<{^_^}>
[nix-pills] @mhwombat opened pull request #91 → Miscellaneous minor clarifications → https://git.io/fp9ir
hlolli has quit [Ping timeout: 246 seconds]
nD5Xjz has quit [Ping timeout: 244 seconds]
<tmplt>
I'm writing a module for my PCI-passthrough setup, and I want to incorporate my libvirt domain XML in my NixOS configuration. Best way to do this would be to import the (eventually generated) file with `virsh define <path-to-xml> --validate`. Can an enabled module run arbitrary commands during system rebuild?
vidbina has quit [Ping timeout: 246 seconds]
nD5Xjz has joined #nixos
<tmplt>
It seems system.activationScripts can manage this.
<gchristensen>
better would be a systemd service wantedBy multi-user.target probably
<tmplt>
How does the two approaches differ?
Yaniel has quit [Excess Flood]
<gchristensen>
it is very dangerous to put any command in an activation script which can fail or block, because it runs very early in boot
Yaniel has joined #nixos
<gchristensen>
putting it in a one-shot systemd service is safe as it won't block if it fails, and executes in a more predictable environment
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/fp9PL
<tmplt>
that makes sense
<gchristensen>
"very dangerous" is over stating it ... but that generation might fail to boot. you can still roll back at the bootloader, of course :)
b has quit [Quit: leaving]
b has joined #nixos
cyounkins has joined #nixos
Yaniel has quit [Excess Flood]
cyounkins has quit [Remote host closed the connection]
Yaniel has joined #nixos
cyounkins has joined #nixos
cyounkins has quit [Ping timeout: 250 seconds]
jperras has joined #nixos
Synthetica has quit [Quit: Connection closed for inactivity]
jperras has quit [Client Quit]
m3lst4d has joined #nixos
jtojnar has quit [Quit: jtojnar]
Supersonic has quit [Disconnected by services]
Supersonic112 has joined #nixos
Supersonic112 is now known as Supersonic
<ldlework>
erm, my blog which is a python app, now fails after upgrading to 19.03 with:
<ldlework>
ImportError: /nix/store/mrfcv8ipiksfdrx3xq7dvcrzgg2jdfsw-glibc-2.27/lib/librt.so.1: symbol __close_nocancel, version GLIBC_PRIVATE not defined in file libc.so.6 with link time reference
<romanofskiWork>
could be that the version of pyobject was linked against a different version of glibc? the one installed seems to not export the close_nocancel function
<romanofskiWork>
you could actually check which symbols are exported.. hmm can't remember the command...
<romanofskiWork>
ah... nm it is
<romanofskiWork>
I think that should be: nm -D /nix/store/mrfcv8ipiksfdrx3xq7dvcrzgg2jdfsw-glibc-2.27/lib/librt.so.1 ... if I got it right
Mr_Keyser_Soze has quit [Ping timeout: 250 seconds]
cyounkins has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @samueldr opened pull request #51812 → Fixes three altcoins packages to use boost165 → https://git.io/fp9M2
orivej has quit [Ping timeout: 244 seconds]
jasongrossman has quit [Quit: ERC (IRC client for Emacs 26.1)]
Taneb has quit [Quit: I seem to have stopped.]
Lears has joined #nixos
Mr_Keyser_Soze has quit [Remote host closed the connection]
cyounkins has quit [Remote host closed the connection]
cyounkins has joined #nixos
[Leary] has quit [Ping timeout: 268 seconds]
<{^_^}>
[nixpkgs] @samueldr opened pull request #51813 → aarch64: ZHF for aarch64 → https://git.io/fp9MF
cyounkins has quit [Ping timeout: 268 seconds]
epicmetal has joined #nixos
b has quit [Quit: Lost terminal]
sgillespie has quit [Ping timeout: 268 seconds]
rhalff has quit [Remote host closed the connection]
hamishmack has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
rhalff has joined #nixos
<rhalff>
not sure what I did, but I get nsenter: cannot open /proc/5886/ns/ipc: Permission denied when I try to root-login into a container.
<rhalff>
I've added an ip and set privateNetwork to true, is that what causes this?
smolboye has quit [Ping timeout: 246 seconds]
<rhalff>
err, ok nevermind, should use sudo.
<ldlework>
yay i got my blog to build
<ldlework>
phew
jmeredith has joined #nixos
iqubic has joined #nixos
<iqubic>
I'm legitamately confused right now. I have a laptop with a dual boot set-up. I can choose Windows 10 or Nixos. If I choose Windows 10, bluetooth works fine. If I choose Nixos, bluetooth fails.
aleph- has joined #nixos
<iqubic>
Like I'm trying to use bluetoothctl to get bluetooth working and I keep getting the unhelpful error: "No Default Controller Available"
<iqubic>
This is really odd, because I was just using Nixos and Bluetooth earlier today.
graphene has quit [Remote host closed the connection]
Mr_Keyser_Soze has joined #nixos
graphene has joined #nixos
Mr_Keyser_Soze has quit [Remote host closed the connection]
Mr_Keyser_Soze has joined #nixos
Baughn_ has joined #nixos
Mr_Keyser_Soze has quit [Remote host closed the connection]
Baughn_ has quit [Read error: Connection reset by peer]
iqubic has quit [Ping timeout: 256 seconds]
iqubic_ has joined #nixos
<iqubic_>
I'm legitamately confused right now. I have a laptop with a dual boot set-up. I can choose Windows 10 or Nixos. If I choose Windows 10, bluetooth works fine. If I choose Nixos, bluetooth fails.
hakujin has quit [Quit: WeeChat 2.3]
aleph- has quit [Ping timeout: 244 seconds]
<ldlework>
iqubic_: what's confusing to me is that when you try to use bluetoothctl it says No Default Controller Available, but when you go to rfkill, you A) see your controller B) it isn't blocked
<ldlework>
which seems odd
<Rusty1>
would dmesg | grep bluetooth show something?
