<bbarker>
I've got a local nix expression (not in nixpkgs), with mkDerivation { src = ./.; ... }. Works great except each time I run the expression, I get a new entry in the /nix/store, even if no files change. Is there a way to 1) avoid this or 2) put the output in a temporary directory instead of the /nix/store (helpful during development)?
DigitalKiwi has quit [Ping timeout: 245 seconds]
<bbarker>
(especially when using nix-shell)
<{^_^}>
[nixpkgs] @samueldr pushed commit from @etu to release-18.09 « nano: 2.9.8 -> 3.0 »: https://git.io/fAotn
graphene has quit [Read error: Connection reset by peer]
graphene has joined #nixos
rprije has quit [Remote host closed the connection]
<ldlework>
Hm how do you configure what browser is used by xdg-open
<tomberekn[m]>
bbarkeryes, nix-gitignore allows you to exclude files like `result` from the derivation.
Cale has joined #nixos
Supersonic has joined #nixos
woodson has joined #nixos
<ogle>
I cant seem to override stdenv in my custom overlay. I have `self: super: { stdenv = super.overrideCC self.stdenv self.gcc6; }` which yields an ininite recursion. I suspect this has something to do with the bootstrapping process, but what is the appropriate way to acheive this?
nD5Xjz has joined #nixos
<ogle>
correction I actually have `self: super: { stdenv = super.overrideCC super.stdenv super.gcc6; }`
<tomberekn[m]>
that looks better
<tomberekn[m]>
what's the error?
mounty1 has quit [Ping timeout: 246 seconds]
<ogle>
[ogle@nixos:~/demo]$ nix-shell
<ogle>
error: infinite recursion encountered, at /nix/var/nix/profiles/per-user/root/channels/nixos/pkgs/top-level/all-packages.nix:8968:3
<ogle>
(use '--show-trace' to show detailed location information)
silver has quit [Read error: Connection reset by peer]
<samueldr>
ogle: self.stdenv is referreing to the new stdenv you're trying to declare
<samueldr>
in an overlay, `self:` is everything *at that point* in the overlay
<samueldr>
to have *that* stdenv, it needs to refer to itself, thus recursion
<samueldr>
wait
<samueldr>
I only read before "correction I actually have"
<samueldr>
(well, I sandwiched the reading, end, then first, missing the meaty middle :/)
yodeler has joined #nixos
ydlr has quit [Read error: Connection reset by peer]
<ogle>
samueldr: yeah not sure how to correct it. I have also tried super.stdenvAdapters.overrideCC, and I still get the recursion
<samueldr>
can I ask what the endgoal is?
Edes has joined #nixos
lassulus_ has joined #nixos
<samueldr>
the recursion seems to be at a line with `nativeTools = ...` where it refers to `stdenv.cc.nativeTools`, so trying to override stdenv tries to use that stdenv
<ogle>
default compiler in nixos 18.03 is gcc7, my packages have deps that require I build with gcc6 unfortunately
<samueldr>
(though weirdly I would have assumed super would have helped?)
<samueldr>
I may be wrong, but you might be forced to not rely on the `stdenv` name for your things, or to use `callPackage ... { stdenv = stdenv6; };`
<samueldr>
(whichever name you pick for your stdenv)
jackdk has quit [Ping timeout: 252 seconds]
<samueldr>
(others more experienced with tricky nix things may help you better)
<ogle>
roger that. I will keep tinkering. hoping I dont have to thread `{stdenv = mystdenv}` through everything
kiloreux has quit [Ping timeout: 240 seconds]
<samueldr>
though, it looks like you're close with your callPackage stuff in default.nix, I don't really know what makeScope does though
<ogle>
yeah, I dont completely grok that just yet. my understanding is that makeScope allows you to create a restricted scope in which callpackages will automatically inject args
lopsided98 has quit [Ping timeout: 240 seconds]
lopsided98 has joined #nixos
<ldlework>
Hmm is there no declarative way to set xdg options?
kiloreux has joined #nixos
<ldlework>
I think my next big project is theme unification across all my tools.
<pyvpx>
I have a potentially silly nix question: is there any automagic way to have nix-env select the version that matches the current kernel? dpdk has numerous versions, tied to kernels, and for some reason I figured doing nix-env --install dpdk would choose the "right" one, not the first one
<srhb>
pyvpx: Generally, things like that are packaged along with a linuxPackages package set such that kernel and any kernel specific items come together.
hotfuzz has quit [Ping timeout: 264 seconds]
<Izorkin>
Mic92_: https://github.com/Mic92/nur-packages/commit/c83e0a3f9a2b973595d904874577733f9a575b80 error install inxi - fixed-output derivation produced path '/nix/store/nydkiaxhs61lc9m2lm6490jw9ghxwavl-source' with sha256 hash '0nhiy752hrwg2pcb5878sknfh7jgs845d941mdz700aww4y047ha' instead of the expected hash '0vpnkw5f8i1cakxxsqbb88bvnpncs87lx110s0ry0mnz4ajq2wi7'
abueide has quit [Remote host closed the connection]
<srhb>
pyvpx: Having one attribute transform based on the kernel sounds like an impurity to me.
abueide has joined #nixos
tzemanovic has quit []
<srhb>
pyvpx: iow on NixOS, you could do something simple like... let myLinuxPackages = pkgs.linuxPackages_4_18; in boot.kernelPackages = myLinuxPackages; ... environment.systemPackages = [ myLinuxPackages.dpdk ]; ...
<srhb>
Thus declaratively specifying that dpdk should match the kernel version.
<pyvpx>
what's the transform there, though? in my specific case I have a fresh nixos install, with kernel 4.14.67. dpdk (from querying) comes in 4.14.67 and 4.17.19 and 4.12.2 and 4.4.153 etc.
<Mic92_>
I should also start with travis ci for nur-packages, like the olders.
<srhb>
pyvpx: Because build purity is a very desirable feature. Things that are or are not present on the system should not, in general, affect packages. Only their functional inputs.
<Izorkin>
Mic92_: inxi updated to ver 3.0.24-2
worldofpeace has quit [Ping timeout: 244 seconds]
<Mic92_>
Izorkin: sorry I cannot keep with their pace, the last version was pushed out 23 hours before that.
<Izorkin>
)
<pyvpx>
srhb: I don't disagree; then again I still don't follow. there are different versions of the package. if nix-env can choose one when given an ambiguious name, why not the one with the matching kernel version in the name?
<srhb>
pyvpx: I'm the wrong person to ask that question, I think that nix-env should never do anything given an ambiguous package :-P
<pyvpx>
hahaha. yes! that would have been the second expectation
<pyvpx>
pick the right one, or do nothing and bail
<pyvpx>
instead it just picked "the first" one
<pyvpx>
which...isn't even the first one listed when queried!
Ariakenom has joined #nixos
<srhb>
pyvpx: Agreed.
<pyvpx>
I'm very very new to Nix (and linux in general. exclusive openbsd user since forever) so I was just trying to get a better understanding
<srhb>
Going by name is often subtly wrong.
<srhb>
pyvpx: Of course! :)
* ekleog
still doesn't understand why nix even allows refering to a package by name
<srhb>
ekleog: I feel the same, mostly.
<srhb>
It's so convenient though!
<srhb>
And also it's hard to uninstall anything without that feature.
<ekleog>
hmm dunno, I'm mostly looking for nix-locate / the /packages.html, usually, not even using package names :°
<Mic92_>
Izorkin: could you send me a pull request for that?
<Mic92_>
maybe also the update
<mpu>
Hey, I'd like to create a small derivation to start with; however, I'd rather not have it integrated in nixpkgs, just storing somewhere else; do you have a short guide for that?
<hyper_ch2>
stupid spice html5 :( just can't get it to run properly
<mpu>
(The manual requires modifying all-packages.nix)
<srhb>
mpu: Instead of specifying every package dependency in the function of your package, try just depending on pkgs to begin with
<srhb>
mpu: That should be all the difference needed.
<srhb>
mpu: all-packages usually just is a small wrapper to name something and callPackage the file, so you can just pull that bit out :)
<mpu>
srhb, thanks; would there be a way to have the package file FOOBAR/default.nix follow the same syntax as the one in nixpkgs and wrap it in another top-level file?
<srhb>
mpu: Yes, exactly
<srhb>
mpu: The "wrapper" is just pkgs.callPackage ./default.nix {}
<srhb>
The main point is that we're calling the function that we import from <nixpkgs> and that callPackage takes a second argument as well (argument overrides to the package)
<mpu>
What are these '{}' doing?
<srhb>
mpu: It's just an empty attrset
<srhb>
mpu: foo {} {} applies the function {} to two arguments (that happen to be empty attribute sets)
<srhb>
applies the function foo*
<srhb>
I really need coffee xD
<srhb>
mpu: The argument set to the nixpkgs function takes all manner of (optional) attributes for configuring nixpkgs
<srhb>
mpu: The argument set to callPackage (its second argument) can override the arguments fed to the called function
<srhb>
mpu: So if test.nix takes eg. { hello }, we can callPackage ./test.nix { hello = pkgs.some-other-version-of-hello; }
<mpu>
srhb, thanks, I got nix to put "hello world" in a file :)
<srhb>
mpu: Hooray :)
<mpu>
srhb, thanks for the explanations about {}, that makes sense.
<mpu>
But I had to fetch pcre,bash,bzip,glibx,gnugrep,gcc,... !
nly has left #nixos [#nixos]
jperras has joined #nixos
<mpu>
srhb, next step is to look into overlays I guess
<srhb>
mpu: fetch? What do you mean?
<mpu>
srhb nix-build fetched all kinds of stuff (probably from stdenv).
<srhb>
Ah
<srhb>
Sure :)
drp` has quit [Remote host closed the connection]
jperras has quit [Ping timeout: 252 seconds]
lawlesseel has joined #nixos
philippD has joined #nixos
<{^_^}>
[nixpkgs] @pSub pushed commit from @hrdinka to release-18.09 « zsh: 5.6 -> 5.6.1 »: https://git.io/fAoRH
<Purple-mx>
<{^_^}>
[nixpkgs] @akamaus opened pull request #46562 → Fix for pytorch-0.4.1 cuda-enabled build. → https://git.io/fAoRh
spear2 has joined #nixos
<mpu>
fetchurl does not seem to like links that point directly to C source, what should I use in that case?
<mpu>
Ah, maybe it's mkDerivation's src which does not like the file...
Boomerang has joined #nixos
<sphalerite>
mpu: yeah it's the unpackPhase that doesn't like it. If you disable that it should work
<sphalerite>
so phases = [ "buildPhase" "installPhase" "fixupPhase" ]; or something
jensens has joined #nixos
jackdk has quit [Ping timeout: 252 seconds]
fendor has joined #nixos
goibhniu has joined #nixos
thc202 has joined #nixos
lawlesseel has quit [Ping timeout: 256 seconds]
<rsa>
is there a standard way to add packages to the store but without adding them to path?
<srhb>
rsa: nix-build with a gc root
mayhewluke has quit [Ping timeout: 240 seconds]
<srhb>
Unless you want it to be gc'ed whenever, in that case just nix-build
mayhewluke has joined #nixos
<sphalerite>
nix-build creates a gc root by default, so you need to specify --no-out-link if you don't want it to be created
<srhb>
Right, I meant "a gc root somewhere safe" because you'll probably end up overwriting your result symlinks if you're anything like me.
