<Lears>
Anyone have any idea why the mpd service doesn't start half the time? I'm having to systemctl start mpd myself. Config: http://lpaste.net/364048
Rusty1_ has joined #nixos
<yorick>
Lears: you could try to set mpd.startWhenNeeded. that's sure to do something different.
<_null_>
is there a way to build a particular package from the unstable channel?
<_null_>
there's an unfree package that doesn't seem to exist online anymore as referenced by the stable 17.09 branch
<yorick>
_null_: NIX_PATH=nixpkgs=https://github.com/NixOS/nixpkgs-channels/archive/nixos-unstable.tar.gz nix-build '<nixpkgs>' -A your_package
<_null_>
oh, thats easy, thanks yorick
<yorick>
Lears: try doing `systemctl cat mpd.service` and seeing what it looks like
<joepie91>
gchristensen: BSD 3-clause, says the package.json
<gchristensen>
ah nice
jrolfs has quit [Ping timeout: 240 seconds]
<ottidmes>
gchristensen: If it was not mentioned, I was wondering whether some tool exists that would tell me, this seemed interesting: https://github.com/nexB/scancode-toolkit
<{^_^}>
→ 88ea7874 by @dtzWill: libbsd: apply upstream patch to fix w/musl
<{^_^}>
→ 969499f3 by @dtzWill: libbsd: more upstream patches
<{^_^}>
→ 515d29c2 by @dtzWill: Merge pull request #37820 from dtzWill/fix/libbsd-musl
<Lears>
The database is 26900 files / 328GB (probably counting a fair bit of cruft). Don't know if that could couse a problem.
<infinisil>
So I'd just do systemd.services.mpd.serviceConfig.ExecStart = "${pkgs.mpd}/bin/mpd --no-daemon /nix/store/wq3allyy1m3jwlj7fh4l8b8flnjilsx3-mpd.conf --verbose"
<infinisil>
oh and add a lib.mkForce to that (so *.ExecStart = mkForce "${...)
<infinisil>
then `journalctl -u mpd -f` to follow the logs, then rebuild
<{^_^}>
[nixpkgs] @dtzWill opened pull request #37831 → celt-0.5.1: disable tests that fail to build due to bad code → https://git.io/vx0OO
<infinisil>
(or just look at the logs whenever it happens)
<Lears>
infinisil: I got it to fail with verbose logging; I've added it to the bottom of: http://lpaste.net/364048
<clacke[m]>
rosa: as long as you don't use more than you have you can address as much as you like!
<Lears>
Still not clear what's happening though
<rosa>
and is SOMEHOW 100% manageable all 16EIB's of it on a 2GB total capacity computer with absolutely no dataloss anywhere at all for any reason unless explicitly requested?
<rosa>
How tf is that even possible
<kerrhau>
infinisil: ah, so how would I update my nixpkgs?
<infinisil>
Lears: Yeah.. Maybe file an issue on mpd and/or nixos
<clacke[m]>
an allocation of bajillion bytes is just a table entry. now, once you start writing to it, that's when you get more table entries.
<rosa>
how much work does it take to modify a package to nix compatibility
<rosa>
non-nix package*
Acou_Bass has joined #nixos
<coconnor>
varies greatly depending on the source package haha
<adisbladis[m]>
rosa: 5. Maybe 3? (Varies)
<clacke[m]>
best case, with a well-behaved configure make make install package that supports --prefix and all deps are known and already in nix: near zero.
Lears has quit [Remote host closed the connection]
Lears has joined #nixos
<adisbladis[m]>
rosa: And a lot of languages have good abstractions (python/golang/rust to name a few) so basically just a few lines of generic boilerplate
<{^_^}>
→ 7151ce7f by @adisbladis: firefox-devedition-bin: 60.0b3 -> 60.0b6
<{^_^}>
→ fab46dc9 by @adisbladis: firefox-beta-bin: 60.0b3 -> 60.0b6
coot has quit [Ping timeout: 240 seconds]
<Hackapepper>
Hello, I'd like to have a bit of feedback. I've been pointed to try Nix for my packaging and deployment needs. I have a robot with a old Gentoo (8years old) which I need to crosscompile stuff as it runs a 32bit OS. This system also has the root user disabled. And I can't hack it, that would void warranty. So I need to build all my software to be deployed in the home folder. I want to run ROS in it. Which implies 100+ dependencies
<kerrhau>
infinisil: sorry to keep bothering you lol, but that command does nothing, it fetches the tar.gz file, unpacks it then just exits
<Hackapepper>
I've already compiled all these dependencies by hand to be able to compile ROS, but it's a big unmaintainable hack with no depedency tracking. So I need a better system
<infinisil>
kerrhau: Ah, i misspelled the package, it's soulseekqt, not soulseekerqt
<Hackapepper>
Can I build and use nix without root privileges? Can I make it deployable in whatever PREFIX of folder I want? Does this imply re-writting lots of packages?
<kerrhau>
infinisil: yeah, i caught that, but even then, it does nothing
<infinisil>
kerrhau: It did for me..
<kerrhau>
hm
<_null_>
What does collision mean? I'm getting "collision between `/nix/store/3d8nq34016rjz5vfy7w36k77nnrvkchk-gst-python-1.10.4/lib/python2.7/site-packages/gi/overrides/__init__.py' and `/nix/store/xir2hwxj8870j311hx8jdi63vcspljs0-python2.7-pygobject-3.24.1/lib/python2.7/site-packages/gi/overrides/__init__.py'"
<_null_>
I'm trying to add services.mopidy.enable = true;
<infinisil>
kerrhau: does nix-env -q list the package?
<kerrhau>
no
<infinisil>
Weird
jrolfs has quit [Ping timeout: 248 seconds]
<infinisil>
kerrhau: If you have a nixpkgs checkout you could just go to the unstable channel and nix-build -A soulseekqt
<adisbladis[m]>
Hackapepper: Typically nix runs with a nix-daemon. You can however run it in single-user mode.