<iqubic>
Why don't I have access to the hciconfig command, even when I have bluez installed?
romanofskiWork has quit [Remote host closed the connection]
gerschtli has quit [Quit: WeeChat 2.2]
gerschtli has joined #nixos
simukis has joined #nixos
<{^_^}>
[nixpkgs] @bobvanderlinden closed pull request #51801 → nixos: kubernetes: use /run instead of /var/run → https://git.io/fp9al
epicmetal has quit [Quit: epicmetal]
endformationage has quit [Quit: WeeChat 2.3]
<fusion809>
Hi folks, I have LibreOffice installed via nix-env (installed as my user), but for some reason LibreOffice doesn't appear in the KDE app menu. ~/.nix-profile/share/applications contains some desktop config files for LibreOffice, which work (I've opened that folder and executed them successfully to check if they work). Any ideas why they don't appear in this menu?
<fusion809>
My version of NixOS is 18.09. I installed LibreOffice from nixos-unstable.
graphene has quit [Remote host closed the connection]
graphene has joined #nixos
<rhalff>
I want to have X access to my nixos containers, currently using xrdp which works ok. But is there a way to get the host's X performance? or should I use something like virtualbox for that (I prefer not).
smolboye has joined #nixos
<rhalff>
My goal is to create several containers with their own vscode and services running, so I can get back each project easily and continue where I left off.
<rhalff>
nah, I guess xrdp is ok enough. :-)
<emily>
xpra is nicer than xrdp
<rhalff>
ok, have to read up on that one, didn't see it mentioned anywhere yet.
<rhalff>
screen for x11 sounds good.
vk3wtf has joined #nixos
<emily>
it can do fairly efficient stuff for same-host forwarding
<emily>
you might also be able to hack something up with virgl
hedning has quit [Quit: hedning]
<rhalff>
ok, but that's qemu, not sure what nixos-containers are using actually.
<ldlework>
Why would a package's deps suddenly not be available in the installPhase?
<eyjhb>
rhalff: from 1 to 10, how hacky is it, to do something like that in a `.bashrc` to get the path?
Lears has quit [Read error: No route to host]
[Leary] has joined #nixos
<rhalff>
eyjhb, you could generate your bashrc so you have access to those vars. or perhaps put it in an environment variable defined in your configuration.nix and then use it in a bash script, which is also hacky I guess :-)
<eyjhb>
rhalff: the environment variable seems like a nice approach tbh
<etu>
Izorkin: Seems to a dependency that failed. On all arches: https://logs.nix.ci/?key=nixos/nixpkgs.51738&attempt_id=e962dc28-34c7-41fd-8d60-81dc3879eaa2
rhalff has quit [Remote host closed the connection]
smrtak has joined #nixos
jasongrossman has quit [Ping timeout: 250 seconds]
epicmetal has quit [Quit: epicmetal]
<teozkr>
is there a way to force nix-shell in interpreter mode (`-i`) to resolve nix files relative to the file folder rather than the working directory?
smolboye has joined #nixos
crmlt has joined #nixos
immae has joined #nixos
drager has joined #nixos
<Izorkin>
etu: please recheck PR
<immae>
Hi everyone! I have an already existing nix installation, to which I did some things (/nix/store contains things), but probably in a non-standard way, since my ~/.nix-profile/bin/ is empty (it only contains nix-foo files). Do I have a way to find out what I did? (like for instance getting a global dependency tree for the whole /nix/store?). I know I ran them as myself (I’m the only user of the machine), but
<immae>
I probably ran blindly nix-build commands or things like that, and I’d like to find out what I did for reference
<srhb>
vdemeester: I see a completely different failure. Do you have sandbox enabled?
<vdemeester>
I think so
<vdemeester>
yep `sandbox = true` in `/etc/nix/nix.conf`
<etu>
Izorkin: I've tried building php73 on my linux machine as well, seems to work fine :)
<vdemeester>
srhb: if you got `find: Only one instance of {} is supported with -exec ... +`, I fixed it :P
<srhb>
vdemeester: That's the one, yeah.
<{^_^}>
[nixpkgs] @markuskowa opened pull request #51826 → [18.09] gnss-sdr: turn off unit tests explicitly → https://git.io/fpHsR
nix` has quit [Ping timeout: 246 seconds]
Ariakenom has joined #nixos
<vdemeester>
yeah my bad, I messed up taking care of comments, it should work now (before the rebase at least)
<vdemeester>
srhb: yeah as soon as I rebase against master, I'm getting this `cgo.a: permission denied` error :(
<tazjin>
did something change in 18.09 with how/when build envs are constructed? Specifically I used to be able to do builds requiring pkgconfig+openssl in my normal user env in 18.03, in .09 it only works in derivations or nix-shell
<{^_^}>
[nixpkgs] @andir opened pull request #51827 → steam: ran update-runtime.py → https://git.io/fpHs1
<{^_^}>
[nixpkgs] @andir opened pull request #51828 → [18.09] steam: ran update-runtime.py → https://git.io/fpHsM
<srhb>
vdemeester: Yep, I can indeed reproduce that.
<srhb>
vdemeester: Time to bisect :)
<clever>
tazjin: pkgconfig will never work when installed, due to its requirement on setup hooks
<srhb>
vdemeester: Seems like a missing input that was maybe propagated before? Not sure.
<tazjin>
clever: it definitely used to, though, so I suppose something was fixed (depending on perspective)
<andi->
Otherwise I don't see how they contradict each other.. My two PRa fix the current state of the expression. Refactoring them is still possible afterwards and should be reviewed by someone that wants to maintain the package :)
<srhb>
Yeah, they definitely don't. :)
jbgi has joined #nixos
liff has joined #nixos
jbgi has quit [Client Quit]
jbgi has joined #nixos
johanot has quit [Ping timeout: 240 seconds]
johanot has joined #nixos
m0rphism has joined #nixos
erictapen has joined #nixos
cyounkins has joined #nixos
pie___ has joined #nixos
pie__ has quit [Remote host closed the connection]
<pie_>
srhb: i dont suppose i can just override the inner rev of the src = fetchFromGithub { ... } ? :P
<pie_>
that would be nice but doesnt sound like it would work...
