2019-08-21

<clever> EsperLily: at one time, i was using a bouncer, and had mirc+irssi sharing the connection, and mirc scripts configured to beep at 15khz
<clever> Linux c2d 3.8.13-gentoo #3 SMP Thu Dec 15 10:33:05 AST 2016 x86_64 Intel(R) Core(TM)2 CPU 6300 @ 1.86GHz GenuineIntel GNU/Linux
<clever> ..... it still runs on an old gentoo box, that i never migrated after adopting nixos.....
<clever> i just use screen + irssi + ssh
<clever> ElectricJanice: nix-env -iA nixos.irssi to get the client i prefer using
<clever> ElectricJanice: a proper irc client is the key, the web clients can have issues when the browser tries to suspend things
<clever> ElectricJanice: ive had no trouble staying online for weeks to months at a time
<clever> its definitely a multiple of itself :P
<clever> you could even do bs=1287651328 count=1
<clever> 8192 works
<clever> > 1287651328 / 8192
<clever> 157184
<clever> and it must be a multiple of the size
<clever> if you make bs bigger, you must divide count by that
<clever> dd if=/dev/sda bs=1 count=size-in-bytes | sha256sum -
<clever> dd can also do it
<clever> gchristensen: you have to stop hashing at the right point
<clever> gchristensen: i dont expect it do, since the usb stick is going to be bigger, and have garbage after the iso
<clever> ElectricJanice: what does dmesg say?
<clever> symphorien: but also, the hash for the partitions arent published
<clever> gchristensen: it will likely read beyond the end of the iso file, if the usb stick is bigger
<clever> so when you nixos-rebuild, it vanishes
<clever> Thra11: it may have been set when building the image, but is not present in the configuration.nix included inside the image
<clever> Thra11: you probably want that util
<clever> ,locate growpart
<clever> exarkun: you have to add zlib to the buildInputs in the shell.nix file stack is told to use
<clever> niso: all i can think of is to clear the cache in /var/lib/hydra/scm/git and try again
<clever> niso: first thing i would check, can you just `curl https://github.com` from the hydra machine, if you ssh in?
<clever> niso: this is the code that downloads the build inputs
<clever> niso: oh, this is in a hydra input, thats not channel related then
<clever> niso: is there more output, what command did you run, what did it print before and after that error?
<clever> it may not have config options for it
<clever> niso: yeah, that has its own timeout
<clever> niso: thats a curl timeout, while downloading channels, not a nix build timeout
<clever> freeman[w]: bios boot must be 1mb
<clever> freeman[w]: you need to delete sda3, then make a 3&4, one is /boot, the other is bios boot
<clever> grub destroyed the /boot because you have the wrong partition type code on /boot
<clever> freeman[w]: and thats seperate from the /boot partition
<clever> freeman[w]: bios boot should be ~1mb, it has no fs, and cant be mounted
<clever> freeman[w]: sda3 is the bios boot, that should never be mounted
<clever> freeman[w]: and `blkid /dev/sda*
<clever> freeman[w]: ls -l /dev/disk/by-partlabel/boot and fdisk -l /dev/sda
<clever> what is the full output when it does fail?
<clever> which is probably no timeout
<clever> [root@amd-nixos:~]# nix show-config | grep timeout
<clever> timeout = 0
<clever> niso: it will use the default timeout of the machine doing the build
<clever> niso: yeah, and in some cases, it doesnt work at all
<clever> nixos-install
<clever> freeman[w]: you can just mount things again, fix the config, and re-run nixos-install
<clever> freeman[w]: what was boot.loader.grub.device set to?
<clever> freeman[w]: what fs did you use for /boot ?
<clever> freeman[w]: you didnt set services.xserver.enable
<clever> freeman[w]: what is the contents of your configuration.nix?
<clever> iqubic: you can file an issue against this repo
<clever> 2019-08-21 02:01:09 < clever> iqubic: i think the JS is crashing while trying to pretty-print the example
<clever> 2019-08-21 01:58:42 < clever> iqubic: hit f12, look at the js console
<clever> time to file a bug!
<clever> iqubic: i think the JS is crashing while trying to pretty-print the example
<clever> iqubic: then try again in `man configuration.nix`
<clever> iqubic: hit f12, look at the js console
<clever> then check the home-manager source
<clever> iqubic: look both up on here, click the declared in link
<clever> dont remember
<clever> i had to use mkAfter to force it to come after
<clever> this would have wound up before the creation of nixos-fw, causing it to fail
<clever> mkBefore and mkAfter can force order
<clever> depends on the order the files are put into imports
<clever> iqubic: the type is lines, so nixos will just join them with \n's
<clever> same, havent looked into why, i try to avoid darwin
<clever> extra-sandbox-paths definitely sounds custom
<clever> i dont think single-user ever makes it
<clever> EsperLily: single-user doesnt create /etc/nix/nix.conf
<clever> EsperLily: exactly :D
<clever> `sudo -i` does the same thing, try it
<clever> pick one tool and stick to it, dont mix identical tools :P
<clever> sudo and su do basically the same thing, its like `sudo sudo`
<clever> use `sudo -i`, not `sudo su`
<clever> and the code you give to home-manager could source the one in etc
<clever> iqubic: your ~/.zshrc could just source /etc/zshrc
<clever> probably
<clever> iqubic: it lands in /etc/bashrc, and gets ran at the usual time by bash

