<clever>
fresheyeball: nix-store is not in the PATH on the remote machine your ssh'ing into
<clever>
fresheyeball: check the journal for nix-daemon?
<clever>
fresheyeball: allowed-users also
<clever>
fresheyeball: also, is trusted-users or other users flags set in nix.conf? can it build anything at all?
<clever>
fresheyeball: sudo -u jenkins -i
<clever>
benley: also, on ext4 there is a max of 65535 hardlinks for a single file, ive seen 2 instances of a file failing to hardlink
<clever>
benley: and checking for a single hardlink in a dir with millions of files can suffer on some filesystems
<clever>
benley: it requires reading and hashing the entire $out twice, once for the narhash, then again for the links
<clever>
benley: ah, maybe some verbosity with it has changed, and nobody else has noticed, because they have it off
<clever>
benley: or auto-optimize?
<clever>
benley: did you use -v?
<clever>
fresheyeball: it should be going thru nix-daemon, and using root to do the ssh
<clever>
eraserhd: since patches is empty, you can also just use patchPhase directly
<clever>
eraserhd: i would expect that to work then, what is it doing when building?
<clever>
eraserhd: phases? buildCommand?
<clever>
eraserhd: run `nix show-derivation` on the drv file, does it have a patchPhase?
<clever>
maralorn: write a derivation that will map over every one of them, and create a text file with every url, then nix-build it and keep the symlink it makes
<clever>
some of the browser flags are setup to not cause a rebuild, that may be a bug in the pepper flash option
<clever>
that would reduce the spamage
<clever>
ah
<clever>
i just run irssi under screen on my own hardware
<clever>
zfnmxt: ah, you can self-host to avoid that spam?
<clever>
so every single user that relays thru matrix disconnects at once
<clever>
matrix went down
<clever>
batzy: what was the solution?
<clever>
batzy: what was the solution?
<clever>
batzy: not sure what else to check
<clever>
batzy: further down, this stack overflow mentions b43, what does `modprobe b43` say?
<clever>
batzy: its not clear which package the driver is in, if any
<clever>
batzy: `modprobe wl` ?
<clever>
batzy: 30 seconds on google says you want the wl driver, not the ath10k driver
<clever>
batzy: nix-env -iA nixos.pciutils
<clever>
batzy: what does lspci say the card is?
<clever>
batzy: try a full reboot?
<clever>
batzy: that wont reload things the same as a reboot
<clever>
cant think of what else to check then, maybe try a reboot anyways?
<clever>
batzy: did you enable wireless in configuration.nix?
<clever>
batzy: then we need to figure out what package provides that file, and install it, google may know what its typically inside
<clever>
batzy: did the error repeat, or not show up again?
<clever>
batzy: you can also boot the installer, mount the existing fs back under /mnt (along with any partitions it needs), then re-run nixos-install to apply changes to configuration.nix
<clever>
batzy: plug in an ethernet cable?
<clever>
batzy: nope, just re-run the modprobe from before
<clever>
batzy: try adding `hardware.enableAllFirmware = true;` to configuration.nix and rebuild
<clever>
batzy: you are still connected
<clever>
any other errors in dmesg?
<clever>
batzy: or the firmware isnt installed
<clever>
batzy: dmesg?
<clever>
batzy: any change in ifconfig? dmesg?
<clever>
hpfr[m]: that should ignore the overlay then, if you comment the overlay out, do you get a different kernel?
<clever>
selfsymmetric-pa: hydra has to rebuild most of nixpkgs before the fix gets into a channel
<clever>
yeah
<clever>
so it depends a lot on how complex the pr is, and how many things depend on it
<clever>
breakage that may happen after your PR gets merged, due to things being missed
<clever>
craige: staging is mostly for when you think it has a chance to break a lot of things, and you want to give hydra time to test things, so you dont break master and hold nixos-unstable back
<clever>
oldandwise: these 2 commands print out a list of -I flags
<clever>
oldandwise: sounds like an anoying bug in pkgconfig, if the dependencies listed in the .pc file are absent, it will silently ignore the .pc file
<clever>
oldandwise: is pkgconfig in the buildInputs?