<sphalerite>
haha
adetokunbo has quit [Quit: This computer has gone to sleep]
__Sander__ has joined #nixos
<rsa>
i would like the packages to be part of the system profile, but not added to the environment
<sphalerite>
rsa: system,extraDependencies
<rsa>
sphalerite: ty!
palo has quit [Ping timeout: 240 seconds]
<sphalerite>
can I specify cores in a store URL somehow? ssh-ng://foo?cores=4 complains "warning: unknown setting 'cores'".
<sphalerite>
oh no… apparently the storage on scaleway's C1s is quite unreliable. Lots of store corruption, and not in the usual "accident" patterns like .pyc files being created
<{^_^}>
cannot read '/nix/store/g0jj3kq7xclad80yf371qdakhcnryvvi-out.json', since path '/nix/store/mqa0v9nn8bigwqxf2xpi6dx9b7fi1c10-out.json.drv' is not valid, at (string):193:1
<sphalerite>
manveru: not sure infinisil is keen on people being able to run arbitrary code on his machine ;)
<manveru>
how is that arbitrary? :P
<teto>
manveru: haha nice ty :)
<colemickens>
I have a nixos module that computes something in the `let ...` at the top. How do I access that from another module?
<sphalerite>
manveru: you could make a fixed-output derivation that never completes building and sends spam emails
<colemickens>
Can I export it from the module, the same way system options are?
<manveru>
sphalerite: without networking?
copumpkin has quit [Ping timeout: 244 seconds]
<manveru>
uhm, i mean, nevermind
<srhb>
colemickens: Yeah, you can just set config.whatever to it
<srhb>
colemickens: Though I'd say it's at least a little weird :-)
<colemickens>
johanot's module computes the kubernetes cniConfig directory.
<colemickens>
I need it to configure containerd with the same configDirectory.
<srhb>
colemickens: Makes sense.
civodul has joined #nixos
<colemickens>
(Though, it may be the case that when using CRI, the CNI config is useless to k8s, but I'm actually not sure of the kubelet internals.)
<srhb>
Ideally that would be a separate module entirely for cni that both modules used the path from
<srhb>
but as a start you can indeed just set config.whatever to the value you care about from the other module.
contrapumpkin has joined #nixos
<srhb>
There's cleaner ways to do it, I think my favourite is factoring it out properly.
<srhb>
(eg a cni module with cniConfigDir)
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « haskell-stm: update overrides that use version 2.4.5.0 to version 2.4.5.1 »: https://git.io/fAogU
<{^_^}>
[nixpkgs] @peti pushed 11 commits to haskell-updates: https://git.io/fAogE
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/fAogw
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « all-cabal-hashes: update snapshot to Hackage at 2018-09-12T08:26:27Z »: https://git.io/fAogb
<{^_^}>
[nixpkgs] @peti pushed to master « all-cabal-hashes: update snapshot to Hackage at 2018-09-12T08:26:27Z »: https://git.io/fAogb
<srhb>
Anyone else seeing a failure with user-runtime-dir@uid with 18.09? "failed to remove runtime directory /run/user/1000 (before/after unmounting): No such file or directory
<{^_^}>
[nixos-org-configurations] @edolstra pushed 2 commits to master: https://git.io/fAowK
patrl has quit [Ping timeout: 252 seconds]
tzemanovic has quit [Remote host closed the connection]
theunknownxy has joined #nixos
<teto>
why gnome apps UI are half of the time missing icons ? for instance I just wanna try this https://wiki.gnome.org/Apps/California but no icons, some panels are half painted (half black)
aristid has quit [Ping timeout: 240 seconds]
michiel_l has quit [Ping timeout: 250 seconds]
<sphalerite>
teto: are you running GNOME?
<sphalerite>
teto: if not I think installing adwaita-icon-theme *might* help
canndrew has quit [Ping timeout: 246 seconds]
canndrew has joined #nixos
lukego has joined #nixos
<teto>
sphalerite: I am not (i3). would a nix-env install be ok ?
<sphalerite>
teto: I think so
michiel_l has joined #nixos
<teto>
sphalerite: yep that fixed it ! ty . Shouldn't it be part of the closure ?
<lukego>
Can somebody tell me how to download the NixOS 16.03 installer & find contemporary installation instructions?
<teto>
parts of the UI are still painted black though but that's fine
<srhb>
lukego: _16.03_ ?
<sphalerite>
lukego: easiest is probably nix build -I nixpkgs=channel:nixos-16.03 -f '<nixpkgs/nixos/release.nix>' -A iso_minimal.x86_64-linux
<sphalerite>
oops, without the -A
<srhb>
Yeah.. The manual should be included, but otherwise you can build it using the same method.
dmc has quit [Quit: WeeChat 2.2]
<lukego>
sphalerite: Oh great idea let me try that
<lukego>
Yes 16.03 :) in my world that is quite a recent release.
mkoenig has quit [Quit: Lost terminal]
<srhb>
lukego: Terrifying! :P
mkoenig has joined #nixos
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « haskell-cabal2nix: update override for hpack 0.31.0 »: https://git.io/fAoiM
Dedalo has joined #nixos
dmc has joined #nixos
<lukego>
srhb: It sucks being stuck on old versions. NixOS does not seem to be that strict on backwards compatibility though and I don't have the time to fowards-port all our infrastructure right now.
<lukego>
I think it's only that we are using a trick to run `sudo` inside builds, and this was disallowed in some version of Nix, and then re-enabled again, but not working in quite the same way as it used to. My head hurts :)
<srhb>
lukego: Ah, I remember your comment on that vaguely
<lukego>
(This is a farm of servers running benchmarks via Hydra. The software being benchmarked has to run as root and can't run inside a VM. So kindof right at the edge of Nix territory.)
xok has quit [Quit: Leaving.]
<srhb>
Ouch.
Ridout has joined #nixos
<lukego>
and I don't have a machine with Nix 2.0 so I am trying to work out how to back-port your 'nix build' command line to something that I have :)
tzemanovic has joined #nixos
<lukego>
Or perhaps I can run nix 2.0 inside docker. Can't install it on NixOS because it's not compatbile with the older daemon.
<symphorien>
replace nix build by nix build, remove -f, readd -A
<symphorien>
by nix-build *
<sphalerite>
lukego: nix-build -I nixpkgs=https://nixos.org/channels/nixos-16.09 '<nixpkgs/nixos/release.nix>' -A iso_minimal.x86_64-linux
<srhb>
(Shouldn't make a huge difference unless someone backported a lot of broken stuff to 16.03 after EOL :-P)
<colemickens>
srhb: I've tried to add myvariable to config, it's right next to some systemd.services.<whatever> definitions, but it just complains that "The option `myvariable` defined in ... does not exist".
<srhb>
colemickens: Yes, it needs an options definition as well.
<colemickens>
oh, I see.
<srhb>
colemickens: And at that point you're about two steps away from having a proper cni module that both kubernetes and whatwasit can use :-P
xenog has joined #nixos
<colemickens>
Yeah, I think there's a bit more too it though. The current flannel module is coupled with kube/pki in the kubernetes module.
<srhb>
But for the quick-and-dirty way you can just define it in either of those modules..
johanot has joined #nixos
<srhb>
colemickens: Yep, it's a tangled web.
aristid has joined #nixos
Neo-- has joined #nixos
<colemickens>
It's just a pain, most (all?) of the relevant CNI plugins need to be able to talk to an etcd instance, securely.
<srhb>
Yeah..
simukis has joined #nixos
ericsagnes has quit [Ping timeout: 252 seconds]
<manveru>
how do i get glibc.static with -fPIE?
<lukego>
Thanks very much for the tips! Building the ISO from source does make total sense. The command worked.
<sphalerite>
\o/
gspia has quit [Quit: Leaving]
Mic92 has joined #nixos
<lukego>
It's still quite comical to me that NixOS hackers are always shocked when I mention what version I am using. To me it seems quite recent :). Maybe this is a bad sign that I am in the RHEL demographic who wants to install an OS and then just keep running it forever...
mounty has joined #nixos
<manveru>
well, it's just gonna be a pain if you ever want to upgrade :)
<sphalerite>
lukego: it's fine for RHEL or whatever, but nixos gets *no* updates, so you're probably runnign stuff with piles of security bugs :/
<sphalerite>
unless you backport all the security patches yourself, in which case good on you :D
<manveru>
can i just add configureFlags in an override?
<lukego>
sphalerite: Yeah. Lack of updates is an issue. I dunno how relevant it is to me on these particular machines though. They are just hydra build slaves and dev boxes. The only exposed service to the outside world is OpenSSH.
<lukego>
but yeah I am looking forward to getting updated some day :)
<lukego>
The main problem for me is that I was really looking forward to Nix 2.0 for ages but now that it arrived it is only causing me compatibility headaches.
<lukego>
but, not a long-term problem.
<manveru>
seems like glibc doesn't recognize -fPIE :(
<manveru>
why does it tell me to compile it with that then
<manveru>
`/nix/store/h0lbngpv6ln56hjj59i6l77vxq25flbz-binutils-2.30/bin/ld: dynamic STT_GNU_IFUNC symbol `memset' with pointer equality in `/nix/store/s58hg86980680l8rr9yp9i8m7czzacvj-glibc-2.27-static/lib/libc.a(memset.o)' can not be used when making an executable; recompile with -fPIE and relink with -pie`
<srhb>
rawtaz: It's similar, but not quite
<srhb>
rawtaz: overlays look like `self: super: { someAttr = definition; ... }` where self is the modified package set and super is the package set before this modification.
<srhb>
rawtaz: For instance, `self: super: { myHello = super.hello; }` is an overlay that adds nixpkgs.myHello as an alias for nixpkgs.hello
<srhb>
rawtaz: They're documented in the nixpkgs manual.
<srhb>
rawtaz: (overrideAttrs is the sort of thing you'd use _in_ an overlay to eg. override the name or source of some derivation)
Neo-- has quit [Ping timeout: 252 seconds]
<rawtaz>
i see
counting1ort is now known as countingsort
patrl has quit [Ping timeout: 246 seconds]
<cyris212>
Does someone have a working nix expression that installs emacs's pdf-tools?
<countingsort>
if my computer insists on fdck-ing the /boot partition and that just doesnt end (i suppose it shouldve finished in an hour) and i dont have anything to liveboot / old profiles to boot into, im boned for now right?
OberstKrueger has joined #nixos
<sphalerite>
countingsort: not entirely, you might be able to fix stuff from the initramfs shell
<sphalerite>
countingsort: if you're using grub or systemd-boot, you can edit the kernel command line by pressing E on the menu entry
jasongrossman has joined #nixos
<sphalerite>
countingsort: then if you add boot.debug1devices you should get a shell before the filesystems are mounted, where you might be able to fix whatever's wrong
tzemanovic has quit [Ping timeout: 240 seconds]
Kelppo has joined #nixos
<sphalerite>
lukego: I've had a look; 16.03 has kernel 4.4.31, and the latest 4.4 is 4.4.155. There are 55 commits mentioning CVE IDs between those two versions. It's a rather primitive way of checking, but still, I'd consider that a cause for concern 😅
lukego has quit [Ping timeout: 252 seconds]
<sphalerite>
oh no, I was too slow to save him!
<countingsort>
sphalerite: how do i add boot.debug1devices? or rather where?
<sphalerite>
countingsort: on the kernel command line
<countingsort>
sphalerite: to which option though? or just append it?
