nixos

If anyone wants to help package COSMIC in nixpkgs (and for NixOS) join the Matrix room :3

I hope this is the right place to ask. I am new to Nix and trying to use it for now as a universal package manager. I installled Stremio using Nix but the app doesn't work properly. I see that it's several versions out of date. The same app works fine using Flatpak. Is there a way to let the maintainer know? I'd also love some documentation that can help me update it myself. I can just use the Flatpak but I just want to have one less package manager to worry about.

This is something that I struggle with. I know how to find top-level packages like git or cowsay. But what about utilities under nested paths? I always spend ages digging through the nixpkgs source code to try to find utilities to use in my nix expressions.

Today I want to use buildRustPackage. It's defined here, and is propagated here. But how do I access it given a pkgs variable? I have no idea!

https://search.nixos.org/packages is no help

nix search nixpkgs doesn't find it

I think I need to search by attribute name, not by derivation name. But I don't know how to do that.

It took me a while to figure this out. I use Home Manager to manage my Gnome settings by setting dconf.settings = { ... }. My settings are non-trivial (for example my paperwm module). So it's helpful for me to check the actual dconf settings that Home Manager produces.

To do that build your configuration with home-manager build, open result/activate, and find a line that looks like this:

$DRY_RUN_CMD $DCONF_DBUS_RUN_SESSION /nix/store/4ab7dx08wx640444m71axlqvbrvz73bv-dconf-0.40.0/bin/dconf load / 
  < /nix/store/0hdnvwx8d9sifd6ib8n2hhblyblq0ccp-hm-dconf.ini

The store path for hm-dconf.ini has the settings.

Edit: added a line break to the script line so you can see the relevant store path

I'm developing a small library which intercepts some library calls from a program installed via home-manager. My library works fine outside of nix, but it fails when trying to intercept calls from a program installed via home-manager. I figured that it is related to hardcoded library paths in my interceptor library.

lld shows

linux-vdso.so.1 (0x00007fff7a341000)
libasound.so.2 => /lib/x86_64-linux-gnu/libasound.so.2 (0x00007fb2f3b72000)
libstdc++.so.6 => /lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007fb2f391c000)
libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007fb2f38f8000)
libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007fb2f3716000)
libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007fb2f3637000)
/lib64/ld-linux-x86-64.so.2 (0x00007fb2f3cac000)

What is the preferred approach here?

• Changing the paths in my binary doesnt seem very portable?
• Setting LD_LIBRARY_PATH seems also a hassle
• Are there linker settings which could help
• Or, should I "just" (no idea, how to do this) create a derivation for my program and install it via home-manager, too?

Any pointers are greatly appreciated. Cheers!

Could be areas of improvement as well

Tried NixOS !

So I tried nixos and I love how things work, I am glad that it exists. Will learn about Flakes today. Any recommended resource to learn it? :)

@nixos@lemmy.ml

I want to call the escapeSystemdPath (defined in nixpgs at nixos/lib/utils.nix) to derive the name of a systemd mount unit from the target path (eg. srv-my-dir.mount from /srv/my/dir), but I can't figure out how I can reference it... any ideas?

I want to add custom commands to my PATH but am unable to get it working. So far I've installed home-manager and added the code below to my nix configuration.

imports = [
  ./hardware-configuration.nix
  
];

Then later on to add the item to PATH

home-manager.users.curtismchale = { pkgs, ... }: {
  home.stateVersion = "23.05";
  home.sessionPath = [
    "$HOME/Workspace/proudcity/proudcity-kubernetes/bin"
  ];
};

I have also tried $HOME/Workspace/proudcity/producity-kubernetes/bin/pc-kube but that doesn't change the issue.

I can see the commands installed at ~/Workspace/proudcity/proudcity-kubernetes/bin/pc-kube but when I type the expected command pc kube $command it says there is nothing installed in path and offers to install some programs that match pc for me.

I'm playing around with nixos in a few VMs and at some point I realized I must have lost the swap configuration in one of my refactorings.