<infinisil>
which is what I would do
<kerrhau>
if I remove the 'A' flag, it errors with 'undefined variable ‘placeholder’ at /nix/store/64i9qa3729gbscsc1i0nfhv4pajj8l46-nixos-unstable.tar.gz/pkgs/tools/misc/colord/default.nix:32:48'
<adisbladis[m]>
Hackapepper: Nix is not like other packaging systems. Every binary will end up in /nix/store/${hash}-name-version
<kerrhau>
?
<infinisil>
kerrhau: Somebody pushed a Nix 2.0 only feature to master and now everybody using Nix1.x can't use it..
<kerrhau>
how does one start using nix 2.0?
<_null_>
You *can* change where nix stores these, but it's not a good idea. At the bare minimum you're going to have to rebuild everything
<_null_>
There's a little line in the guide about doing this / why you shouldn't
<Hackapepper>
_null_: I already need to rebuild everything almost for sure, but if the recipes for rebuilding accept a different root folder, I'm all good
<kerrhau>
infinisil: I dont know why i wouldnt be on nix 2.0, I installed nix just a few days ago
<kerrhau>
i dont think it's been a week yet
<infinisil>
Ah I missed Hackapepper, didn't see ya there
<infinisil>
Hackapepper: yeah nix can use any directory and doesn't require root for it, but you do need to build a lot of stuff from source for that
<rosa>
Hackapepper: you would probably need to do something like this: LD_LIBRARY_PATH=./application/ ./application /bin/application $ LD_LIBRARY_PATH=/ system_application, tho that is ASSUMING application does not fork and exec* anything, wich applies to all other applications aswell
jrolfs has joined #nixos
<clacke[m]>
infinisil: Glad to hear this. I tried a home-directory build for guix once, and it had a bunch of assumptions that broke it.
<adisbladis[m]>
rosa: Or you know let nix deal with all the build steps and then use something like `nix-shell --run` to invoke your program.
<adisbladis[m]>
LD_LIBRARY_PATH is a hack
<rosa>
as if it does fork/exe anything them LD_LIBRARY_PATH will obviously fail unless the environment is NOT kept (or unless the application explicitly detects LD_LIBRARY_PATH and resets it to NULL before exec'ing from fork
<rosa>
then*
<infinisil>
rosa: he doesn't have root, can't even use /nix/store
<rosa>
which is also why i dislike LD_* anything as it is probogational
<rosa>
except LD_PRELOAD as it would be the only thing someone WOULD want to probagate
<Hackapepper>
I have no root, and the system is old as hell, I needed to compile my own tar cause the one in the robot is provided by busybox and has a bug in it that won't decompress big .tar.gz (like my custom compiled GCC)
<infinisil>
Hackapepper: You need a custom build of the nix package manager that uses something like /home/<youruser>/nix as the store directory
<_null_>
oh there you found it infinisil, I was scouring the docs to find that
<infinisil>
:)
<Hackapepper>
infinisil: sounds good to me, any link to get me started?I'm just going thru the downlaod page
<infinisil>
Hackapepper: You don't want to download that version
<infinisil>
because you need a custom one, your starting point is to clone nix
<Hackapepper>
infinisil: I'm checking the from source, the prerequisites are quite a few that I have not compiled yet, I hope they aren't too much into the dependency hell
<infinisil>
Hackapepper: Do you not have a proper machine you could build this on?
<Hackapepper>
infinisil: I have a virtual machine that represents the robot, is where I do the work
<infinisil>
Hackapepper: Well can you copy files to the virtual machine? If so just build it on your actual machine, then copy over
<Hackapepper>
so I'm gonna try to do an initial build of nix straight away and see if it works with the stuff I already have compiled
<Hackapepper>
I need to crosscompile if I build it in my machine, so I jsut build it in the VM
<Hackapepper>
I went over the bootstrap.sh error, but then I have a ./configure error that doesnt make much sense ./configure: line 6009: syntax error near unexpected token `OPENSSL,'... the internet says I'm missing pkg-config but I have it, and it's actually the same version in my host and vm
_null_ has quit [Remote host closed the connection]
<infinisil>
adisbladis[m]: Nice!
<infinisil>
Hackapepper: That's problems you wouldn't have if you were using nix
<Hackapepper>
infinisil: haha I want it so bad already then!
<infinisil>
Just install a normal nix on your host and then cross compile a nix with custom store path for your VM
<Hackapepper>
Ok, installing nix, can you guide me on the crosscompile bit when I'm ready?
<adisbladis[m]>
Yeah. Probably the easiest way to get bootstrapped :)
<infinisil>
Hmm maybe, I think i should know how to cross compile it
<rosa>
whan a package is ported to nix is EVERYTHING redirected from / to it is nix path?
<rosa>
its*
<infinisil>
rosa: yea, builds can't install outside of their nix store path
<rosa>
or are there some exceptions
<Hackapepper>
ok, nix installed, I'm ready to rock infinisil
<rosa>
like for example, opening files in /dev or /etc or so on
orivej has quit [Ping timeout: 264 seconds]
<infinisil>
Hackapepper: whew, hold on, need to assemble a command for that
<rosa>
how do u decide what should and what should not be redirected to the nix install path at runtime of the application
<infinisil>
Hackapepper: So clone github.com/nixos/nix for a start
<Hackapepper>
infinisil: fast, I'm all excited about all this being automated haha
<rosa>
like does ALL of /etc/* paths need to be redirected? do only some need to be redirected?
<Hackapepper>
infinisil: cloned
<rosa>
What if an application tries to open a file part of another application?
<krey>
what's a convenient way of nix-prefetching a url with variables in it?
<infinisil>
rosa: Ah yeah i didn't mean it like this
<rosa>
what if that file is present in multiple applications and not present in the application attempting to open it?