<sphalerite>
yeah just add it at the end
tzemanovic has joined #nixos
<countingsort>
added fsck.mode=skip to get into a decent environment and fix from there. got lightdm (which i switched away from 2 weeks ago) and xterm....
tzemanovic has quit [Ping timeout: 240 seconds]
orivej has joined #nixos
geier1993 has joined #nixos
<geier1993>
hey there, can anyone help me with CUPS, adding a local PPD file. Using the webinterface I always get the error "Unable to copy PPD file. Read-only file system"
<{^_^}>
[nixpkgs] @dotlambda merged pull request #45644 → pythonPackages.rlp: 1.0.1 -> 1.0.2 (with some new deps) → https://git.io/fAmV7
<mpu>
can one make a derivation for the head of a git repo?
<Dezgeg>
I wonder if your fsck stuff actually works and doesn't get hung when trying to ask the user some questions
<Dezgeg>
s/your/our/
<Dezgeg>
because I have seen something similar
<countingsort>
sphalerite: managed to boot into my setup, now fixing it. thanks
<sphalerite>
mpu: not really, since it varies over time
<sphalerite>
mpu: you can use builtins.fetchgit to nondeterministically fetch a git repo, but it won't be a derivation (though you can use it in many of the same ways)
iyzsong has joined #nixos
<mpu>
sphalerite, thanks; another question: where is the "version" business documented (e.g., it looks like what is after the - in a derivation name is treated specially).
<sphalerite>
oh hm that doesn't mention builtins.fetchgit/builtins.fetchtarball either yet
<sphalerite>
mpu: updating that wiki page just now :)
<{^_^}>
[nixpkgs] @globin pushed to master « acme module: fix self-signed cert with openssl 1.1 »: https://git.io/fAoH5
Dedalo has joined #nixos
<{^_^}>
[nixops] @Chakerbh opened pull request #1002 → Make sure that the machine is UP before trying to destroy it in GCP → https://git.io/fAoHd
thekolb has joined #nixos
<{^_^}>
[nixpkgs] @globin pushed to release-18.09 « acme module: fix self-signed cert with openssl 1.1 »: https://git.io/fAoHb
<thekolb>
I can ssh into a box just fine (pubkey auth) but nixops says root@<ip>: Permission denied (publickey,password,keyboard-interactive)., what could I be missing?
tzemanovic has joined #nixos
Kelppo has quit []
<sphalerite>
mpu: I've updated the page, hopefully this is enough to give you an idea
<mpu>
great, thanks!
<thekolb>
How do I find out which key nixops tries to use?
<Purple-mx>
can we use a precompiled nix tree? avoid parsing
tzemanovic has quit [Remote host closed the connection]
<sphalerite>
^ just updated by me to mention builtins.fetchTarball and builtins.fetchGit :)
<aminechikhaoui>
teto: nothing automatic from configuration.nix ?
lopsided98 has joined #nixos
<hyper_ch2>
what a sad day the EU parliament gave green light to upload filters
<sphalerite>
aminechikhaoui: hm as far as I understand it should start automatically if it's in systemPackages
<sphalerite>
aminechikhaoui: which desktop environment are you using?
<sphalerite>
hyper_ch2: we have #nixos-chat for off-topic. You say you don't believe in off-topic channels, then don't discuss off-topic at all please :)
hxrts has joined #nixos
<hyper_ch2>
sphalerite: channel topic doesn't say anything about offtopic in here
npmccallum has joined #nixos
<sphalerite>
hyper_ch2: it lists #nixos-chat
<hyper_ch2>
so? it's custom that channels do set in the channel topic if offtopic should be somewhere else
<{^_^}>
[nixpkgs] @globin pushed to master « grafana module: allow path for extraConfig vals »: https://git.io/fAoFc
<{^_^}>
[nixpkgs] @globin pushed to release-18.09 « grafana module: allow path for extraConfig vals »: https://git.io/fAoF8
<sphalerite>
gchristensen: nixcon cfp is no longer open, could you maybe adjust the channel topic? And maybe in a way to convince hyper_ch2 to use #nixos-chat too, if you want ;)
<Myrl-saki>
lol
<srhb>
sphalerite: I think he's still on vacation :)
<sphalerite>
oh he is?
<sphalerite>
I missed that
tzemanovic has joined #nixos
tzemanovic has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @xeji pushed commit from @teto to staging « iproute: add $dev output (#46558) »: https://git.io/fAoxT
<srhb>
So now you see a problem
<srhb>
You want to be using overrideAttrs, but then, you're destroying the existing unpackPhase
<srhb>
Meaning the src tarball won't ever be unpacked and the build will fail
<srhb>
tobiasBora: Makes sense?
<tobiasBora>
srhb: yes that was another problem, I was not sure how to "concat" both phases
<srhb>
tobiasBora: foo.overrideAttrs (oldAttrs: { attribute = oldAttrs.attribute + "newvalue" ; } but I would suggest simply using a different phase if you go down that route
<srhb>
eg. postInstall
<srhb>
Ah, that's already taken.
<srhb>
:-)
hotfuzz_ has quit [Ping timeout: 252 seconds]
<tobiasBora>
srhb: ahah ^^ Do I need to add an additional new line then?
<srhb>
You can just concat as you were suggesting
<srhb>
It's good exercise at least :-)
tzemanovic has joined #nixos
<srhb>
I would also suggest trying the other approach where you simply wrap the existing emacs derivation, but you can try that after you get this to work ;-)
lopsided98 has quit [Ping timeout: 264 seconds]
<tobiasBora>
srhb: yes, it will be my next exercice ;)
lopsided98 has joined #nixos
<tobiasBora>
ahah, it's funny, unpack phase in fact does not exist ^^
<srhb>
tobiasBora: In which case it uses the generic builder's default unpackPhase.
<tobiasBora>
I can do something like (oldAttrs.unpackPhase or "")
<srhb>
Yeah, but it won't do what you expect..
<tobiasBora>
oh...
<tobiasBora>
so how could I refer to default unpackPhase?
<tobiasBora>
something like (oldAttrs.unpackPhase or default.unpackPhase)
<srhb>
tobiasBora: It's a bash function from stdenv, you can call unpackPhase yourself if you really want to. But instead I suggest using postUnpack or something
<srhb>
tobiasBora: search for unpackPhase in pkgs/stdenv/generic/setup.sh
<tobiasBora>
srhb: postUnpack does not have default value?
<srhb>
tobiasBora: iirc none of the post/pre-hooks do
<tobiasBora>
oh, so it's an actuall bash function. i can just write "unpackPhase()" to call it right? (even if I think I'll go for postUnpack)
<srhb>
tobiasBora: If you look at setup.sh, you'll see that each fooPhase has a preFoo and a postFoo at the beginning and end.
<srhb>
tobiasBora: Just unpackPhase :)
<sphalerit>
samueldr: was it you who had the manuals for all the channels up on your website?
<tobiasBora>
srhb: ok thanks. So I tried postUnpack, and I get an error "cp: cannot create regular file '/nix/store/5zh7hgh2f09zfx591i01y75r41a3lgn8-emacs-25.3/share/applications/': No such file or directory". So i tried to add on the first line "mkdir -p ${desktopItem}/share/applications/", but same error
<samueldr>
sphalerit: nope, no manual shenanigans on my site
<mpu>
Hmm, in mkDerivation, if I set: unpackPhase = ""; it does not seem to prevent the default from happening.
<samueldr>
uh, no nix, nixos or nixpkgs manual shenanigans
<srhb>
tobiasBora: Did you mean to mkdir -p $out/... ?
<samueldr>
but yes, sphalerit, I had an experiment with manpages
<srhb>
tobiasBora: (I think it's weird to do this so early in the build! Normally installPhase is the first to create $out)
tzemanovic has quit [Remote host closed the connection]
<Myrl-saki>
samueldr: Rust made me give up on Nix on my tablet. :(
<samueldr>
Myrl-saki: because of the compile times?
<Myrl-saki>
samueldr: Pretty much, yeah.
<tobiasBora>
srhb: With mkdir -p ... it seems to work, but it tries to recompile emacs completely... Which I'd like to avoid of course
<tobiasBora>
so I can try postInstall
<srhb>
tobiasBora: And this is where the second suggestion, simply wrapping the existing derivation comes in
<srhb>
tobiasBora: That won't help.
<tobiasBora>
ok
<srhb>
tobiasBora: With overrideAttrs or override you are changing the existing emacs' hash, so it must be rebuilt.
<tobiasBora>
i see
<srhb>
tobiasBora: Which is the downside of this approach. :)
<mpu>
When I set, eg buildPhase, do I need to use writeScript or can I just dump the commands directly?
<{^_^}>
[nixpkgs] @jtojnar pushed 23 commits to gnome-3.30: https://git.io/fAohP
<srhb>
mpu: Just a string.
<tobiasBora>
srhb: so now I understand how this first method works, I can go for the second method
<mpu>
I get a bunch of warnings when sourcing stdenv/setup; is that to be expected?
<mpu>
E.g., line 312: pkgBuildAccumVars: readonly variable
<tobiasBora>
srhb: but I'm not sure to see how I should wrap the existing derivation
orivej has quit [Ping timeout: 252 seconds]
tzemanovic has joined #nixos
jperras has joined #nixos
<srhb>
tobiasBora: One way would be to make a new derivation that symlinks everything from ${emacs}/* into $out/ and adds your extra share stuff.
<srhb>
tobiasBora: I think that method is fairly simple to understand :)
OberstKrueger has left #nixos [#nixos]
vaibhavsagar has joined #nixos
<clever>
srhb: buildenv should be able to do that
<srhb>
yup. Trying to get there from first principles ^^
jperras has quit [Ping timeout: 252 seconds]
<srhb>
tobiasBora: But if you want to skip ahead, ^
<exarkun1>
Can the rust support build libraries (crates?)? Seems like it must but when I try building this rust library (blake2-rfc) I get a package nothing nothing (except an empty bin/) in it.
<symphorien>
the point is: why do you want to build this crate ? it would be easier (if less computation-efficient) to just use buildRustPackage on all dependent executables
iyzsong has quit [Ping timeout: 252 seconds]
<exarkun1>
I don't actually know much about Rust. blake2-rfc is a dependency of another Rust library, libzcashrust, which is a dependency of Zcash. What I actually want to do is build a recent version of Zcash. :) That depends on a recent version of libzcashrust which depends on a git+https version of blake2-rfc.
<symphorien>
just use buildRustPackage on zcash
<exarkun1>
So I thought I was doing something like what you suggest ... use buildRustPackage to build blake2-rfc so that I could pass it as a buildInput (I guess?) to a buildRustPackage for libzcashrust
<exarkun1>
Sure. But "just use" it with what arguments?
<symphorien>
hum, just to be sure: is zcash a rust app ?
<exarkun1>
Actually, no. I forgot about that. Instead, libzcashrust builds a .a for zcash to link against.
<symphorien>
either use carnix (I have no idea how it works, but I suspect it only builds .rlibs) or patch buildRustPackage to install .so and .a as well
<symphorien>
in any case, if you build a crate with buildRustPackage you never need to package dependencies of this crate which are also crates
<exarkun1>
_Also_ kind of separately from that, when I try to patch the libzcashrust dependencies w/ `cargoPatches`, the patch is not actually applied and I see no explanation about why.
<octe>
with home-manger, is it possible to mix release-channel packages and unstable packages?
<octe>
home-manager*
<bitmapper[m]>
Hello
carlosdagos has quit [Quit: Connection closed for inactivity]
<bennofs>
exarkun1: you shouldn't need to package blake2-rfc separately
<bennofs>
we don't have a nix expression per rust crate. instead, each rust "package" (application or thing that installs a .so/.a) builds all the dependencies itself during the build phase
<bennofs>
exarkun1: rust does not really have the concept of "locally installed dependencies" for single crates
<bennofs>
there is no way to have it use prebuilt crates as deps without a lot of hacks
fendor has quit [Ping timeout: 250 seconds]
<exarkun1>
Okay. So I just have to get both source packages near each other somehow, patch libzcashrust to point at the local blake2-rfc source, and then build libzcashrust?