2019-08-20

<clever> hpfr[m]: run nix-build on this, and see how it differs from yours
<clever> > linuxPackages.v4l2loopback
<clever> 86 #extraModulePackages = [ config.boot.kernelPackages.v4l2loopback ];
<clever> hpfr[m]: boot.extramodulepackages should be adding it, confirm that the drv has the module in the right path
<clever> hpfr[m]: boot.kernelmodules is only obeyed on bootup, you need to manually `modprobe` it if you dont want to reboot

2019-08-19

<clever> tilpner: the path you gave isnt cached
<clever> 404
<clever> and/or the kernel parsing it wrong when executing nix-shell
<clever> tilpner: possibly related to nix-shell having to parse its own #! line
<clever> tilpner: '' always eats the indent that is common to every line
<clever> > '''''''
<clever> freeman42x: you forgot .meta.description
<clever> evanjs: never actually used the new search util
<clever> $ nix eval nixpkgs.hello.meta.description
<clever> "A program that produces a familiar, friendly greeting"
<clever> __monty__: did you update the hash after changing it?
<clever> lol-md5: lib/maintainers.nix i believe
<clever> Shyim[m]: (mkIf (cfg.server.type == "nginx") {
<clever> cinimod: do you have a pkgutil in your overlays or config.nix?
<clever> i dont see pkgutil in that version
<clever> freeman[w]: 2019-08-19 13:23:08 < cinimod> $NIX_PATH is empty
<clever> cinimod: what git rev of nixpkgs are you on?
<clever> cinimod: what about pkg, then tab twice?
<clever> cinimod: its listing all 10,000 packages
<clever> cinimod: what about pkg, then tab twice?
<clever> cinimod: then you want `nix repl ../../nixpkgs`
<clever> Taneb: 2019-08-19 13:23:08 < cinimod> $NIX_PATH is empty
<clever> which made it very difficult for him to read the directions
<clever> infinisil: a few days ago, somebodies irc client ate all <things> because it was html-y
<clever> its likely the value of $NIX_PATH
<clever> ah, then its not your irc client
<clever> cinimod: what do you see when i say <nixpkgs> ?
<clever> or nix repl ../../nixpkgs
<clever> cinimod: you can also just: nix repl '<nixpkgs>'
<clever> iqubic: if you want the new systemd features
<clever> iqubic: systemd updated
<clever> iqubic: you need to reboot your system
<clever> prior to placeholder, you had to read the $out env var, when the builder is running
<clever> > hello.builder
<clever> and those are args bash expects
<clever> > :p hello.args
<clever> > hello.args
<clever> ah, good point
<clever> Guest16: also, touch is a symlink, so it matters what argv[0] is
<clever> Guest16: i think args includes argv[0], not entirely sure though
<clever> Guest16: if your doing this in nix repl, its using the previous value of d
<clever> Guest16: but the hash of d.outPath depends on d.args
<clever> Guest16: i would expect d.outPath to cause infinite recursion, what happens if you rename the first d and not the 2nd?
<clever> pbb: was just thinking it might be logs from a prior boot
<clever> pbb: does the timestamp match up right in the logs?
<clever> pbb: not sure what the error is, try restarting sshd?
<clever> pbb: does sshd exist in /etc/passwd ?
<clever> pbb: how did you enable sshd?
<clever> hpfr[m]: you can just use fetchzip directly, it already copies the thing to its $out
<clever> thats how you load out-of-tree modules
<clever> boot.extraModulePackages = [ config.boot.kernelPackages.v4l2loopback ];
<clever> hpfr[m]: all modules land in /run/booted-system/kernel-modules/
<clever> hpfr[m]: yeah
<clever> hpfr[m]: hardware.firmware in configuration.nix
<clever> which is just a text file
<clever> but behind the scenes, --add/remove/list operate on ~/.nix-channels
<clever> nope
<clever> iqubic: read --list, then --add and --remove
<clever> both booted-system and firmware are symlinks
<clever> hpfr[m]: nixos looks in /run/booted-system/firmware
<clever> craige: https://nixos.org/nixos/manual/index.html#sec-replace-modules and/or packageOverrides, to pull in the changes from master
<clever> craige: the fix will likely need to be backported to 19.03
<clever> > builtins.typeOf [ "foo" ]
<clever> > builtins.typeOf "foo"
<clever> vika_nezrimaya: line 3 of your pastebin is trying to do string ++ list
<clever> vika_nezrimaya: makeFlags can be both a list or a string, and in thi case, its a string
<clever> > uwimap.makeFlags
<clever> vika_nezrimaya: can you pastebin your nix expr?
<clever> vika_nezrimaya: list
<clever> tazjin: nix-env resolves it via ~/.nix-defexpr/
<clever> vika_nezrimaya: you gave something a list, when it expected a set
<clever> you want nix-env -f '<nixpkgs>' -iA cachix
<clever> from nix-channel --list
<clever> tazjin: that tells it to use the channel called nixpkgs
<clever> tazjin: that doesnt tell it to use NIX_PATH
<clever> tazjin: what args are you using with nix-env?
<clever> i think thats mainly dealing with split output logic
<clever> behind the scenes, builtins.derivationStrict is the true primop
<clever> > builtins.removeAttrs hello ["type"]