<infinisil>
rosa: packages can only *install* into their nix store locations, but they can of course read and write whatever they want at runtime, we can't restrict that
<infinisil>
rosa: (but /nix/store is read only of course)
vidbina_ has quit [Ping timeout: 264 seconds]
<krey>
even worse, this thing isn't even a url,
<rosa>
so how does it work then?
<rosa>
is it like gentoo and is just symlinked to /
<infinisil>
Hackapepper: (am on it, hold on)
hiroshi- has joined #nixos
<adisbladis[m]>
krey: Either by passing `--name` to `nix-prefetch-url` or by accessing the attribute `nix-prefetch-url '<nixpkgs>' -A hello.src`
<rosa>
if so how do u know what should be symlinked and what should not be symlinked
<kandinski>
I'm trying to figure out which nixos channel I should add. I installed with the 18.03 installer, so I should add https://nixos.org/channels/18.03 as a primary, right?
<infinisil>
Hackapepper: i686 doesn't need cross compiling doesn't it
<infinisil>
It's just 32 bit
<Hackapepper>
well -m32 flag needs to be set somewhere
<infinisil>
Yeah yeah no problem
<Hackapepper>
which is kinda cross compiling, it's a binary for a different arch, right?
<unlmtd>
does nixops use my env's `NIX_PATH` by default?
gmarmstrong has quit [Quit: Quit]
vidbina has joined #nixos
<kerrhau>
so, how does one upgrade from nix 1.x to nix 2.0?
rosa has quit [Quit: rosa]
<sorbet>
On unstable, I think if you just update the channel and rebuild you'll have 2. (At least I do.)
<sorbet>
Not sure about stable (or the versioned one).
<kerrhau>
sorbet: how do I switch to unstable from stable?
<Hackapepper>
So, let's say I end up building my awesome nix, and then I need to build my packages of nix... once I've done that, can I have a repository (you call it channel I think) for my deployments to pull from in a easy way?
<Hackapepper>
my bad, just found it "You can easily share the Nix store of a machine via HTTP. This allows other machines to fetch store paths from that machine to speed up installations. "
<Hackapepper>
awesome
vidbina has quit [Ping timeout: 240 seconds]
<Hackapepper>
infinisil: I'm getting very excited while reading the docs, how is that going mate?
<\\\\\\\\\\\>
The libinput package doesn't enable trackpoint middle button scrolling by default?
<{^_^}>
→ a148686a by @adisbladis: Merge pull request #37727 from enumatech/swagger
<{^_^}>
→ 72e89870 by @peterromfeldhk: nodePackages_8_x.swagger: init at 0.7.5
<infinisil>
Hackapepper: It shouldn't matter, I'd go for the file :P
<infinisil>
But make sure to change it to your own home directory
phI||Ip has joined #nixos
Intensity has quit [Ping timeout: 260 seconds]
<infinisil>
Oh and my build just succeeded, only took a couple minutes
<clever>
infinisil: the main limit i can see with your variant, is that the nix it initially builds still lives in the store of the nix that built it, not the paths you passed to it
Intensity has joined #nixos
<infinisil>
Oh..
alphor has joined #nixos
<infinisil>
Right so that wouldn't work
<clever>
NIX_CONF_DIR, NIX_LOG_DIR, NIX_STORE all have to be set when running nix-build, to modify the config of where the nix its making will live
<clever>
and the .override, so that nix will target the same place
<infinisil>
Yeah
<clever>
and then NIX_REMOTE to re-root that, so you dont need write to that final destination
<clever>
allowing you to build it for a path you cant make
<Hackapepper>
clever: thanks,then going with your version
<infinisil>
We can be glad clever is here!
<kandinski>
I just rebuilt nixos after updating the 18.03 channel, but I can't tell if a certain fix (https://github.com/NixOS/nixpkgs/pull/36283) is what I'm running or not. Can anyone help me diagnose? This is the vagrant I'm running: /nix/store/32zly5yhz0hcqgk6w5y0ish2rssfc6y9-vagrant-2.0.2/bin/vagrant
<clever>
kandinski: compare the files modified in the PR to what you have in ~/.nix-defexpr/channels_root/nixos/
<kandinski>
clever: thanks
<kandinski>
why would it be local to my user?
<clever>
the channels_root directory points to the channels managed by root
<kandinski>
ah, I see, ta
<clever>
for root itself, its just ~/.nix-defexpr/channels/nixos/
<infinisil>
clever: Hackapepper Oh but hold on
<infinisil>
Hackapepper: You need to replace `nix.override` with `pkgsi686Linux.nix.override`
<clever>
yeah, that will switch it over to a 32bit nix
<Hackapepper>
infinisil: yeah I was just double checking that too
<Hackapepper>
clever: can it be NIX_CONF_DIR=/home/$X/etc/nix --> NIX_CONF_DIR=/home/$X/nix/etc instead? so I have everything isolated in the nix folder?
<kandinski>
clever: I'm not running what's in that PR. How do I find when a given PR goes into a channel like 18.03?
<clever>
Hackapepper: probably, youll also want to change the confDir in the override
<kandinski>
clever: s/when/whether/
<neonfuz>
so I've never used zfs before, and I'm wondering why in this example they create rpool/root and rpool/root/nixos https://nixos.wiki/wiki/NixOS_on_ZFS
<kandinski>
clever: thanks. what happens if I add one of those channels? Is it possible to just add a channel for one package and its dependencies?
tertle||eltret has quit [Quit: Connection closed for inactivity]
schoppenhauer has joined #nixos
<neonfuz>
clever: that makes sense, I was wondering why he creates another level, IE rpool/root/nixos rather than just rpool/root
<clever>
kandinski: the simplest thing right now would probably be to just run nixos-unstable for a month, its a lot more stable then the name sudjests
<clever>
neonfuz: probably just personal preference
<kandinski>
clever: it's like running Debian Sid, right?