To my surprise, however, the VMs do use the swap partitions I had set up.

There is no mention on "swap" in my nix configuration (or in fstab) and no .swap units in /etc/systemd/system; I do however have a swap partition labelled "swap".

Turns out there is a systemd unit (albeit not a corresponding file) that sets up swap:

[root@vm1:~]# free -hw
               total        used        free      shared     buffers       cache   available
Mem:           2.8Gi       664Mi       955Mi       4.0Mi       3.0Mi       1.3Gi       2.0Gi
Swap:          3.7Gi          0B       3.7Gi

[root@vm1:~]# systemctl list-dependencies swap.target 
swap.target
● └─dev-disk-by\x2ddiskseq-1\x2dpart3.swap

I'm wondering where the unit comes from? Can I rely on this and never configure swap ever again?

I have a workaround so this isn't exactly a problem for me. I'm just curious about what is going on, and what best practices are.

I'm setting up Arion. I think it will be helpful for my development flow in a project where I have several services that need to run and communicate with each other. Docker-compose is a nice way to handle this, but you have to have a Docker image to run, and it's a pain to create a new image after each code change. OTOH Arion will run an arbitrary command, and creates Nix-friendly images automatically. Very promising!

The Nix expression for the service I'm developing is exported from a flake, while the arion executable reads its configuration from a Nix expression that is not a flake. There is an example configuration that recommends importing a flake using builtins.getFlake which you can see here: https://github.com/hercules-ci/arion/blob/main/examples/flake/arion-pkgs.nix

The problem is that builtins.getFlake is slow. It adds >20s to every arion command I run. That includes starting services, reading logs, removing stopped containers, etc.

The example config includes a fallback that loads the flake using flake-compat instead of builtins.getFlake. When I use flake-compat loading the flake is nearly instant.

So I'm using flake-compat, and that seems to be working. (Many thanks to the flake-compat author!) But I'm curious why builtins.getFlake is so slow.

I have an option that must be left with the default value when a certain flag (another option) is false.

I didn't find any example (let alone documentation) on how to implement this, so I've come up with two ideas:

option-that-errors-out-if-set-when-flag-is-false =
let
  default = if config.some-flag
          then "some default value for when flag is true"
          else "value that should not be changed when flag is false";
in lib.mkOption {
  type = lib.types.str;
  inherit default;
  apply = v: assert assertMsg (config.some-flag || v == default) "Do not set this option unless 'flag' is true";
          v;
};

option-that-ignores-value-when-flag-is-false =
let
  default = if config.some-flag
          then "some default value for when flag is true"
          else "value that should not be changed when flag is false";
in lib.mkOption {
  type = lib.types.str;
  inherit default;
  apply = v: if config.some-flag then v else default;
};

Which one do you think is "best" (cleaner, more idiomatic, etc..)?

Is apply the "right" place to validate options? Should I make a custom type instead? Should I approach this in some different way?

I'd like to set a "global" option from within a submodule, but the config I return is grafted into the "global" under the submodule "path" rather than at the root... any idea if it's somehow possible?

Er... I guess I didn't make a great job at explaining what I want to do... Some code will hopefully help.

In mymodule.nix I have:

{ lib, config, ... }: {

  options.myoption = lib.mkOption {
      type = lib.types.attrsOf (lib.types.submodule (
        import ./mysubmodule.nix
      ));
  };

}

and mysubmodule.nix is:

{ name, lib, config, ... }: {

options.mysubmoduleoption = {
  type = lib.types.str;
};

config = {
  # here I want to set a "global" option, say "systemd.mounts"
  # based on the value of "mymodule.name.mysubmoduleoption"
  # but it seems I can only set values under "mymodule.name" 
};

}

Instead of getting plugins through nixpkgs I prefer to use my neovim-specific plugin manager. (In my case that's lazy.nvim.) Mostly this works without problems - but some setup is required when a plugin needs to compile something. The plugin that has given me the most trouble is Treesitter which wants to compile grammars. Here is how I got that working.

tl;dr: Configure Treesitter to compile grammars with gcc instead of clang.

