2019-12-25

<clever> when i first login on linux, i would see the WINDOWS WALLPAPER on the desktop, lol
<clever> the previous drivers did all kinds of wonky things, it mostly looked like use before set bugs
<clever> and has been that flat for the full 3 years that i can see in the online billing page
<clever> ashesham`: its fairly flat year-round, between 600 and 800 kWh
<clever> ashesham`: .... i leave 3 desktops, 1 rack case, and a laptop on, 24/7
<clever> notgne2: my ups failed a while back, so ive been at the whims of random 2 second blips lately
<clever> ashesham`: 07:05:42 up 53 days 15:04, 2 users, load average: 0.00, 0.00, 0.00
<clever> 01:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Bonaire XTX [Radeon R7 260X/360]
<clever> i never put the machine into sleep, and i rarely reboot
<clever> it may vary from card to card
<clever> ive been using amdgpu lately and havent had any issues with it

2019-12-24

<clever> mpickering: are you using the correct sha256 when you fetch the src?
<clever> dirkx__: test never makes a result symlink, only build does
<clever> that should just be working...
<clever> dirkx__: what does `echo $NIX_PATH` return?
<clever> and the result symlink is updating when you `nixos-rebuild build` ?
<clever> dirkx__: what args did you pass to nix-store?
<clever> (or not changing)
<clever> dirkx__: how are you checking httpd.conf is changing?
<clever> dirkx__: if you use :b you can make nix repl build the drv an expr points to
<clever> just re-launch and re-eval
<clever> dirkx__: nix repl randomly coredumps, its a known issue
<clever> dirkx__: warnings like that can happen if you try to read the entire attrset, your reading deprecated options
<clever> dirkx__: and check services.httpd.configFile ?
<clever> dirkx__: try `nix repl '<nixpkgs/nixos>'` then try to eval things like config.services.httpd.virtualHosts and see if values come out right
<clever> dirkx__: check the module source, what does it to with extraModules?
<clever> dirkx__: what does `nix-store --verify --check-contents` return?
<clever> dirkx__: nope, its likely an issue with the config your giving it, what did you try changing?
<clever> dirkx__: try `nixos-rebuild build` and then dig the httpd.conf out of the result symlink, perhaps with `nix-store -qR ./result | grep httpd.conf` ?
<clever> doesnt need the quotes
<clever> dirkx__: how is it getting included in?
<clever> roxirc: can you pastebin the entire file?
<clever> :)
<clever> orloge: what about `nix-channel --list` ?
<clever> orloge: what did nix-channel --update print out?
<clever> orloge: nix-channel --update, as root
<clever> orloge: you may need to update your channel
<clever> or just move that remaining line (24) into installPhase
<clever> mehlon: you can delete lines 21-23, and rename it to postBuildHook i believe
<clever> mehlon: the default configurePhase will run cmake for you, and the default buildPhase runs make for you
<clever> mehlon: is enableParallelBuilding set on the drv?
<clever> mehlon: and `nproc` ?
<clever> mehlon: what about `nix show-config | grep cores` ?
<clever> then its not the nixbld group to blame
<clever> mehlon: what does the `id` command output?
<clever> bbl
<clever> dirkx_: this lets you nix-env -iA foo.hello
<clever> [clever@amd-nixos:~]$ cat .nix-defexpr/test/foo/default.nix
<clever> import /home/clever/apps/nixpkgs
<clever> duairc: for example, an x86 qmake binary, with arm qt libraris
<clever> duairc: i think just buildInputs, and the package should be specially modified to provide some native tools in the cross-compiled package
<clever> bennofs[m]: another fun one, is that ssh:// accepts another store uri, so you can chain them
<clever> bennofs[m]: i grepped my irc logs for every uri i could find, and then documented some
<clever> one min
<clever> bennofs[m]: you want local?root=/foo
<clever> Yaniel: should be fine
<clever> u1f320: nixos-unstable-small still gets tested, but updates more often, it doesnt wait for things to compile fully
<clever> u1f320: nixpkgs-unstable doesnt have tests ran on it fully, and can brick a nixos machine
<clever> ,stateVersion
<clever> duairc: postUnpack = "sourceRoot+=/socket-io; echo source root reset to $sourceRoot";
<clever> try adding pkgconfig to nativeBuildInputs?
<clever> what error is it failing with?
<clever> jakobrs: add buildInputs = [ qt5.qtbase ];
<clever> which packages did you need to install?
<clever> that will run aclocal for you
<clever> jakobrs: nativeBuildInputs = [ autoreconfHook ];
<clever> the default buildPhase just runs `make`
<clever> lsattr and the man pages show which flag it immutable, chattr can change the flags
<clever> but that also blocks deletion
<clever> the immutable bit has been set, to ensure it will ALWAYS be empty
<clever> pistache: lsattr and chattr