2019-08-18

<clever> or set.modify { b = 43; } to add an attr
<clever> you can now do set.modify { a = 43; } to mutate it
<clever> > let original = { a = 42; }; set = original // { modify = new: original // new; }; in set
<clever> last nixcon
<clever> so you can re-run builtins.derivation with a modified set
<clever> then it will use // to modify the return value, and add a .overrideAttrs function, which an reference the original set
<clever> behind the scenes, stdenv.mkDerivation will modify the set some, then pass it to builtins.derivation
<clever> > let set = { outPath = "foo"; }; in "${set // { a = 42; }}"
<clever> > let set = { outPath = "foo"; }; in "${set}"
<clever> but that wont impact the build
<clever> and you can add whatever you want to it
<clever> the return value is also a set
<clever> your using // on its return value
<clever> pie_: the derivation is the arguments to builtins.derivation
<clever> pie_: // doesnt change the derivation
<clever> arcnmx: its using editline
<clever> pie_: you probably want to set it to [], rather then use remoteAttrs
<clever> pie_: id say start by trying to reproduce it when building nix from git, and then find a version that works and doesnt
<clever> pie_: ive had that issue as well, havent looked into why yet
<clever> lambda-11235: .overrideAttrs { src = ./.; }
<clever> lambda-11235: -I nixpkgs=/path/to/nixpkgs
<clever> ah, that will do it
<clever> tokudan: the supplied configuration.nix does that
<clever> tokudan: you need to include kexec.nix in the imports
<clever> Gilfoyle-: yep
<clever> Gilfoyle-: https://nixos.org/nixos/manual/index.html#sec-replace-modules is still your answer
<clever> and then pnsd is not powerdns
<clever> one powerdns is clearly powerdns, the other is pdns
<clever> that one is powerdns
<clever> wut
<clever> > pdns-recursor.src.urls
<clever> isnt helping matters :P
<clever> > pdns-recursor.meta.description
<clever> with conflicting names
<clever> they look to be entirely different programs
<clever> > powerdns.src.urls
<clever> > pdnsd.src.urls
<clever> > powerdns.meta.description
<clever> > pdnsd.meta.description
<clever> > pdns.meta.description
<clever> pdns runs ${pdnsd}/bin/pdnsd
<clever> powerdns runs ${pkgs.powerdns}/bin/pdns_server
<clever> Gilfoyle-: powerdns does completely different things, and yes, --setuid=nobody
<clever> what was the other paste link? i cant find it
<clever> extraConfig gets inserted on line 21 of the file i linked
<clever> Gilfoyle-: that looks nothing like the config nixos generates
<clever> Gilfoyle-: how did you enable pdns?
<clever> Gilfoyle-: what about the config file?
<clever> Gilfoyle-: what channel are you on?
<clever> Gilfoyle-: the nix expr says it should be running as pdnsd
<clever> ah, but thats currently not an option, and it will likely fail to bind to port 53 if you set serviceConfig.User
<clever> Gilfoyle-: but the service appears to already support dropping root on its own, after startup
<clever> Gilfoyle-: no need for an override, nixos will automatically merge the definitions
<clever> also, have fun re-implementing modprobe, its more complicated then you think
<clever> infinisil: if you re-implement nix, fdisk, mkfs and dhcp in haskell, you could do it with 1 binary :P
<clever> infinisil: this is a haskell binary, that runs as pid 1, in the initrd
<clever> infinisil: you could also go the insane route
<clever> Gilfoyle-: add .serviceConfig.User = "foo"; to its definition