<clever>
luckily, the pkgconfig files are still in the standard lib/pkgconfig/foo-1.pc, so pkgconfig can be modified to find them
<clever>
but it cant deal with non-standard things like include/foo-1
<clever>
nix will iterate over everything in buildInputs, and put the include/ dir into -I for you
<clever>
and now you have no way to predict what the hash is
<clever>
but then nix comes along, and puts the headers in /nix/store/hash-foo/include/foo-1/
<clever>
and they expect users to be able to just blindly gcc -I/usr/include/foo-1/
<clever>
iqubic: cairo and a lot of gtk based things, typically install headers to /usr/include/foo-1, so it wont conflict with foo-2
<clever>
iqubic: pkgconfig is more to auto-detect the -I and -l flags you need to build, to deal with things being in non-standard places
<clever>
ive done that before, and it works nicely
<clever>
tobiasBora: no idea, i dont really use home-manager, nor kde
<clever>
tobiasBora: which will do the same job as .xprofile (it just gets inserted on line 100)
<clever>
tobiasBora: nixos also has a services.xserver.displayManager.sessionCommands
<clever>
tobiasBora: or launch certain things like an ssh agent
<clever>
tobiasBora: nixos will also source ~/.xprofile, so you can inject env vars before letting the normal flow resume
<clever>
tobiasBora: i think most of them do similar
<clever>
tobiasBora: but, line 112 will just execute ~/.xsession if it exists, so ~/.xsession can hijack the whole process, and bypass whatever nixos had configured
<clever>
tobiasBora: line 120 is what is responsible for running the DE you select at the login screen
<clever>
cransom: but currently, halvm only works under xen, one of my pipe dreams is to make a variant that works on bare metal
<clever>
cransom: another choice is halvm, that runs haskell code as the kernel, so you can just drop linux entirely!
<clever>
cransom: but then your letting the impure languages win!
<clever>
cransom: you could also give up, and just bring in the dhcpcd binary :P
<clever>
cransom: then you only need to deal with configuring an ethernet interface via ffi, and bake the ethernet driver into the kernel
<clever>
how can haskell do raw gfx on the hardware, without xorg!!
<clever>
cransom: and then gpu drivers and xorg....., do you want to implement xorg (or similar) in haskell? :D
<clever>
cransom: the hard part, is that modprobe has to be an external binary, so you either need to bake modules into the kernel, or let dirty c code into your initrd :P
<clever>
cransom: due to forkIO, you could potentially turn haskell-init into a full os, that doesnt even have processes, lol
<clever>
cransom: and not-os is a nix based os, that is not nixos, but does borrow a number of modules, it compiles down to a 47mb squashfs
<clever>
cransom: but you may also be interested in things like haskell-init and not-os, which just make smaller rootfs's
<clever>
cransom: ahhh, yeah, this wont really solve that situation, and a full AMI is best for auto-scaler
<clever>
and generation 1, will be your nixops config, rather then a template from the AMI
<clever>
so you can replace nixos-install with nixops
<clever>
that will let you deploy to /mnt on a remote machine
<clever>
the magic part, that nixops can take advantage of, is phase 3: nix copy --to ssh://root@target?remote-store=local?root=/mnt /nix/store/hash-nixos
<clever>
colemickens: phase 1 of this ticket, is just a way to boot the nixos installer on a remote machine, so you could ssh in, fdisk, mkfs, nixos-install, as normal
<clever>
infinisil: yeah, nixpart could be used for phase 2
<clever>
if you can get root and the kernel supports kexec, the above can be done
<clever>
and it will support any cloud provider, with most of the above features
<clever>
it makes the ami smaller, and pushes the big copy time until the deploy phase, which can take advantage of binary caches
<clever>
it will deal with problems like wireguard never working on the first deploy, because you changed the kernel version
<clever>
cransom: this will allow custom FS's for the rootfs, custom partition layouts, including fun things like a zfs stripe across many high io/sec ELB's if you wanted
<clever>
cransom: have you seen my recent nixops issue about custom fs's and such?
<clever>
infinisil: and there is an auto-optimize, what will hardlink things after the build has finished, every time
<clever>
while squashfs's dedup only has that cost once, when making the squashfs
<clever>
cransom: zfs's dedup is expensive, because it has to store the hash table, to be able to find dups when writing
<clever>
since it lacks write support, the files are just made in one piece, and left that way
<clever>
squashfs will also also never suffer from fragmentation
<clever>
some of this isnt documented, and ive just memorized chunks of the source
<clever>
but once it deploys successfully once, it will remember that it has installed the key
<clever>
nixops will only record that its own key is working, if the switch-to-configuration passes with no errors, so issues in your systemd services can force you to keep relying on the ssh agent
<clever>
then nixops can get in, and install its own key
<clever>
so install a temporary key on the remote machine thru any means you like, and add the matching private to your agent with ssh-add
<clever>
the keys in your agent will leak into nixops, and still work
<clever>
angersock: if you are using the none backend, an ssh agent is the simplest option
2019-09-10
<clever>
Zer000: callHackage or callHackageDirect
<clever>
nhill: those let you modify the pulse config files, and follow the other guides online
<clever>
nhill: specifically, look at configFile, extraConfig, and extraClientConf
<clever>
nhill: did you look at the nixos options for pulseaudio?
<clever>
nhill: the only other fix is to learn more about how mpd and pulseaudio interact, and fix it better
<clever>
nhill: id say give up and just manually run mpd in a shell :P
<clever>
nhill: if you keep pavucontrol open, it will force the proper pulseaudio daemon to keep working, and give you better feedback when things are working