2019-12-23

<clever> evanjs: and due to the missing -f, we didnt see what
<clever> evanjs: something with building the derivation, but we are building that outside nixos-enter
<clever> evanjs: that will build the whole thing within /mnt, but not update the bootloader, so youll need to nixos-rebuild at the end
<clever> evanjs: outside of then nixos-enter, try doing `nix-store -r /nix/store/hmql87dbw1qa9g4x70r1fc9h21dv7kzw-nixos-system-sekka-19.09.1647.2e73f72c87e.drv --store local?root=/mnt/` ?
<clever> evanjs: nixos-enter can break that with its namespacing
<clever> evanjs: it might be garbage root collection stuff
<clever> evanjs: did you include -f in strace?
<clever> evanjs: its failing when it starts the build, so --show-trace wont do anything
<clever> evanjs: both would help
<clever> evanjs: can you pastebin more output?
<clever> evanjs: `strace -o logfile -f nixos-rebuild boot --install-bootloader` then `grep EPERM logfile` ?
<clever> evanjs: which file?
<clever> evanjs: try `nixos-rebuild boot --install-bootloader` ?
<clever> evanjs: how is it supposed to be booting? efi? legacy?
<clever> gchristensen: in cases like that, i would just manually execute it, under `sudo -u foo`
<clever> evanjs: switch will never work inside nixos-enter, you want boot
<clever> evanjs: did you run it as root?
<clever> the_pumpkin_man[: you probably need --install-bootloader when changing the ESP mount point
<clever> juhe: try nix-store --add-fixed sha256 flash_player_npapi_linux.x86_64.tar.gz ?
<clever> and then check if the timestamps changed again
<clever> the_pumpkin_man[: `nixos-rebuild boot --install-bootloader`
<clever> ah
<clever> one sec
<clever> id say try rebooting and see what happens
<clever> thenn it looks like boot0008 is the most recent one
<clever> the_pumpkin_man[: what does `ls -lh` report for both?
<clever> the_pumpkin_man[: what about the last-mod timestamp on both?
<clever> juhe: is pkgs.flashplayer.src using fetchurl or fetchzip?
<clever> the_pumpkin_man[: its loading boot0008 first
<clever> the_pumpkin_man[: look closer, at the boor order
<clever> the_pumpkin_man[: and has that file in /boot been updated?
<clever> the_pumpkin_man[: grub can use efibootmgr without adding it to the global PATH
<clever> the_pumpkin_man[: youll need to install it with nix-env, nix-shell or nix run
<clever> the_pumpkin_man[: we will also want to inspect efibootmgr -v, before we reboot
<clever> juhe: just let flash die? lol
<clever> __monty__: yeah, the above is how to mount it from the livecd for repair work
<clever> that might have been udev configured to activate lvm automatically
<clever> then you need to luksOpen, then pvscan and vgchange -a y
<clever> is the lvm inside the luks?
<clever> why is it not possible?
<clever> nixos-install is just a script that runs nixos-rebuild under a chroot
<clever> the_pumpkin_man[: just boot the livecd, mount things at the "right" place, maybe tweak the config, and re-run nixos-install
<clever> the_pumpkin_man[: you dont need to reinstall anything if you break it
<clever> the_pumpkin_man[: can you try again, but before you reboot, pastebin the output of `lsblk ; blkid ; fdisk -l ; efibootmgr -v` ?
<clever> the_pumpkin_man[: you might be better off just making /boot the ESP, and not having it encrypted
<clever> i'm not sure how you would force grub to embed the keymap into its stage1 file
<clever> the_pumpkin_man[: looks like you need to generate a keymap file, and then load it within grub, but that would normally load after it mounts /boot
<clever> the_pumpkin_man[: ah, so you have the ESP at /boot/efi, no fs at /boot, and /boot is a regular dir on the encrypted / ?
<clever> infinisil: depends on if only / is encrypted, or if /boot is also encrypted
<clever> the_pumpkin_man[: is it actually grub asking? is /boot encrypted? does the prompt happen before the grub logo?
<clever> tbenst[m]: allowUnfreePredicate passed "cudatoolkit-10.1.243" to you, lib.getName then stripped off the version info
<clever> tbenst[m]: theres the problem
<clever> nix-repl> lib.getName "cudatoolkit-10.1.243"
<clever> "cudatoolkit"
<clever> tbenst[m]: can you pastebin the entire nix expression?
<clever> tbenst[m]: thats the name your matching against
<clever> "cudatoolkit-10.1.243"
<clever> nix-repl> cudatoolkit_10.name
<clever> bots dead
<clever> ,
<clever> > cudatoolkit_10.name

