<fresheyeball[m]>
Overall I have learned that explicit sha256 is not a good thing. It's better to use the fetchGit primitive instead. Since a hash is a hash.
cr4y1 has quit [Ping timeout: 240 seconds]
<mightybyte>
Also notice that the above shas are not the same as the one in default.nix.
<fresheyeball[m]>
Yeah, which is odd
<cole-h>
Yeah, that's really strange...
<mightybyte>
The explicit sha is what gives you bit-for-bit reproducibility.
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/JfdMw
<iqubic>
So, this ilspy thing is pissing me off.
<iqubic>
I can't figure out how to make it run on Nixos.
<worldofpeace>
iqubic: reviewed
<worldofpeace>
But anyone with PRs that are approved that could be merged?
<iqubic>
That's my first time adding a package to nixpkg. I hope I did it correctly.
<DigitalKiwi>
post-installation fixup
<DigitalKiwi>
/nix/store/ycbm0z3lvinvbi1x8kq7j4mkmmyd8ky1-mlv-app-1.11/Applications/MLV: fixing dyliberror: otool: can't open file: /nix/store/ycbm0z3lvinvbi1x8kq7j4mkmmyd8ky1-mlv-app-1.11/Applications/MLV (No such file or directory)
<DigitalKiwi>
it's supposed to be MLV App.app
m0rphism has quit [Ping timeout: 260 seconds]
<DigitalKiwi>
my workaround is make it MLVApp.app and rename it MLV App.app in postFixup >.>
<{^_^}>
#87579 (by cole-h, 5 weeks ago, open): nixos/sudo: default rule should be first
<worldofpeace>
ahh goodness and mkBefore, always looks like magic numbers 🤣
<cole-h>
:D
<worldofpeace>
maybe put a reason why it's mkBefore 600 in the file? (or is that contextually apparent?)
<cole-h>
It's contextually apparent when you read the commit message... ;D
<iqubic>
Where is libstdc++.so.6 located?
<cole-h>
,locate libstc++.so.6
<{^_^}>
Couldn't find in any packages
<iqubic>
I actually want a store path, as I'm trying to do some patchelf stuff.
<worldofpeace>
stdenv.cc.cc?
<iqubic>
I'm doing this imperatively.
sigmundv_ has quit [Ping timeout: 256 seconds]
sheeldotme has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<worldofpeace>
cole-h: I guess I'm suggesting a comment?
<iqubic>
I installed dotnet-sdk then ran "sudo dotnet tool install ilspycmd ~/ilspy" which installed an executable for me, but I can't run it because: "/home/avi/ilspy/ilspycmd: error while loading shared libraries: libstdc++.so.6: cannot open shared object file: No such file or directory"
codygman has quit [Read error: Connection reset by peer]
<iqubic>
Trying to patchelf this.
codygman has joined #nixos
pamplemousse has quit [Ping timeout: 265 seconds]
<iqubic>
This is being a pain.
<cole-h>
worldofpeace: Sorry, I made a sandwich :D Yeah, I'll add that in a sec
<worldofpeace>
cole-h 2020 @@@@@ "Sorry, I made a sandwich :D" - Simply iconic, no problem 😁
gustavderdrache has joined #nixos
gueorgui_ has joined #nixos
<cole-h>
worldofpeace: Done. Now, to resume eating said sandwich :D
<iqubic>
So, why is it hard to get patchelf to work?
<cole-h>
worldofpeace: I was under the impression that the "port to stable" label is for PRs that should be backported, and finally have been... But now I don't know, because it makes sense either way...
<worldofpeace>
wtf, there's a .zip in the archive
<cole-h>
Archive of an archive? :o
<samueldr>
cole-h: needs port to stable for those PRs taht need a port to stable
<DigitalKiwi>
that's from the git one
pamplemousse has quit [Ping timeout: 264 seconds]
<cole-h>
samueldr: Right. I think I saw veprbl change the label to "has port to stable" once the port-PR had been opened... Maybe I'm losing my mind.
<iqubic>
So, I'm trying to figure out ilspy, and it's not working.
<DigitalKiwi>
my first attempt was to try and build it (most of the code is still there but commented) and then i got to that dpaste error and switched to just using their .app >.>
<{^_^}>
[nixpkgs] @worldofpeace pushed 2 commits to staging: https://git.io/JfdDE
<gustavderdrache>
now i'm very confused
<gustavderdrache>
what problem are you trying to solve?
<red[evilred]>
cole-h (IRC): thank you
marcusr has joined #nixos
<iqubic>
This is the original error I got.
<wolfjb>
Hello, I have installed home-manager and added the "oh-my-zsh" configuration, also installed fzf package and the module for zsh in the oh-my-zsh configuration in my home.nix file. However, starting a shell gives the message "Please add `export FZF_BASE=/path/to/fzf/install/dir` to your .zshrc" - but the .zshrc file is not writable. Nor is the .zshenv file, matter of fact those files are symlinks to files owned by root. So, how does one
<gustavderdrache>
i think you just need to `export DOTNET_HOME` with whatever the install location is
<{^_^}>
[nixpkgs] @mmilata opened pull request #90700 → [20.03] json_c: add patch for CVE-2020-12762 → https://git.io/JfdD2
<iqubic>
the install location is "~/ilspy" which is a directory I created to hold the files.
<iqubic>
So just export DOTNET_HOME?
<gustavderdrache>
i believe so
<gustavderdrache>
again, not really a dotnet expert so you'll probably need to find some more docs
<iqubic>
What do I set the value equal to? "~/ilspy"?
<gustavderdrache>
yeah, i think so
<iqubic>
Actually, it's called DOTNET_ROOT
<cole-h>
tbh, I'd try setting it to `${pkgs.dotnet-sdk}`
<iqubic>
Yeah.
<{^_^}>
[nixpkgs] @nh2 opened pull request #90701 → consul.passthru.tests: Fix failure on current consul versions, add more tests → https://git.io/JfdDo
user_0x58 has joined #nixos
<iqubic>
cole-h: I'm not sure which folder that is, as I installed dotnet-sdk with nix-env.
<msirabella>
Perhaps, but most of the literature I've found on nixops has been for provisioning and deploying new machines
<cole-h>
For continuous deployment, you could add a post-commit hook that deploys :D
<cole-h>
btw, nixops isn't only for new machines...
<cole-h>
I use it in place of nixos-rebuild for my machine
<msirabella>
right, my question is basically what should "deploys" do?
<cole-h>
Build the system and switch to the new config?
<cole-h>
At least, that's what the deployment tools, like nixops, morph, and nixus, do
waleee-cl has quit [Quit: Connection closed for inactivity]
<msirabella>
Would nixops need the full configuration file to do the "nixos-rebuild"? I think that's what I'm confused about, I have a standalone repository for just the nginx content of my machine
<cole-h>
Yes, it needs all the configuration that a normal machine would.
emacsoma1 is now known as emacsomancer
ok2` has joined #nixos
<{^_^}>
[nixpkgs] @JohnAZoidberg pushed 3 commits to master: https://git.io/JfdSO
<msirabella>
Thanks cole-h ! I'm not sure if I want to bundle in my entire nixos configuration with a specific package I want to deploy to my server, but I'll take a look
<cole-h>
I mean... If you're deploying to a server, you'll need to setup a configuration.nix for the server, right?
<sheeldotme>
cole-h thanks for sharing! Are you able to add the home-manager channel and configure it on rebuild with nixops too? I'm not sure where that would be found.