<clever> which is better then static, because glibc isnt duplicated
<clever> so you get the dyn libs, binaries, and nothing else
<clever> this will copy a bunch of bin/foo files, then recursively copy all libs they depend on, and re-patchelf them
<clever> the initrd tools would be a better model
<clever> not sure about static, that can hurt if you have a lot of binaries
<clever> so as long as you have a basic shell, nix, some partitioning utils, and mkfs.*, you can install nixos
<clever> of note, this lets you copy a closure to a chroot on a remote machine
<clever> nix copy --to ssh://root@target?remote-store=local?root=/mnt /nix/store/hash-nixos
<clever> make something highly specific, that can only install nixos
<clever> i mean not-os'ing the image :P
<clever> right now, its just a full blown nixos install
<clever> something else i should look into, is a more specialized kexec image
<clever> nice
<clever> though there is also a squashfs within there, that still remains compressed
<clever> kexec is likely undoing the initrd compression
<clever> 776mb with zfs compression for me, 972mb without, yeah, thats pretty tight
<clever> infinisil: ah, you want --max=0
<clever> --apparent-size to disable that
<clever> though du obeys zfs compression by default
<clever> then do: du -h --max=1 -c $(nix-store -qR ./result) | sort -h, i think
<clever> instead of aiming -A at kexec_tarball, aim it at config.system.build.toplevel
<clever> infinisil: guess you cant trust the error codes, also check the closure
<clever> /etc/NIXOS_LUSTRATE just automates it for you, by moving it all to /old-root/
<clever> `rm -rf /etc /bin /usr` fixed it, lol
<clever> nixos was obeying things like the old dbus config, and using binaries from gentoo
<clever> i basically just ran switch-to-configuration on gentoo, then debugged and fixed the problems it caused
<clever> and i forgot it was even an option
<clever> infinisil: funny thing about that lustrate method, is that its based on tales of how i manually did it to a gentoo netboot, lol
<clever> if ram is the issue
<clever> ah, that should also work
<clever> and it will overwrite the bootloader and assume full control of the box
<clever> infinisil: just create that file (it can be zero bytes) and run the `switch-to-configuration boot` for a nixos build
<clever> infinisil: ohhh right, nixos-infect is using the lustrate method
<clever> infinisil: how does the current nixos-infect stuff work then...
<clever> infinisil: you can also check the source, both to see how the checks are done, and if any undocumented checks exist
<clever> this doesnt exist on my machine either, so its probably nothing
<clever> ls: cannot access '/sys/firmware/edd': No such file or directory
<clever> ah
<clever> and it shouldnt be a ram issue
<clever> ENOMEM Could not allocate memory.
<clever> that should be easy to check
<clever> EINVAL The file referred to by kernel_fd or initrd_fd is empty (length zero).
<clever> is it the same nixpkgs rev? the closure size of the image may have grown
<clever> infinisil: you might not have enough ram for the initrd
<clever> infinisil: how much ram?
<clever> infinisil: uname -a?
<clever> infinisil: and what does strace say failed?
<clever> (or modify and rebuild)
<clever> infinisil: read the kexec script, and run kexec with --debug added to it
<clever> infinisil: dmesg say anything?
<clever> and it just wont answer until it has checked its own caches, and then report what it couldnt get