<clever>
kandinski: the biggest difference is that nixos has rollbacks, so you can undo any change
<clever>
while debian cant really undo updates so seamlessly
_null_ has joined #nixos
<neonfuz>
also, any way to expand a zfs pool?
<_null_>
so, I tried to rebulid upgrade to 18.03, and kde is saying it can't find a power daemon
<kandinski>
clever: fair point, kinda the reason why I have moved over. Thanks for the patient advice.
<_null_>
I explicitly added services.upower.enable = true and restarted the display manager but that doesn't seem to help
<adisbladis[m]>
neonfuz: You can add a vdev to a pool
<clever>
neonfuz: expanding is easy, shrinking is imposible
<infinisil>
Hackapepper: There's still $HOME/etc at the very end
<infinisil>
And I think those env vars are for the build machine, right clever?
<clever>
infinisil: yeah, the env vars are to affect the un-modified nix, to make it act on a non-standard store
_null_ has quit [Remote host closed the connection]
<krey>
i'm trying to install a package from GitHub in a temporary way, but I'm not really getting there
<clever>
and the overrides are to make the nix within that non-standard store act on the same non-standard store, during its own runtime
<Hackapepper>
clever: should I change it all around to that I guess? btw, this is sucking a lot of disk space, my disk was getting full, I need to switch a my other disk, lol
<clever>
Hackapepper: this will cause the final nix binary to try to act on $HOME/nix/store, using the value of $HOME from the build host, not the target machine
<adisbladis[m]>
krey: Because you are not actually building pypi2nix. You are just fetching the source.
<infinisil>
Hackapepper: you changed it the other way, didn't mean it like this
<krey>
adisbladis[m]: but it's a build input! isn't that what nix-shell is all about? building the inputs?
<infinisil>
Hackapepper: clever: I think this should be it: https://git.io/vx0Cq
<Hackapepper>
clever: infinisil refresh my gist, I made it back
<adisbladis[m]>
krey: Gimme a sec
<clever>
Hackapepper: looks good
<infinisil>
Oh
<infinisil>
Ok get it, never mind my gist it's not correct
<infinisil>
clever: You sure could use a bit more newlines in your scripts though!
<neonfuz>
so I messed something up and now the root .nix-profile points to /nix/var/nix/profiles/default which doesn't exist
<adisbladis[m]>
krey: `fetchFromGitHub` fetches and unpacks the source for you, it does not know how to build anything (how could it). That's what `stdenv.mkDerivation` and friends are for.
<clever>
neonfuz: thats what .nix-profile is supposed to point to
<clever>
neonfuz: if you have never installed something with nix-env as root, then that wont exist
<adisbladis[m]>
krey: So either you make your own `pypi2nix` derivation or you use an override like https://ptpb.pw/gsMO
<clever>
krey: also, pypi2nix already has a default.nix you can just use
<infinisil>
Hackapepper: Looking good
<krey>
clever: yeah, I noticed! I'm trying to figure out how to give it all the arguments it needs
<infinisil>
I won't be there anymore once it's finished compiling though
<clever>
krey: pkgs.callPackage
<Hackapepper>
fixed-output derivation produced path '/home/nao/nix/store/yv06f35xnkxi42sa4srhhapaps21cfr6-busybox' with sha256 hash '1i5yj20j1j85swmamxhdqhmw1kizf9020as2rk4swh1cw3vvwxks' instead of the expected hash '18fbjl8pdwwzz9ln9i34jdvdc172sbi0mnggakvf8mxfqzk1nk7g' cannot build derivation '/home/nao/nix/store/7ixg9al347x35mnfvddpgm6gkvyjgrb2-bootstrap-tools.drv': 1 dependencies couldn't be built
<Hackapepper>
is not building after my last change, I'm working on a NTFS disk now instead, may be that?
<neonfuz>
it's making nixos-install say error: file 'nixpkgs/nixos' was not found in the Nix search path (add it using $NIX_PATH or -I)
<clever>
Hackapepper: run "file /home/nao/nix/store/yv06f35xnkxi42sa4srhhapaps21cfr6-busybox"
<clever>
Hackapepper: ah, NTFS will likely break a lot of things
<krey>
clever: ooh
<Hackapepper>
clever: /home/nao is the virtual environment of the robot, I can't do that command. Bummer, my disk is too full for building it... damn I need to cleanup stuff
<neonfuz>
ah, nix-channel --update and nix-env -u fixed it
<infinisil>
Hackapepper: Um, maybe just rm -rf /media/sam/DATA/rootfs and try again, I have no idea why that occurs
<clever>
Hackapepper: i suspect that the ntfs doesnt support +x, so everything will fail
<Hackapepper>
clever: infinisil here we are again, I cleaned up 7GB of disk... I have like 15GB free now, I hope is enough. It's working if I don't use the NTFS disk as build space
<infinisil>
Huh alright
<infinisil>
I think that should be plenty
<infinisil>
Probably
<Hackapepper>
Do you have a estimate for how long this should take? I have slave machines I could send the job to if it's gonna take hours
<infinisil>
Hackapepper: Yeah this might take hours
<clever>
Hackapepper: the slaves wont really work right with the store being in the wrong place
<Hackapepper>
clever: what do you mean? I mean running exactly the same thing in a different machine, instead of my work-laptop
<clever>
Hackapepper: i meant the work-sharing build-slave logic within nix, to split the load up between many
<clever>
but if you run the whole thing on a single remote machine, it wont matter
magnetophon has quit [Remote host closed the connection]
<clever>
krey: ah, he added a weird argument, you need pkgs.callPackage /path/to/pypi2nix { nixpkgs = pkgs; }
<infinisil>
Hackapepper: But you should also be able to view it with `find path/to/the/rootfs/nix/store -name '*.drv'`
<infinisil>
.drv files represent derivations which are like recipes for building stuff, and since that nix store was previously empty, you'll only have the stuff of your build in there
<Hackapepper>
clever: Ok, its gonna build GCC 7.3.0 (wow, thats more modern that I find comfortable with haha) this is gonna take at least 4-5h just with that step
<clever>
Hackapepper: its probably going to build 2 gcc's, and a glibc
<infinisil>
Hackapepper: It is pretty nice to watch all the nice compiling without you having to do anything :P
<infinisil>
and the chances of a build failure are pretty much 0%
<Hackapepper>
clever: my god haha I'm gonna move this job to a 12 CPU machine then, it's older CPU but faster in theory by GHz xD
<Hackapepper>
infinisil: it's actually neat, yeah
<infinisil>
Alright I'll go to sleep now, Hackapepper if you have problems with it just ask here (and ping me if you want)
<Hackapepper>
Do I need to do anything to enable "use all the cores you find pretty please" for this compilation?