<srhb>
pie_: Unfortunately not, it's "gone" since that's a regular function application
<pie_>
yeah
<vdemeester>
srhb: hum :(
bvdw has joined #nixos
<pie_>
srhb: has anyone thought of changing things so something like that *can* work? xD
<srhb>
pie_: Many people. The simplest way that people tend to hack around this is to make it an argument of the outer function, such that it's overridable via .override
Rusty1 has joined #nixos
<srhb>
I don't know if I like that. i feel like src should just be a smarter type that supports override by itself. :)
<{^_^}>
[nixpkgs] @georgewhewell opened pull request #51830 → cool-retro-term: fix on darwin → https://git.io/fpHcY
<{^_^}>
#51630 (by cyounkins, 3 days ago, open): nixos: Decide and document scope of configuration
nore has joined #nixos
<Izorkin>
etu: any ideas to fix build in Darwin?
Guanin has joined #nixos
<Guanin>
Hi, I'm trying to package some application that is distributed as uncompressed binary only and have some problems. When trying to use unpackPhase = "cp $src dummy", it basically works when doing nix-build, but within nix-shell, it fails with "do not know how to unpack source archive"
<etu>
Izorkin: Was it broken? Didn't look that close.
<clever>
Guanin: `unpackPhase` always runs the function by that name, it wont eval $unpackPhase
<clever>
Guanin: you need to `phases="unpack" genericBuild` i believe
<smrtak>
srhb: thx, rebuild finally w/o any warnings \o/
<{^_^}>
[nixpkgs] @c0bw3b merged pull request #51826 → [18.09] gnss-sdr: turn off unit tests explicitly → https://git.io/fpHsR
<Izorkin>
etu: error - clang-5.0: error: no such file or directory: ...
<Guanin>
clever, ah, I understand. Doing eval $unpackPhase works fine, so I guess I just mixed that up
jbgi has quit [Ping timeout: 252 seconds]
Phlogistique has quit [Remote host closed the connection]
<vdemeester>
srhb: alright, currently bisecting the problem :)
<Izorkin>
etu: please check only php73. maybe not working php-embed and php-unit
<etu>
Izorkin: I've managed to build php71 php72 and php73 on my linux. I don't have a darwin system.
<{^_^}>
[nixpkgs] @flokli pushed commit from @colemickens to release-18.09 « google-chrome-dev: add at-spi2-core dependency »: https://git.io/fpH8g
<Izorkin>
etu: buildInputs = [ if isPhp73 then pkgs.pcre else pkgs.pcre2 ]; - how to fix that variant?
qyliss has quit [Quit: bye]
jperras has joined #nixos
qyliss has joined #nixos
<clever>
Izorkin: you need () around the elemtns in the list
iyzsong has quit [Ping timeout: 252 seconds]
<averell>
pie_: looking at that CMakeLists, i don't think CXX flags will help at that stage. it searches wxWidgets_INCLUDE_DIRS. so either modify that or wxPYTHON_INCLUDE_DIRS directly
<pie_>
averell: i dont know much cmake, I suppose this isnt the correct way to do it: "-DwxPYTHON_INCLUDE_DIRS=${wxPython}/include/wx-3.0"
<pie_>
somehow i managed to get "error: anonymous function at /nix/store/sn2i9faimsbgxbmi3nrs4lpsqia5vcgv-nixos-19.03pre160749.3a393eecafb/nixos/pkgs/development/tools/build-managers/cmake/default.nix:1:1 called with unexpected argument 'majorVersion', at /nix/store/sn2i9faimsbgxbmi3nrs4lpsqia5vcgv-nixos-19.03pre160749.3a393eecafb/nixos/lib/customisation.nix:69:12" , not sure if thats unrelated
<pie_>
ok i commented out my change, its something else i did :)
<pie_>
yup im lost :I
ThatDocsLady has joined #nixos
* pie_
runs some diffs
<pie_>
ok...fixed..this is what happens when you randomly save an old nix script form a random git commit
<pie_>
averell: thanks a bunch, I got it to work
<pie_>
waiting for build :)
<averell>
good luck :)
smrtak has left #nixos ["thx4support"]
<Guanin>
I have another problem while trying to package that application. While using buildFHSUserEnv and supplying mono and sqlite in there, I am able to run that binary. But when I try to make a package without using buildFHSUserEnv, patching the binary seems to break it. it seems that there are some checks to prevent reversing (it e.g. reads back the last few bytes of the binary). Is there a sane way to package it with buildFHSUserEnv to
<Guanin>
make patching unnecessary?
<clever>
Guanin: have you tried steam-run yet?
<Guanin>
No
<Guanin>
I can try it, but the application is not related to steam in any way
<clever>
steam-run is just an fhs env helper, that includes all of the deps of steam based games
<clever>
if it works, you can look into how it works and adapt the code to your app
<pie_>
averell: at 40% "/build/source/pcbnew/swig/python_scripting.h:45:18: fatal error: wx/wxPython/wxPython.h: No such file or directory #include <wx/wxPython/wxPython.h>" xD
<pie_>
well, maybe if i have the cxx flags as well...
<qyliss^work>
cyounkins: channels are, imo, broken out of the box
<qyliss^work>
I plan on trying to fix it at some point but haven't found the time yet
graphene has quit [Remote host closed the connection]
<cyounkins>
Could you briefly describe how they are broken?
graphene has joined #nixos
<qyliss^work>
Sure. When you install Nix multi-user, the only channels in your NIX_PATH are root's.
<qyliss^work>
So adding channels with `nix-channel --add` or whatever won't actually work for you
<qyliss^work>
even though you're the one doing the build, not root
<qyliss^work>
`sudo nix-channel` doesn't even help, because sudo preserves $HOME, which is what nix-channel uses
<qyliss^work>
You have to `sudo sh -c 'nix-channel ...'` to be able to manipulate the channels in the default NIX_PATH
<qyliss^work>
This is totally counter-intuitive and really makes no sense at all
<__monty__>
It doesn't preserve HOME everywhere though but I agree it's annoying when it does.