As has been reported in https://github.com/nvim-treesitter/nvim-treesitter/issues/1449 Treesitter will try to use clang to compile Treesitter grammars, and on NixOS for some reason clang is not able to locate necessary C++ libraries. The fix that works for me is to configure Treesitter to use gcc instead. Here is the relevant part of my plugin config:

return {
  'nvim-treesitter/nvim-treesitter',
  build = ':TSUpdate',
  config = function()
    -- Set compiler to get grammar installation working in NixOS. See
    -- https://github.com/nvim-treesitter/nvim-treesitter/issues/1449
    require('nvim-treesitter.install').compilers = { 'gcc' }
    require('nvim-treesitter.configs').setup {
      ensure_installed = 'all', -- "all", or list of languages
      ignore_install = { 't32' }, -- t32 is failing to download for me
    }
  end,
}

I still had a problem with the t32 grammar, but I don't need that one so I disable it.

Of course you need to make sure that gcc is available. You could put it in your user profile, but I prefer to make sure by using the extraPackages option from Home Manager's neovim module. Here's my full config:

programs.neovim = {
  enable = true;
  defaultEditor = true;
  withPython3 = true;
  extraPackages = with pkgs; [
    fd
    gh # for github integration
    ripgrep

    # needed to compile fzf-native for telescope-fzf-native.nvim
    gcc
    gnumake

    # language servers
    nil # Nix LSP
    lua-language-server

    nixpkgs-fmt # I have nil configured to call this for formatting
  ];
};

I'm trying to debug a module I'm writing with builtins.trace, but it's being more complicated than I anticipated.

Let's say I have a module:

{ config, lib, pkgs, modulesPath, ... }:

{

  config =
  let
    some-list = lib.attrsets.mapAttrsToList (n: v: {
        some-attr = "${n} ${v}";
    }) { n1 = "v1"; n2 = "v2"; };
  in {
    users.mutableUsers = builtins.trace (some-list) false;
  };

}

This will print

trace: [ <code> <code> ]

because builtins.trace (for whatever reason?) evaluates its first argument only shallowly.

Changing the trace expression to:

builtins.trace (builtins.toJSON some-list) false;

helps a lot, but as soon as one tries to print a long list or a structure with some complexity the output is completely unreadable, and it's not like it can easily be piped into jq (I mean... &| grep ^trace: | sed 's/trace: //' | jq works*, but there must be a "better" way?)

(*) in fish shell, IDK about bash

edit: It's not like I specifically want JSON output: any format will do (ideally, nix would be nice)

Edit: Solved at https://lemmings.world/comment/1719409

Hi there! I'm trying to make php and composer work. I have this in environment.systemPackages:

    (pkgs.php82.buildEnv {
      extensions = ({ enabled, all }: enabled ++ (with all; [
        xdebug
        redis
      ]));
      extraConfig = ''
        memory_limit=2G
        xdebug.mode=debug
      '';
    })
    php82Extensions.redis

The problem is that while running php -m correctly prints that redis extension is installed, composer does not, because it uses a different php:

• file $(which php) prints the path /nix/store/igx8j4qjxy9jyj8kjyccwarnzqq5vsml-php-with-extensions-8.2.9/bin/php
• cat $(which composer) shows that it's a wrapper for '/nix/store/lv4prxa52zifr54ws56iz3b9kdhs1b5w-php-with-extensions-8.2.9/bin/php' --add-flags '/nix/store/avqj0662f4gg2s875zlbbjajx6fm6bl0-php-composer-2.5.5/libexec/composer/composer.phar'

Note that the path to php is different. Is there any way to correct it on my side? I'd like to avoid having to install composer manually

I need to generate a number of scripts in my configuration and make them into a single package (for ease of reference, because there are a lot of them).

So far, I'm creating the scripts via writeShellApplication, making them into packages via an overlay, merging them with buildEnv and then adding the resulting package to `systemPackages.