<exarkun1>
bennofs: I guess we can pick a channel, eh
<bennofs>
:D
<bennofs>
exarkun1: i am not 100% up to date on rust support on nixpkgs, but it surprises me: does it really not support git dependencies?
jperras has joined #nixos
<andi->
octe: yes like anywhere else. You just have to import those packages.
<octe>
but i still get "error: file 'nixos-unstable' was not found in the Nix search path" when i try to import it in home.nix
<exarkun1>
bennofs: I don't think it does. Someone pointed me at an open PR yesterday that might fix this. I don't understand Rust or Nix well enough to understand the PR though.
<octe>
unstable = import <nixos-unstable> {};
<octe>
is this not how i import it?
<LnL>
bennofs: exarkun1: cargo --frozen doesn't support git IIRC
<bennofs>
LnL: o.o did not know about cargo --frozen. I only know that cargo-vendor should support git deps
<exarkun1>
My build currently fails for trying to update Cargo.lock when looking at the git+https dep
<LnL>
bennofs: oh right there's a pr open to use that I think, that might fix this
<octe>
oh, i needed to add unstable channel as root
<exarkun1>
So .. how do I get two sources checked out for one build? Is this what `srcs` is for?
<{^_^}>
#46362 (by symphorien, 4 days ago, open): Non broken fetchcargo
<bennofs>
exarkun1: you can just add an attribute to your derivation (like blake2src = fetchgit ...) and then access it as $blake2src in build scripts
contrapumpkin has quit [Quit: My MacBook Pro has gone to sleep. ZZZzzz…]
<LnL>
symphorien: yeah, that's what I was thinking of
<bennofs>
I was just surprised since this was already in discussion like a year ago or something :D
<exarkun1>
bennofs: great, thanks
Henson has joined #nixos
copumpkin has joined #nixos
<Henson>
is there a way to get nix-build to output more detailed build information? Even with a bunch of "-v" options, it just says "builder for ... failed with exit code 1" but doesn't provide any information as to why it's failing. Is there some stdout and stderr log file stored somewhere?
<aminechikhaoui>
Henson: nix log <drv>
<clever>
Henson: if -Q is off, then it should always give stdout and stderr
<aminechikhaoui>
oh you're using nix-build not `nix build` sorry
<bennofs>
symphorien: why do we need to normalize the output of cargo-vendor?
<symphorien>
because the order of the sources is not well defined, nor the spacing, etc.
<bennofs>
symphorien: would it be possible to generate the vendor configuration at build time (with a dry-run-like option to cargo-vendor that only produces the config but does not download anything?)
<symphorien>
and it still must be a fixed output derivation
<bennofs>
then the vendor config would not even be part of the fixed output
Kelppo has quit [Ping timeout: 240 seconds]
<bennofs>
not sure if we can get cargo to generate a vendor config without network access?
<symphorien>
I tried
<symphorien>
it does not work
work_ has joined #nixos
<bennofs>
hmm ok
<bennofs>
is that a fundamental issue or could it be implemented in cargo-vendor in the future?
<symphorien>
no idea
<symphorien>
anyway, current PR works fine, is moderately hacky, and reimplementing what cargo vendor already does/forking cargo vendor to so this would be much more work
<symphorien>
and as you pointed out, a fix for this issue is long overdue.
<bennofs>
symphorien: yeah I think we should do your PR for now
<bennofs>
just wondering if there is a solution that would perhaps require less maintenance on our part in the long term
<bennofs>
if we can get the feature to produce a vendor config without internet into upstream then it would make the implementation in nixpkgs much simpler
<bennofs>
and we wouldn't depend on the vendor config format
<bennofs>
but that might be impossible with the way cargo works
<symphorien>
honnestly, I doubt there is more risk of the format changing in the config than in the vendored output
Conqueror has joined #nixos
orivej has joined #nixos
<exarkun1>
bennofs: Do you know if a source replacement saves me from having to fix up Cargo.lock?
<symphorien>
I mean, we go great lengths to normalise a 10 lines toml file when we trust the whole vendor dir content
humanoyd has joined #nixos
<bennofs>
exarkun1: in theory yes. source replacement might require that you create some cargo-internal files though, not exactly sure about that
<bennofs>
exarkun1: i think there needs to be a checksum file in the replacement source, but you should be able to generate that with cargo-vendor and just include it in the package
<bennofs>
it's not nice but it should work right now
wpcarro has joined #nixos
patrl has joined #nixos
<bennofs>
exarkun1: changing Cargo.lock manually is probably a pain. it includes transitive dependencies (it's a representation of the full dep graph)
kiloreux has quit [Ping timeout: 244 seconds]
<bennofs>
plus it also has checksums for everything, not sure if they change if you use a local checkout vs git url
<exarkun1>
bennofs: I agree (it's a pain).
spear2 has quit [Remote host closed the connection]
spear2 has joined #nixos
<bennofs>
exarkun1: i mean in theory I think the only thing missing from the current rust support to work with git repos is a proper ".cargo/config" for the vendored sources
<bennofs>
exarkun1: so an easy solution should be to just manually generate a proper vendoring .cargo/config and patch that before building
<bennofs>
that wouldn't require any Cargo.toml/Cargo.lock patching
<exarkun1>
It might be a simple solution but I'm not exactly a Rust programmer so I'm not sure it's easy. :)
dbmikus has joined #nixos
<exarkun1>
Are there docs about ".cargo/config"?
<bennofs>
exarkun1: just run cargo-vendor manually and copy the config it generates to .cargo/config
sbdchd has joined #nixos
<exarkun1>
"just" ;)
<exarkun1>
cargo-vendor doesn't appear to be packaged for my OS ... so I get to figure out how to build it, I guess.
<exarkun1>
and then I'm about 11 yaks away from my real goal
<bennofs>
oh. i can generate one for you if you want :)
patrl has quit [Ping timeout: 240 seconds]
<exarkun1>
I feel like it would be good if I knew what were going on. But I don't feel that strongly enough to say no to that offer.
<clever>
sphalerite: this allows you to take a directory of .nar files, and just mount it to /nix/store
<exarkun1>
can't cargo install cargo-vendor, cargo-vendor doesn't build with whatever version of rust I have
<clever>
sphalerite: and once you have that, you just need a 404 handler that downloads things on-demand, and you have what you linked above, with caching
<sphalerite>
clever: niksnut mentioned that nix mount-store would be useful for getting the netboot images smaller, by only fetching stuff on demand. But without caching I suspect it would be unusably slow
<sphalerite>
clever: pulling haskell stuff into the netboot image will not make it smaller >_<
<clever>
sphalerite: precompiled haskell, not a full compiler
<clever>
sphalerite: its only 1 or 2mb compiled
<sphalerite>
clever: and has no references to ghc?
<clever>
thats trivial to do
<clever>
static haskell linking gets rid of ghc references
<sphalerite>
great
<{^_^}>
[nixpkgs] @Ericson2314 opened pull request #46574 → elf-header: Init at <libc version> → https://git.io/fAKmv
<bennofs>
exarkun1: here's how the config looks for me: http://ix.io/1mya
<Myrl-saki>
What happens if you use hint and getContents? :P
<bennofs>
exarkun1: you probably need to replace the directory at the very bottom with whatever location nix uses for the vendored sources
<clever>
sphalerite: the main reason i wrote narfuse, was actually for ipfs
patrl has joined #nixos
<clever>
sphalerite: if you keep the storepaths as nar files, then you can share those over ipfs or bittorrent, without having to pay double on the disk, to hold the unpacked and packed
<exarkun1>
bennofs: And the idea is to patch this in to the librustzcash source (.cargo/config) right?
<bennofs>
exarkun1: yeah, patch it before the build
<clever>
sphalerite: the closure for narfuse is 63mb right now
<clever>
sphalerite: glibc is to blame
<bennofs>
exarkun1: in fact, it looks like that's exactly what the parity beta derivation in nixpkgs does if you
<bennofs>
're looking for an example
<exarkun1>
bennofs: And also drop the blake2-rfc source into .../vendor? Or just change that path to point at the parent of the blake2-rfc checkout?
Alling has joined #nixos
<Alling>
How can I make printing possible on NixOS? I have a Dell 3115cn and Gnome 3.
<bennofs>
exarkun1: I don't think you even need any of that. the vendor directory should already contain the sources, it just doesn't generate the correct config for cargo to find it
<Alling>
I'm almost sure it worked before, but I have reinstalled NixOS now.
<bennofs>
exarkun1: if you look at the build log, there should be some references /nix/store/...-vendor and if you check what's in there blake2 should be already contained
<Alling>
I have enabled services.printing, but I can't add the printer in the GUI. I get a notification about "a printer needs drivers" or something, but when I click "Find in software", nothing happens.
<exarkun1>
bennofs: Why would the vendor directory contain the source if the upstream package is declaring a dependency on a git repo? In any case, the upstream package doesn't have blake2-rfc vendored (it has nothing vendored, in fact).
<bennofs>
s/nix vendors/the rust support in nixpkgs
<sphalerite>
clever: yeah
orivej has quit [Ping timeout: 246 seconds]
<Alling>
symphorien: I tried to follow the Printing page in the wiki, but I didn't know what do to. Should that driver work even though I have a different printer?
<clever>
sphalerite: i didnt think of that back then, but it could be along the lines of, when it fails to find a path, check the binary caches and run `nix-store -r` on it, and the narfuse backend in nix would just download a .nar file, and register it as being valid
<symphorien>
Alling: I doubt that
<clever>
sphalerite: and narfuse is then responsible for the "unpacking"
<symphorien>
but you lose nothing trying
<symphorien>
if not, you will have to package the driver dell provides yourself
<Alling>
symphorien: I only get the "Failed to add printer" message. Tried gutenprint and gutenprintBin. I'm gonna try rebooting.
<sphalerite>
clever: narfuse backend in nix?
<symphorien>
Alling: I don't know about gnome. cups has a web ui, maybe it will give more details
hyper_ch2 has quit [Quit: Page closed]
<Alling>
symphorien: I don't know how to package a driver myself. And I'm almost sure it used to work before reinstalling NixOS. :s
<Alling>
Yes, maybe it will! I'll check that out as well.
<clever>
sphalerite: nix would need to be modified, to not unpack the .nar when downloading, and just dump it into a directory, and notify narfuse to rescan the dir
<clever>
sphalerite: and then nix downloading things from the binary cache, would just dump .nar's into a dir, and narfuse would make it look like it was unpacked
<sphalerite>
clever: oh right, you're still thinking nix should download the nars
<clever>
yep
<exarkun1>
bennofs: Okay ... so I just need to find the vendor path.
