (FreeIPA to be precise). In order to do this I've set "services.samba.enable = true;" and "services.samba.extraConfig = ''...passdb backend = ldapsam:ldap://<my-freeipa-instance>...'';. Additionally I've added openldap to environment.systemPackages. When I try to run this config however, nixops can't run the service because
<tobi_>
winbindd complains that "no builtin nor plugin backend found for ldapsam found". I guess this is because samba was built without ldap support. Furthermore, I can see that in pkgs/servers/samba/4.x.nix, enableLDAP is a method parameter, alas I don't understand how I can pass it.
sridIrcCloud has quit [Quit: Connection closed for inactivity]
<tobi_>
does anybody know what I have to do to enable ldap support for samba?
<betawaffle>
huh, why is nixos.netcat == libressl?
drakonis has joined #nixos
<fps>
tobi_: maybe something like pkgs.servers.samba.override { enableLDAP = true; }; in the configuration.nix
<tobi_>
ok, I _think_ I have figured it out: I've defined an overlay in which I did "samba = super.samba.overrideAttrs (oldAttrs: rec { enableLDAP = true; });", now I can see that it is compiling samba from source
<scott>
is that true for all versions? i'm not sure how to find things in the hydra interface (my current nixos-unstable has julia_07, julia_10, julia_11, and julia_13)
<zeta_0>
when using lorri, do i have to have the mkShell function in my shell.nix file, or can i go without it? my ihaskell `let in` expression is not working inside of mkShell: https://dpaste.org/nneL
<cole-h>
Only reason I did the nixos-homepage hack was because you said "current stable channel"
<cole-h>
`curl nixos-unstable/.version` will give you the next version, but there's no way to get the current version
knupfer has joined #nixos
<zeta_0>
this ihaskell let in expression works fine in home.nix, but when i try to place it in a shell.nix file to work with lorri, it does not work?
ixxie has joined #nixos
<pie_>
cole-h: oh right doh
<pie_>
zeta_0: define does not work
magnetophon has quit [Ping timeout: 264 seconds]
rogue_koder has quit [Ping timeout: 260 seconds]
<pie_>
cole-h: ah well whatever i guess i just wont do this
orivej has joined #nixos
<zeta_0>
pie_: emacs throws this error: https://dpaste.org/Dqbf and the ghc binary is not being recognized either
o1lo01ol1o has quit [Remote host closed the connection]
o1lo01ol1o has joined #nixos
<zeta_0>
when i tried to set some environment variables in shellHook yesterday errors were thrown, but i'll try one more time to double check
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
EdLin has joined #nixos
zupo has joined #nixos
noudle has joined #nixos
oborot has quit [Ping timeout: 260 seconds]
cr4y1_ has quit [Ping timeout: 272 seconds]
<zeta_0>
this ihaskell script is supposed to install ghc883, but when i try to set up these environment variables, when i go into a nix-shell, it tells me ghc not found, how strange: https://dpaste.org/M1wE
evanjs has quit [Read error: Connection reset by peer]
proofofkeags has quit [Remote host closed the connection]
<zeta_0>
pie_: agreed, but i don't know why when i go into a nix-shell with this shell.nix no errors are being thrown, also, i thought the `compiler = ghc883;` in that ihaskell expression, ihaskell will install it's own ghc, that's why i had that lib.prio stuff setup in my home.nix, so i'm guessing that i need to install ghc and python/jupyter separately in that ihaskell let in expression
<zeta_0>
pie_: thanks, i'll try nix-shell --pure first, and if that does not work, i'll try installing ghc and jupyter in that ihaskell let in expression of the shell.nix file to see if that fixes it, oh and that github issue looks very similar to my issue, so i'll try those steps out as well
<{^_^}>
[nixpkgs] @jonringer pushed commit from @r-ryantm to master « mednafen: 1.24.1 -> 1.24.2 »: https://git.io/JfZp1
AmandaC has joined #nixos
gagbo has quit [Ping timeout: 260 seconds]
EdLin has quit [Quit: Going offline, see ya! (www.adiirc.com)]
iqubic has joined #nixos
MawKKe[m] has joined #nixos
<MawKKe[m]>
Hello
chagra_ has joined #nixos
alp has quit [Ping timeout: 246 seconds]
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<MawKKe[m]>
On the NixOS download page, the server provided .iso filenames are generic, but the respective SHA256 files contains some build-dependent names
<clever>
WinchellsM: `grub.device = "/dev/sda";` is only used for legacy booting
linarcx has quit [Ping timeout: 264 seconds]
<clever>
and if efibootmgr doesnt work, there is a chance you booted without efi
rogue_koder has quit [Remote host closed the connection]
<WinchellsM>
So all of my efi declarations above aren't needed?