<msirabella>
My current workflow is just plain old `nixos-rebuild`, configuration exists on the server in /etc/nixos
<sheeldotme>
oh i see, you clone and install it
<cole-h>
sheeldotme: I use channels as little as possible (so, only for `nix repl '<nixpkgs>'` and nix-shell)
<cole-h>
But I do use home-manager, pinned through niv
pr06lefs has quit [Ping timeout: 260 seconds]
<sheeldotme>
cole-h what's the reasoning behind that? Is it to maintain reproducibility?
<cole-h>
Yep.
<msirabella>
If I'm understanding correctly cole-h, you would theoretically have a "nixops deploy ..." in any post-commit hook on your personal machine?
<cole-h>
msirabella: Right. Currently, I don't (because I don't commit as frequently as I should), but that's probably what I would do
Pidgeotto has joined #nixos
<cole-h>
and I might even be tempted to do it in a pre-commit hook, so that I don't commit a broken configuration.
<colemickens>
Is anyone using nvim-lsp with nixos or hm's vim infra?
<msirabella>
Makes sense
spacefrogg has quit [Quit: Gone.]
aw has quit [Quit: Quitting.]
aw has joined #nixos
<sheeldotme>
Makes sense to me cole-h, how do you currently handle secrets do you just have them in another git repo? Thinking about setting up something with vault.
spacefrogg has joined #nixos
felixfoertsch23 has joined #nixos
<cole-h>
I have a secrets repo, where the contents are encrypted with git-crypt (so they're readable once I authenticate with GPG after setting up my machine, but total garbage on GH or a new clone)
<sheeldotme>
Very cool. Hope you don't mind all the questions but I'm right in the thick of setting up my environment right now and it seems you've thought about a lot of the things I'm about to get to.
<cole-h>
No problem, I'm always happy to help :^)
<sheeldotme>
thanks cole-h!
<msirabella>
Unfortunately I don't think the process of bundling all my configuration together for independent projects with continuous deployment is going to work, but this conversations has given me the idea of maybe making my own channel for my build (even though you avoid channels :P)
<cole-h>
You don't necessarily need to bundle all your configuration... Just need to have enough configuration for each host that it will boot and run the services you want :P
<msirabella>
That is a definite solution, if I want to stick to one host would I run all the things in docker containers?
<cole-h>
NixOS has native containers, but I don't know if people prefer one or the other.
<msirabella>
Ah, that's probably something to look in to
<cole-h>
If you run it on one host, you'll still need to rebuild your system every time you change something, so that wouldn't change whether you use nixops or nixos-rebuild
ogkloo has joined #nixos
dfordvm_ is now known as dfordvm
domogled has quit [Quit: domogled]
<msirabella>
It's just the idea of where I'm keeping the configuration files for my server, and keeping an unrelated package decoupled from my actual server configuration
<cole-h>
Sorry, that does not compute for me... If it's decoupled from your server configuration, how are you deploying it to the machine?
pr06lefs has joined #nixos
<msirabella>
Right now in the depths of my configuration.nix I have `nginx...root = import website.nix` with website.nix being a derivation pointing to a git url with a hardcoded hash
<msirabella>
I *basically* just want a way to bump that hardcoded hash from gitlab when I do a push
<cole-h>
Ohhhhhhhhhhhhhh
<msirabella>
Like you were saying, that could really mean just redeploying a machine
<cole-h>
I mean, at that point, why not just point it to a master tarball or something? :P
<msirabella>
Sorry, I wasn't being very specific :P
<MtotheM>
What is the reason for being so inconsistent when it comes to configurations in nixOS? I understand that home-manager exists. but why only allow configuration declarations for a handful of programs and not everything?
<msirabella>
You are right, but with nixos' purity I have to specify a sha256 of the tarball as well and its the same problem?
<cole-h>
Not really
<multun>
MtotheM: home-manager configures programs for your home only
<cole-h>
builtins.fetchTarball doesn't *need* a sha256.
<MtotheM>
yes, but you can't configure programs globally with nix from my limited experience. apart from the handpicked ones. like vim, tmux, xmonad etc
<cole-h>
It's great for reproducibility
<msirabella>
Give me a minute, I need to read about this, i am dumbfounded at how simple the solution is
<multun>
MtotheM: it mainly manages your dotfiles, which means that for a program to be supported, somebody has to have written some code to generate it from the configuration you give. sometimes, along with the dotfiles, it sets up a systemd user service and installs a program
<multun>
MtotheM: well yes you can
<cole-h>
^
<MtotheM>
Mind pointing me in the right direction? cause I can't find any information on it.
<cole-h>
Look at any of the already-present modules in NixOS
<multun>
MtotheM: global packages can be added in /etc/nixos/configuration.nix, in the environment.systemPackage key
<multun>
_unless_ you only use nix and not nixos
<MtotheM>
I know I can declare my packages in there, but I meant the configuration files.
<cole-h>
You can configure user packages the same way you configure it on any other machine: if there's no module for it, just modify ~/.config/somepackage.
<msirabella>
cole-h: thank you so much, it looks like this is exactly what I am looking for
<msirabella>
And I learned a lot along the way, too
<cole-h>
Need to configure system packages without a module? Just use `environment.etc."somepackage/package.conf".text = "here's some configuration lol";`
<cole-h>
msirabella: Happy to help, glad we got you sorted :D
<msirabella>
fetchTarball must re-fetch the resource on the internet on every rebuild, correct? It can't know if it's changed otherwise
<multun>
MtotheM: configuration for globally installed programs are in /etc/nixos/configuration.nix as well, and configuration for stuff in your home is handled by home-manager
<cole-h>
Governed by tarball-ttl: "The number of seconds a downloaded tarball is considered fresh."
<MtotheM>
`environment.etc` gonna look at this
<msirabella>
ah I see, I'll have to look in to how to manually force it to re-get it
<cole-h>
"If the cached tarball is stale, Nixwill check whether it is still up to date using the ETag header. Nix will download a new version if theETag header is unsupported, or the cached ETag doesn't match."
<multun>
both can install good old text files, but have convenient settings to generate those
<cole-h>
msirabella: You can just do `--option tarball-ttl 0` if you need to force a re-fetch.
<cole-h>
"Setting the TTL to 0 forces Nix to always check if the tarball is up to date."
<msirabella>
Where are you quoting this text from? I can't find such comprehensive documentation on fetchTarball
<cole-h>
`man nix.conf`
<msirabella>
oh, ty
<cole-h>
It's not limited to fetchTarball
<cole-h>
Anything that uses a .tar.gz underneath follows these settings, AFAIK
<cole-h>
Of course, not many things accept a tarball without a sha256, so fetchTarball is the only real user-facing place this occurs, to my knowledge
<notgne2>
sib81: it doesn't look like there's any modules in that PR, just packages, so it should be pretty simple
<notgne2>
you can import it with like, `somePkgs = import (builtins.fetchGit { url = ""; rev = ""; }) {};` and just use `somePkgs` as you would use `pkgs`
<notgne2>
I think you can also do it in a channel, and there's an info page on doing this (it's essentially the same steps as using unstable on a stable install) though I don't know where to find it
<sib81>
notgne2: the PR says it allows one to specify mutable vscode extensions that will be installed on a per project basis, but I don't know what's the syntax for doing that configuration. I apologize for being clueless, still trying to learn my way around nixos
Hedgework has joined #nixos
<notgne2>
oh, I missed that it's actually been merged and you're just trying to utilize it, in that case I'm not sure
<smiles>
I am trying to install nixos on a raspberry pi2 using qemu to install it on a sd card since doing it on the rpi2 taking way to long and trying to figure out what command to use with qemu to accomplish this I already dd the image to the sd. I get a window to pop up but no terminal or anything pops up if any one got any suggestions this is the command I am using
<angerman>
Ericson2314: gcc 9 (cross: mac -> linux at least) is broken; complains about cycles from lib -> out; we'd need to patch the rpath in the sanitizer libs if we move them out. https://www.irccloud.com/pastebin/8taOMZgQ/
atriq is now known as Taneb
sheeldotme has joined #nixos
cfricke has joined #nixos
xvnvx has quit [Remote host closed the connection]
xvnvx has joined #nixos
xvnvx has quit [Remote host closed the connection]
xvnvx has joined #nixos
noudle has joined #nixos
NeoCron has joined #nixos
mallox has quit [Quit: WeeChat 2.8]
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<{^_^}>
[nixpkgs] @uvNikita opened pull request #90719 → kdenlive: fix core dump on file open → https://git.io/Jfdxa
o1lo01ol1o has joined #nixos
<Ariakenom>
I've used fetchurl to get the source in mkDerivation. but the builder should use a different dir than the root in the download. how do I do that?