<acowley>
I think this vulkan-related PR is ready to merge. I'd appreciate it if anyone can do so in order to clear the way to subsequent version updates. https://github.com/NixOS/nixpkgs/pull/51672
<{^_^}>
[cabal2nix] @peti pushed to master « travis: update test build script to use ghc 8.6.3 »: https://git.io/fpHzd
Cale has joined #nixos
<gchristensen>
__monty__: preserving HOME / not preserving HOME is a annoyance based on the host's `sudo` config
<gchristensen>
I'd love to excise all mentions of "multi-user" since that is the smallest benefit imo of Nix daemon
<__monty__>
gchristensen: Yes and there's arguments in favor of preserving HOME *but* I like thinking of sudo as "do this as root."
<gchristensen>
yes me too
<gchristensen>
Apple is doing everything they can to make `root` a thing of the past on macOS
THFKA4 has joined #nixos
<acowley>
Which isn't entirely misguided. It is a bit troubling how much we train linux users to re-try a failing command with 'sudo' before thinking
<gchristensen>
I agree
Mr_Keyser_Soze has joined #nixos
apaul1729 has joined #nixos
<gchristensen>
it makes some things about the Nix installation more complicated
<clever>
also of note, `sudo nano` is generally a bad idea
<clever>
you wind up with root owning the state files in the wrong $HOME
<ldlework>
lol
<clever>
and then the editor gives permission warnings every time its ran without root
<clever>
thats what sudoedit is for
<apaul1729>
hey guys - anybody have some experience building nix expressions for software that depends on ocaml and opam libraries? I'm working on an expression for facebook's static analyzer infer, and I'm currently using a sort of hacky call of `opam init` and an opam lock in the postUnpack phase. Wondering if there's a better way that people have heard abou
<apaul1729>
t (I did recently hear about opam2nix-packages and I'm taking a look at that too)
<clever>
but sadly, sudoedit only updates the file when you exit the editor, so you cant just spam save and retry things, you must fully quit the editor
<ldlework>
am i the only one who finds themselves wanting to know less and less of that kind of stuff as i get older
<{^_^}>
[nixpkgs] @vbgl pushed to master « coqPackages.simple-io: init at 0.2 »: https://git.io/fpHg5
<mdash>
ldlework: alas, what we want don't enter into it much.
<ldlework>
true
<pie_>
ldlework: let the bullshit flow through you :D
endformationage has joined #nixos
<hodapp>
just got a 2015 Macbook... have a 2008 aluminum Unibody one that I love, but didn't want to deal with the keyboard complaints on the 2016 and newer ones
<hodapp>
too hard to type reasonably fast, and touchpad is just full of little annoyances
<ldlework>
i've had so many mbp keys fall off where you can't put them back
<ldlework>
so annoying
<hodapp>
mine haven't fallen off, but there is just not enough key travel on the 2015
<hodapp>
and it's hard to type fluidly
ninjin has quit [Remote host closed the connection]
<apaul1729>
i've gotten used to the key travel on a '17 macbook for work, now i'm slower on my mech at home :/
ottidmes has joined #nixos
<hodapp>
I still use the 2008 sometimes (I just gave it to my girlfriend) but every time I type on it I go "gaaaah" at how much easier it is comparatively
Ariakenom_ has quit [Ping timeout: 240 seconds]
qqlq has joined #nixos
<qqlq>
I am trying to deploy a local virtual machine to libvirtd using nixops. It keeps giving the following error: "error: value is null while a set was expected, at /nix/store/.../nixos/pkgs/top-level/default.nix:63:5"
jonaswouters has joined #nixos
<clever>
qqlq: what is line 63 of that default.nix file?
<gchristensen>
my recommended solution is to exclusively type on Kinesis Advantages. then you don't get disappointed by substandard keyboards on on your otherwise favorite laptops.
<qqlq>
It's in the attribute localSystem = lib.systems.elaborate (builtins.intersectAttrs { platform = null; } config // args.localSystem);
dbmikus has quit [Ping timeout: 252 seconds]
<clever>
qqlq: is your nix file setting anything related to that?
<qqlq>
clever: the expressions are minimal. The machine itself is defined as helloserver = {config, pkgs}: {}; and the network has one attribute helloserver = {deployment.targetEnv = "libvirtd";};
<qqlq>
Initially, the expressions were much more elaborate, but even without setting anything, it fails.
<clever>
qqlq: might be a bug due to changes in either nixpkgs or nixops
<clever>
qqlq: using `nixops deploy -I nixpkgs=...` you can change the nixpkgs version used, see if some older ones work, maybe bisect it
Taneb has joined #nixos
<Taneb>
How does nixops decide which version of nixpkgs things come from?
<gchristensen>
it uses NIX_PATH
<acowley>
Does -I work with nixos-rebuild? I couldn't make it work the other day when I tried, so fell back to NIX_PATH=... nixos-rebuild, but I may have screwed it up.
<clever>
Taneb: by default, whatever <nixpkgs> maps to in NIX_PATH
<clever>
acowley: it should work
<acowley>
Okay, thanks. I thought it was probably user error, but the help didn't mention supporting it, so I wasn't sure.
<clever>
Taneb: but you can also do `nixops create ... -I nixpkgs=foo` or `nixops modify ... -I nixpkgs=foo` to bake a -I arg into the state file, and it will persist to all deploys
Mr_Keyser_Soze has quit [Ping timeout: 244 seconds]
betaboon has quit [Quit: WeeChat 2.2]
<{^_^}>
[nixpkgs] @alyssais opened pull request #51835 → Backport "geckodriver: unstable-2018-02-24 -> 0.22.0" to 18.09 → https://git.io/fpHwq
Mr_Keyser_Soze has joined #nixos
dbmikus_ has joined #nixos
bbarker has joined #nixos
<bbarker>
I've seen some repos on using MATLAB in nixos, and have used it myself (hackily) from a nix expression on a non-NixOS system, though it has been a while. Has anyone satisfactorily used MATLAB in recent times in NixOS (before I embark on a fool's errand)?