rogue_koder has joined #nixos
<DamienCassou>
hi. On NixOS, I have both `programs.ssh.startAgent = true` and `programs.gnupg.agent.enableSSHSupport = true` (nixos-option confirms) but `/nixos/modules/programs/gnupg.nix` forbids this with an assertion. What's wrong?
<clever>
WinchellsM: depends on if you want to boot with efi or legacy
<clever>
DamienCassou: you can only enable one ssh agent
<dmj`>
Is there a derivation for Facebook's docusaurus ?
<{^_^}>
[nixpkgs] @hedning pushed commit from @r-ryantm to master « gnome3.gucharmap: 13.0.0 -> 13.0.2 (#86885) »: https://git.io/JfZjC
<DamienCassou>
clever: I know this and I want gnupg as my ssh agent. But the SSH_AUTH_SOCK environment variable is not populated so I try to debug. I had hope to crash nixos-rebuild by setting both variables to true but it accepted it. So now I don't understand anything anymore
marsh has quit [Remote host closed the connection]
hoijui has quit [Quit: Leaving]
marsh has joined #nixos
<clever>
DamienCassou: ive been just setting SSH_AUTH_SOCK manually
<{^_^}>
[nixpkgs] @hedning pushed commit from @r-ryantm to staging « gnome3.adwaita-icon-theme: 3.36.0 -> 3.36.1 (#86766) »: https://git.io/JfZjR
<WinchellsM>
clever: Is there a way to determine which way I booted?
<DamienCassou>
clever: gnupg.nix sets SSH_AUTH_SOCK automatically. I would prefer not repeating that in my configuration. Moreover, I would like to understand why an nixos-rebuild can build a configuration with a failing assertion
<WinchellsM>
I'm in NixOS doing this now, so if `efibootmgr -v` shows up as "EFI variables are not supported on this system", does that mean I booted in legacy mode?
<clever>
DamienCassou: ive also found that the gnupg module in nixos is much worse, it re-binds gpg to the "active" tty every time i ssh into the machine
<clever>
DamienCassou: it then asks for the pw prompt on the ssh session, even if i'm back at the local keyboard
<clever>
DamienCassou: so pw prompts just silently "hang" because they opened in an ssh window on another box
marsh has quit [Ping timeout: 272 seconds]
<WinchellsM>
clever: Ok. I'm going to live dangerously and try removing that line :)
<pR3>
I have a problem that sometimes i try to get an env-variable with builtins.getEnv in my configuration.nix it does not work. It is pretty weird: builtins.getEnv "HOSTNAME" does work, but builtins.getEnv "NIXOS" does not work. I did export NIXOS beforehands of course. Does anybody have an idea?
<prusnak>
yorick: i have nix 2.3.4 and this works for me as expected - builtins.fromJSON "00023"
<cole-h>
Yeah, it seems like a regression in nixFlakes
<{^_^}>
[nixpkgs] @jonringer opened pull request #87134 → openblas: also export unversioned libraries for linux → https://git.io/Jfnvq
<esclear>
I thought that pulseaudio is a user service now, however I can't start it anymore:
<esclear>
~> systemctl start --user pulseaudio
<esclear>
Failed to start pulseaudio.service: Unit pulseaudio.socket not found.
<esclear>
However, there *is* a .socket unit next to the .service file. Any hints? I'm running 20.3, although this has been an issue since before the upgrade.
<clever>
niksnut: having some issues with rust cross-compiling, do you know about the area where it gets x86_64-pc-windows-gnu and x86_64-w64-windows-gnu mixed up?