Chiliparrot has quit [Quit: My iMac has gone to sleep. ZZZzzz…]
orivej has quit [Quit: No Ping reply in 180 seconds.]
eoli3n has quit [Ping timeout: 246 seconds]
orivej has joined #nixos
<o1lo01ol1o>
What does: syntax error, unexpected $end, at foo.nix mean? The file is a simple import of nixpkgs from a json file. The syntax is correct but I can't shake this error when I try to build or enter a shell.
xelxebar has quit [Remote host closed the connection]
<srhb>
o1lo01ol1o: Typically some closing structure is missing
xelxebar has joined #nixos
<srhb>
> let foo = 42; in
<{^_^}>
error: syntax error, unexpected ')', at (string):318:1
<txt-file>
I have a problem to install nvidiaLegacy340 on nixOS 20.03. It tells me that 1 dependency could not be build. https://paste.debian.net/1152644/
<srhb>
o1lo01ol1o: No, there are more lets than ins, so something is wrong there :)
<o1lo01ol1o>
aye aye aye. I copy pasted that from another expression and erased one to many ins
<o1lo01ol1o>
ok, thanks.
<o1lo01ol1o>
srhb++
<{^_^}>
srhb's karma got increased to 112
<o1lo01ol1o>
sshow++
<{^_^}>
sshow's karma got increased to 1
orivej has quit [Ping timeout: 256 seconds]
orivej has joined #nixos
waleee-cl has joined #nixos
<Ariakenom>
im trying to build a project, SMTPlan. I got some compilation errors so I tried using gcc8Stdenv. It managed to compile but I get link errors. I got rid of one by overriding that package's stdenv but that didnt work for boost
<LambdaDuck>
Is it possible to only download and install the binaries for a haskellPackage, and not the library and all the build dependencies?
magnetophon has quit [Read error: Connection reset by peer]
<dminuoso>
LambdaDuck: What's your goal here?
<the_pumpkin_man>
LambdaDuck: I don't think it is, but you could get the binaries by looking the source download in the derivations file on nixpkgs
magnetophon has joined #nixos
the_pumpkin_man has quit [Quit: WeeChat 2.8]
<notgne2>
is there a good-practice way of getting a path to a file you keep alongside your NixOS nixfiles? I'm using `builtins.toFile "myfile.txt" (builtins.readFile ../myfile.txt)` right now but not sure if there's a better way
<tobi_>
hello, is it possible in nixos to automatically add a vlan tag to all traffic which is coming from a container?
<qyliss>
notgne2: why not just ../myfile.txt?
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<notgne2>
qyliss: huh, didn't realize toString could turn those into a full path
<srhb>
pingiun: Whereas the nixos channels check (Linux) NixOS functionality
<pingiun>
srhb: then I'll change my channel back to just nixos-20.03
<srhb>
pingiun: So I should have said "it's the same nix code, but checked at different commits with different criteria for success -- the darwin criteria might not be relevant or sufficient to check nixos functionality"
<pingiun>
srhb: ah right
dingenskirchen has quit [Quit: dingenskirchen]
dingenskirchen has joined #nixos
<txt-file>
I want to add a file to /etc/tmpfiles.d/. How can I do that with nixOS in a way that it "nix-rebuild" does not throw it away?
<LambdaDuck>
dminuoso: To not download a bunch of things I don't need. Plenty of haskell packages build binaries. Most of the time, that's all I need, unless I'm wrinting a Haskell program. For example: Use a by hydra precompiled pandoc without downloading everything that is needed to build the program (especially GHC).
<betawaffle>
is there any way to have multiple Go builds share the same downloaded dependencies? ie. two different derivations that depend on the same go module, to only download/build that module once
<b42>
is there an easy way to list packages that build-depend on a library?
zupo has joined #nixos
bbb1 has quit [Read error: Connection reset by peer]
bbb1 has joined #nixos
<LambdaDuck>
Oh, I was unclear. I meant I just want the executable of an existing haskell package and not the library and all its dependencies. For example, something corresponding to the hypothetical: `nix-env -f '<nixpkgs>' -iA haskellPackages.executableOnly.pandoc`. The goal is to reduce the closure size.
<LambdaDuck>
Wrong channel, sorry
orivej has quit [Ping timeout: 256 seconds]
orivej has joined #nixos
<infinisil>
Ideally don't ask the same question in multiple channels at the same time
<LambdaDuck>
Sorry!
gxt has quit [Remote host closed the connection]
gxt has joined #nixos
tobi_ has quit [Quit: leaving]
<Raito_Bezarius>
why nix-env -i firefox works but nix-env -i ns-3 does not work?
Strubbl[m]2 has joined #nixos
Rusty1 has joined #nixos
asymptotically has quit [Quit: Leaving]
<infinisil>
LambdaDuck: (I'm answering in #haskell.nix)
pr06lefs has joined #nixos
orivej has quit [Ping timeout: 265 seconds]
orivej has joined #nixos
<LambdaDuck>
Raito_Bezarius: What errors do you get?
<Raito_Bezarius>
LambdaDuck: no matched derivation
<{^_^}>
[nixpkgs] @oxalica opened pull request #90976 → rust-analyzer: 2020-06-01 -> 2020-06-15 and fix version display → https://git.io/JfFkx
<Raito_Bezarius>
My question is really, why is there derivations globally available?
<Raito_Bezarius>
vs why like ns-3 is only namespaced in nixpkgs?
<Raito_Bezarius>
Is there a rationale or something I'm missing?
<LambdaDuck>
Raito_Bezarius: Try nix -i ns-3-30
<Raito_Bezarius>
*Ah*
<LambdaDuck>
Raito_Bezarius: If you run `nix search ns-3`, you'll see that the attribute is named nixpkgs.ns-30, but the package is named ns-3-30.
domogled has joined #nixos
<Raito_Bezarius>
but why firefox works with only firefox?
<Raito_Bezarius>
rather than firefox-77.0.1 ?
<LambdaDuck>
Raito_Bezarius: That I do not know. Perhaps the source or commit history of the nix-definition for it will give some clue.
<Raito_Bezarius>
But someone asked me why it didn't work and I got curious
jakobrs has joined #nixos
meh` has joined #nixos
<jakobrs>
I think I've discovered an interesting issue: If you terminate gparted using ^C (as opposed to alt-f4), you can no longer restart the nix daemon, and if it dies, you can't start it.
<gchristensen>
wat
<jakobrs>
The issue sort of makes sense because, I believe, gparted masks the mount units when run but I'd expect them to be unmasked with ^C.