<clever>
Hackapepper: -j 10 --cores 10
<clever>
the first runs 10 derivations in parallel, but it cant do much early on
<clever>
the 2nd runs 10 gcc's within each derivation
<clever>
together, it would run up to 100 procs, but the 2nd also limits itself to keeping the loadavg under 10
<Hackapepper>
I have 12 cores available in that machine, does that sound like a good param?
<clever>
yeah
<Hackapepper>
and those are flags to add to nix-build, right?
<clever>
yep
<Hackapepper>
It only has 16GB of RAM I'm concerned that's gonna be a problem. I do remember GCC taking by himself gigabytes... if that happens, does the nix-build try to do its best to not fail compilations or get extremely slow?
<Hackapepper>
what does happen if a compilation fails for... out of RAM?
<clever>
Hackapepper: the `--cores 10` part tells it to keep the load avg under 10
<clever>
so if things start to lag some, it will throttle itself
jrolfs has joined #nixos
<clever>
but it will entirely ignore ram usage, and may kill itself
<clever>
but you can always re-start it with smaller number, and it will resume at the last derivation
<infinisil>
But 16GB should be enough really
<infinisil>
I doubt it's gonna use more than 1 or 2GB
Biappi has quit [Remote host closed the connection]
<Hackapepper>
The only horrible thing of that machine is that... it has no SSD
<Hackapepper>
so every disk step is slow as hell
<infinisil>
Glad I was able to convince somebody to use nix even though the benefit you're using it for is probably one of the least used ones :P
<infinisil>
There's a lot of nice things on Nix and NixOS
<infinisil>
s/use/try
<Hackapepper>
infinisil: We did look in my lab to use nix last year, but we found some problems with it, that I don't remember anymore, and I was in a big hurry so I just manned up and did everything by hand... 100+ packages... hell
<infinisil>
Yeah there are some problems (as with every technology), but the benefits far outweigh those imo
<Hackapepper>
if this actually works as expected (please let it be so) I'll do my best to make easy having ROS building based on the nix packages. and if so... anyone with my robot could use it
<Hackapepper>
could you explain me some of those problems? if you are awake enough haha
<infinisil>
Well for one you can't really use any dependency unless it's specifically packaged for nix
<infinisil>
But it's kinda turned around, because packaging stuff for nix is really fixing packaging in general
<Hackapepper>
if you have 'build-essentials' and everything up to 'x11', even more, to 'qt' I think it's pretty good already
<Hackapepper>
where can I look for what packages are available?
<{^_^}>
→ f623858a by @dtzWill: glib: no one wants --witih-libiconv=gnu
<{^_^}>
→ 9204fa26 by @dtzWill: propagate libiconv header from musl too
<{^_^}>
→ b6b94f0b by @dtzWill: man-db: disable tests w/musl as simplest solution
<infinisil>
Hackapepper: You can get an up-to-date list by running `nix-env -qaP`
<Hackapepper>
so, for example, I see you hace PCL 1.8.0, but I need PCL 1.7.2... what's the approach then? make my own recipee (I don't know what you call them) to compile it with nix?
<infinisil>
Hackapepper: That's another slight problem with Nix (well not nix really, more the nixpkgs packaging), which is that there's pretty much only 1 version
<Hackapepper>
but for example, I'm compiling two different GCC, how can that be??
<infinisil>
You can no problem use any number of versions, but all packages in nixpkgs use a single one in general
reinzelmann has joined #nixos
<infinisil>
The best way with nix is to just use 1.8.0 and only explicitly use 1.7.2 when you need it
<Hackapepper>
That's pretty straightforward
<Hackapepper>
so I just declare my nixpkg thing to use 1.7.2... having created my nixpkg dependency before stating it's version 1.7.2 and that's it, right?
<infinisil>
You can override packages to change stuff of them, e.g. to set the source to use the 1.7.2 version from github or so
<infinisil>
Hackapepper: eh, not really. I shall take a minute to explain sandboxing
<infinisil>
So when a package gets built, the only thing it can write to is a path in /nix/store/<hash>-<name>
<infinisil>
And the only thing it can read is everything in /nix/store
<infinisil>
It doesn't have networking, it can't download anything
<Hackapepper>
just so you know the crazy way that was my last try... I was getting the source of ubuntu packages (which are declared as dependencies in ROS packages), I was generating a default build script, trying it, if it didnt work, it was my turn to manually tune it, and so on... for every package.. but in the end I was almost spending the same time doing that than fully manual. And no way to compile glibc/boost
<infinisil>
Lol that's pretty horrible indeed
<infinisil>
Continuing.. If we *do* know the result of a build, aka we can give it a sha256 hash, network access is allowed, because whatever it fetches, it will have to result in that hash, so it's pure and secure
<Hackapepper>
better than my last year approach, all by hand, 100%, google the dependency that was missing/crashed. download source code, compile on my machine, works? try on the VM. doesnt work, needs more dependencies, repeat this loop until finding all the dependencies... run the tests, they crash on runtime, look for dependencies, wohoo, one depency done... keep going
<infinisil>
So in the end this means that in order to download stuff from the internet we will have to give it a sha256 hash
<infinisil>
This includes the sources for the 1.7.2 version
<infinisil>
So what you do to get version X of package Y is more or less this: `Y.overrideAttrs (old: { src = fetchurl { url = "https://..."; sha256 = "lrcuaoh5l4lchoeu..."; }; })`
<infinisil>
That is you override the package to use a different source instead, giving it something that fetches an url and checks for the hash of it
<infinisil>
this is a nix expression, which are used to declare derivations, which are the recipes for packages
<infinisil>
Well it's finally time to sleep, see ya :)
<Hackapepper>
infinisil: thank you for the explanation
<Hackapepper>
sounds pretty doable, specially when you think where I come from haha
<Hackapepper>
and safe
orivej has joined #nixos
blankhart has quit [Ping timeout: 256 seconds]
robstr has joined #nixos
<neonfuz>
clever: how do I expand a zfs pool?