2019-12-22

<clever> yeah, any program can now benefit from the modesetting
<clever> prior to kms, the panic was printed to the text console, but xorg had a strangle-hold on the gpu, so you never saw the error, and the machine just "hangs" silently
<clever> so you can actually see the panic msg
<clever> another benefit of KMS (in theory) is that the kernel can switch back to text mode when it has a panic attack
<clever> Fare: -I nixpkgs=/path/to/nixpkgs
<clever> ah, correction, it runs the shell with -n, not shellcheck
<clever> thomashoneyman: writeShellScriptBin will stick a #! at the top, and shellcheck it for you
<clever> thomashoneyman: thats what you get from `ls $out`
<clever> thats the temp dir that your cd'd into when the script starts
<clever> yeah
<clever> it never did a cd, so the ls is in $NIX_BUILD_TOP
<clever> thomashoneyman: and the --query --binding?
<clever> thomashoneyman: also, try running `nix-store --query --binding buildCommand /nix/store/foo.drv` on the drv that fails to build
<clever> ah, that behaves a bit differently from the normal checkPhase, but it should run it
<clever> thomashoneyman: can you paste your nix expr?
<clever> thomashoneyman: i dont think writeTextFile runs checkPhase normally

2019-12-21

<clever> Guest13048: what about file-roller?
<clever> Guest13048: what happens if you just run unzip on the executable?
<clever> Guest13048: can you pastebin it?
<clever> > pythonPackages.binwalk
<clever> Guest13048: try running binwalk on the installer, what does it find?
<clever> Guest13048: check to see if it has a --extract flag, or if you can extract the files without running it?
<clever> srhb: the quotes on the outside have the same effect
<clever> srhb: that will copy the entire ./nix dir, then tack a file on the end
<clever> srhb: i think `mySource = '${./nix}/some-specific-file.nix"; is much much simpler
<clever> mehlon: if you want it in $PATH, use nix-env
<clever> sphalerite: yeah, thats what i said earlier, the ratio of inodes to bytes of disk is set when you format
<clever> sphalerite: but what about inode count?
<clever> and once created, the ratio is fixed
<clever> betaboon: i believe you can only change it when creating the fs
<clever> Raito_Bezarius: just expand the EBS and reboot it, and your done
<clever> Raito_Bezarius: the aws image will resize its own rootfs on bootup
<clever> is the server in aws? nixops?
<clever> assuming its ext2/3/4
<clever> same as any other distro, expand the partition, then run resize2fs on it
<clever> and it cant be changed
<clever> Raito_Bezarius: when you format an ext filesystem, the ratio of disk space to inodes is set
<clever> Raito_Bezarius: make the disk bigger
<clever> nothing in /tmp at all?
<clever> manually delete non-nix things first, to let it run
<clever> that will delete as few things as possible
<clever> also, nix-collect-garbage --max-freed 10m
<clever> delete some things!
<clever> you have too many files
<clever> df -i /
<clever> try copying the resulting file back up to the server?
<clever> Raito_Bezarius: can you run the same query that fails in the error?
<clever> which filesystem?
<clever> and if all went right, deletes the journal when its done
<clever> it creates a journal when it begins a write, copies blocks to the journal before it overwrites them
<clever> it has journal files to prevent such things from corrupting it
<clever> the entire db engine is designed to be hardened against corruption
<clever> Raito_Bezarius: how do people keep corrupting a sqlite db? lol
<clever> Raito_Bezarius: how did the db get lost?
<clever> Raito_Bezarius: not really, it would be simpler to just delete /nix and re-run nixos-install from the livecd
<clever> so you have to manually dump things, and know your in such a state
<clever> nixos partially supports it, but will just boot normal nixos in the event of a crash
<clever> so it can then dump the ram of the dead kernel upon crashing
<clever> the crash kernel is meant to kexec over to a pre-loaded kernel, and run within a pre-reserved area
<clever> b: intentionally crash, so you run within the reserved region!
<clever> which can be used to either a: exclude ram
<clever> you can configure how much ram is reserved for it
<clever> another hacky option is the crash kernel
<clever> thats the one i was thinking of
<clever> 2373 mem=nn[KMG] [KNL,BOOT] Force usage of a specific amount of memory
<clever> mb, but its not spelled right
<clever> gchristensen: i think its maxmem=256 on the kernel cmdline
<clever> jakobrs: but the cpu can run 32bit binaries just fine, so its simpler to use a 32bit->32bit compiler
<clever> jakobrs: crossSystem is more for when you cant run the binaries your building
<clever> jakobrs: when you run nix-build, you can just add `--argstr system x86_64-linux` or `--argstr system i686-linux` to force 64 or 32bit mode when building
<clever> jakobrs: with nix, you just describe how to build it for any arch, and then nix can build a 32bit or 64bit version, as needed