<jakobrs>
especially since alt-f4 works
<jakobrs>
Might have got something to do with #87818
<jakobrs>
Also, when booting, I sometimes get a "Failed unmounting /nix/store" message at boot. It's probably unrelated, though.
knupfer has joined #nixos
nate48423 has joined #nixos
<nate48423>
america government and denaro are getting my human brain i am nathan biernatt and i dont wan the america government to get it please tell your leaders and my brain goes through things when they do and i never told them i wanted them to have it
<jakobrs>
Huh, it happened again, and I don't remember running gparted *this time*.
<jakobrs>
I'll try rebooting into Linux Mint and seeing if gparted does the same thing there.
drakonis has quit [Read error: Connection reset by peer]
drakonis_ has joined #nixos
<lemsip>
A rather beefy oneliner to test this would be `echo 'fn main() { std::net::TcpListener::bind("0.0.0.0:81").expect("failed to bind to port 8"); }' | rustc - && sudo chmod u+s rust_out && ./rust_out`
acarrico has joined #nixos
<cransom>
lemsip: the file isn't owned by root in your oneliner
<{^_^}>
[nixpkgs] @datafoo opened pull request #91042 → nixos/acme: execute a single lego command → https://git.io/JfFY4
bennofs has joined #nixos
agsdheidjd has quit [Quit: WeeChat 2.8]
<gueorgui>
Hi! I'm running NixOS on a Thinkpad X1 Carbon (8th gen). The platform specs say that it supports video output at 5K via thunderbolt, however when I plug in my 5K thunderbolt display, it can only be used at 4K. I know for a fact that this display can work at 5K on Linux as it does work on another machine in Arch, but I can't figure out how to get NixOS to recognise the full resolution. Any ideas?
<infinisil>
gueorgui: This very much depends on your DE/WM
<infinisil>
I heard that sway works well with external displays and high dpis
<numkem>
I've started looking into taking my config and dividing it into modules (sort of like what cole-h did). If I want to add packages to systemPackages in 2 files is it possible? I currently get a `error: value is a function while a set is expected`
detran has joined #nixos
<gueorgui>
Hmm I use i3 in both cases without any DE
<numkem>
or I have to do `config = (mkMerge [{ envionment.systemPackages = []; }]);`
janneke_ has joined #nixos
<gueorgui>
On Xorg, not Wayland
kenran has joined #nixos
<{^_^}>
[nixpkgs] @flokli opened pull request #91043 → nixos/buildbot-master: support reporters, migrate away from status → https://git.io/JfFOJ
<infinisil>
numkem: Just assign environment.systemPackages as you would in configuration.nix in the alternate files
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
ashesham` has quit [Ping timeout: 272 seconds]
<infinisil>
And do `imports = [ ./alternate-file.nix ]` in configuration.nix to import it
<numkem>
infinisil: it will do the merge on it's own?
<infinisil>
Yup
<numkem>
ok, must be something else then because that's what I did already
janneke has quit [Ping timeout: 264 seconds]
EdLin has joined #nixos
<EdLin>
is the latest nixos kernel and software compatible with a Comet Lake i7 system with a Navi (RX 5700XT) graphics card?
<EdLin>
specifically I want to know if the 10th generation i7's graphics are compatible, actually, as the rx 5700 xt will be passed through to a VM
mallox has joined #nixos
<AmandaC>
if mainline is, then nixos is, EdLin. NixOS doesn't do custom kernels unless you make it do that
kenran has quit [Ping timeout: 256 seconds]
<AmandaC>
by default it uses the latest mainline. But for example, I'm currently on the latest LTS because main line totally ganked my iGPU
<EdLin>
what igpu do you use?
<AmandaC>
An intel one, in a couple-years-old i7
<EdLin>
9th generation or older?
<EdLin>
I think 10th uses the same igpu as 9th
<AmandaC>
NOt sure, how would I check? Hardware's not really my thing
<Ericson2314>
angerman: thanks. Can you file an issue? I think I saw something similar recently fwiw
<hexa->
EdLin: i'm using a RX 5700 on kernel 5.6 on nixos
<EdLin>
cool...
<infinisil>
numkem: It's hard to know the problem without seeing the code
<EdLin>
once I have the passthrough VM set up, the iGPU will be more relevant though I guess, as the GPU will be only a vfio device as far as the host OS is concenred
<AmandaC>
EdLin: /proc/cpuinfo says this for the model name: Intel(R) Core(TM) i7-7500U CPU @ 2.70GHz
<EdLin>
AmandaC oh, I'm using a desktop. I don't know if that'll be relevant.
<AmandaC>
I'm on a laptop, so I'm nto sure eithr. :)
<AmandaC>
My only desktop tower is just a win10 box plugged into a TV which I basically treat as a games console running Win10
<EdLin>
yeah, it uses HD Graphics 620
<EdLin>
I don't know if that's relevant to my system's compatibility, I'm on 630 or 640, and desktop version I guess could be different
<angerman>
Ericson2314: hmm what did i write exactly? Stupid mobile app.
<EdLin>
HD 630, yeah
<Ericson2314>
angerman: about gcc references breaking mac -> linux cross
<angerman>
Ericson2314: so we cant cross compile gcc to Linux on macOS at least because it does lib output splitting.
<EdLin>
I need to go to a meeting. Thanks everyone.
EdLin has quit [Quit: Going offline, see ya! (www.adiirc.com)]
<Ericson2314>
ok
<angerman>
And we end up with the asan, tsan, ... libraries in the lib output.
<angerman>
They however reference GC
<angerman>
Gcc
<angerman>
So we end up with gcc -> gcc-lib -> gcc and nix gives up due to dependency cycles.
<angerman>
The cherry on top: you only learn this once gcc was built.
<angerman>
But then fails in the last step.
<panos14[m]>
που θα βρω αυτο το αρχειο ? Alternate_Installer
<numkem>
infinisil: understandable, my bad, wanted to figure it out on my own but I got into a meeting. It's essentially an import of a file from configuration.nix that imports another the same way. It's complaining about the last file within the [] of environment.systemPackages = with pkgs; []
User1821 has left #nixos [#nixos]
<leungbk>
Microsoft's python-language-server has a Nix derivation, but after I run "nix-shell '<nixpkgs>' -A python-language-server" and perform the unpackPhase, I find that configurePhase and buildPhase do nothing since python-language-server doesn't have a configure script or Makefile. This happens despite the fact that buildPhase is clearly defined in the derivation.
<xwvvvvwx>
How would I trigger a full from source rebuild of all packages on my system?
<leungbk>
Is there any way to make Nix perform the specified buildPhase for me?
<immae>
I guess the correct answer yo your question is "no there isn’t", although you could use an environment variable and wrap the output program with that enviornment pointing to a folder full of fonts
<glittershark>
yeah I'm thinking that's probably gonna be the thing to do
<glittershark>
that and open an issue upstream asking them to do something less silly
<glittershark>
aka "just use fontconfig, dangit"
<{^_^}>
[nixpkgs] @flokli opened pull request #91048 → [20.03] systemd: apply patch for CVE-2020-13776 → https://git.io/JfFne
griff_ has quit [Quit: griff_]
drakonis has joined #nixos
dermetfan has joined #nixos
<immae>
It’s a reasonnable issue although I don’t expect it to be as simple as "just" use that :p (but maybe I’m wrong and fontconfig is just dead simple)
<chipb>
immae: ick. it's almost certainly easier to use than manually implementing a scan for an appropriately matching font based on something other than static file name.