<clever>
sphalerite: and nix would flag things as being valid in db.sqlite, after it finishes /nix/nars/hash-hello-1.2.3.nar
<sphalerite>
clever: I was thinking you could mount cache.nixos.org with httpfs or whatever
<clever>
sphalerite: then narfuse mounts /nix/nars to /nix/store and it looks like a normal store
<Alling>
symphorien: The web UI worked! Thank you so much! <3
<clever>
sphalerite: narfuse could have an option to also try to download everything from a cache upon failing to find it locally
<bennofs>
exarkun1: alternatively, you can also just append the source for the github repo to the existing .cargo/config (nixpkgs buildRustPackage generates a .cargo/config in the postUnpack phase, so if you append it in prePatch which runs after that it should come out correctly)
<sphalerite>
clever: well the key difference between narfuse and nix mount-store is that the latter can mount any store, including remote ones
<symphorien>
Alling: usually I use system-config-printer, it is more aesthetic and may still work
<clever>
sphalerite: narfuse would rely on nix to fetch the .nar via some method (including a remote store), and drop the .nar into a local dir
<clever>
sphalerite: so the "design limit" of narfuse, winds up giving you caching
<betaboon>
hi #nixos, i am currently looking into continuous build/integration/delivery and just looking into hydra. are ppl using something else than hydra ?
<betaboon>
it looks very new (judging only by date/commit#)
<manveru>
betaboon: it is :)
ihar has joined #nixos
<manveru>
i need to release more info about it though, plus the service module
<sphalerite>
but it is in production use, right manveru ? :D
<manveru>
atm it's mostly for trial in our company and has like 5% of the functionality of hydra :P
<sphalerite>
but is it supposed to have more than 5% of the functionality of hydra? :D
<manveru>
not really :)
mayhewluke has joined #nixos
<manveru>
anyway, it's for when all you want to do is run nix builds from github PRs
<sphalerite>
clever: disadvatnage of the copy-the-whole-nar approach is that you won't necessarily need all the files in a store path
<avn>
clever: by the way, I thinking in opposite direction -- are possible to re-use files existing in /nix/store/.links when downloading new packages
<betaboon>
just to give some context to what I'm doing and what i (ideally) want to achieve: we run a microservice-system. service-per-repo. a repo which holds a nixpkgs-overlay, and a repo for nixops-deployments. idealy i want to have automatic builds on PRs for all the service-repos (with pinned nixpkgs and pinned nixpkg-overlay) (step one). lateron i would love to have automatic deployments (just to a
<betaboon>
staging/development environment) on PRs in the nixops-deployments-repo
<clever>
sphalerite: due to the way the nar fileformat works, you need to download every file before X to get the position of X in the stream
reinzelmann has quit [Quit: Leaving]
<sphalerite>
clever: oh ok, never mind then :p
<avn>
betaboon: in this layout I found that I miss tool which allow me grab all (optionally signed) branches from gits, and deploy all services in single command. (not for final deployment, but for development)
<clever>
sphalerite: its just a giant stream of length-prefixed blobs, some of them are filenames, some are file contents
<bennofs>
clever: is this also true if you take .ls listings into account? I thought they have file offsets?
<clever>
bennofs: the compression also comes into play
<bennofs>
I agree compression makes it hard
<clever>
you dont know what offset in the compressed stream, matches to a given offset in the uncompressed stream
<bennofs>
probably impossible with xz?
<betaboon>
avn: i have that figured out. i deploy the whole system (around 30 services) including all aws-resources with a single command from pinned commits of nixpkgs and my overlay
<clever>
which is why narfuse doesnt work on .nar.xz files
<avn>
betaboon: I deploying only 3-4 components (+ system itself) on localhost, and I tired of update each hash manually ;)
<betaboon>
avn: i maintain a nix-expression with all the package-definitions which take src from a seperate expression. that seperate expression contains all the fetchFromGitHub expressions with hashes etc. that file is generate with a script i wrote using nix-prefetch
<sphalerite>
clever: hm this feels like a massive bikeshed tbh for nixos-channel-scripts#21
<sphalerite>
ugh, it seems the firmware of these machines won't netboot if they're powered on by WoL? >_>
<sphalerite>
even if it's configured as the first entry in the boot order and netboots if they're powered on by the power button
<clever>
sphalerite: minor correction to one of your comments, the rootfs/nixstore, is in the squashfs(compressed) which is in the initrd(also compressed)
<avn>
betaboon: yep, I have some plan to make a tool, to build separate expression file for sources (with ability to use even just-in-time snapshots) and pass it to following nix-build automatically.
<clever>
sphalerite: so it first unpacks the initrd to ram, including root.squashfs, then it mounts that squashfs, and uncompresses its files on-the-fly
grp has joined #nixos
spear2 has quit [Remote host closed the connection]
spear2 has joined #nixos
dbmikus has quit [Ping timeout: 240 seconds]
erasmas has joined #nixos
<sphalerite>
clever: you wouldn't happen to know if this is in any way expected behaviour, for a machine to netboot when powered on with the button and to boot from hard disk when WoLed?
<sphalerite>
the bios config is stock, except that the boot order has been changed to have netboot first
<clever>
sphalerite: i think that would mostly be up to the bios config
<clever>
sphalerite: but i have setup stuff that would allow you to remotely customize that
<sphalerite>
clever: how is that supposed to work?
<sphalerite>
oh I guess you'd have to put ipxe on the hard drives
<sphalerite>
I don't want to do that :/
<clever>
i was thinking put ipxe on the network, and have no local MBR
<sphalerite>
oh hm I wonder what happens if I zero the MBR
<clever>
or remove legacy from the boot menu
<clever>
sanboot --no-describe --drive 0x80
<clever>
sphalerite: this ipxe command, will boot the local hdd, even if ipxe was netboot'd
<clever>
and if your ipxe script comes from a server-side program, it can dynamically change the ipxe commands it serves
<sphalerite>
well if the boot order is ignored for wol that won't help I'm guessing
<sphalerite>
what do you mean by removing legacy?
<clever>
take the local hdd out of the boot order
<clever>
and/or dig around in the bios config near network or boot
<sphalerite>
oh pff
<sphalerite>
it has a "Remote wakeup boot source" option that only allows hard drive or network server
<clever>
ah
Alling has quit [Ping timeout: 252 seconds]
Ariakenom has joined #nixos
Boomerang has quit [Quit: WeeChat 1.9.1]
<sphalerite>
damn. Now I could really use a USB stick to save the config to, but I don't have one >_>
<sphalerite>
and modern android phones only do MTP afaik, not mass storage, so I can't use that either…
<clever>
sphalerite: i have an android app for mass-storage, but it needs root
<clever>
i believe it maps to a .img on the sdcard, not the sd itself
<betaboon>
is anyone using hydra and give me some pointers? do i create a project for each repo that i want to build from? or can i create a project that has jobs that build from several repositories? why do i have to define an input in the project aswell as the job?
<sphalerite>
clever: oh that sounds good. Although I don't seem to have a micro usb cable with me either >_>
<clever>
betaboon: you can also create one jobset for each repo, and put them all in the same project
<rawtaz>
ehm. what's the difference between samba and samba4Full? not finding a relevant "full" in the expression
<Dezgeg>
a project doesn't need inputs
<Dezgeg>
only jobsets
<clever>
betaboon: the inputs in the project, are for declarative jobsets, which is where hydra creates the jobsets automatically based on json and nix files
waleee has joined #nixos
<betaboon>
clever: so in that case i just select "Boolean" as "Declarative input type" ?
<sphalerite>
clever: any idea if the BIOS NVRAM might be accessible from linux as well? :D
<clever>
sphalerite: sometimes, depends on the bios
<betaboon>
clever: would you suggest using declarative jobsets over imperative ?
<clever>
betaboon: declarative jobsets help a lot, by letting you define the config in your git repo
<clever>
betaboon: and you can easily reconfigure it again later if you loose the hydra
<sphalerite>
clever: it's not a very modern one, pre-EFI
<clever>
sphalerite: if you `morprobe nvram` then you get this chardev
<clever>
sphalerite: with one of my laptops from the 486 era, i was able to remove a bios password by dd'ing into that character device
<clever>
sphalerite: in that case, the bios had 2 passwords, the general config, and a boot pw
<betaboon>
clever: i guess i will just play around with imperative jobsets for now (as i'm just running a local virtualbox before deploying a aws-machine and using it for our whole system) and switch to declarative later on :D
<Dezgeg>
speaking of hydra, what to try when the queue is full but it isn't building anything?
<clever>
the config pw was set, which restricted what i could do
<clever>
but i still had permission to change the boot pw
<clever>
and by diff'ing the nvram, i was able to see how big the hashed pw was, and generally where it was
<sphalerite>
clever: how convenient!
<sphalerite>
clever: whoa it was hashed?
<clever>
Dezgeg: do you have build slaves configured? with the right features?
<clever>
sphalerite: i'm guessing it was hashed, it wasnt cleartext
<Dezgeg>
yes, it build few hundred steps successfully, then stopped
<clever>
sphalerite: i made an educated guess as to where the 2nd password was, dd'ed over a single byte, and then the crc failed, so the bios factory reset itself
<clever>
Dezgeg: find the .drv for the job thats not building, and try running `nix-store -r --dry-run` on it, on the hydra master, what does it output?
<Dezgeg>
on some particular user account?
<sphalerite>
clever: YAY it seems to work! at least dding /dev/zero to one makes it forget all its settings
<clever>
Dezgeg: any user should work
<Dezgeg>
let's see..
<Dezgeg>
it does start building just fine
<clever>
Dezgeg: whats important, is what derivation it started to build first
<clever>
Dezgeg: and what features that needs
<Dezgeg>
it shouldn't be a requiredFeatures thing
patrl has quit [Ping timeout: 240 seconds]
<Dezgeg>
perhaps I just nuke the postgres database and start from scratch :P
<clever>
Dezgeg: try restarting hydra-queue-runner.service first
<clever>
it can sometimes hang
<Dezgeg>
tried that
<clever>
what does the journal for queue-runner say?
<Dezgeg>
eh, now it started doing something
<sphalerite>
Hm, is there some sort of "un-hexdump"?
<sphalerite>
clever: ooh it's even on f-droid! Awesome!
<sphalerite>
Dezgeg: with output from hexdump -C?
<Dezgeg>
journal's just full of 'loading build 62855' stuff
<Dezgeg>
I think xxd has pretty configurable output (and input) formats
<clever>
Dezgeg: when it says that, it will nix-store -qR the drv, look in the store to see what is missing, then look in the binary caches to see what can be downloaded and what has to be compiled
<Dezgeg>
yeah, I have this substitution thing on... maybe I can turn that off
<clever>
sphalerite: i need to root this tablet to get that stuff working
bennofs has quit [Quit: WeeChat 2.0]
xenog has joined #nixos
<Dezgeg>
I would imagine plenty of androids will lack the kernel driver for the mass storage nowadays
<sphalerite>
I hope lineageos does have it :p
<Dezgeg>
ok, I'd imagine they have enough foresight to enable it
xok has joined #nixos
<Dezgeg>
btw I wonder what's the state of the ADB gadget on mainline... would be cool to have that for the various arm boards with hard-to-access serial ports
bennofs has joined #nixos
JonReed has joined #nixos
<samueldr>
sphalerit: drivedroid works with it, so I'm pretty sure it does have it
jperras has quit [Quit: WeeChat 2.2]
<samueldr>
(never heard about usbmountr before today, will be replacing drivedroid with it)
johanot has quit [Quit: leaving]
xok1 has joined #nixos
xok has quit [Read error: Connection reset by peer]
<rawtaz>
where in the Nix manual is the "Nix daemon" explained?