<clever> jakobrs: what part of the project needs multiple arches at once?
<clever> jakobrs: my latest project involves 3 different cross-compilers, and mixing things up like that is trivial
<clever> myme: might be an issue with 19.03 then, a reboot sounds good
<clever> myme: and what does `nix-channel --list` return?
<clever> myme: what config did you try last, and how did it fail exactly?
<clever> elvishjerricco: yeah, youll have to check the bios options to try and convince it to netboot
<clever> ,unstable myme
<clever> myme: boot.kernelPackages = pkgs.nixos-unstable.linuxPackages_5_3; for example, plus
<clever> elvishjerricco: yeah, thats just general idle traffic
<clever> elvishjerricco: and i dont see any trace of dhcp in that wireshark
<clever> elvishjerricco: grub shouldnt be coming up if its network booting
<clever> elvishjerricco: screenshot?
<clever> yeah
<clever> elvishjerricco: what does wireshark show in the dhcp packets?
<clever> pie_: thats why modules have { ... }:, to ignore any extra things you add
<clever> pie_: and that will ruin all error messages
<clever> that problem, is why we used options instead
<clever> pie_: which requires the if statement to be eval'd
<clever> pie_: which requires knowing every key of config.
<clever> pie_: but to read extra, you must read config._module.args.extra
<clever> pie_: and imports gets moved up a level when it wraps
<clever> pie_: if options and config are both missing, nixos will wrap it in a { config = ...; } for you, to fix things
<clever> pie_: all modules contain options, imports, and config
<clever> pie_: and extra itself may cause the same problem, try making isVM a let block
<clever> pie_: try wrapping it as { config = if ... ; }
<clever> elvishjerricco: only if you run justdoit, you can also just ignore that script and run anything else
<clever> pie_: the problem, is that it must know if your set contains a .options, before it can compute the options
<clever> pie_: `{ config = if condition then {} else {}; }` would fix it
<clever> pie_: `if condition then {} else {}` may cause problems
<clever> pie_: can you pastebin your example?
<clever> fuzen: try creating a new user?
<clever> fuzen: only other thng that could play a factor is your home folder
<clever> elvishjerricco: you point wan towards the real router and internet, and lan towards a private ethernet card, that only has the client(s)
<clever> elvishjerricco: yep
<clever> elvishjerricco: which is why netboot_server.nix uses 2 cards, the ethernet port is isolated from the router and in full control
<clever> elvishjerricco: or plug the client into a seperate network
<clever> fuzen: that wouldnt help, the --verify already confirmed that all files nix can reinstall are already perfect
<clever> elvishjerricco: yeah, its usually best to only have one dhcp server
<clever> fuzen: it would have still made a generation for each upgrade
<clever> not sure what else to check then
<clever> that lets you just run the steam from any generation
<clever> fuzen: /nix/var/nix/profiles/system-123/sw/bin/steam
<clever> have you tried older generations? you can test them without a rollback
<clever> what changed?
<clever> or maybe its already fixed in unstable?
<clever> fuzen: you could try an older nixpkgs version
<clever> fuzen: it might be a but with the chroot env stuff, not mounting /nix into the chroot
<clever> fuzen: does that file exist?
<clever> [pid 20090] execve("/nix/store/1vsf7dahgpwacy1ab3h0hx1w2ng9dqwx-lutris-init", ["/nix/store/1vsf7dahgpwacy1ab3h0h"..., "/home/aizuzu"], 0x7fff3186a578 /* 71 vars */) = -1 ENOENT (No such file or directory)
<clever> elvishjerricco: start by opening up wireshark, and filter to dhcp and tftp packets, what is happening when it boots?
<clever> elvishjerricco: the dhcp server should report a boot filename
<clever> fuzen: can you pastebin `strace -f steam` ?
<clever> fuzen: then its not missing binaries
<clever> fuzen: nix-store --verify --check-contents ?
<clever> fuzen: what are you running to get the error?
<clever> fuzen: run the file command on the binary
<clever> fuzen: what does file say about the binary?
<clever> fuzen: are the apps properly patchelf'd?
<clever> fuzen: depends on what the args to execvp are, strace would show them
<clever> so it can figure out what a sibling of foo is, and worry about foo later
<clever> the lazyness works by transforming `mkIf condition {foo.bar = 42;}` into `foo = mkIf condition { bar = 42;}`
<clever> pie_: then you dont depend on the config tree, and dont get inf rec
<clever> pie_: use options.virtualisation.cores instead
<clever> how are you detecting that its a vm?
<clever> and normal if cant refer to config if it sets config
<clever> you need normal if
<clever> i dont think you can set those options with mkIf
<clever> pie_: and config.foo is invalid, so fail
<clever> pie_: mkIf translates turns `config = mkIf false { foo.bar = 1; }` into `config.foo = mkIf false { bar = 1; }`