<{^_^}>
[nixops] @grahamc pushed to lint-docs « README: add a bit of a splainer on docs »: https://git.io/JfnJj
nabzi has joined #nixos
nabzi has quit [Max SendQ exceeded]
<infinisil>
I think the answer to this is no, but:
<infinisil>
Is there a way to have a derivation's passthru value be queriable from its drv path, without it actually influencing the drv?
cosimone has quit [Read error: Connection reset by peer]
<clever>
infinisil: no real way, because the passthru can contain anything that cant be serialized, and the drv path must always be a hash of its own contents
<infinisil>
Hm yeah, and if it's not a passthru, but an env var of the build, this causes a rebuild
cosimone_ has quit [Ping timeout: 244 seconds]
<{^_^}>
[nixops] @grahamc pushed to lint-docs « live-docs.py: automatically build docs once on startup »: https://git.io/JfnUn
dsg has quit [Remote host closed the connection]
dsg has joined #nixos
Nilgeist has quit [Quit: ZNC 1.7.2+deb3 - https://znc.in]
spacefrogg has quit [Remote host closed the connection]
MarcWeber has quit [Remote host closed the connection]
<clever>
cab404[m]: what is line 89 of /nix/store/rfk5j99kh4ny0jfi0m88yapv3pbw19nr-nixos-20.03.1619.ab3adfe1c76/nixos/nixos/lib/make-options-doc/default.nix ?
<hr[m]>
If I wanted to map a `types.attrsOf types.path` to source each path, how would I do it? I'm looking at using `lib.attrsets.mapAttrs` but that can only map the value of each attr. Given an attrset of `{"bash/source.bash" = ./file-to-source.sh }` I'd like to map it to `xdg.configFile."bash/source.bash".source = ./file-t-source.sh`
<kuznero>
I am wondering if it is possible to have symlink to current dynamic linker in place instead of patchelf'ing custom binaries? There are a lot of tools and processes that freely download binaries and execute them. E.g. prost in Rust (there is a PR, but that's it).
<{^_^}>
[nixpkgs] @snicket2100 opened pull request #87145 → dnscrypt-proxy2: service restart on failure → https://git.io/JfnI8
<kuznero>
It just sounds impossible to stop everybody from ad-hoc downloading stuff and using... My dev process is affected by this more and more...
<infinisil>
kuznero: You can of course still use the PR in your local system
<kuznero>
In prost e.g. they agreed to take PROTOC from env var. But that is a workaround really
<infinisil>
kuznero: That's not a workaround, Nix pushes these projects to fix their assumptions!
rardiol has joined #nixos
<infinisil>
And that's one of the main arguments against using something like above PR
<infinisil>
Namely that if everybody then used it, nobody would bother making binaries pure anymore
<simpson>
kuznero: I am not at all a Rust person, but https://github.com/danburkert/prost this library? Should *not* download binaries without letting somebody know. I would expect a PB compiler to do only maybe two things, and neither of those things involve fetching and running code from the Internets.
<kuznero>
That's happening nevertheless unfortunately. There is a PR for prost here that allows it otherwise through env var: https://github.com/gabrielrussoc/prost
<infinisil>
kuznero: I don't see how Nix would have problems with the env var at build time
ambroisie has joined #nixos
numkem has joined #nixos
julm has quit [Quit: Lost terminal]
<Cadey>
did cache.nixos.org just enable ipv6?
<numkem>
I'm getting just now a weird problem with go. When I do a `go run` I end up with an error like: `fork/exec /tmp/go-build861032060/b001/exe/cert: no such file or directory`
<kuznero>
infinisil, you are right, it's just getting difficult to replace prost under several layers of subdependencies, e.g. tonic and tonic-build while generating gRPC code.
julm has joined #nixos
zxcvluy has joined #nixos
<infinisil>
kuznero: I guess you meant the elf interpreter (like /nix/store/*-glibc-2.30/lib/ld-linux-x86-64.so.2) for binaries in your original question?
hlisp has joined #nixos
<kuznero>
yes
<infinisil>
Yeah so for that you can use the PR I linked, but it's not officially supported and I wouldn't recommend it
<infinisil>
Making stuff work under Nix sometimes takes a bit more work, but I think this usually pays out in the long run
chagra has quit [Ping timeout: 256 seconds]
proofofkeags has quit [Remote host closed the connection]
<kuznero>
I understand the principle and agree with it completely. Then I would like to ask this: is it possible to somehow catch the moment when binary is about to be executed (outside of /nix/store) and patchelf interpreter on the fly?
chagra has joined #nixos
<simpson>
There's binfmt_misc, and IIRC somebody wired it up somewhat, but I'm not sure how to make it do this.
