2018-07-17

<clever> iqubic: hydra would allow caching things between builds, while travis would loose that cache on every build
<clever> iqubic: not that hard
<clever> it merges 2 sets, overwriting any attributes that collide
<clever> so its imposible to modify a after rec has copied it
<clever> > rec { a = 42; b = a; } // { a = "but it takes place before any other operator"; }
<clever> now it ignores the old a, and uses the a within the set
<clever> > rec { a = 42; b = a; }
<clever> there was an a already defined elsewhere
<clever> > { a = 42; b = a; }
<clever> it allows you to refer to other attributes in the current set
<clever> rec is almost never what you want
<clever> kalbasit: line 44 refers to the value on line 44 directly
<clever> kalbasit: also, because of the rec on line 7, your not even referencing the value defined elsewhere
<clever> kalbasit: all list based options append automatically
<clever> dhess: nope
<clever> dhess: not that i know of
<clever> dhess: nixos wont restart containers upon making declarative changes
<clever> dhess: so it will now be pre-building (and blowing all my free space) both machines every time nixpkgs updates
<clever> dhess: it has also done the same with the router
<clever> dhess: the nas has successfully built itself on hydra, no job took over 5 seconds, and it was mostly recreating config files without the secrets
<clever> dhess: this maps over the config.containers set, and creates a containers/${name}.conf for each key in the set
<clever> dhess: i cant see where that name truncation is coming from
<clever> dhess: i sometimes do when i run services for a friend and dont fully trust the service's security
<clever> lol
<clever> dhess: and if you check in this dir?
<clever> EnvironmentFile=-/etc/containers/%i.conf
<clever> dhess: yeah, it looks like it is being truncated
<clever> and uses the value after the @ to compute the full path
<clever> it should contain a script that refers to another dir
<clever> dhess: read that .service file
<clever> dhess: does it have the same shortened name in /etc/systemd/system/ ?
<clever> dhess: ah yeah
<clever> Jul 17 00:19:00 nas hydra-evaluator[13174]: /nix/store/fxlvm8c0dv46ciq6cbfrm6kkg12c2vwk-hydra-2017-11-21/bin/.nix-prefetch-git-wrapped: line 187: awk: command not found
<clever> dhess: still failed?!
<clever> dhess: hydra has been re-deployed, now eval'ing
<clever> you can just pick the rev from a channel, and fetch it from nixpkgs
<clever> rotaerk: that only matters if you reference a branch
<clever> rotaerk: nixpkgs has to load the libGL in /run/opengl-drivers/ to match what your gpu needs
<clever> thats one file it doesnt seem to cache
<clever> dhess: nixos will automatically repair it at the next rebuild-switch or boot
<clever> dhess: sudo rm /etc/nix/machines
<clever> [clever@system76:~/nixos-configs]$ nixops deploy -d house
<clever> iqubic: i use nixops to manage my NAS and router, both are within the house
<clever> dhess: my hydra-configs also support PR building
<clever> iqubic: i can configure hydra to auto-build anything i want, against the latest version of any (and even every) channel
<clever> dhess: i have git submodules in nixos-configs, and this version of hydra doesnt support that
<clever> iqubic: the CI system for nix
<clever> dhess: now i create a project in hydra, direct it towards https://github.com/cleverca22 and nixos-configs/spec.json, and wait
<clever> dhess: preparing a 2nd commit now...
<clever> dhess: most of these where origianlly normaly nixos machines, that i migrated to nixops
<clever> dhess: yeah, check the deployments directory
<clever> iqubic: yes
<clever> and thats also how you corrupt files
<clever> thats a firmware feature
<clever> iqubic: i have done that on gentoo before, but i dont remember how
<clever> rotaerk: no need for hard
<clever> rotaerk: pressing the power button normally triggers a normal reboot
<clever> rotaerk: same thing happened to me, it killed something dbus related and broke the entire gui
<clever> dhess: i also have notes on that
<clever> dhess: let me throw one together...