<Dezgeg>
i am pretty sure it compiles things like adb from source
<sphalerite>
Dezgeg: http://android-rebuilds.beuc.net/ is the only place I know of trying to build the SDK from source properly, and it's all using prebuilt compilers and stuff
xok has joined #nixos
<Dezgeg>
yes, it would be with the prebuilt compilers
<LnL>
antoinerg: I think you need to set nixpkgs explicitly, eg. -I nixpkgs=channel:nixos-unstable
<avn>
Thank you folks ;) I actually need only adb/fastboot so curious if it possible to have them w/o pulling all other crap ;)
<floop>
hello all, I want to do a garbage collect, but I don't want to have to rebuild dependencies for my particular project. I do all of my development in a nix shell. Does nix treat all of the dependencies of a shell session as GC roots as long as the nix-shell is running?
<floop>
I'm concerned that if I do a garbage collect that many/all of the dependencies which I built for this shell will be removed and the next time I enter the shell they'll have to be redownloaded or rebuilt
<floop>
In another topic: after upgrading to nixos 18, every time I run a nix command I'm getting warnings about failing to set locale. Does anyone know how to fix this?
<stphrolland>
I want to upgrade to 18.03, but the latest time I read the release not it was not clear what was the correct way to go from 17.09 to 18.03 ? Are there several steps to do, or simply subscribing to the 18.03 is totally sufficient? I don't know what to do with the system.stateVersion for example.
<clever>
stphrolland: stateVersion must not be changed
<clever>
stphrolland: just add the channel as root with nix-channel --add, ensure its listed with the name nixos, and then nixos-rebuild boot --upgrade
semilattice has joined #nixos
sbdchd_ has quit [Ping timeout: 272 seconds]
sbdchd has joined #nixos
mupf has joined #nixos
<mupf>
hello
sbdchd has quit [Remote host closed the connection]
<stphrolland>
clever: thx, my first nixos version upgrade, I was a bit frightened even with the rollbacks possibility
<rawtaz>
oh never mind.
<rawtaz>
i see the expression files are two different files.
<rawtaz>
it would probably make sense to have more specific descriptions for these packages, so one can tell the difference
<tobiasBora>
srhb: Sorry I needed to go and I had no time to notify you. So you can symlink things between derivations? Funny!
<stphrolland>
THe other day I saw a process consuming a lot of CPU% which owner was the "nobody user". I had not configured in my .nix files. Do you know a way to disable this user in nixos? Maybe it is not necessary ?
orivej has quit [Ping timeout: 252 seconds]
<tobiasBora>
I'm very new to derivations so I don't even know what buildenv means ^^' But let's try. So if I understand, I want to do a new derivation that just symlink to emacs. So for what I know, there is "derivation" and "mkDerivation" that adds default values to "derivation". So, as I don't mind default derivation, shouldn't I use "derivation", and just add "emacs" as input?
Neo-- has joined #nixos
winem_ has quit [Ping timeout: 240 seconds]
graphene has quit [Read error: Connection reset by peer]
<sphalerite>
tobiasBora: typically you'll still want to use some of the features in the stdenv, so still mkDerivation — or runCommand, which is a wrapper over stdenv.mkDerivation
sbdchd has quit [Remote host closed the connection]
kenshinC1 has quit [Ping timeout: 246 seconds]
graphene has joined #nixos
Thra11 has joined #nixos
<tobiasBora>
sphalerite: runCommand will be run just once at every "nixos-switch", or it also create like a real "package"?
<sphalerite>
tobiasBora: it creates a package, where the commands are what create the output
<sphalerite>
tobiasBora: so e.g. runCommand "hello" {} "echo hello > $out" will create a store path with the name "hello" and containing the text "hello"
<tobiasBora>
sphalerite: interesting... and can it have also inputs?
<tobiasBora>
it's the {}?
<sphalerite>
tobiasBora: iirc you wanted to create a desktop file? You'll probably want to use makeDesktopItem, a mkDerivation wrapper conceived specifically for that purpose :)
rfold has joined #nixos
<sphalerite>
,find make-desktop-item/default.nix
<{^_^}>
Couldn't find any such files
<vaibhavsagar>
tobiasBora: yes you can put buildInputs in the {}
<sphalerite>
,find make-desktopitem/default.nix
<tobiasBora>
sphalerite: oh, you mean that this is already a wrapper?
graphene has quit [Remote host closed the connection]
<tobiasBora>
I was trying to use it inside a derivation
<vaibhavsagar>
tobiasBora: you can change the buildPhase inside a derivation to do whatever you want to the derivation output
graphene has joined #nixos
<sphalerite>
^ so you can put (pkgs.makeDesktopItem {name = "emacs"; exec = "${pkgs.emacs}/bin/emacs"; desktopName = "Fancy Emacs";}) in your systemPackages for instance
<vaibhavsagar>
but you can't make changes to anything outside the output path
<tobiasBora>
vaibhavsagar: that's what I did before, but it makes it recompile everything
<vaibhavsagar>
yes, that sounds reasonable, what did you expect?
<sphalerite>
tobiasBora: for a thorough introduction, I recommend the nix pills
<rawtaz>
ok guys; best practice question: for some applications, e.g. falkon, that isnt a service, i just add them to the pkgs list to have them installed. but programs that are services, e.g. gnome3.tracker, i can skip adding to the pkgs list and instead just add the gnome3.tracker.enable=true to have it automatically installed. which do you prefer, to mix and match these types of install ways, or do you always add to pkgs even if you also have the .enable dire
<tobiasBora>
sphalerite: I will definitely go through the nix pills
<ldlework>
rawtaz: I always use the nixos modules when available
<tobiasBora>
So I tried your method, and indeed it creates a new derivation with the file: /nix/store/8ysaqhx336fnymp698b1a2p3gd0xd8y5-EmacsEditor.desktop/share/applications/EmacsEditor.desktop
<tobiasBora>
but i don't know why, it does not appear in KDE stuff
<rawtaz>
ldlework: so when there's e.g. .enable which will auto install the application, you only add that, and for those that doesnt have this, you add to pkgs? in essence, you mix.
<rawtaz>
ldlework: i have no problem with that per se, but it becomes a matter of not having one single list of "these things are installed"
<rawtaz>
as it's spread out in at least two places
<ldlework>
rawtaz: I encapsulate ALL my configuration in custom modules
<ldlework>
So in order to do anything I have to enable my own config.mine.* module options
<ldlework>
So I have a full list of things installed
<rawtaz>
oh, ok.
<rawtaz>
ldlework: inside your modules, if it's an app that also has a service, do you just add the .enable or also the alias to pkgs ?
work_ has quit [Quit: Connection closed for inactivity]
<avn>
tobiasBora: you need to add it systemEnvironment (and I idk what need to do, to KDE re-read list of desktops)
<ldlework>
rawtaz: I always use the nixos modules if I can
<ldlework>
(from within my own modules)
<tobiasBora>
avn: I don't get it, I should put pkgs.makeDesktopItem in both systemPackage and systemEnvironment?
<rawtaz>
ldlework: im sorry but im not sure how that answers my question. for e.g. gnome3.tracker, would you add gnome3.tracker to the pkgs list alongside the services.gnome3.tracker.enable=true , or would you only add the latter?
<vaibhavsagar>
the service configuration already adds the package to environment.systemPackages
<ldlework>
only the latter, because the latter adds the package
<ldlework>
rawtaz: ^
<rawtaz>
ldlework: okay, great. thats what i asked all along :P
<{^_^}>
#46509 (by kalbasit, 1 day ago, open): bazel-watcher: init at 4d5928e
<vaibhavsagar>
it'll tell you everything that's installed and every file that is involved
<avn>
tobiasBora: well. My knowledge ends after desktop file hit /run/current-system. I don't know anything about kde
rihards has joined #nixos
<tobiasBora>
avn: ok thank you for your help, will try to make the two files looks the same (with actual emacs), and see where things get wrong ;)
<goibhniu>
tobiasBora: does running `kbuildsycoca5` help at all?
georges-duperon has joined #nixos
<goibhniu>
(it may also fail, complaining about stuff in ~/.cache)
Mateon1 has quit [Ping timeout: 240 seconds]
Mateon2 has joined #nixos
stphrolland has quit [Quit: leaving]
<tokudan[m]>
is there a way to opt out of these stupid badges in discourse?
doyougnu has joined #nixos
Mateon2 is now known as Mateon1
<rawtaz>
vaibhavsagar: thanks, tried it
<tobiasBora>
goibhniu: great, it appears! Thanks for the `kbuildsycoca5` trick. Any idea why restarting the session was not enough?
<rawtaz>
i have a bigger problem though; i've added services.gnome3.tracker.enable = true; and services.gnome3.tracker-miners.enable = true; and even tried adding gnome3.tracker-miners to systemPackages as well, but no matter what i do tracker gets installed but there are seemingly no miners, as per `tracker daemon --list-miners-available` and `tracker status` reporting no files indexed even though i also tried `tracker daemon -s`. any idea why miners arent
<rawtaz>
or arent even available
<goibhniu>
tobiasBora: great! I think that's supposed to be run when plasma is started ... was there an error when you ran it manually?
<tobiasBora>
goibhniu: yes, but about another service: The desktop entry file "/run/current-system/sw/share/applications/org.kde.systemmonitor.desktop" has Type= "Application" but no Exec line
<goibhniu>
hm
<vaibhavsagar>
rawtaz: have you run `nixos-rebuild switch`? If so, look at the service configuration for those services in nixpkgs to see what is actually happening
Izorkin has quit [Read error: No route to host]
<vaibhavsagar>
or you can do `sudo journalctl -u <service-name>.service` and scroll to the bottom to see if there are any errors preventing it from starting
Izorkin has joined #nixos
<rawtaz>
vaibhavsagar: yes, totally. i edit the configuration and then run that, i even tried rebooting just for kicks
<grp>
clever: ping
<clever>
grp: pong
<rawtaz>
checking the expressions again
<tobiasBora>
By the way, is anyone here using emacs? I just hate the nix-mode on emacs, can I somehow make it usable? The indentation is just nonsense for me.
<vaibhavsagar>
rawtaz: I would recommend using journalctl and systemctl to check that the services started correctly
<avn>
tobiasBora: goibhniu: actually I feel kbuildsycoca5 should be invoked from user's systemd, and this action should be triggered by `switch`
<rawtaz>
`systemctl --all|grep track` yields nothing either
<avn>
As I grepped -- it should placed to activation script, but idk why it haven't effect
<vaibhavsagar>
rawtaz: is tracker-miners in the $PATH?
<goibhniu>
avn that'd be nice
sbdchd_ has joined #nixos
<rawtaz>
vaibhavsagar: no
<tobiasBora>
line 24: why does it indend left this line, and all the subsequent lines? Line 51, why isn't the "(epkgs:" indented more on the right?
_ris has joined #nixos
<grp>
clever: I'm trying to make a generic builder for a scripting-language's libs. What I need to do is somehow append stuff to an env variable whenever those libs are used as buildInputs. I got very close to get it right but I'm stuck: I've defined a setupHook script that does the job, but the $out variable is not the package's, but the new build env's out... so when I try nix-shell ... I get the shell's $out
<grp>
instead of the package. Been searching through the manual for hours now but still haven't found how to achieve this. I've also read perl's and lua's modules generic builder and mine is pretty much the same. I figured I have to use addEnvHooks, but it fails and I have no clue how to stop fixupPhase's whatever substitution is failing...
<tobiasBora>
Or still funnier, look at this slighly modified example:
jasongrossman has quit [Ping timeout: 252 seconds]
* grp
sighs
<sphalerite>
tobiasBora: I think someone was working on improving nix-mode recently. Not sure who it was though
<tobiasBora>
sphalerite: let me know if he has some nice update ;)
sbdchd has joined #nixos
<sphalerite>
tobiasBora: nix-mode#42 I think this might be it
<grp>
hmmm... just found about envHooks, maybe...