Something like:

nixpkgs.overlays = [ (final: prev: {
  my-hello-1 = final.writeShellApplication {
    name = "my-hello-1-script";
    text = "echo my hello wolrd 1";
  };
  my-hello-2 = final.writeShellApplication {
    name = "my-hello-2-script";
    text = "echo my hello wolrd 1";
  };
  my-hello-scripts = final.buildEnv {
    name = "my-hello-scripts";
    paths = [ final.my-hello-1 final.my-hello-2 ];
  };
}) ];

environment.systemPackages = [ pkgs.my-hello-scripts ];

This works, but I don't really need the my-hello-1 and my-hello-2 packages... can you think of a way to make do without needing them?

I'm migrating my NAS to nixos, and I got to the point of setting up my restic backups.

services.restic.backups is great, but -- on top of the systemd timers/services -- I also want some helper scripts (eg. one to easily mount the backups, stuff that with ansible I currently generate into /usr/local/sbin).

These scripts would be entirely generated from the services.restic.backups config and would reference sops secrets also from configuration.nix, so... I don't think it would make sense to make a package out of them?

What should I use to make these scripts? Should I use nixpkgs.writeShellApplication and then alter the PATH?

I've been searching for a way to do this, but I haven't found anything. After I have refactored my Home Manager configuration is there a way I can test the changes in a shell before I switch?

From what I understand the next-best option is to switch, and then find and run the activate script of the previous generation to switch back.

I was configuring DWM, among other things, for the last 3-4 days, and every single rebuild switch caused a new generation to appear. There were too many Systemd-boot entries so they couldn't even fit on the screen and continued down to Gen 41. It's just crazy.

Edit: This post: https://feddit.uk/post/1454174 shows the rest of the boot entries

Im just starting out in my nixos journey. This thought came to me in the shower: would nixOS lend itself for a phone OS?

I couldn't find a nix community, so I'm hoping it is ok that I'm posting on the nixos one instead. I'll switch to mailing list/discord if necessary, but I have a lemmy app on my phone, and it is much easier to have an ongoing conversation from here, so I decided to give it a shot. Here goes!

While I have a NixOS laptop, I primarily use other systems (e.g. OpenSuse Tumbleweed) as of now.

I love the ability to define the packages I want installed, with home-manager managing my command line utilities (e.g mtr, dig, protobuf etc).

I've been playing around a bit with protobuf recently, and after generating some c++ code using protoc, I loaded up the generated code in vscode, which understandably wasn't able to find the development headers for protobuf (since they are in the nix store - /nix/store/h2h5fs8iv2a8rmlkfhr6id6y4jxwd5i1-protobuf-3.21.12/include/google/protobuf/io)

I tried to compile the code anyways on the command line, and got some errors.

I might need an OS specific protobuf install just for the development headers, but I'm pretty sure I should be able to, and just don't know how. Here's what I get when I try to compile:

$ g++ searchReq.pb.cc
In file included from searchReq.pb.cc:4:
searchReq.pb.h:10:10: fatal error: google/protobuf/port_def.inc: No such file or directory
   10 | #include 
      |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
compilation terminated.

Any tips/pointers would be appreciated. Thanks in advance!

===========================================

Edit: Thank you all, particularly @Xephopiqua@lemmy.ml and @chayleaf@lemmy.ml for the help. After setting the include path (compile) and LD_LIBRARY_PATH (link), things work great.

I ended up writing a small Makefile for convenience in the short run:

INC_FLAGS:=-I$(HOME)/.nix-profile/include -Icpp
LD_FLAGS:=-L$(HOME)/.nix-profile/lib -l protobuf

run: task
	LD_LIBRARY_PATH=$(HOME)/.nix-profile/lib ./task

task:
	g++ $(LD_FLAGS) $(INC_FLAGS) main.cpp cpp/searchReq.pb.cc -o task

regen:
	protoc --python_out=python --cpp_out=cpp searchReq.proto

That's enough to get me going for now.

TODO - read up the NIXOs Wiki C page in more detail