<jluttine[m]>
I have a Haskell project which I build with nix. However, how can I enter a nix shell where `cabal` would be available so that I could try how `cabal install` works for this project? Here's my project: https://github.com/jluttine/dbus-listen
<{^_^}>
[nixos-homepage] @github-actions[bot] pushed commit from GitHub Actions to master « Update flake.lock and blogs.xml [ci skip] »: https://git.io/JfFcV
<glittershark>
chipb: problems like this are what nixos was basically invented to solve, haha
<glittershark>
jluttine[m]: nix-shell -A env is the usual way, without reading your source
<jluttine[m]>
glittershark: thanks, i guess my `shell.nix` is doing that already. but i tried `nix-shell default.nix -A env` but still there's no cabal in PATH. so maybe cabal isn't a build dependency of haskell packages..?
<glittershark>
that would be surprising to me
<glittershark>
jluttine[m]: I'm not sure if it's the issue, but can you try replacing callCabal2Nix with haskellSrc2Nix ?
<glittershark>
I don't remember what the difference is only that my projects use the latter and I can run cabal in my shells
<chipb>
based on the fc-match utility, I'd be surprised if it were a terribly difficult API.
<chipb>
I mean, you can match the best ttf with e.g. "fc-match monospace:bold"
<chipb>
of course, it was good enough for the upstream author as-is. there's something to the simplicity of not having to pull in another dependency...though it's already using SDL, and anybody running the program almost certainly already has fontconfig installed...
nuncanada has joined #nixos
nuncanada2 has joined #nixos
nuncanada has quit [Read error: Connection reset by peer]
<pbogdan>
jluttine[m]: runhaskell ./Setup.hs install from within your shell should work I think?
<pbogdan>
the haskell packages builder uses Setup.hs by default IIRC so doesn't need cabal-install
kleisli has quit [Remote host closed the connection]
kleisli has joined #nixos
<JameySharp[m]>
Firefox has a bunch of problems with OpenGL unless I run it with LD_LIBRARY_PATH pointing to libglvnd, anyone know why? I have a bunch more details of the strace crap I did to get this far if that would help
is_null has joined #nixos
ebopp has quit [Ping timeout: 246 seconds]
asheshambasta has quit [Ping timeout: 272 seconds]
<infinisil>
jluttine[m]: cabal is indeed not a dependency, you need to add that manually
<infinisil>
I usually do it with something like `.overrideAttrs (old: { buildInputs = old.buildInputs ++ [ haskellPackages.cabal-install ]; })`
ebopp has joined #nixos
nuncanada2 has quit [Read error: Connection reset by peer]
sheeldotme has joined #nixos
lilpea has quit [Ping timeout: 258 seconds]
<pbogdan>
might be possible to do it with haskellPackages.shellFor too I think
kreyren has quit [Remote host closed the connection]
kreyren has joined #nixos
ashesham` has joined #nixos
<JameySharp[m]>
apparently there are multiple open nixpkgs issues for my problem so I guess I have more reading to do
buckley310 has joined #nixos
lord| has quit [Remote host closed the connection]
lord| has joined #nixos
<{^_^}>
[nixpkgs] @kalbasit opened pull request #91055 → nanopb: include C sources → https://git.io/JfFC3
<{^_^}>
[nixpkgs] @puzzlewolf opened pull request #91056 → [WIP] sequoia: unbreak on darwin → https://git.io/JfFCc
mallox has quit [Quit: WeeChat 2.8]
<JameySharp[m]>
guess I needed firefox-wayland rather than firefox, which I didn't know was a thing. that's kind of a nasty surprise
<{^_^}>
[nix] @Ericson2314 opened pull request #3720 → Prevent '%' in URL from causing crashes → https://git.io/JfFCB
<gchristensen>
I thinke firefox-wayland just sets an environment variable and runs the regular ff
kamid has quit [Ping timeout: 258 seconds]
mallox has joined #nixos
<qyliss>
It sets the wrong environment variable too, apparently.
<gchristensen>
hah
<qyliss>
JameySharp[m]: I think the Wayland experts in here have told me before that you're supposed to set MOZ_ENABLE_WAYLAND
<gchristensen>
btw qyliss are you still up to date on ruby things?
<qyliss>
fraid not
<qyliss>
wish I'd been able to keep up better :(
<JameySharp[m]>
that's true if you're going to set it in your general environment, but setting GDK_BACKEND works fine if you only set it for firefox and not, say, any Electron app
bvdw has joined #nixos
<qyliss>
I think there were reports of FF bugs that only manifest if GDK_BACKEND was set
<qyliss>
but I could be wrong
<JameySharp[m]>
and yeah, firefox-wayland does exactly two things differently AFAICT: it sets that environment variable, and it adds libglvnd to LD_LIBRARY_PATH
<JameySharp[m]>
but the latter makes a really big difference
<JameySharp[m]>
without it, some Firefox subprocesses can find libEGL and others can't, resulting in broken WebGL and no support for more experimental things like WebRender
<muronglizi>
Is it legit to re-ask a question here if it didn't find an answer on discourse for a day?
boegel has joined #nixos
<eyJhb>
Is there any reason for NixOS just having the users group?
<{^_^}>
[nixpkgs] @ruuda opened pull request #91057 → fedora-coreos-config-transpiler: init at 0.6.0 → https://git.io/JfFWL
<eyJhb>
Thinking of setting the default group to the username, is there something I should be aware of?
nuncanada has joined #nixos
dedz has joined #nixos
<infinisil>
muronglizi: Sure, ideally update discourse (or other places where you asked) with the new things you learned though
dedz has quit [Remote host closed the connection]
ashesham` has quit [Ping timeout: 272 seconds]
dedz has joined #nixos
<qyliss>
eyJhb: I do that, works fine
<qyliss>
Still useful to also have the users group though
dedz has quit [Remote host closed the connection]
ckauhaus has quit [Quit: WeeChat 2.7.1]
cosimone has quit [Quit: Quit.]
o1lo01ol1o has quit [Remote host closed the connection]
o1lo01ol1o has joined #nixos
o1lo01ol1o has quit [Ping timeout: 246 seconds]
fendor has quit [Remote host closed the connection]
sheeldotme has quit [Read error: Connection reset by peer]
nuncanada has quit [Quit: Leaving]
endformationage has joined #nixos
cript0nauta has quit [Ping timeout: 240 seconds]
cript0nauta has joined #nixos
pr06lefs has quit [Ping timeout: 264 seconds]
cript0nauta has quit [Remote host closed the connection]
muronglizi has quit [Ping timeout: 244 seconds]
ebopp has quit [Remote host closed the connection]
pr06lefs has joined #nixos
cript0nauta has joined #nixos
cript0nauta has quit [Remote host closed the connection]
janneke has joined #nixos
boxscape has joined #nixos
cript0nauta has joined #nixos
<boxscape>
hmm if I'm in a nix shell with agda installed, and I try to run an external command in neovim with :!, I get "bash -c: option requires an argument". If I'm not in that nix shell, I can run commands in neovim just fine. Any idea what might be causing that?