<neonfuz>
clever: were you talking about actually extending the partition and expanding the pool? or adding more vdevs to the pool?
<infinisil>
neonfuz: am still here and i think you mean export
<infinisil>
Ah
<infinisil>
Nvm
<neonfuz>
no clever said I can expand a zfs pool and I'm not sure exactly what he means, I can't really find info on this
<neonfuz>
the only thing is I have about 200gb on a 256 gb drive that's formatted ext4, and my only other large storage is ntfs which would lose permissions
<infinisil>
I don't think you can expand a zpool other than adding more vdevs
<infinisil>
There's #zfsonlinux too
<neonfuz>
I mean I'd rather not just arbitrarily have multiple permissions but it's not the worst
<neonfuz>
I'll check there
* infinisil
definitely 100% absolutely goes to sleep now for real this time no bamboozle
<neonfuz>
lul
<{^_^}>
[nixpkgs] @ElvishJerricco opened pull request #37842 → WIP: GHCJS: Fix GHCJS with haddock-library-1.4.3 → https://git.io/vx0lW
<adisbladis[m]>
neonfuz: zpool add $pool $vdev
<adisbladis[m]>
neonfuz: You cannot grow a vdev though (so raidz is fixed-size after creation)
<Hackapepper>
clever: oopsie, after 1h30m compiling... it has stopped with builder for '/home/nao/nix/store/vrqml7yrn4dbhkl8k9qixcvk8sm68zgk-perl-5.22.patch.drv' failed with exit code 1
<Hackapepper>
I've tried to re-run it a couple of times, but it stops there
<Hackapepper>
I'm trying without the 10 jobs & 10 cores thing to see if I get a more clear error output
<unlmtd>
how do I find out why a certain package is pulled in by `nixos-rebuild`?
<unlmtd>
nix why-depends ?
jrolfs has quit [Ping timeout: 276 seconds]
<garbas>
krey: if you find some bug/errors with pypi2nix please open an issue. i'd love to help but follwoing #nixos is sometimes hard for me since it gets busy
<adisbladis[m]>
neonfuz: That's usually handled by your login session
<Hackapepper>
I've found a mirror for the patch I'm missing to build nix (as anonscm.debian.org is still down, and no one is answering me if thats normal in the #debian channel)
<Hackapepper>
Is there a way for me to add a mirror for the patch?
Ariakenom has quit [Read error: Connection reset by peer]
pallav has joined #nixos
xy2_ has joined #nixos
dingotux has quit [Quit: Konversation terminated!]
<pallav>
Hi everyone. I need some help. I have been using NixOS for a long time now, but after the upgrade to nixos-18.03, my `nixos-rebuild` command just stalls. (Log of the last few lines: https://www.hastebin.com/veduxafaga). I have had no progress for a long while, and its not letting me upgrade the system anymore. Since this is my primary laptop, this is a blocker for most of my work. Any help would be very appreciated!
iyzsong has quit [Ping timeout: 264 seconds]
<pallav>
I believe that the problem is related to the switch to HTTP/2 instead of HTTP/1.1, but I am not sure. Is there any way to revert to the old behaviour?
_null_ has quit [Remote host closed the connection]
winem_ has joined #nixos
m0rphism has joined #nixos
[Leary] has joined #nixos
xy2_ has quit [Ping timeout: 276 seconds]
<elvishjerricco>
pallav: I don't know why it would present as a stall, but this definitely seems like a problem: `don't know how to build these paths:`
Lears has quit [Ping timeout: 240 seconds]
<elvishjerricco>
pallav: I picked one of those paths (specifically /nix/store/g1jqxcqiswca72yr8mcj8lbd0w9fz6zh-git-2.16.2) and checked cache.nixos.org for it.
<vaibhavsagar>
there were issues with an older revision when building qtwebengine-5.10.1
<pallav>
Hmm, I think I understand. I'm using a different cache from my institute server. This is also a recent change. How do I override cache for the build (apart from changing in configuration.nix which would only affect subsequent builds i think)
<vaibhavsagar>
so maybe you could do a `sudo nix-channel --upgrade`
<vaibhavsagar>
`--update`, sorry
<vaibhavsagar>
and then see if that makes a difference?
<elvishjerricco>
pallav: You might be able to do `nixos-rebuild --substituters https://cache.nixos.org`
<kandinski>
hi, vaibhavsagar
<elvishjerricco>
if nixos-rebuild doesn't take that argument, you can definitely build it yourself with nix-build.