<rawtaz>
vaibhavsagar: this is one file i found matching "*tracker-miner*": /nix/store/a9l81xk85ihwiqf7xjhw5nswspnds4bp-system-path/lib/systemd/user/tracker-miner-fs.service - suggests to me that the service should be named tracker-miner-fs
<vaibhavsagar>
rawtaz: I just tried `nix-build -E '(import <nixpkgs> {}).gnome3.tracker-miners' and it didn't have a /bin directory so I don't think that's what's wrong
<vaibhavsagar>
rawtaz: what are the contents of that service?
<rawtaz>
vaibhavsagar: could it be that this is something that should be started when i *log in*? if so, it doesnt seem its triggered by my logging into i3
mayhewluke has quit [Ping timeout: 252 seconds]
<vaibhavsagar>
rawtaz: what is the output of `ps aux | grep tracker`
<sphalerite>
rawtaz: it lives in lib/system/user, not lib/systemd/system, so it's a systemd *user* service
<sphalerite>
rawtaz: so you can only see it using systemctl --user, if at all
mayhewluke has joined #nixos
<vaibhavsagar>
sphalerite: I didn't know about this distinction, thanks!
<rawtaz>
vaibhavsagar: only /nix/store/ikbjv9k4havn3chh0yv89ilysj5pnjyv-tracker-2.0.3/libexec/tracker-store is running
sbdchd has quit [Ping timeout: 252 seconds]
<rawtaz>
sphalerite: i see. trying to hunt it down now
<tobiasBora>
sphalerite: ok thank you. I home it will solve lot's of problems :D I need to wait that the dev push this to melpa, or melpa is supposed to automatically triger this changes?
<sphalerite>
tobiasBora: update instructions are at the top of ^
sbdchd_ has quit [Ping timeout: 244 seconds]
<tobiasBora>
sphalerite: Oh... Maybe that's because I'm on nix 18.03?
<tobiasBora>
(and it looks pretty close to the date I have melpa)
<tobiasBora>
so I think that I should move to unstable
<sphalerite>
tobiasBora: oh hm yeah nixos-18.03's melpa doesn't seem to get updated at all
<sphalerite>
but even on unstable it's still at 2018-07-22
<rawtaz>
vaibhavsagar: and `journalctl -u tracker-miner-fs.service` just reports no entries in the log. i dont think this service was ever started. trying to start it now.
<tobiasBora>
sphalerite: ok, so I guess I'll still wait a bit.
<tobiasBora>
Or wait for them to update melpa cache
<sphalerite>
tobiasBora: you can update it yourself, it doesn't look too complicated
<vaibhavsagar>
rawtaz: what does `systemctl status tracker-miner-fs.service` tell you?
neonfuz has joined #nixos
<neonfuz>
Hello
<vaibhavsagar>
hi there neonfuz
<neonfuz>
so does bitlbee use the proprietary nvidia driver?
<sphalerite>
neonfuz: bitlbee has nothing to do with nvidia..?
<neonfuz>
not nouveau?
<sphalerite>
you mean bumblebee?
<neonfuz>
pff, yeah
<neonfuz>
sorry
<neonfuz>
I was looking up both
<sphalerite>
It does default to the proprietary one iirc, yes
<tobiasBora>
sphalerite: it means that I'll need to maintain a fork of nixpkgs?? Not sure that I can do that reliably ^^'
<neonfuz>
okay, I'm trying to get nvenc working
<sphalerite>
tobiasBora: no, just fork it, update it, PR it, and it'll go upstream :)
<neonfuz>
I'm using obs studio, I ran it through optirun, but it seems to fail when I try to record
<neonfuz>
with it set to nvenc rather than software encoding
<rawtaz>
vaibhavsagar: https://pastebin.com/FZSYEyFA - do you know if i should start it as user or as root (it asks when i try to start it, which i want to start it as)?
<tobiasBora>
sphalerite: oh good idea. I though that dev would do that kind of automatiquely. But I will try tomorrow!
<sphalerite>
rawtaz: user systemctl --user start <name>
<rawtaz>
ok
<sphalerite>
tobiasBora: looking at the git log, it looks like various people do it whenever they need it updated
<vaibhavsagar>
rawtaz: try `sudo systemctl start`, I would recommend starting as root
Dedalo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<tobiasBora>
sphalerite: is it applied like quickly? Because last time I tried a PR, I just added one line in a nodeJs link, and it took a few weeks to be accepted upstream, so I don't want to add lot's of PR if someone already did it
<vaibhavsagar>
oops, thanks sphalerite
<sphalerite>
tobiasBora: yeah PRs can sometimes take a while, usually just pinging the maintainer or someone with merge rights who could be interested helps
<sphalerite>
tobiasBora: e.g. ttuegel was the last person to update melpa-packages, so I'd just ping him in the PR
<avn>
tobiasBora: here patches usually accepted fast. If not, you always can poke someone here to review
<rawtaz>
vaibhavsagar, sphalerite: okay i got a start attempt and an error in the journal, will parse it to see what it's about (and paste it for your reference)
<tobiasBora>
Ok good, I'll try tomorow then. thank you!
<neonfuz>
goibhniu: on nixos? btw, is window capture also broken for you?
<neonfuz>
(on obs)
<grp>
clever: just got closer to get it right: got addEnvHooks to work, but it feeds the custom bash function with a lot of unrelated paths. I guess I can filter them and done, but I'd like a cleaner solution. It'd be perfect if I could get the pkgs' $out right.
ixxie has joined #nixos
<neonfuz>
hmm, it errors 'cannot init cuda'
<grp>
I hit infinite recursion whenever I try anything fancy
<neonfuz>
my gpu isa bit older, I wonder if it doesn't even have that? or I don't have drivers
<grp>
clever: nevermind, got it working...
<goibhniu>
neonfuz: yep, on NixOS ... window capture is working fine too
<neonfuz>
weird, I've never had window capture work on nixos
<neonfuz>
on either of my machines
<tobiasBora>
ouf, seems to be still in elpa
<neonfuz>
are you on nixos stable?
rauno has quit [Ping timeout: 250 seconds]
<goibhniu>
I'm on unstable
<neonfuz>
me too
lopsided98 has quit [Ping timeout: 250 seconds]
tmaekawa has joined #nixos
lopsided98 has joined #nixos
tmaekawa has quit [Client Quit]
semilattice has quit [Ping timeout: 252 seconds]
<goibhniu>
neonfuz: hrm ... did you try with a fresh config? ... e.g. by renaming ~/.config/obs-studio
semilattice has joined #nixos
<clever>
grp: oh, you might want @out@
<neonfuz>
goibhniu: yeah when I try to add a window capture it crashes with [VGL] ERROR: in init-- [VGL] 43: Invalid argument
orivej has joined #nixos
Ericson2314 has joined #nixos
<goibhniu>
that's no fun :/ ... I wonder if it's a bumblebee thing, or a hardware thing
sbdchd has joined #nixos
mounty has quit [Quit: Konversation terminated!]
<{^_^}>
[nixpkgs] @uskudnik opened pull request #46577 → tsung: Add package tsung version 1.7.0 → https://git.io/fAKiR
acarrico has quit [Ping timeout: 252 seconds]
<grp>
clever: yes!! that's it!
<grp>
hmm, how was that karma stuff?
<grp>
clever+
<grp>
clever++
<{^_^}>
clever's karma got increased to 24
<{^_^}>
[nixpkgs] @costrouc opened pull request #46578 → libxnd, libndtypes: refactor add support for darwin → https://git.io/fAKiw
sbdchd has quit [Ping timeout: 246 seconds]
<clever>
grp: when nixpkgs is generating the setuphooks, it runs substituteAll
<clever>
grp: which will replace tokens like @out@ with the value of $out, at the time its building the hook
<clever>
other env vars can also be embeded, if they are needed
<grp>
mhmm
realrokka has quit [Ping timeout: 252 seconds]
sbdchd has joined #nixos
sbdchd has quit [Remote host closed the connection]
sbdchd has joined #nixos
patrl has quit [Quit: WeeChat 2.0]
patrl has joined #nixos
xok has quit [Quit: Leaving.]
Ericson2314 has quit [Ping timeout: 240 seconds]
jedahan has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<neonfuz>
I know I could make changes to my config and rebuild, then boot between those two, but it'd be nicer to be more clear about it like having multiple names
* hodapp
looks back and forth between option 1, "Recompiling OpenCV with V4L support", and option 2, "The nearest cliff"
<samueldr>
LnL: going with the infos in the evals, listed at the top: https://hydra.nixos.org/eval/1478459 still shows "optima" as a build with failed dependencies
<LnL>
samueldr: hydra has disabled t2m, but it caused a bunch of build failures yesterday
WilliamHamilton[ has joined #nixos
<WilliamHamilton[>
when trying to use steam-run, I get the error "relocation error: /usr/lib/libc.so.6: symbol _dl_exception_create, version GLIBC_PRIVATE not defined in file ld-linux-x86-64.so.2 with link time reference"
<samueldr>
and *that* lists happy as failed
<samueldr>
I can't reliably parse them out until the builds don't say they have a failed happy :/
<WilliamHamilton[>
but I remember using steam-run and not having this problem on the same files. Can you help me to debug this?
<LnL>
samueldr: I think you should only look at the last step for each build, optima failed again but because of attoparsec-time this time
<samueldr>
LnL: if happy was failing for real, and attoparsec too, wouldn't they both show there?
<samueldr>
(I don't know hydra that well)
Neo-- has quit [Ping timeout: 240 seconds]
<LnL>
not sure what you mean
<samueldr>
if build A depends on BCDE, and both B and D fail, would they show "Failed build steps | B | D" or only one of the two? (assuming BCDE don't depend on each-other)
<samueldr>
would A show "Failed build steps | B | D "**
<LnL>
only the first failed one
<ldlework>
I have aspell aspellDicts.en installed in NixOS but emacs still complains that (Error: No word lists can be found for the language "en_US".)
<ldlework>
Anyone know about that one?
<samueldr>
I don't understand build steps :/
<LnL>
samueldr: it's the list of derivations that where not cached yet but needed to build the job
<LnL>
if one fails it stops there similar to a local nix-build
exarkun_ has quit [Read error: Connection reset by peer]
exarkun_ has joined #nixos
<samueldr>
okay, so even if B and D are tried on hydra and fails, only one will show there?
<LnL>
yeah and if B doesn't depend on D it will be a random one
<LnL>
so in this case D failed (because of a bad machine) and the second attempt failed with B
<samueldr>
hm, this is upsetting to me :/
<LnL>
:p
<samueldr>
(as in bothering)
<LnL>
also if another job A stole all the build steps you depend on the list will be empty, even tho your build also needs them
<LnL>
but that doesn't matter to find failures
<patrl>
has anyone had any success with the rust crate ggez
<samueldr>
!! maybe I'll need to figure out those depdency stuff another way
<samueldr>
dependency*
<patrl>
I can compile a minimal example, but `cargo run` triggers a weird SDL error
<samueldr>
it could pad the numbers considerably, but having all failures listed may be better?
antoinerg has quit [Ping timeout: 252 seconds]
barb has joined #nixos
<LnL>
depends, if both A and B cause 100 failures B might show up as fixing only 1 issue
<LnL>
so if you're lucky counting older steps might bump B up higher
fragamus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<ldlework>
figured it out
<samueldr>
I'm thinking of not using the info on the build page, but using the attribute names from the eval page, and doing the dependency checking using a local eval?