<Baughn>
I've been trying to make sure UIDs of users match across a couple of machines, none of which have the exact same users, but ran into trouble: https://www.irccloud.com/pastebin/m5TKZzto/
<Baughn>
Rovanion: Not entirely impossible, almost certainly a bad idea.
<Baughn>
There's no code for it; you'll have to build your own.
<symphorien>
Rovanion: you can build the config and get the systemd unit files, but there is no guarantee they would work on ubuntu
<Baughn>
(I've done just that. But that was as proof of concept, with conversion to NixOS in the future.)
<symphorien>
the firewall thing will most certainly not work
jomik has joined #nixos
<clever>
Rovanion: heh, i just answered almost exactly that question on google hangouts
<Izorkin>
etu: updated phpPackages, please check
<clever>
nix-build '<nixpkgs/nixos>' -I nixos-config=./configuration.nix -A 'config.systemd.units."prometheus.service".unit'
<clever>
Rovanion: this will generate a result/prometheus.service file, that you can then configure systemd to use
<clever>
the firewall part of the config will be entirely ignored
<Rovanion>
clever: Heh, I will try this folly!
<clever>
and if prometheus makes any nixos-y assumtions, things can fail
<jomik>
Fresh install of NixOS... Not sure which WM to pick D: This is probably not the correct place to ask for opinions though :P I am looking for a tiling window manager. Split between Awesome, dwm and qtile atm.
<clever>
Rovanion: you may also want to build grafana and alertmanager.service files as well
<apaul1729>
i run bspwm but have been thinking of trying dwm
<thefloweringash>
went poking at nvidia+gdm, didn't find anything yet, but I did find a curious assertion. If you try to turn off wayland in gdm, the nvidia drivers prevent you from doing so, because they don't support wayland.
<clever>
Rovanion: and if you want to clean it up, write another nix file, that imports nixos, and does a buildEnv over the 3 .unit's
<apaul1729>
jomik dwm looks like the best out of your three suggestions imo
<jomik>
apaul1729: Yah, seems NixOS has pretty nice support for patches too :D
<apaul1729>
tru
<jomik>
patching dwm that is.
<jomik>
I just see a lot of "awesome" pop up.
<Rovanion>
clever: I'm not sure I understand how that is cleaner? So that it becomes one step in the terminal to produce the systemd-files?
<clever>
Rovanion: yeah, one step to make a single dir with 3 .service files
<apaul1729>
yeah i think awesome is a bit more popular partly because of the higher time to setup dwm
<Rovanion>
jomik: I use awesome, it tiles my windows. Before that I used ratpoison, but it hated the mouse more than I do.
<gchristensen>
it is called ratpoison :)
<Rovanion>
It was not a misnomer.
<jomik>
Rovanion: Lol! I saw some WM sold itself on being "fully mouse-controlled" with no keybindings... Jeez :P I'd prefer avoiding the mouse too tho :P
jbgi has quit [Ping timeout: 260 seconds]
<Rovanion>
My only complaint with awesome is that they reinvent the configuration format a little too often. Other than that it tiles my windows well and deals with corner cases well.
MisterOutofTime_ has quit [Ping timeout: 256 seconds]
dbmikus_ has joined #nixos
<Rovanion>
I mostly just use the one where I can decide how many columns there are and how many windows there should be in each column.
<Ankhers>
Wht approaches are there to getting a language to support Nix? I know Haskell has a tool that converts I believe Stackage to a giant Nix expression which lives in the nixpkgs repo. I believe golang has a tool that will just dump a deps.nix file for it's dependencies. Are there other strategies? Is there an approach that is more suggested, or is it really up to the language for what would be most ergonomic?
<jomik>
Hmmm. How do I enable dwm-git as my window manager in NixOS? I can only see dwm in the windowManager object ? dwm is version 6.1 which is from 2015 :/
<jomik>
I guess I can make my own expression to enable it.
alp has quit [Ping timeout: 245 seconds]
sysadmin_wannabe has quit [Ping timeout: 240 seconds]
<{^_^}>
#51630 (by cyounkins, 3 days ago, open): nixos: Decide and document scope of configuration
<gchristensen>
cyounkins: post on discourse?
<cyounkins>
Yeah I guess I should do that.
<__monty__>
Hmm, I'm using a ghc in a myEnvFun. Normally `ghc file.hs` produces a binary called `file`. Now it doesn't though. It produces the .hi and .o files but no binary. Sound familiar to anyone?
Mr_Keyser_Soze has joined #nixos
<cyounkins>
gchristensen: You didn't direct me to the RFC RFC?? lol
<jomik>
__monty__: don't you need `ghc -o` or `ghc --make` ?
<__monty__>
jomik: No. Turns out ghc doesn't produce a binary unless Main.main exists. I'm in the habit of naming my modules something else (and not exporting main) and that works fine when I use the nix-shell shebang but not when compiling apparently. Also a little surprised ghc doesn't complain about the #! lines.
<Ankhers>
gleber_: ping
pxc has quit [Ping timeout: 240 seconds]
<{^_^}>
[nixpkgs] @ottidmes reopened pull request #51797 → allow extending lib with an overlay → https://git.io/fp9gb
pxc has joined #nixos
<jomik>
Haha. Yeah okay. That makes sense then. Good to know :)
<{^_^}>
[nixpkgs] @spacekookie opened pull request #51838 → railcar: init at 1.0.4 → https://git.io/fpH9w
<jomik>
Huh. My KDE System Settings keeps unchecking Reverse Scrolling..?
<{^_^}>
[nixpkgs] @MMesch opened pull request #51839 → python: add geographiclib and update geopy → https://git.io/fpHHz
<qqlq>
So, I was having a discussion earlier with clever about how nixops generate an error when deploying a network using nixos-unstable. I am wondering if I should file a bug against nixos or nixops.
