<clever>
hyperfekt: if the source was kept as an external module, it could be built on any kernel version
<clever>
infinisil: neat
<clever>
hyperfekt: yeah, i see it now, its using a fork of the linux kernel, with the BCACHEFS_FS flag enabled
<clever>
hyperfekt: thats changing out the entire kernel package set, and the kernel version
<clever>
hyperfekt: oh wait, no, misread that nvm
<clever>
everything else looks great
<clever>
hyperfekt: its best to use a kernel package from config.boot.kernelPackages, which has to be added to the right region of all-packages.nix, or it will wind up being of the wrong kernel version
<clever>
hyperfekt: this line will likely cause problems
<clever>
WilliamHamilton[: callPackage and callCabal2nix will then import that with the right args
<clever>
WilliamHamilton[: cabal2nix will return a nix file containing mkDerivation
<clever>
WilliamHamilton[: either run cabal2nix on that repo (either url or ./.) and then haskellPackages.callPackage it, or use `haskellPackages.callCabal2nix "name" (pkgs.fetchFromGithub { ... }) {}`
<clever>
lukego: that option is a list type, so it will merge with all other values automatically
<clever>
11 my $dbPath = "/nix/var/nix/profiles/per-user/root/channels/nixos/programs.sqlite";
<clever>
last i looked, command-not-found has that path hard-coded
<clever>
ottidmes: you can also use nix-index and nix-locate directly to update your own db
<clever>
ottidmes: you need a channel called nixos, on the root user, that came from nixos.org
<clever>
-r--r--r-- 1 root root 3.3M Dec 31 1969 /nix/var/nix/profiles/per-user/root/channels/nixos/programs.sqlite
<clever>
ah, the new one is also using programs.sqlite
<clever>
ottidmes: programs.sqlite is the older command-not-found system
<clever>
hyperfekt: command-not-found uses a database in the channel
<clever>
hyperfekt: are you using nix-channel and one of the proper channels?
<clever>
azazel: syncplay tends to catch the exception though, and then resume things, so that may not work
<clever>
and upon reconnecting, it tries to list the same dir, and fails again
<clever>
in my case, the exception isnt caught, and it unravels the stack hard enough to break my connection to the server
<clever>
so you must then spend hours looking over strace output to find out why, lol
<clever>
and if any filename fails, it throws an exception, and doesnt tell you what file caused the problem
<clever>
aleph-: i have a different python program, which runs a function to recursively list a directory, and it tries to parse every single filename as utf8
<clever>
aleph-: and it uses pythonPackages as an arg, so you can mopidy = super.mopidy.override { pythonPackages = super.python3Packages; };
<clever>
Drakonis: not sure, but i do remember that line from somewhere
<clever>
Drakonis: i'm a dude :P
<clever>
andi-: i was using spot instances for all testing, and its currently hard-coded to only do spot, i plan to make it a nix option in the deployment file
<clever>
andi-: and i need to redo that entire nixos config layer, to fetch what is pre-generated on the machine
<clever>
andi-: probably, i havent tested deploy yet
2018-11-02
<clever>
:D
<clever>
possibly
<clever>
azazel: and you need to use uwsgi.service, not uwsgi
<clever>
azazel: you probably also want to set after
<clever>
hyperfekt: you want -I nixos-config=/mnt/etc/nixos/configuration.nix
<clever>
hyperfekt: what args did you run it with?
<clever>
Mic92: :O
<clever>
now i need to loop and check the state until it gains an ip
<clever>
Mic92: so i can now create and destroy sshkeys, and machines, but the machine doesnt have an ip immediate after creating, causing nixops to break
<clever>
"string", lol
<clever>
the docs also dont tell you what it expects in any fields
<clever>
with zero type checking
<clever>
and the library is just a thin wrapper around the REST api
<clever>
Mic92: that string is directly from the json that the REST api spit out
<clever>
my first idea, was that i lacked permission to create devices
<clever>
Mic92: how would you parse this error?
<clever>
Error 422: users: "7e24caef-6317-4b2b-8830-d5e9b77f27d2" are not collaborators on this project
<clever>
that didnt help with the api giving vague errors
<clever>
Mic92: i already patched the python library to print all request and response bodies
<clever>
Mic92: so what looked like a permission error, was actually a 404 for the ssh keys listing! lol
<clever>
i assumed it was my userid, but no, it was the unique id i put in to the ssh keys list
<clever>
it then gave an error, saying .... is not a colaborator
<clever>
i assumed that the ssh keys your authorizing, are a list of unique id's for the ssh keys, from the api to create them
<clever>
half of my problem is just a lack of documentation with this cloud provider
<clever>
Mic92: heh, i'm currently adding a new backend to nixops
<clever>
ottidmes: its a somewhat recent addition, ive also been messing with service names in the copy to make them not conflict, prior to discovering that
<clever>
ah yeah, i did see it
<clever>
xourt: one of the /proc/*/maps files refers to that path, check them all with a grep
<clever>
`-I nixos-config=./configuration.nix` on most nix commands, to make them use a different configuration.nix
<clever>
imports = [ ./foo.nix ]; to load a custom module in your configuration.nix
<clever>
hyperfekt: `nix repl '<nixpkgs/nixos>'` and then eval `config` and `options` to inspect the current state of all modules in your current config
<clever>
hakujin: if recursive == false, then the path is a file, and does not support directories (fetchurl for ex)
<clever>
locallycompact: you also dont have to fork nixpkgs if you just want to build one thing
<clever>
locallycompact: running your own hydra is about the only way to build a reasonable chunk of nixpkgs
<clever>
jonge: but ive noticed that traversing the massive /nix/store/.links/ dir does get worse over time, and then even adding a new link becomes a cost
<clever>
jonge: its limited to just the storepath thats being built/downloaded, so it can optimize faster
<clever>
auto-optimise-store = false
<clever>
jonge: there is also a nix.conf flag to do it after every build/download
<clever>
i just read the nix files, all the time
<clever>
locallycompact: i consume source code for breakfast, lol
<clever>
)
<clever>
locallycompact: when you import <nixpkgs/nixos>, also pass it `system = "i686-linux";` as an argument (near the `configuration=`
<clever>
cinimod: installing libraries with nix wont make them findable by any compiler, and only the compilers within nix have been modified to be able to find them correctly
<clever>
,library
<clever>
and the /stuff isnt needed anymore
<clever>
locallycompact: and depending on what you want, also get rid of the autologinUser
<clever>
locallycompact: you can get rid of everything under virtualisation, kernelParams, systemd
<clever>
that will allow it to use up to 8g of ram, and the build should pass
<clever>
locallycompact: mount /tmp -o remount,size=8g
<clever>
locallycompact: tmpfs?
<clever>
locallycompact: df -h $TMP
<clever>
locallycompact: nixos or other?
<clever>
locallycompact: what about the host drive?
<clever>
locallycompact: your drive is full
<clever>
cinimod: use haskell.ghcWithPackages i believe
<clever>
locallycompact: that seems to have been removed in july of 2018, as it didnt actually do anything by then
<clever>
locallycompact: it needs a `let` near the start, and a `in` between lines 44 and 45
<clever>
and then you can freely edit the config on lines 5-34
<clever>
locallycompact: youll need to cut lines 2-45 into its own file, and then refer to the rootDisk attr
<clever>
locallycompact: line 43 sets the size of the disk image, 41 makes it have a partition table, 42 makes it bootable, and 35 loads a configuration.nix value (held in dom0_config this time)