2019-12-20

<clever> and the stdenv will take care of the rest
<clever> nativeBuildInputs = [ autoreconfHook ]; i believe
<clever> gustavderdrache: there is also `autoreconfHook` which will give you the autoconf binaries, and automatically run them for you, to generate a new configure script
<clever> anything you use at runtime goes into buildInputs (libraries, binaries ran at runtime)
<clever> anything you use at build time (cmake, pkgconfig) goes into nativeBuildInputs
<clever> there are system-wide proxy settings
<clever> looks normal
<clever> NixOsFun: ls -l /dev/vbox*
<clever> NixOsFun: do you see the vbox character devices in /dev/ ?
<clever> NixOsFun: does the error say what is wrong?
<clever> not sure what else to check
<clever> NixOsFun: what does `id` output?
<clever> boogiewoogie: environment.systemPackages = [ (pkgs.runCommand "thing" {} "mkdir -pv $out/share ; cp ${./foo.xkb} $out/share/foo.xkb") ];
<clever> boogiewoogie: so if the package has a $out/share, your already set
<clever> boogiewoogie: anything you put into systemPackages gets merged into /run/current-system/sw/
<clever> not sure then
<clever> coco: wifi? wired? how much bandwidth does the video consume?
<clever> coco: nixos.org/nixos/options.html and search for firewall
<clever> where did the .desktop come from?
<clever> then youll know if its even worth trying to configure
<clever> just stopping the firewall is a quick way to confirm if that is indeed the problem
<clever> coco: you could also just `systemctl stop firewall.service` short-term, to confirm if that is the issue
<clever> coco: of note, tcpdump bypasses the firewall