<gchristensen>
for some reason the zoom-us package takes forever to "build" and I'm not sure why, bash has been at 100% CPU for ages
jbgi has quit [Ping timeout: 250 seconds]
dermetfan has joined #nixos
pxc has quit [Ping timeout: 245 seconds]
<{^_^}>
[nixpkgs] @jappeace opened pull request #51840 → Fix someone getting powned trough bad docs → https://git.io/fpH7e
<clever>
jomik: on lines 7, 16 and 25, change dwm to dwm-git, and then it should work in imports
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]
<jomik>
Ah, that's what I intended to do. Perfect - I'd like to pass an extra `patches` argument and add that as an option tho. I would have to use .override on pkgs.dwm-git I assume. IIRC I can't just pass the argument to pkgs.dwm-git
asymmetric has joined #nixos
<clever>
jomik: you would want overrideAttrs there
<jomik>
so line 33 would be environment.systemPackages = [ pkgs.dwm-git.overrideAttrs { inherit patches; } ];
<clever>
and lists are wonky in nix, thats a list containing a function and an attrset
<clever>
not the result of applying a function to a set
<clever>
throw some () around it to make things right
jonaswouters has quit [Quit: Connection closed for inactivity]
Radivarig has quit [Ping timeout: 272 seconds]
jbgi has joined #nixos
<ottidmes>
clever: I kind of wished they would just have required ; after each item, including for destructering attrsets { config; lib; }, it might be a little less pleasant on the eyes, but thats a matter of getting used to it, but would allow for easily moving code snippets around without punctuation gettig in the way (common issue with buildInputs / argument list)
jeremejevs has quit [Quit: Connection closed for inactivity]
waleee has quit [Quit: WeeChat 2.3]
pxc has joined #nixos
<{^_^}>
[nixpkgs] @veprbl opened pull request #51841 → pythonPackages.pyjet: fix for python37 → https://git.io/fpHFz
<jomik>
Oooh, right, that does make sense :D
<jomik>
ottidmes: destructuring attrsets need commas, right?
troydm has joined #nixos
orivej has quit [Ping timeout: 244 seconds]
<{^_^}>
[nix-pills] @liff opened pull request #92 → Update Attribute sets example and remove obsolete note → https://git.io/fpHbW
rawreraw has quit [Quit: WeeChat 2.3]
<ottidmes>
jomik: yep, as you can see at top of most modules, for example: { config, pkgs, lib, ... }: (the ... allows it to be open ended, more attributes are allowed to exist in the passed attrset)
<ottidmes>
jomik: yes, it is being used in patchPhase as defined in nixpkgs/pkgs/stdenv/generic/setup.sh
<jomik>
OH! That's where it hid. I remember reading it in the manual/pills... Thank you.
<jomik>
So is types.path suitable, or should it be types.str?
<ottidmes>
jomik: conceptually it could be considered a types.listOf types.path, but the type system is only being used within the module system of NixOS, so it does not apply to Nix packages
<ottidmes>
jomik: you can use paths, they will be converted to strings at some point though (to be used in the script)
azdle has quit [Remote host closed the connection]
<Izorkin>
etu: need fix only php couchbase
<jomik>
Ooh, so it wouldn't generate an error if I didn't give it a path? I guess I like specifying it as a path as that is what it has to be ultimately, whether it is enforced or not. Thank you.
<clever>
jomik: if you quote the path, it will depend on if the build is local or remote, and if the user doing the build has read access to the thing
<clever>
jomik: the nix sandbox also blocks access to such paths, which makes such errors a lot easier to find
<clever>
jomik: it tries to install to /usr/local/ by default
<clever>
that line forces it to instead install to $out
<jomik>
Oooooh!
<ottidmes>
clever: that is why you generally want to use paths, because they will get copied to the store, right? strings are ok, but only if already absolute store paths (in sandbox at least) (just to verify that my understanding is correct)
elgoosy has joined #nixos
<clever>
ottidmes: even storepath like strings ("/nix/store/hash-foo" for ex) arent "safe"
<clever>
ottidmes: only the magic strings with context, from "${foo}" are truely safe
<bbarker>
Is there a way to add user-level environment info to configuration.nix? sorry if I missed it in the appendix - didn't see anything likely under users.users.<name?> - end goal is to have something like environment.loginShellInit at user-level granularity
<clever>
"${./foo}" will also have context
<jomik>
I guess the remaining issue is being able to apply these patches in any order... I guess that isn't really possible, because they are applied in turn by patch, correct?
philippD has joined #nixos
<ottidmes>
clever: right
<jomik>
And I can't really apply multiple patches in a row, when the source file will change with each patch - it'll throw line numbers off?
<clever>
jomik: yeah
<jomik>
Penis :D
<clever>
jomik: thats what the context in a patch is for
<jomik>
Oh, so it can actually apply them correctly?
<clever>
yep
<jomik>
By using the context
rawreraw has joined #nixos
<jomik>
I see! Time to test then ~
<gchristensen>
is there a reason to not allow arbitrary users to sign up for a grafana instance?
Guanin has quit [Ping timeout: 250 seconds]
<ottidmes>
jomik: BTW its just using regular `patch`, so more details can be looked up just by checking: man patch
<simpson>
gchristensen: They'd be able to create resources on the server, like dashboards, and also run arbitrary queries to your metrics. I think instead you want to allow anon viewing?
<__monty__>
Does the nix-shell shebang call runghc with a main-is argument?
kim0 has joined #nixos
drakonis has joined #nixos
<jomik>
switching from overrideAttrs to overrid seems to let it build. Are you certain I should use overrideAttrs here?
<jomik>
clever ^
<vaibhavsagar>
monty: I don't think it does anything you don't specify in the shebang line
<jomik>
Well, I can't boot into none+dwm-git now :D
<__monty__>
vaibhavsagar: It's weird. I have a haskell program that runs fine *unless* I add a module declaration. And it's only a problem for the more difficult version of the problem the code solves.