<{^_^}>
→ 8163358c by @etu: testssl: 2.9.5-2 -> 2.9.5-3
<{^_^}>
→ 78ab1f17 by @adisbladis: Merge pull request #37849 from etu/testssl-update
dan_b has joined #nixos
vaninwagen has joined #nixos
edwtjo has quit [Ping timeout: 256 seconds]
MichaelRaskin has left #nixos [#nixos]
ma27 has joined #nixos
ma27 has quit [Client Quit]
seanparsons has quit [Read error: Connection reset by peer]
<{^_^}>
[nixpkgs] @peterhoeg merged pull request #37847 → home-assistant: backport of a number of python packages → https://git.io/vx0En
<{^_^}>
[nixpkgs] @peterhoeg pushed 8 commits to release-18.03: https://git.io/vx0gN
<{^_^}>
→ c4e37974 by @peterhoeg: pythonPackages.yahooweather: init at 0.10
<{^_^}>
→ af9b83af by @peterhoeg: pythonPackages.deluge-client: init at 1.2.0
<{^_^}>
→ 05e0dc4c by @peterhoeg: pythonPackages.sseclient: init at 0.0.19
ma27 has joined #nixos
ThatDocsLady has joined #nixos
seanparsons has joined #nixos
<adisbladis[m]>
infinisil: Ping?
periklis has quit [Ping timeout: 260 seconds]
Tucky has joined #nixos
pie_ has joined #nixos
pie__ has quit [Read error: Connection reset by peer]
asuryawanshi has joined #nixos
asuryawanshi has quit [Remote host closed the connection]
asuryawanshi has joined #nixos
NinjaTrappeur has quit [Quit: WeeChat 2.1]
patrl[m] has joined #nixos
LouisDeFunes294 has joined #nixos
zzamboni has quit [Quit: Leaving.]
zzamboni has joined #nixos
ma27 has quit [Quit: WeeChat 2.0]
ma27 has joined #nixos
<garbas>
is anybody experiencing problems with pulse? for the last month i'm listening from time to time "flaky" audio. killing pulseaudio (pulseaudio -k) helps, for 30min. i'm also seeing this on release-18.03 (as well as on master)e
<goibhniu>
hi garbas, have you got anything custom in ~/.config/pulse?
<joko>
Hello, is it possible to run nix-shell using sudo -u other_user? I get error: Nix database directory ‘/nix/var/nix/db’ is not writable: Permission denied
humanoyd has quit [Ping timeout: 265 seconds]
reinzelmann has joined #nixos
<symphorien>
export NIX_REMOTE=daemon (might not be true with nix 2, though)
<LnL>
joko: set NIX_REMOTE=daemon
ma27 has quit [Ping timeout: 276 seconds]
ashgillman has joined #nixos
<joko>
symphorien LnL: thanks, it seems to be working!
rosa has quit [Quit: rosa]
<LnL>
yeah 2.0 will just figure it out on it's own :D
ashgillman has quit [Ping timeout: 240 seconds]
ashgillman has joined #nixos
logzet has joined #nixos
humanoyd has joined #nixos
Mateon3 has joined #nixos
freeman42x]NixOS has joined #nixos
katona has joined #nixos
Mateon1 has quit [Ping timeout: 240 seconds]
Mateon3 is now known as Mateon1
<{^_^}>
[nixpkgs] @peti merged pull request #37834 → Request older haddock library for haddock-api in GHC 8.0 → https://git.io/vx03w
<Nyanloutre>
ok thanks @clever I looked at tgtd before, so I need to create a systemd service to configure it at boot ?
<Nyanloutre>
but hasn't STGT been replaced by LIO ?
<clever>
i could never get LIO to work when i was playing with things
ma27 has joined #nixos
roberth has joined #nixos
Hackapepper has quit [Ping timeout: 260 seconds]
jophish_ has joined #nixos
jophish_ is now known as Guest53753
martingale has quit [Ping timeout: 276 seconds]
rosa has quit [Quit: rosa]
orivej has joined #nixos
<roberth>
MoreTea: would you mind having a look at https://github.com/NixOS/nixpkgs/pull/37760 ? It's a PR by BjornMelgaard to make dockerTools simulate docker build a bit more
zzamboni has quit [Quit: Leaving.]
zzamboni has joined #nixos
vaninwagen has quit [Ping timeout: 264 seconds]
tv has quit [Ping timeout: 264 seconds]
vaninwagen has joined #nixos
<ij>
sphalerite, Hey, last time you told me your nixos-rebuild was fast(~30s). Was that a rerun from a compiled configuration or on a changed conf with added package or something?
tv has joined #nixos
rosa has joined #nixos
<sphalerite>
ij: let me just upgrade and see how long that takes
<sphalerite>
ij: it took 3min because it had to download a bunch of updated stuff, so I'll need to try again with smaller changes
pkill9 has joined #nixos
<sphalerite>
ij: adding hello to systemPackages and nixos-rebuild switch took 20 seconds, including downloading hello
<sphalerite>
ij: a no-op switch that doesn't involve any derivation building takes… 5.8s
ashgillman has quit [Ping timeout: 256 seconds]
freeman42x]NixOS has quit [Ping timeout: 264 seconds]
<{^_^}>
[nixpkgs] @abbradar closed pull request #37369 → Move to libglvnd for OpenGL → https://git.io/vxsZq
kier_ has quit [Quit: This cannot continue.]
kier has joined #nixos
spear2 has joined #nixos
leat has joined #nixos
pkill9_ has joined #nixos
pkill9 has quit [Ping timeout: 276 seconds]
<fearlessKim[m]>
is is possible to have sthg like `nix-shell -i bash -p 'python.withPackages(ps: [ ps.myPackage ])' nix-prefetch-scripts` on 2nd line of shebang, nix keeps telling me "unexpected $undefined"
<ottidmes>
fearlessKim[m]: I experimented with nix-build in shebangs a while ago and I came to the conclusion you should not try and do so, instead point to some script that in turn calls what you want, so add an indirection. The program in the shebang will be given the series of arguments as just one argument (one big string) and depending on the length of the shebang, it might become cut off (clipped?)
cx405 has joined #nixos
<monotux>
so is cache.nixos.org down or something?