<boxscape>
I had the same problem when trying to start programs with gdb a while back, but without agda - I didn't realize it was nix related then, and didn't manage to fix it
<Nafai>
Who can I talk to about transferring ownership of https://github.com/travisbhartwell/nix-emacs to the Nix org? Because of work and other things, I don't have time to actively work on NixOS related things, and I've been missing some issues that have been filed and such. I think it would be better if this were supported by the Nix community so it can be kept up to date.
knupfer has quit [Remote host closed the connection]
knupfer has joined #nixos
<jwaksbaum[m]>
boxscape: I don't have this issue: running :!echo hi from neovim inside nix-shell -p hello gave me no issue. Can you give some more info on your setup?
gentauro has joined #nixos
<boxscape>
jwaksbaum[m] First of all, I do get the same issue with "nix-shell -p hello". Though actually I just realized that since I'm on a machine right now I should probably update before debugging this...
boxscape_ has joined #nixos
<boxscape>
(brb rebooting)
boxscape99 has joined #nixos
meh`_ has joined #nixos
boxscape has quit [Ping timeout: 256 seconds]
mallox has quit [Quit: WeeChat 2.8]
mallox has joined #nixos
takataka has joined #nixos
janneke has quit [Quit: janneke quits Mes'sing]
boxscape99 has quit [Quit: Connection closed]
boxscape has joined #nixos
<takataka>
Hi, what to overwrite to modify nvidia package pulled in by services.xserver.videoDrivers = [ "nvidia" ]; ? I have tried nvidia_x11 from linuxPackages_latest and kernelPackages but it must be something else. Any ideas?
<boxscape>
I'm using zsh as my shell - if I run the nix-shell from sh instead, I apparently don't get the problem. That's a workaround at least. (Apart from that I'm using the unstable channel and should have the current version from unstable of everything relevant)
janneke has joined #nixos
boxscap8 has joined #nixos
mgdm has left #nixos ["WeeChat 1.9.1"]
basilbunting[m] has joined #nixos
<boxscape>
uh wait nevermind I get the issue from sh as well
boxscape_ has quit [Ping timeout: 260 seconds]
<boxscape>
ah, but only if I start sh from a nix-shell I started in zsh, otherwise it's fine
leotaku_ has joined #nixos
kenran has quit [Ping timeout: 258 seconds]
kreyren_ has joined #nixos
kreyren has quit [Remote host closed the connection]
<boxscape>
is there a way to tell nix-shell not to run .bashrc? Or alternatively go into sh instead of into bash?
<jwaksbaum[m]>
boxscape: this reminds me of an issue with prompts getting mangled in nix-shell when started from zsh, but now I can't remember what that was about... I know I don't have that issue any more. Might have had something to do with locales for some reason, but could be wrong.
<takataka>
When I uncomment services.xserver.videoDrivers = [ "nvidia" ]; I see how it pulls in an additional and unpatched nvidia_x11. I have no idea how and why.
<takataka>
That ''nvidia'' is an alias for kernelPackages.nvidia_x11 and kernelPackages is an alias for linuxPackages and here for boot.kernelPackages = pkgs.linuxPackages_latest;
<nobrak>
If I'm using dockerTools.fromImage, should I expect to lose configuration from the source image? I'm looking to add on to an image,b ut my resulting image no longer has the env vars and entrypoint that the base image has.
<takataka>
I have a working override for nvidia_x11 but somehow that services.xserver.videoDrivers uses unpatched version.
<jwaksbaum[m]>
jco: so that means there's no more issue? Also, if it's helpful, I think there's a special function just for the purpose of making development environments like this one, instead of using stdenv.mkDerivation with no src: mkShell (https://nixos.org/nixpkgs/manual/#sec-pkgs-mkShell). Also, if you want it to just be used as a shell, you can put it in shell.nix instead of default.nix.
<boxscape>
I'm trying to get a nix-shell to open sh instead of bash (which runs zsh via .bashrc). If I run nix-shell --command sh *from* sh, it opens zsh. If I run the same command from zsh, it opens sh. But I need it to open sh from sh. I
<boxscape>
not sure why that happens in the first place
<thequux[m]>
hazel: I've got it running on my apu4; the default USB worked fine. I don't remember if I had to pass console=ttyS0 though
<hazel>
alright
kleisli has joined #nixos
<boxscape>
well, commenting out the relevant line in .bashrc temporarily works as a workaround...
<hazel>
did you have to do anything to get serial working on a host machine, if your host ran NixOS thequux[m]?
<LambdaDuck>
How would you install just a closure and not all the build dependencies from a package on cachix? The closure is about 59 Mb and the full install that you get with a default install is gigabytes (I'm not sure how to check the exact size).
<thequux[m]>
One second while I look at my config to be sure; I do seem to recall that there was something necessary to get it to fully work
buckley310 has quit [Quit: Connection closed for inactivity]
sbetrium_ has quit [Read error: Connection reset by peer]
<hazel>
that's *on* the APU, yeah?
orivej has quit [Ping timeout: 240 seconds]
<hazel>
I'm talking about being able to connect to the APU
orivej_ has joined #nixos
<hazel>
albeit that's appreciated!
<thequux[m]>
Oh; no, just use your serial terminal of choice.
sbetrium has joined #nixos
<hazel>
hm yeah but I'm not getting a /dev/ttyUSB0...
<hazel>
i might be missing some kernel module
boxscape_ has quit [Ping timeout: 260 seconds]
<hazel>
or not in some group
<hazel>
dunno
<thequux[m]>
What chip is in the adapter you're using? Some show up as ttyACM0
<infinisil>
LambdaDuck: Nix only builds derivations if they aren't already in your store and it can't download them from a cache either. And if it builds a derivation, it needs all dependencies of it, which again are first looked up in your local store, then in caches, and otherwise are also built, and so on.
<hazel>
i'm using the pcengines usbcom1a, and it isn't showing up at all
sbetrium has quit [Read error: Connection reset by peer]
<thequux[m]>
Huh, that implies that there's a hardware problem. At the very least you'd get a message saying that a new device was enumerated
<jco>
what do you give to the src field if you want to build a package which source is in the current working directory?
<jco>
there's fetchurl but the source are stored on the machine
Maxdaman1us is now known as Maxdamantus
<hazel>
`modprobe cp210x` works but doesn't do anything
<hazel>
great
<hazel>
lemme try another box
boxscape_ has joined #nixos
<LambdaDuck>
infinisil: It doesn't seem to build anything for me. It just downloads all the build dependencies before downloading the package. I think it is some kind of IFD problem. This is the program I'm refering to: https://github.com/srid/neuron/blob/master/project.nix Here's a log of it downloading a bunch of unnecessary things:
<hazel>
..." The USB ID is NOT serialized so adapters can be exchanged without undue hassles / repeated driver install on mildew-brained operating systems."
<hazel>
what does this mean
knupfer has quit [Quit: knupfer]
knupfer has joined #nixos
<infinisil>
LambdaDuck: Nix wouldn't be downloading them if they wouldn't be needed
<infinisil>
Doesn't matter whether it's IFD or not
knupfer has quit [Client Quit]
knupfer has joined #nixos
<LambdaDuck>
infinisil: They aren't needed for the final product. At least not according to nix-store. Maybe the construction of the derivation for the final product is not cached?
knupfer has quit [Client Quit]
knupfer has joined #nixos
<cransom>
hazel: i assume it means that it doesn't add it as 'ttyUSB0', ttyUSB1, ttyUSB2, etc.
<hazel>
well, if it's not even showing up in dmesg, how do i interface with it?
<hazel>
is it just DOA?
<infinisil>
LambdaDuck: They might be needed to build the result, or needed to evaluate the result derivation (that's IFD)
<thequux[m]>
I think that Windows does something stupid where if the serial number is different, it needs to reinstall the driver
<hazel>
oh ew
<cole-h>
Nafai: I think adisbladi or etu have permissions to add stuff to tyhe nix-community org -- maybe get in touch with one of them?
<{^_^}>
[nixpkgs] @snglth opened pull request #91060 → utahfs: init at 1.0 → https://git.io/JfFBR
<cransom>
if you re-plugged after modprobe, you could try a full power down. I have some embedded stuff with cp210x chips that will crash my mac if i pull them, or cause the usb buson linux machines to disappear until a hard power off.