<kuznero>
I mean patchelf if it is not already pointing to the right location, @infinisil
<simpson>
(To be totally blunt, it is really surprising to me that the Rust toolchain can't be hacked apart to behave properly with Nix. It can't be worse than Go or C.)
<kuznero>
simpson, it is no different when it comes to IDEs and toolchain. E.g. coc.nvim with coc-rust-analyzer extension downloads rust-analyzer that needs to be patchelfed. That is then simply a manual process.
EdLin has joined #nixos
<cab404[m]>
anyone's interested in nix ide with ability to actually debug functions and see how expressions got evaluated? (and is there one already/)
<cab404[m]>
* anyone's interested in nix ide with ability to actually debug functions and see how expressions got evaluated? (and is there one already?)
<kuznero>
infinisil, did it sound like a completely crazy idea to patchelf --set-interpreter on-the-fly for binaries that are getting executed outside of /nix/store?
<cab404[m]>
nope, I mean like run thing which uses nixpkgs on some nix-trace-build, which would work like gdb
<simpson>
kuznero: Yikes, but *why*? I'd hope that that all can be built within Nix directly. I wonder to what degree this is a failure of the text editor package managers (especially VSCode's walled garden) vs. failure of packaging.
<{^_^}>
[nixpkgs] @snicket2100 opened pull request #87147 → ums: added snicket2100 as maintainer → https://git.io/JfnLT
proofofkeags has joined #nixos
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<kuznero>
simpson, this is just the design of certain extensions that check for updates and download them when needed. Those updates are coming directly from github or other places, but definitely not out of Nix packager.
<simpson>
cab404[m]: Have you seen builtins.trace? A full debugger for lazy languages is quite difficult, and unlikely to happen for Nix without a serious plan and support.
<simpson>
kuznero: Sounds horribly broken to me; I would rather not let various packages update themselves without letting Nix know first.
<infinisil>
cab404[m]: Yes, very interested! I've thought a bit about how it would work before too (just with some notes)
<kuznero>
simpson, this is what this part of the industry looks like.
jkachmar has quit [Quit: Connection closed for inactivity]
<numkem>
I've dug a little more about my `go run` problem and it seems like it's 1.14.2 for some reason that causes the issue. I've done it as an override with the same values as the current PR to upgrade Go from 1.14.1 to 1.14.2. It used to work just fine until this morning when I did a full upgrade after the latest channel bump
<infinisil>
kuznero: With Nix, you either commit to it fully, or you don't. Anything inbetween is nasty
griff__ has quit [Quit: griff__]
<kuznero>
infinisil, I understand, but proposing to patch in-flight so to say
<energizer>
for a while i was using `nix-env` as just another apt, that worked pretty well
<simpson>
kuznero: Yes, I am aware of the horribly broken state of the industry. I grew up in a Microsoft monoculture, FWIW, so I'm no stranger to this sort of terrible nest of anti-patterns.
<simpson>
I'm not saying that it's *your* fault that things are this way. No one person could be blamed.
<kuznero>
I am simply trying to find a way to relatively painlessly work with existing tools
<cab404[m]>
<simpson "cab404: Have you seen builtins.t"> it's not that difficult, something like firefox debugging tools on js objects would be already worlds better than ":p" in nix repl
<Ashy>
i'm trying to merge some folders into $out/foo/bar during a mkDerivation installPhase script but `cp -r ${package}/foo/bar $out/foo/bar` errors out with permission errors when that $out/foo/bar directory path exists, so only the first set of files end up there, how should i do this?
<infinisil>
kuznero: But no, I don't think there's a way to automatically patch things on-the-fly, mainly because the elf interpreter is the entrypoint for a binary, without a correct one, it can't even read the file so to say
<Ashy>
is there a section of the nixpkgs manual that covers the $out directory and permissions behaviour?