<LnL>
samueldr: ah yeah that's just to get nix-shell -p style behaviour
<{^_^}>
[nixpkgs] @Mic92 pushed commit from @worldofpeace to release-18.09 « eolie: 0.9.35 -> 0.9.36 »: https://git.io/fAK9O
<LnL>
hrm, I have a query-attributes.sh that's almost the same
<rawtaz>
is James Earl Douglas in here?
lopsided98 has joined #nixos
semilattice has quit [Ping timeout: 240 seconds]
<JonReed>
Is there a way to write a multiline string without newlines? For example, I'm writing an abort message that I want to split into two lines. Is `abort (replaceStrings ["\n"] [""] "my lengthy message...")` the way to do it?
mizu_no_oto has joined #nixos
<clever>
JonReed: "foo" + "bar" and let the nix parser ignore newlines between values?
<{^_^}>
[nixpkgs] @xeji pushed commit from @r-ryantm to master « star: 2.6.0c -> 2.6.1a (#46176) »: https://git.io/fAKHS
<{^_^}>
[nixpkgs] @costrouc opened pull request #46583 → quantum-espresso, siesta init packages for density functional theory calculations → https://git.io/fAKHN
<{^_^}>
[nixpkgs] @xeji pushed commit from @r-ryantm to master « radarr: 0.2.0.995 -> 0.2.0.1120 (#46185) »: https://git.io/fAKQU
ma27 has joined #nixos
<andi->
Has anyone successfully used the `allowKeysForGroup` option in the acme configuration? To me it seems like there is a chmod missing AFTER issuing the real certificates :/
<{^_^}>
[nixpkgs] @xeji pushed commit from @r-ryantm to master « miniupnpc_2: 2.0.20180203 -> 2.1 (#46250) »: https://git.io/fAKQV
xenog has joined #nixos
<adamantium>
Hi, i'm trying to use vaapi hw video decoding and it isn't working, i've been following the arch wiki on it, can't figure out what the hangup is. "vaainfo" output: http://termbin.com/sh13
<{^_^}>
[nixpkgs] @xeji pushed commit from @r-ryantm to staging « gdbm: 1.17 -> 1.18 (#46285) »: https://git.io/fAK5Q
<patrl>
I *think* i've narrowed down some problems I was having with a rust crate under NixOS to my openGL version
<patrl>
glxinfo tells me my openGL version is 3.0
<patrl>
I have an intel 5500, nothing special in my configuration.nix
<patrl>
I know that this card supports higher openGL versions. I suspect I could work around this if I could bump up the version of mesa that NixOS uses
<patrl>
does anyone know how I could go about doing this?
<fpletz>
Aleksejs: our skype package is probably out of date and skype deleted the old version :(
<adamantium>
makefu: sphalerite got me sorted out, in my case i needed to do it this way: hardware.opengl.extraPackages = with pkgs; [ vaapiIntel libvdpau-va-gl ];
<witchof0x20>
Is there a way to get more debug information from nix-env than -v?
<{^_^}>
[nixpkgs] @xeji pushed commit from @r-ryantm to master « mbuffer: 20180318 -> 20180625 (#46241) »: https://git.io/fAKdQ
<adamantium>
and i install vainfo unde systemPackages
<makefu>
adamantium: nice, matches with the nixos-hardware entry for the intel cpus
jtojnar has quit [Quit: jtojnar]
<goibhniu>
patrl: ah, ok. Do you know if the the version you need is already in unstable?
<witchof0x20>
I symlinked firefox-overlay.nix from https://github.com/mozilla/nixpkgs-mozilla into ~/.config/nixpkgs/overlays and using nix-env, home-manager as my non-root user just freeze
jtojnar has joined #nixos
<patrl>
goibhniu: good question
<patrl>
goibhniu: unfortunately I'm not familiar enough with the video drivers ecosystem to know how mesa versions correlate with openGL versions on particular cards etc etc
<witchof0x20>
`nix-env -v -qa firefox` for example, seems to stop at `evaluating file '/nix/store/l3ix0j0cyw3cl7s1w6gbs0khc8rk60v6-nixos-19.03pre151837.ca2ba44cab4/nixos/pkgs/top-level/aliases.nix'` Another symptom of both nix-env and home-manager is nix-env (or nix-build when calling home-manager) uses an entire CPU's full power. I'm not sure if it's trying to build firefox, but I'm just trying to install the nightly
<witchof0x20>
binaries.
simukis has quit [Quit: simukis]
tertl3 has joined #nixos
<goibhniu>
patrl: you could always test it out, and rollback if you want to stay on 18.03 ... 18.09 isn't far off anyway
<patrl>
goibhniu: I was hoping to avoid doing a full switch, but it sounds like the most sensible idea! thanks for the advice
<goibhniu>
patrl: if it's not in unstable either yet, you could fork nixpkgs, update mesa (no idea how tricky that is) ... and submit a PR ... I think mesa gets updated regularly though
<sphalerite>
witchof0x20: you can add more -v switches
<samueldr>
(nix-top is now in nixpkgs)
<sphalerite>
witchof0x20: but it probably won't make a difference
<sphalerite>
samueldr: isn't nix-top for running builds?
<samueldr>
yes
<sphalerite>
it sounds like witchof0x20's stuff isn't completing evaluation
<sphalerite>
since it happens with nix-env -qa
<samueldr>
that was mainly for manveru though :)
cement has joined #nixos
<manveru>
:)
<sphalerite>
oooh right
<manveru>
what about `nix search``?
<sphalerite>
I'm guessing the same
<sphalerite>
witchof0x20: how long did you leave it running for?
<witchof0x20>
30 mins or so
<sphalerite>
huh ok that shoudl really be enough :')
<sphalerite>
how much RAM do you have? Do you have any swap?
<cement>
I've slacked off in updating nixos and now nixos-rebuild switch --upgrade isn't actually updating nix. It's currently on 1.11 and I'd probably like it to be on version 2.1
<samueldr>
depends on what's the exact issue you're hitting
<sphalerite>
cement: then the same thing but for 18.03
<samueldr>
and as sphalerite said
<samueldr>
skipping versions can work, but it isn't as often tested
Dedalo has joined #nixos
<sphalerite>
cement: alternatively you can skip 17.09 by booting into an 18.03 installer, mounting everyhting and just running nixos-install again (still read the release notes for both though)
silver has quit [Ping timeout: 244 seconds]
patrl has quit [Quit: WeeChat 2.0]
<samueldr>
so if you value neither your time, nor your data
silver has joined #nixos
<witchof0x20>
`nix search` does the same thing. high cpu usage, low (and constant) memory usage
<samueldr>
(that was a joke)
<LnL>
skipping a version isn't a problem, but I'd definitively use nixos-rebuild boot then
<samueldr>
yep!
<sphalerite>
LnL: iirc the nix version causes problems
<LnL>
don't remember anything like that
<sphalerite>
oh wait that's for 17.09 -> 18.09
<LnL>
with 17.03 -> 17.09 there's the sudo thing IIRC
contrapumpkin has joined #nixos
<LnL>
but not live switching means you won't hit that
<sphalerite>
,releasenotes = Before upgrading nixos, read the release notes for the release you're upgrading to at https://nixos.org/nixos/manual/release-notes.html to make sure you're aware of any backwards incompatibilities that may break your config.
<{^_^}>
releasenotes defined
<cement>
yeah, my jank setup actually didn't have anything broken
<cement>
which I'm really quite surprised about
<cement>
so from 17.03 -> 17.09 there might be a sudo thing that requires rebooting?
<sphalerite>
,upgrade = To upgrade nixos, read the release notes (this is important! Check ,releasenotes for details), then run (as root) nix-channel --add https://nixos.org/channels/nixos-xx.yy nixos && nixos-rebuild boot --upgrade , then reboot.
<{^_^}>
upgrade defined
<sphalerite>
cement: you should generally reboot for release upgrades
<LnL>
cement: yeah, PATH changed so you'd "loose" sudo for existing processes like your window manager
<cement>
I'd already started this one with switch
<cement>
so I guess I should ctrl C?
<LnL>
that's fine, just reboot after
<sphalerite>
probably won't be disastrous :p
copumpkin has quit [Ping timeout: 244 seconds]
cement_ has joined #nixos
<sphalerite>
right, I need to go to bed earlier today
<sphalerite>
gnight all!
<cement_>
and it's 17.09 -> 18.09, right?
<LnL>
wow, what's going on with my clock lately
<symphorien>
cement_: 17.09 -> 18.03
<cement_>
ty
<symphorien>
18.09 is still beta
rfold has quit [Ping timeout: 245 seconds]
<LnL>
18.09 will be released at the end of this month
cement has quit [Ping timeout: 240 seconds]
<adamantium>
Okay guys, for vaapi and vdpau on my intel integrated gpu i use hardware.opengl.extraPackages = with pkgs; [ vaapiIntel libvdpau-va-gl vaapiVdpau ]; ... for my amdgpu workstation, should i change vaapiIntel to something other?
Dedalo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
silver has quit [Read error: Connection reset by peer]
silver has joined #nixos
<cement>
so, is using nix-env -iA still the way things're done?
erasmas has quit [Quit: leaving]
<cement>
because that ain't workin' for me right now. I get this error: error: Nix database directory ‘/nix/var/nix/db’ is not writable: Permission denied
<cement>
the fix that I got to through googling was the export thing
silver has quit [Read error: Connection reset by peer]
silver has joined #nixos
<{^_^}>
[nixpkgs] @elitak opened pull request #46586 → factorio: download using token, not password → https://git.io/fAKji
xenog_ has quit [Remote host closed the connection]
xenog_ has joined #nixos
<jtojnar>
cement: what does nix-env --version print?
tzemanov_ has joined #nixos
jluttine has quit [Ping timeout: 244 seconds]
tzemanovic has quit [Ping timeout: 252 seconds]
jluttine has joined #nixos
jedahan has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<cement>
error: Nix database directory ‘/nix/var/nix/db’ is not writable: Permission denied
jasongrossman has joined #nixos
<clever>
cement: what does `nix-env --version` return?
<tobiasBora>
Hum maybe not in fact
<cement>
[cement@ChapTopTwo:~]$ nix-env --version
<cement>
error: Nix database directory ‘/nix/var/nix/db’ is not writable: Permission denied
<clever>
cement: what about `type nix-env` ?
<cement>
nix-env is /nix/var/nix/profiles/default/bin/nix-env
<clever>
cement: thats not right, `sudo nix-env -e nix`
<clever>
somebody installed nix using nix-env as root
<tobiasBora>
I installed some other programs and I got in the list this stuff "/nix/store/9d6qh345zv4nbfwm06jisbzv02xhahh1-kernel-modules/lib/modules/4.17.19/kernel/drivers/ata/ata_piix.ko.xz", so it looks like that the kernel I will have on reboot is 4.17.19... And of course I don't want it as the system is supposed to be usable only for kernel >= 4.18
<tobiasBora>
how could I make sure I'm on the good kernel?
<tobiasBora>
(without rebooting, and having a risk to lose it)
<clever>
cement: yep, so thats fixed, is there anything else not working?
<cement>
that was it
<cement>
tyvm
Dedalo has joined #nixos
<clever>
your welcome
Dedalo has quit [Client Quit]
Kelppo has joined #nixos
<tobiasBora>
ok, I confirm, I will be on a bad kernel version: ls -al /run/current-system/kernel ==> ... -> /nix/store/6mcp0misy3fknl4xalgx4srpm49xf0d1-linux-4.17.19/Image
<tobiasBora>
Hum, it's like if my channel has not been updated