<thequux[m]>
hazel: Do you have any other USB-serial adaptors? You might just have a dead adapter
knupfer has joined #nixos
<hazel>
nope
<hazel>
I have an FTDI adapter but I'd have to do some jank
boxscape_ has quit [Read error: Connection reset by peer]
<wrl>
hazel: are you using a null modem cable?
<wrl>
apu serial port needs one
<hazel>
shouldn't need to, it's a pcengines adapter
<hazel>
pretty sure unstable syncs with nixpkgs master on regular intervals
<qyliss>
chloekek: no, it's the same as the Nixpkgs unstable channel (the nixpkgs-unstable branch)
<LambdaDuck>
infinisil: So, what workarounds are possible? It would help a lot if I could download 22 Mb instead of 2Gb to the CI servers and elsewhere if possible. Could the problem be that the derivation files themselves are not uploaded to cachix? It does say things like "building '/nix/store/jb3kx6m6b6lm4bn2q8mqjkd75g485g11-cabal2nix-aeson-1.4.5.0.drv'". Or is the problem related to `callHackageDirect`?
<rschulman>
Does anyone else have any experience with this package?
<qyliss>
chloekek: but generally you don't want to be installing packages from master
<hazel>
what's the matrix channel for this IRC
<qyliss>
often they won't be cached yet, and if they have tests those won't have been run
<chloekek>
I always pick the latest commit on master when pinning a Nixpkgs for a new project. :)
<qyliss>
You should pick the latest unstable instead
<hazel>
check out nmattia/niv
<qyliss>
It's only a couple of days behind, but it's passed certain automated tests
<hazel>
it's good for doing things like pinning unstable nixpkgs
<qyliss>
Unstable is just the last "known good" version of master
<qyliss>
And it's almost never more than a week old
<infinisil>
LambdaDuck: Looking through the logs, there's nothing unnecessary there I'm pretty sure
<infinisil>
So there is no workaround, because there is no problem
sbetrium_ has quit [Quit: sbetrium_]
knupfer has joined #nixos
<wrl>
hazel: found it. shows up immediately
<infinisil>
I guess cabal2nix is rather bloated, most likely coming from haskellPackages.callCabal2nix. In recent nixpkgs versions this is reduced considerably
<{^_^}>
#81125 (by madjar, 16 weeks ago, merged): cabal2nix: split into a lightweight version and a wrapper
<LambdaDuck>
infinisil: What do you mean? I only want the closure for the final product. Not the build dependencies. I don't want cabal2nix.
<hazel>
damn
<hazel>
let me try rebooting with "cp210x" enabled
mallox has quit [Quit: WeeChat 2.8]
<hazel>
brb
<infinisil>
LambdaDuck: You need to because of IFD
hazel has quit [Remote host closed the connection]
<infinisil>
IFD essentially means "I need to first build these things to even know what the final product should be"
<LambdaDuck>
So the only solution would be to not IFD?
<infinisil>
If possible that would improve it yes
<LambdaDuck>
Is there any possible workaround that would force it to say "I know this is the product form the IFD, you don't need the IFD dependencies"?
<infinisil>
Not really
<infinisil>
Pretty sure
NeoCron has quit [Ping timeout: 256 seconds]
<LambdaDuck>
So the only other solution would be to host the nar elsewhere and download that?
<LambdaDuck>
Btw, building it with -j0 fails, despite everything being in a cache. Is this expected?
sbetrium has joined #nixos
<srid>
LambdaDuck: There is a way to 'strip dependencies' in Nix. I tried that approach once. Let me find that ...
<hazel>
okay well I rebooted and the serial adapter's fine
<LambdaDuck>
Huh, why is that set?
<hazel>
now the problem is that the NixOS installer isn't broadcasting over serial for some reason
<infinisil>
LambdaDuck: "allowSubstitutes: If this attribute is set to false, then Nix will always build this derivation; it will not try to substitute its outputs. This is useful for very trivial derivations (such as writeText in Nixpkgs) that are cheaper to build than to substitute from a binary cache."
<LambdaDuck>
Ah, right. Except when they aren't.
<infinisil>
And -j0 fails because it tries to build something, this ^ (as seen by the "building ..." lines in your log)
<srid>
LambdaDuck: I was told that `enableSeparateDataOutput` would ditch those dependencies, but it caused Nix to complain of cyclic dependency between the binary and the data derivation. There was something like propagated-build-inputs or something that unnecessary referred the cabal library package.
<srid>
infinisil: ^
<thequux[m]>
hazel: You need to enable add the console=ttyS0,115200 kernel parameter to get console output on serial
<infinisil>
LambdaDuck: The derivation is very cheap to build, in this case it's just that the dependencies aren't already present, meaning it's more expensive in the end
<CFS-MP3>
I have a question about right way to go about a package that generates two libraries from the same .tgz. It relates to this diff: https://github.com/NixOS/nixpkgs/pull/90706
<CFS-MP3>
that PR builds libprom. In the same tgz however, just in a different directory, they also added libpromhttp
<srid>
I'm not sure how allowSubstitutes is relevant, because when you install neuron from cache nothing (including neuron) is built from source.
<thequux[m]>
hazel: If you have a NixOS machine handy, you could also try building your own installer. I've done it many times and it's ridiculously easy
<CFS-MP3>
I'm discussed it at work a bit and there's no consensus, some people prefer to just have two different derivations even if the source is the same, and other people prefer to have one derivation with multiple outputs
<LambdaDuck>
infinisil: Exactly.
<srid>
LambdaDuck: Also keep in mind that neuron is using a way older nixpkgs (via reflex-platform), but maralorn is planning to fix that.
<infinisil>
srid: And that with an empty store?
<infinisil>
Because IFD won't need to download anything if you already have the necessary derivations locally
<maralorn>
srid: I must say getting reflex to build in nixpkgs was a walk in the park compared to what I still need to do to your default.nix.^^
<hazel>
thequux[m]: how do I do that? edit it in GRUB?
janneke has quit [Quit: janneke quits Mes'sing]
<hazel>
also, how do I send input? namely, keys on a keyboard nor keys in minicom navigate GRUB...
<thequux[m]>
yeah
<LambdaDuck>
infinisil: But why is everything after the cabal2nix necessary? The final closure still doesn't contain the library dependencies. The final step just contains a whole bunch of copying.
travelion787 has joined #nixos
<hazel>
alrighty
<srid>
maralorn: Once reflex and reflex-dom is in nixpkgs, I'd think it would be a piece of cake to port neuron? What am I missing?
janneke has joined #nixos
<infinisil>
LambdaDuck: You're Nix, and I'm telling you "You need to determine the dependencies for this project, here's how you can get the tools for that". The first thing you need to do is actually get the tools before you can do anything
anotherAevin has joined #nixos
<maralorn>
srid: Nothing in principle. In practice I don‘t have an overview over what features the nix files of neuron do offer. Just writing a default.nix that compiles the binary, would probably work.
<thequux[m]>
hazel: IIRC, you can edit the syslinux configuration on your installer device as well
anotherAevin has quit [Remote host closed the connection]
<hazel>
I just need to be able to *do things* in GRUB, but I don't know how to send input
<infinisil>
LambdaDuck: I think if `allowSubstitutes` were true, it would instead be "Here's where you can download the result of somebody else having determined all the dependencies"
<infinisil>
The tools in this case is just cabal2nix
<infinisil>
But also, if you were to use a later version of nixpkgs with #81125 in it, cabal2nix wouldn't be so big to download
<LambdaDuck>
infinisil: Yes, I understand why cabal2nix and all its dependencies are needed, but I don't quite get why all the dependencies after line 339 are needed.
cole-h has quit [Ping timeout: 240 seconds]
<infinisil>
Runtime dependencies?