<infinisil>
kuznero: Again, https://github.com/NixOS/nixpkgs/pull/69057 should work though, and while I wouldn't recommend it, it's the best way to have impure binaries work on NixOS relatively painfree
<simpson>
cab404[m]: It took what, two decades before Haskell got a line-by-line debugger? Lazy languages have some serious issues doing line-by-line debugging. Although it sounds more like you want graphical tools?
<bpye>
Does it make sense to run stable NixOS with an unstable user channel? And if I added a user channel would I add NixOS or nixpkgs?
<cab404[m]>
i want a bad lsp server
cr4y1 has quit [Remote host closed the connection]
<simpson>
Ashy: Like, you want an environment built from multiple packages, or you want to copy some stuff from $src to $out? It seems curious that you only have $out/foo/bar existing conditionally.
<cab404[m]>
which would try running each thing in nix repl for a second or two, and showing type it got. in a worst and simplest case
<Ashy>
simpson: an environment built from multiple packages yes, this is for a language interpreter to include multiple addon packages from different github repos
<kuznero>
I do not want that at all, infinisil. Essentially in NixOS we have to set right interpreter before executing anything. For properly packaged software it is easy and shoudl be done once when packaging more or less. But for binaries distributed in ad-hoc manner it does not work. One other approach to me perhaps would be to try to persuade upstream devs to build fully statically linked binaries then?
<infinisil>
kuznero: That doesn't help with the elf interpreter problem
<infinisil>
Packaging things with Nix is the only "correct" way really
<Ashy>
simpson: as is that works, but if i add in the package general_dirtrees too the `cp -r ...` fails because $out/share/j/addons/general already exists
<Ashy>
simpson: the intention is that i end up with: $out/share/j/addons/general/dirutils/... and $out/share/j/addons/general/dirtrees/...
<simpson>
Ashy: You may want `linkFarm`, a trivial builder which tries to do some of that logic for you. But I don't think that there's a general resolution for conflicts like that; one of the paths is going to get shadowed or ignored.
<Ashy>
one works but it seems that because the second one also attemps to add files under the "general" directory its' erroring out
<kuznero>
infinisil, it actuall works: ./testssl: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), statically linked, BuildID[sha1]=bce12538bcf95db0d425b0c2e66cdb8ccdfddb77, with debug_info, not stripped
oborot has joined #nixos
<kuznero>
when I run `file ./testssl`
<Ashy>
they don't have any files that clash, it's two separate directories that should nest cleanly under "general/"
<charukiewicz>
I'm on a NixOS server trying to build a Haskell project, with a local-to-the-project *.nix file that defines which nix channel I should be using. For the development environment, I'm using nix-shell. I can enter the nix-shell as root, but not as a non-root user, even though the non-root user has sudo privileges. Is there any reason why this would be happening?
<charukiewicz>
This is the error I get as the non-root user when entering nix-shell in the project: error: build of '/nix/store/...-glibc-locales-2.30.drv', '/nix/store/...-stdenv-linux.drv', '/nix/store/...-bash-4.4-p23.drv', '/nix/store/...-ghc-8.6.5-with-packages.drv', '/nix/store/...-hpack-0.33.0.drv' failed
shibboleth has joined #nixos
<charukiewicz>
Do any of those packages require root access to build?
<charukiewicz>
Running `sudo nix-shell` seems to work by for the non-root user, but a few weeks ago that wasn't necessary, and I don't understand why it is now.
cole-h has quit [Quit: Goodbye]
zeta_0 has quit [Quit: rcirc on GNU Emacs 26.3]
Reszenic has quit [Quit: Leaving]
<infinisil>
charukiewicz: Sounds like you have a single-user installation of Nix, i.e. no nix-daemon is running, and only the user Nix was installed with (presumably root) can do stuff
<infinisil>
Oh wait, that's NixOS
cole-h has joined #nixos
<infinisil>
No idea then, but posting the full logs might be helpful
philr__ has joined #nixos
mir100 has quit [Ping timeout: 264 seconds]
justanotheruser has quit [Ping timeout: 256 seconds]
felixfoertsch has quit [Ping timeout: 260 seconds]
<bpye>
Well apparently nixpkgs-unstable as your user channel and NixOS-20.03 as your system channel breaks OpenGL on user packages
mir100 has joined #nixos
<bpye>
Is that always going to happen with that sort of combination or should I have used NixOS for both?