<clever> dhess: something i was thinking of today, is to make a nix expression so hydra builds my entire deployment
<clever> infinisil: still has the same problem that backport commits have a different hash
<clever> nixos-18.03 lacks it
<clever> now just change the branch in the UI
<clever> the state of this file, at the commit you linked
<clever> so this trick only works for things that where commited before the fork
<clever> the backport commit would have a different hash
<clever> rotaerk: its currently only present in nixos-unstable, but there might be a backport commit in a stable branch
<clever> rotaerk: open that commit on the nixpkgs-channels github, if you paste a rev, i can give an example
<clever> dhess: ah, utils is just a random collection of functions uses commonly in nixos, but not nixpkgs
<clever> nixpkgs/nixos/modules/misc/extra-arguments.nix: utils = import ../../lib/utils.nix pkgs;
<clever> dhess: i no longer have every line of nixpkgs memorized!!
<clever> dhess: i notice that the networking module is refering to a argument option by the name of utils!
<clever> yeah
<clever> dhess: can you pastebin the backtrace from --show-trace ?
<clever> dhess: that only gets ran if you refer to the pkgs_i686 argument in any nixos module
<clever> dhess: overlays dont set nixpkgs.pkgs
<clever> rotaerk: for gui programs, the version of nixpkgs you built the project from has to be fairly similar to the nixpkgs the host OS was built from
<clever> rotaerk: the libGL stuff is a bit crazy, due to how it has to support multiple GPU's
<clever> dont know then
<clever> ah
<clever> dhess: the patch probably didnt fix containers
<clever> dhess: i think your doing a darwin build of the container, not a linux build
<clever> kalbasit: the path refered to by ../.. must not begin with a .
<clever> and treating it as a string will copy it into the nix store, and return the path of that copy
<clever> > "${./.}"
<clever> toString returns the input path
<clever> all paths are absolute automatically
<clever> > toString ./.
<clever> sslh is probably creating its own copy of every connection
<clever> or you have a serial killer in the other room :P
<clever> so its coming from inside your machine
<clever> the input is 127.0.0.1
<clever> and the server port is fixed
<clever> the client picks a random port for its end
<clever> tobiasBora: thats how the network stack works
<clever> tobiasBora: youll have to catch one in the established state to know what exactly is causing both ends
<clever> tobiasBora: time_wait means the connection was already closed
<clever> kalbasit: nixos will search for <nixos-config> in $NIX_PATH
<clever> revskill: it appears to be currently building atlas, and it should start the next build when that finishes
<clever> revskill: thats what a pastebin is for
<clever> revskill: its already building something, according to `ps aux`
<clever> revskill: what is the content of `/etc/nix/nix.conf` ?
<clever> kalbasit: yes
<clever> revskill: that is not an error from `stack build`
<clever> i dont even see an elementary set in my end
<clever> jtojnar: oh, maybe elementary-greeter doesnt accept a stdenv argument
<clever> revskill: can you also pastebin the full output when it fails?
<clever> revskill: can you pastebin the output of `ps aux` ?
<clever> revskill: nixos or normal linux?
<clever> revskill: do you have build slaves configured?
<clever> revskill: what command did you run?
<clever> revskill: what command did you run?
<clever> jtojnar: try removing eself:
<clever> i was thinking of keepDebugInfo
<clever> nvm, that is the correct way to use enableDebugging
<clever> 369 enableDebugging = pkg: pkg.override { stdenv = stdenvAdapters.keepDebugInfo pkg.stdenv; };
<clever> 368 # intended to be used like nix-build -E 'with <nixpkgs> {}; enableDebugging fooPackage'
<clever> jtojnar: i think your using enableDebugging wrong, not overrideScope
<clever> kalbasit: nixpkgs.overlays = [ (import ./overlay1.nix) ];
<clever> kalbasit: you have to put a list of (already imported) overlays in nixpkgs.overlays
<clever> rename it?
<clever> tobiasBora: it might be that your own sslh.nix is being disabled by the same flag?