<LambdaDuck>
infinisil: Why are they not in the closure according to nix-store then?
hazel has quit [Remote host closed the connection]
<infinisil>
LambdaDuck: How did you determine this?
<quinn>
haskell people: i'm trying to add brittany to my shell.nix for a haskell repo (that uses developPackage) and i'm not sure what the "elegant"/"correct" way to do it is
<LambdaDuck>
Or nix-store -q -R "/nix/store/pz08vmgv1y1vlvx0jfjwdwjvhs06246d-neuron-0.5.5.1"
<infinisil>
LambdaDuck: And the last one gives you what result?
<srid>
maralorn: if the new default.nix will compile the binary, in addition to working on the CI, that's all we need. then we can nuke the project.nix stuff and reflex-platform including the ./deps/* stuff.
<thequux[m]>
hazel: check if minicom has hardware flow control enabled. If so, turn it off
<infinisil>
You have /nix/store/pz08vmgv1y1vlvx0jfjwdwjvhs06246d-neuron-0.5.5.1, CI has /nix/store/gjz0jdxq64zri4kszp2bm5ncajwxjjcd-neuron-0.5.5.1
<LambdaDuck>
Right, this is on macos. I'll upload the command to ci.
<srid>
LambdaDuck: what's under "/nix/store/j6w8a63v99z8hhyay1di6k8f7chxmzqq-source" ?
<infinisil>
Pretty sure macos does dynamic linking differently
<infinisil>
Which can explain the difference
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<srid>
I think now I remember. THe haskell binary seems to have a hardcoded path to the cabal library path under a different derivation, which in turn pulls in the whole slew of Haskell deps.
<srid>
infinisil: Oh yea. But why-depends spits out "does not depend on". Very odd
<stevenxl>
Folks. I do not know how to use nix, but I am working on a project that does use nix. (https://github.com/NorfairKing/hastory/pull/29). When I try to run `nix-shell` and then `nix-build`, I get a failure. I believe that the failure stems from the fact that nix does not know where to find a Haskell library called "base64". However, I do not know how to fix this problem.
<{^_^}>
NorfairKing/hastory#29 (by StevenXL, 1 week ago, open): Multi-User Server
* stevenxl
The particular error I get is :error: anonymous function at /nix/store/ax59027mi351rzq6wqir45yxcm7x2y6v-cabal2nix-hastory-server-data/default.nix:1:1 called without required argument 'base64', at /nix/store/3c0k1lhzwlr9l5bzilpfdrmjsnyid6li-source/pkgs/development/haskell-modules/make-package-set.nix:87:27
<infinisil>
srid: If `nix-store -qR <path>` shows <otherpath>, then `nix why-depends <path> <otherpath>` should work
<srid>
Oops, my error
<infinisil>
Oh you need `$(realpath result)`
<srid>
I didn't type the full path to the store, including the "slug" at the end
<cole-h>
flokli: One use case where a string-path is useful (re: ^) is when, for whatever reason, something needs sudo in its path. If you add the package, it won't have suid, so you'd want to add /run/wrappers/bin
<cole-h>
This is something somebody recently ran into (though they may have been going about it wrong...)
<gregf>
trying to print the current generation in my motd, I tried ${system.generation} but that errors out on build. Can someone tell me the right way to do this?
<jnorkus>
When nix learning resources state that you can have multiple versions of the same package installed, do they generally mean multiple versions installed under /nix/store, instead of meaning multiple versions available in the current environment?
pjt_014 has joined #nixos
<hazel>
multiple versions and then you can switch between them with nix-shell yes
<jnorkus>
hazel++ thanks
<{^_^}>
hazel's karma got increased to 1
<hazel>
i have it set up for rust projects (for example) that nix-shells in my project directories requiring nightly use nightly rust, but it's stable system-wide
<hazel>
you don't actually need nix for that exact use-case (rustup could do it) but it's an example
justanotheruser has quit [Ping timeout: 246 seconds]
neeasade has quit [Remote host closed the connection]
<gchristensen>
how are you copying it to the USB device, hazel?
dansho has joined #nixos
* srid
published today's investigation of cabal Paths_* mess and using remove-references-to fix it, in https://www.srid.ca/04b88e01.html
<{^_^}>
[nixpkgs] @rnhmjoj opened pull request #91072 → [20.03] mutt: patch for CVE-2020-14093 → https://git.io/JfF2k
<hazel>
dd gchristensen
<gchristensen>
dd if=./my.iso of=/dev/sda ?
Baughn_ has joined #nixos
Baughn_ is now known as Baughn
rajivr has joined #nixos
<bigvalen>
How do you get "nix build" to use multiple cores ? It doesn't seem to have a --max-jobs option
<hazel>
correct
agsdheidjd has quit [Ping timeout: 272 seconds]
<hazel>
well, it's an SD, so /dev/mmcblk0
<hazel>
but
civodul has quit [Quit: ERC (IRC client for Emacs 26.3)]
cjpbirkbeck has joined #nixos
asymptotically has quit [Quit: Leaving]
dermetfan has quit [Ping timeout: 272 seconds]
<gchristensen>
as long as you sent it to the raw device and not to a partition
<gchristensen>
mmc though you might need to add more kernel modules
<cjpbirkbeck>
hello, how do you add keys to ssh-agent?
<hazel>
it's over a usb reader for a memory stick gchristensen
<hazel>
but whatever
<hazel>
any other ideas? obviously this isn't working
<hazel>
what even is /dev/root
<cjpbirkbeck>
i've enabled `programs.ssh.startAgent = true;`, but it does not seem to remember that i've entered my passphrase
liori has joined #nixos
<DigitalKiwi>
cole-h: abathur i can do what i want now. i had to change the shell to bash and add a .bashrc that sourced the nix.sh and nix-daemon.sh (i don't know if i need both but i have both...). can't figure out what file i need to make it work with zsh but w/e
<DigitalKiwi>
cole-h: abathur so now i can do this [kiwi@mvp-nixos pr-nixpkgs]$ nix-build . -A mlv-app --argstr system x86_64-darwin # from my macbook pro, that runs nixos, that has a qemu kvm of macos, that has nix in it... :|
<cole-h>
:D
cjpbirkbeck has quit [Quit: cjpbirkbeck]
<abathur>
changed the user shell, or?
<DigitalKiwi>
chsh -s /bin/bash
<DigitalKiwi>
now it just whines at me that i'm supposed to use zsh
<abathur>
and that's on the system you're issuing from, or the vm?
<DigitalKiwi>
The default interactive shell is now zsh.
<DigitalKiwi>
To update your account to use zsh, please run `chsh -s /bin/zsh`.
<cjpbirkbeck>
okay, it seems like ssh-agent isn't working despite being enabled and running. it never saves my keys, so i always have retype my passphrase. i feel like i'm missing something obvious but i don't what it is.
<hazel>
i highly doubt that is it but i will do it anyway
<cjpbirkbeck>
but with the ssh-agent, entering the passphrase once should put it automatically into the agent, so that you don't have to enter it again, right?
<hazel>
not if it doesn't have somewhere to store that data
leungbk has joined #nixos
<cjpbirkbeck>
hazel: so do you enable it store the data in given location?
<hazel>
nvm I was doing dd bs=128K
<hazel>
do not do that
hazel has quit [Remote host closed the connection]
m0rphism has quit [Ping timeout: 240 seconds]
timemage has quit [Quit: .]
werner291 has quit [Remote host closed the connection]