fdert has quit [Read error: Connection reset by peer]
<Guest44214>
Hello, say I'm on NixOS 19 and I want to run a nix-build command with 20.03, how do I do to specify the channel's as the <nixpkg> to load via the CLI
<Guest44214>
I was trying to override NIX_PATH="nixpkgs=<channel-url>" nix build ...
<Guest44214>
but that doesn't work
<Guest44214>
also nix-build -I 'nixpkgs=<channel-url>' and that also does not work (requires a local path)
rasmusm has quit [Remote host closed the connection]
felixfoertsch has joined #nixos
<infinisil>
charukiewicz: Oh, it wants to compile bootstrap tools, which should normally be cached. What's in your /etc/nix/nix.conf?
<charukiewicz>
This was happening both yesterday on a 19.09 installation I had running, and today I deployed a fresh 20.03 installation and it's still ocurring there. Both were deployed as AWS EC2 instances using the community AMIs.
justanotheruser has joined #nixos
<infinisil>
charukiewicz: What's the output of `nix-store -q --outputs /nix/store/3q99kcf568dz01ck1d8rxzxxymnyh18i-bootstrap-tools.drv`
chagra has quit [Ping timeout: 260 seconds]
Guest44214 has quit [Quit: leaving]
<charukiewicz>
infinisil: It prints out another path: /nix/store/vxjwm492zk4hq00jhi8vj490lcxc1yp7-bootstrap-tools
<infinisil>
Okay and doing `nix-store -r /nix/store/vxjwm492zk4hq00jhi8vj490lcxc1yp7-bootstrap-tools` then tells me "don't know how to build these paths", which means that it's not in cache.nixos.org
fdert has quit [Read error: Connection reset by peer]
<infinisil>
charukiewicz: Which means that something in the Nix files must be changing stdenv
<charukiewicz>
infinisil: So if it's a problem with the project, why is it that a root user can build it, but not a non-root user?
<infinisil>
charukiewicz: Did it actually finish building with root?
<charukiewicz>
Yes, I can nix-shell both as root and with sudo as the non-root user
<charukiewicz>
and it works
<infinisil>
Hm, then it sounds like an impurity in the evaluation
<infinisil>
charukiewicz: Does `nix-instantiate shell.nix` differ between root and non-root?
<infinisil>
Oh, I kind of suspect you have something in ~/.config/nixpkgs/overlays or config.nix
<infinisil>
Which wouldn't apply for root
ixxie has quit [Quit: Lost terminal]
<charukiewicz>
yeah it outputs /nix/store/h1crnf5av087d5xjghjmywsawx0cp94c-ghc-shell-for-project-0.0.1.drv with a different hash for each
<charukiewicz>
The only config file on the machine I edited was /etc/nixos/configuration.nix
<infinisil>
charukiewicz: Does the project do something like `import nixpkgs {}` in its files?
<infinisil>
If so, try changing that to `import nixpkgs { config = {}; overlays = []; }`
<infinisil>
And then checking again if the nix-instantiate hash differs
<pistache>
hi! I'm having some trouble understanding the behaviour of the hicolor-icon-theme in NixOS
<pistache>
for other icon themes, there is a symlink in ~/.nix-profile/share/icons/THEME_NAME to the Nix store
<pistache>
however, for the hicolor-icon-theme, there is no symlink, but just a directory with the same subdirectories as the hicolor-icon-theme derivation
<{^_^}>
[nixpkgs] @matthewbauer opened pull request #87152 → Dont make lazy default on kde → https://git.io/JfnOB
<pistache>
I'm going through this as I wish to understand why is the "theme.index" file missing for this icon theme, and I can see that this file is indeed present in the derivation built in my store
u_u has joined #nixos
<infinisil>
charukiewicz: Gonna go to bed now, but if that doesn't help, I suggest complaining to the project authors, or maybe somebody here can find the impurity if you post the source