<jomik>
How do I start dwm from command line? I need to see the error it produces :( I can't just run the binary, as it requires x to be running. Instructions says to put it in my xinitrc... But that's not applicable.
<clever>
jomik: modify your nix file to run `dwm > /tmp/dwm-stdout 2>/tmp/dwm-stderr &
<jomik>
clever: I get this error : XIO: fatal IO error 2 (No such file or directory) on X server ":0" after 12 requests (12 known processed) with 0 events remaining.
<clever>
not sure what the cause could be there
techieAgno2tic has joined #nixos
<jomik>
Sadly doesn't seem like google is much help D:
<{^_^}>
[nixpkgs] @vbgl opened pull request #51843 → ocamlPackages: fix a few evaluation issues → https://git.io/fpHjf
orivej has joined #nixos
<ottidmes>
jomik: could it perhaps have to do with ~/.Xauthority (that is the only file issue I have ever had in the context of X server)
<jomik>
There is only some nixos MIT-MAGIC-COOKIE in there. I dunno. I can try deleting it.
<clever>
jomik: the display manager should auto-create that file, with the current token, when starting X
drakonis_ has joined #nixos
<jomik>
HUH, xsession-errors.old says that $ is not a PID or valid jobspec, so maybe my `waitPID` is what fucks up?
<jomik>
I forgot a bang.
drakonis has quit [Ping timeout: 252 seconds]
<jomik>
Perfect :3 Thanks for hinting at that ottidmes .
hedning has left #nixos [#nixos]
<jomik>
And here I was just about to desert dwm for xmonad!
<clever>
jomik: oh, so waitPID wasnt set right, so the bash script failed to wait for dwm
<clever>
jomik: and then it terminated X while dwm was running normally!
<jomik>
Yup~! I suppose that's it :P
<techieAgno2tic>
i have a shell.nix that does some image magick stuff. when i run it on my laptop with --pure, i get one image, and on my server with --pure i get a slightly different one. i thought it might be that it didn't have the font, so i added that to the overrides and it installed fine, but im still getting a different image. is there something special i have to do to install fonts?
drakonis has joined #nixos
drakonis_ has quit [Read error: Connection reset by peer]
<symphorien>
you may have different fontconfig system-wide configuration.
<symphorien>
even with --pure, software still read /etc/
<clever>
techieAgno2tic: try using pkgs.runCommand, and nix-build
<clever>
then the nix sandbox can fully enforce purity
drakonis_ has joined #nixos
<arianvp>
for some reason pythonPackages doesn't seem to be cached anymore
<{^_^}>
[nixpkgs] @vbgl opened pull request #51844 → ocamlPackages.atdgen: init at 2.0.0 → https://git.io/fpQe7
Lisanna has quit [Quit: Leaving]
jomik has quit [Quit: Page closed]
<techieAgno2tic>
clever: how do i use pkgs.runCommand? ive never seen that before. whenever i go into the shell i just run ./result/bin/program and then exit
<clever>
techieAgno2tic: i gave an example above
<techieAgno2tic>
oh thanks
ntd has quit [Quit: ntd]
baum_ has joined #nixos
dbmikus__ has quit [Ping timeout: 250 seconds]
dbmikus__ has joined #nixos
<baum_>
Hi, same question again since i can't seem to figure out the solution: executing "nix-shell" works, executing "echo $PWD | xargs nix-shell" doesn't. Someone said that it works on his system, however i tried it on both, stable and unstable without success. (i did also try it with different terminal emulators and different shells). While nix-shell is executed the shell "doesn't start". the verbose output looks
<baum_>
pretty much the same.
m3lst4d has joined #nixos
jbgi has quit [Ping timeout: 264 seconds]
<baum_>
by chance anyone has an idea why that's the case / what i could check next
<cransom>
what are you expecting nix-shell to do with your input?
fendor has joined #nixos
<baum_>
cransom: i'm expecting it to handle it as a "path" arg
hamishmack has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<baum_>
cransom: thus, opening a shell based on the referenced nix-expression
<asymmetric>
is there a way to convert an .ini config file to nix syntax?
<asymmetric>
e.g. ini2nix?
mulderr has joined #nixos
<cransom>
baum_: if you are piping to nix-shell, you can't get stdin back from the xargs so there's nothing for nix-shell to run. it invoked the shell, but exited because it has no input
<asymmetric>
i could probably use an .ini to .json converter
<cransom>
asymmetric: nix manual says there's a `builtins.fromTOML` as of 2.1
m3lst4d has quit [Remote host closed the connection]
<baum_>
cransom: uhm are you sure? it seems like some kind of input is recognized since the env is built (assuming a newone was created)
<cransom>
baum_: yes, the nix-shell would get the input, but there's no way for you to interact with it. xargs gave it input and it's over. i don't appear any different to me than say `echo true | bash`. i don't expect bash to hang around.
<baum_>
cransom: makes sense, thanks! by chance you have any idea how i could work arround this?
<cransom>
baum_: bash for loop? if you are expecting interactive output, `for f in some list here; do nix-shell $f; done` ?
fusion809 has quit [Remote host closed the connection]
Radivarig has joined #nixos
<baum_>
hmm alright, i'll write it in bash. I was hoping on only chaining it with pipes. :P
<cransom>
pipes are slower :(
<cransom>
`xargs nix-shell --run "echo my command that only works in this shell passed"` works too if you know whats going to be run, but there's no interactive there.
<asymmetric>
cransom: i'd need something like fromINI
<asymmetric>
there's generators.toINI but couldn't find a fromINI
<cransom>
maybe i missed the memo. i thought ini was toml
dermetfan has quit [Ping timeout: 256 seconds]
<baum_>
cransom: thanks for the help :)
cx405 has joined #nixos
<cx405>
hello! How can I add stuff to existing attribute within same scope, without using dedicated file/import? For example: boot.initrd.kernelModules = [ 1 2 3 ]; and somewhere below in configuration add more modules. I am asking to help modularize the configuration. Thank you!
graphene has quit [Remote host closed the connection]
boogiewoogie has quit [Remote host closed the connection]
<samueldr>
do you know if the bluetooth interface is on the intel wireless interface?
seqizz has joined #nixos
<iqubic>
I have no idea.