<clever> samueldr: i think thats recursively resolving every single attribute in the set
<clever> tobiasBora: when disabledModules works correctly, it will fail claiming services.sslh.enable is not a valid option
<clever> > pkgs.pkgs.pkgs.pkgs.pkgs.pkgs.hello
<clever> tobiasBora: what exactly did you do to disable the modules?
<clever> kalbasit: super will give you the version without all the changes this overlay is doing
<clever> kalbasit: id also recomend using self, not super in line 32
<clever> tobiasBora: the lack of an error is odd, can you double-check the files on the rpi and see if rsync is to blame?
<clever> kalbasit: super and self are both package sets, nearly identical in function to pkgs itself
<clever> kalbasit: you can just do `with super;` to start with
<clever> tobiasBora: systemd actively ignores all things in systemPackages
<clever> tobiasBora: you also dont need to put it into the systemPackages
<clever> tobiasBora: and what is the contents of configuration.nix?
<clever> tobiasBora: what is the content of deploy.sh?
<clever> tobiasBora: what output did it print?
<clever> tobiasBora: did you nixos-rebuild?
<clever> dhess: only ec2 and none
<clever> you are returning a new set of packages to override
<clever> super is an argument being passed in, and it cant be modified
<clever> samueldr: missing a {} on the callPackage
<clever> kalbasit: an overlay is still needed to make nixpkgs actually read the new package

2018-07-16

<clever> lostman: and share is $out/share/
<clever> lostman: the docker image scripts automatically put the $out/bin/ stuff in $PATH for you, so you dont have to think about /usr
<clever> lostman: can you gist the nix file your using?
<clever> lostman: if you do echo $buildEnv inside the nix-shell, what do you see?
<clever> but then you cant set the rev
<clever> `builtins.fetchGit "ssh://git@github.com/nixos/nixpkgs"` is identical to what i gave before
<clever> it can accept both a string and a set containing a url field
<clever> ah, much better
<clever> oh, how does this render? `builtins.fetchGit { url = "ssh://git@github.com/nixos/nixpkgs"; }`
<clever> dmj`: in addition to a better OS, you may also want an irc client that doesnt mangle everything :P
<clever> dmj`: i resent the above msg over hangouts
<clever> that
<clever> 2018-07-16 19:32:40 < clever> dmj`: builtins.fetchGit { url = "ssh://git@github.com/nixos/nixpkgs"; }
<clever> refer to the example i just gave
<clever> dmj`: you need to set the user to git
<clever> dmj`: builtins.fetchGit { url = "ssh://git@github.com/nixos/nixpkgs"; }
<clever> dmj`: it does
<clever> dmj`: Git not git
<clever> dmj`: builtins.fetchGit
<clever> dmj`: nixpkgs != builtins
<clever> dmj`: its very clearly not
<clever> dmj`: that is not from builtins.fetchGit
<clever> dmj`: and what is the content of /nix/store/8b5d4v9qilhdpn698qw3jq8n1x5iq0j5-postal-4ce8f5e.drv ?
<clever> dmj`: can you pastebin the output?
<clever> so you can just ignore the -m32 stuff
<clever> buckley310: the above will give you a gcc that is 32bit only, along with 32bit glibc
<clever> buckley310: nix-shell -E 'with import <nixpkgs> {}; pkgsi686Linux.stdenv.mkDerivation { name = "name"; }'
<clever> dmj`: what if you run nix-build with -vvvvv ?
<clever> dmj`: switch to a better OS? :P
<clever> dmj`: you have to reboot into the recovery OS to do that
<clever> you need to turn off system integrity protection
<clever> dmj`: execsnoop is a darwin tool
<clever> dmj`: run execsnoop in another terminal while its fetching, and pastebin the output
<clever> dmj`: builtins.fetchGit doesnt use nixbld
<clever> dmj`: yeah, i'm not sure why it isnt working for you
<clever> dmj`: this is how you allow the ssh agent to work inside a nixbld user
<clever> then try the nix build again
<clever> dmj`: run ssh-add on the path to your private key
<clever> dmj`: what about `ssh-add -l` ?
<clever> dmj`: what is the content of ~/.ssh/config ?
<clever> dmj`: and if you `ssh git@github.com` what does it say?
<clever> the whole point of the builtin one is that you dont have to do that
<clever> dmj`: no
<clever> dmj`: thats why i said to use the builtins one
<clever> dmj`: that one is more complex to use
<clever> dmj`: builtins.fetchGit does not use <ssh-config-file>
<clever> dmj`: builtins.fetchGit can use your ssh agent
<clever> tobiasBora: read the .service files in /etc/systemd/system/
<clever> jtojnar: i have noticed xfce sometimes restores my xterm and pavucontrol windows on login
<clever> dmj`: you can run the unpackFile function on the tarball to unpack it
<clever> dmj`: the default unpackPhase will unpack it for you
<clever> dmj`: why do you need fetchurl to unpack?
<clever> aanderse: youll probably want to look at deployment.keys in the nixops manual then
<clever> aanderse: i would just switch to lets encrypt, let the machines auto-generate their own certs
<clever> dmj`: all attributes of a derivation become env vars during the build
<clever> dmj`: if IN_NIX_SHELL isnt set, then its probably a nix build
<clever> ah, just lagging a bit
<clever> infinisil: and that part of your bot is dead
<clever> <patchelf-hints.sh>
<clever> infinisil: a script to do most of this already exists
<clever> typetetris: nix-store -qR
<clever> LnL: yep
<clever> [root@amd-nixos:~]# nix-store --query --hash /run/current-system
<clever> sha256:0mbiapybw5sscf854pjik5k8f6j25aar2qaf7k45ny4l0d4a9z3j
<clever> LnL: oh yeah, let me check
<clever> gchristensen: next thing to do would be to open db.sqlite in the sqlite3 repl and confirm the hash in the db
<clever> gchristensen: try it on the local system as well, just incase
<clever> typetetris: haskellPackages just worked with the latest nixos-unstable
<clever> gchristensen: did you run it on the source or destination?
<clever> typetetris: you can also switch to haskell.ghc824.packages.ghcWithHoogle to change the ghc version
<clever> gchristensen: the hashes are missing from that export, you need to `nix-store --verify --check-contents` to hash everything
<clever> gchristensen: was this db.sqlite loaded from a closure expore?
<clever> typetetris: this should work
<clever> typetetris: nix-shell -p 'haskellPackages.ghcWithHoogle (ps: [ ps.gi-gtk ])'
<clever> and the .env at the end drops me into an env suitable for building the package described by that cabal file
<clever> typetetris: this is an example of something i did last week, loading a cabal file, with overrides to disable version constraints on tasty
<clever> nix-shell -E 'with import <nixpkgs> {}; ((haskellPackages.override { overrides = self: super: { tasty = haskell.lib.doJailbreak super.tasty; }; }).callCabal2nix "serokell-util" ./. {}).env'
<clever> typetetris: can you pastebin more of the error?
<clever> typetetris: which derivation fails tests?
<clever> typetetris: -p will insert the argument into the buildInputs of a derivation, so you can use it
<clever> typetetris: nix-shell -p 'with import <nixpkgs> { }; haskell.packages.ghc843.ghcWithHoogle(p: with p ; [ gi-gtk haskell-gi-base base ])'
<clever> typetetris: so you need to put that whole ghcWithHoogle into the buildInputs of another derivation, and wrap some of the haskell packages with haskell.lib.dontCheck to turn off the tests
<clever> typetetris: that drops you into a shell suitable for building ghc with packages, not using ghc with packages
<clever> i think the bot is just offline
<clever> ,library
<clever> i think it was ,foo=bar
<clever> ,
<clever> its normally not an issue, because nix-build/nix-instantiate dies shortly after
<clever> gchristensen: that desyncs the nix protocol
<clever> gchristensen: did you ctrl=c at one point?