<iqubic>
Sorry.
graphene has joined #nixos
<iqubic>
I'm using a dell latitude 5480 right now.
<samueldr>
in my dmesg output I get "Bluetooth: hci0: Intel Bluetooth firmware file: intel/ibt-hw-37.7.10-fw-1.80.1.2d.d.bseq", so this at least looks plausible that it somehow is not loading the firmware
<iqubic>
That's the brand and model of my current laptop.
<seqizz>
Hi, do we have any option to skip pytests while installing python modules (installing pylint takes ages and even fails on 18.09 channel)? Or do I have to override?
<iqubic>
I just searched my dmesg output for the phrase `bluetooth`: http://termbin.com/4nik
<{^_^}>
[nixpkgs] @pmiddend opened pull request #51847 → epsxe: add libGLU_combined to buildInputs → https://git.io/fpQI4
<simpson>
seqizz: You can put `doCheck = false;` on failing Python packages. Ideally we would fix the tests, but some Python packages are really bad at writing hermetic tests.
<iqubic>
samueldr: Take a look at the last line of that termbin link I posted.
<iqubic>
It says that something or other failed.
<samueldr>
iqubic: using `uname -r` which version of the kernel are you using?
<{^_^}>
[cabal2nix] @peti merged pull request #394 → Include cabal-version field inferred by hpack → https://git.io/fpyM3
<samueldr>
and iqubic, do you have something similar to the following in your configuration.nix? imports = [ <nixpkgs/nixos/modules/installer/scan/not-detected.nix> ];
<iqubic>
4.14.86
<iqubic>
No i have nothing at all like that import statement there.
<samueldr>
it may very well be in hardware-configuration.nix
<iqubic>
Let me check.
<iqubic>
samueldr: Yeah. It is in my hardware-configuration.nix file.
<samueldr>
:/ so it's not an obvious thing like "the firmware isn't made available". I would try using the latest kernel instead of the LTS one, it often help for hardware things on laptops, but a ~2016 (AFAICT) laptop shouldn't really need it
<samueldr>
something like: boot.kernelPackages = pkgs.linuxPackages_latest; # in your configuration.nix
<iqubic>
because that weird behavior is exactly what I'm seeing.
<samueldr>
it's possible there's a regression somewhere for a few hardware combinations
<samueldr>
(and with upstream projects mainly)
<iqubic>
should I try using the most bleeding edge kernel?
<srhb>
iqubic: It's usually one of the cheapest and fastest tests you can do :)
<samueldr>
_latest isn't bleeding edge, but yeah, current new stable, at least it will either give results or not
<seqizz>
simpson: Thanks. Could you show me a simple way to override in configuration.nix directly? I am dazzled by the options :\
<samueldr>
and yeah, as srhb said
<samueldr>
and from your dmesg output, I'm pretty sure "Bluetooth: hci0 command 0xfc05 tx timeout" is the thing that's causing issues, my dmesg | grep -i bluetooth is similar up until that point (with intel bluetooth)
Neo-- has quit [Ping timeout: 272 seconds]
<ivan>
does anyone use mozc and actually prefer the behavior where it starts in `Direct input` instead of `Hiragana`, I might want to PR a patch to fix it by default
<samueldr>
software is kept as upstream designs it as much as possible
pkx3 has quit [Read error: Connection reset by peer]
pxc has quit [Ping timeout: 240 seconds]
Ariakenom has quit [Quit: Leaving]
pxc has joined #nixos
dsx has joined #nixos
sysadmin_wannabe has quit [Quit: Cya soon]
<dsx>
How to make zfs utils to be available in docker's PATH?
<gchristensen>
whoa, what for?
pkx3 has joined #nixos
<dsx>
I want it to use its zfs driver
<dsx>
dockerd[17107]: time="2018-12-10T18:02:33-05:00" level=warning msg="could not use snapshotter zfs in metadata plugin" error="exec: "zfs": executable file not found in $PATH: …
<dsx>
And it seems that kubernetes is not happy about this too
pkx3 has quit [Read error: Connection reset by peer]
spacekookie has joined #nixos
pkx3 has joined #nixos
<simpson>
seqizz: Sorry, was busy. I normally wouldn't go through configuration.nix for this, since this sort of problem needs to be fixed in nixpkgs. I'm not sure what to recommend doing if you just want it fixed for you.
jabranham has quit [Quit: ERC (IRC client for Emacs 27.0.50)]
<dsx>
gchristensen: it works!
<gchristensen>
yay!
doyougnu has quit [Ping timeout: 246 seconds]
pkx3 has quit [Read error: Connection reset by peer]
<seqizz>
simpson: np, I was playing with some options, and realized however I set doCheck = false, it doesn't care at all. I guess I need to read more about overlays.
<Mic92>
dsx: I think we enabled zfs automatically for docker
pkx3 has joined #nixos
<dsx>
Mic92: I'm running 18.09, it doesn't seems to be enabled
pkx3 has quit [Read error: Connection reset by peer]
fragamus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<dsx>
Most probably.
pkx3 has joined #nixos
<GlennS>
Anyone know how to use Nginx modules in NixOS? I'm guessing I need to make an override of the Nginx package and put something into the `modules` parameter, then pass the overriden package to `services.nginx.package`?
pkx3 has quit [Read error: Connection reset by peer]
<GlennS>
(Actually there is a general problem here that I have a lot where I can find some Nix expressions in the nixpkgs repository, but don't know how to get from that to an attribute path I can use.)
pkx3 has joined #nixos
<asymmetric>
is it good practice to set "doCheck = false" for packages that don't have tests? and if so, why? what's the advantage?