<clever> you could either make a custom iso for it, use the netboot server (in the same dir), or just manualy add it to the imports of /etc/nixos/configuration.nix after booting a normal iso, and nixos-rebuild to make it work
<clever> that file is setup as a nixos module, meant to be added to the imports section of the installer media
<clever> tenten8401: it always wipes it when you run justdoit
<clever> tenten8401: this is a script i had made to automate some of my installs, luks is one of the options it supports
<clever> Lisanna: i think hydra will just not start A
<clever> tenten8401: is ssh enabled?
<clever> tenten8401: nix-store -qR /run/current-system | grep nvidia
<clever> zenspider: not really
<clever> zenspider: i dont remember how its done on darwin, i dont do much with macs
<clever> zenspider: it auto-creates that dir the first time you install something as the ryan user
<clever> zenspider: the default profile will have all of the programs, and when you run `nix-env -iA nixpkgs.something`, it will create the ryan dir
<clever> zenspider: that is expected to be empty on a multi-user install
<clever> disasm: yeah
<clever> zenspider: ah, so you simply didnt have nix-info installed yet
<clever> zenspider: what paths is it downloading?
<clever> disasm: i'm guessing it uses the default ~/.ssh/id_rsa
<clever> zenspider: try restarting the daemon with launchctl and then see if that fixes nix-info
<clever> zenspider: can you pastebin the entire output from the installer?
<clever> Lisanna: yeah, that would explain things
<clever> Lisanna: i also suspect its just a ui frontend, and if you can find the http request made when you restart failed jobs in a different eval, you can manually make that api call in the browser
<clever> Lisanna: yeah, it probably should be filed on the issue tracker
<clever> Lisanna: not really, youll need to push a change of some kind that triggers a new eval, or just manually restart everything
<clever> Lisanna: only the 2nd eval onward can have bulk restarts
<clever> Lisanna: there is a bug that the first eval in a jobset lacks the option to restart things
<clever> zenspider: i'm not sure then, you may need to delete /nix/ and re-do the entire install
<clever> Lisanna: click on the eval#, and then open the actions menu
<clever> zenspider: even if you spawn a new terminal window and source it there?
<clever> it may only be a problem in nix.sh, just re-source it
<clever> zenspider: i suspect that you have something set that causes nix.sh to abort half way thru running, so you need to stop that MANPATH error from happening
<clever> zenspider: can you pastebin all of those files
<clever> zenspider: i'm guessing that you have some non-standard bash options set in your .bashrc or .bash_profile or such
<clever> it said to source nix.sh
<clever> the installer doesnt tell you to run nix-shell
<clever> nix-shell reads the shell.nix or default.nix file
<clever> thats normal
<clever> zenspider: try adding that to PATH
<clever> zenspider: what binaries are in /nix/var/nix/profiles/default/bin/ ?
<clever> zenspider: what output did the install script give?
<clever> zenspider: and does `which nix-env` show something now?
<clever> zenspider: 1
<clever> zenspider: what about: source /nix/var/nix/profiles/default/etc/profile.d/nix.sh ?
<clever> tobiasBora: no idea, but you could ask #toktok
<clever> zenspider: try source ~/.nix-profile/etc/profile.d/nix.sh
<clever> tobiasBora: yeah, toxvpn also has an option to change the port, but it also needs to reach the port the remote systems are using
<clever> zenspider: are you using bash?
<clever> tobiasBora: and its basically just PM'ing ethernet frames thru the tox framework, lol
<clever> tobiasBora: you would run toxvpn-remote on both peers, and tell them to connect to the other party
<clever> tobiasBora: it doesnt really have a client or server, its fully p2p
<clever> tobiasBora: android itself does support vpn's without rooting, but i havent modified toxvpn to use that
<clever> tobiasBora: i dont have a proper android port yet, but if you happen to have linux in a chroot, it will probably work there
<clever> tobiasBora: give each machine a unique ip, preferably in an unused private range, and then link the peers with toxvpn-remote