<cx405>
Hi! I want to propose adding link to Nixos Wiki to support page. Its fine if its "unofficial", however it would reduce great amount of frustration and bugs for newcommers(and not only).
<cx405>
I am posting here, because I couldn't figure out where this belongs in github.
<fearlessKim[m]>
ottidmes: do you know what cuts off the parameters ? is that a limit in nix-shell ? It's for a script I want to submit to nixpkgs so adding an intermediate script seems bad.
<ottidmes>
fearlessKim[m]: I could look up the test program someone else gave on the IRC that clearly showed this, if you want it. I believe it was inherent to shebang, not nix-build
<ottidmes>
fearlessKim[m]: But why is having 2 files with no shebang magic, worse than 1 file?
<ottidmes>
fearlessKim[m]: There you go, inherit limit of shebang
<fearlessKim[m]>
ottidmes: I dunno, my first line is just `#! /usr/bin/env nix-shell` so it should be fine, the second line is read by nix-shell I suppoose `#!nix-shell -i bash -p 'lua5_2.withPackages(ps: [ ps.luarocks-nix ])' nix-prefetch-scripts bash `
<fearlessKim[m]>
I have a hard time with the shebangs, sometimes they work, sometimes the exact same doesn't. Also the `-i` is sometimes recognized, sometimes not :/
<cx405>
Any chance someone responsible for Nixos Wiki is here?
<cx405>
I'd hate to make direct edit to the main page and something also tells me if I add proposal on the talk page - nobody would read it for years.
<ottidmes>
BlessJah: I know its possible, I explicitly meant a call to nix-build with multiple arguments
<BlessJah>
then no, impossible
<{^_^}>
[nix] @dezgeg opened pull request #2010 → libexpr: Make unsafeGetAttrPos not crash on noPos → https://git.io/vxEJR
<joepie91>
cx405: hop into #nixos-wiki :)
<ottidmes>
fearlessKim[m]: Can't you just supply nix-shell with a multi-line --run command?
<{^_^}>
→ 1c7e358e by @ryantm: blender: 2.79a -> 2.79b
<{^_^}>
→ e9fc2756 by @edolstra: Merge pull request #37782 from ryantm/auto-update/blender
<ottidmes>
BlessJah: Is that because those additional shebangs under nix-shell are a feature of nix-shell? Normally you cannot have multiple shebangs, right?
<fearlessKim[m]>
looking at the tracker, nix-shell looks like a nest of bugs xD
<BlessJah>
ottidmes: it is because of that, however nix-shell should be fine with exactly same command you've used in shell
<{^_^}>
→ cb9c1c63 by @offlinehacker: nixos/tor: expose control socket
<{^_^}>
→ 1c889be4 by @joachifm: Merge pull request #37827 from oxij/pull/28938-tor-control-port
<{^_^}>
[nixpkgs] @joachifm pushed 2 commits to release-18.03: https://git.io/vxEmk
<{^_^}>
→ 3d007c88 by @joachifm: tor-browser-bundle-bin: 7.5 -> 7.5.1
<{^_^}>
→ a585401f by @joachifm: tor-browser-bundle-bin: 7.5.1 -> 7.5.2
pkill9_ is now known as pkill9
<ij>
Hm, this one project uses Gemfile with "ruby '2.3.3'", which nixpkgs doesn't use. It seems that specifying an arbitrary version is not possible with the ruby's nixexpr. :( I guess I'll've to use rbenv.
DenisLehmann[m] has joined #nixos
<{^_^}>
[nixos-homepage] @edolstra merged pull request #195 → Add nixos wiki to support → https://git.io/vxETb
<{^_^}>
[nixos-homepage] @edolstra pushed 2 commits to master: https://git.io/vxEmS
<{^_^}>
→ 39f7cde5 by @cx405: Add nixos wiki to support
<{^_^}>
→ f2324271 by @edolstra: Merge pull request #195 from cx405/patch-1
freeman42x]NixOS has joined #nixos
peacememories has joined #nixos
rauno has quit [Ping timeout: 240 seconds]
civodul has joined #nixos
<zybell_>
ottidmes: if you want to unify bash and python please note that """:" opens a docsting in python and is a noop in bash. """:"<<SNIP comments out the following lines both in bash and python, but the end is marked differently.
<ottidmes>
zybell_: Thanks for the tip, but I doubt I will ever use it myself, not a fan of Python. I just found it an interesting workaround to having to quote most of the file as a string
zzamboni has quit [Quit: Leaving.]
coot has quit [Quit: coot]
generic has joined #nixos
abathur has quit [Client Quit]
obadz has quit [Ping timeout: 268 seconds]
obadz- is now known as obadz
<{^_^}>
[nixpkgs] @teto opened pull request #37865 → gnulib: symlink include/ to lib/ → https://git.io/vxE3d
<srhb>
This branch isn't dynamic -- it takes human intervention to modify it.
<AMDmi3>
but I'll have to track these changes in repology
<srhb>
But the channel (nixos-18.03) will always lag behind a bit, due to tests etc.
<srhb>
Rigt..
Sonarpulse has joined #nixos
xy2_ has joined #nixos
<srhb>
AMDmi3: Someone has already been doing a lot of work on repology stuff for nixpkgs, I don't recall who though.
<srhb>
Could be ryantm maybe?
javaJake has joined #nixos
<infinisil>
AMDmi3: ah you want a channel that's always on the lates release, so 17.09 -> 18.03, etc.?
zzamboni has quit [Remote host closed the connection]
<shlevy>
FWIW right now the "latest" should be considered 17.09
<shlevy>
18.03 is still in pre-release stabilization
\\\\\\\\\\\ has joined #nixos
<shlevy>
I wouldn't be opposed to a release-latest concept somewhere
lexos has joined #nixos
ryanartecona has joined #nixos
javaJake has left #nixos ["leaving"]
vidbina has joined #nixos
Xyoff has joined #nixos
ckauhaus has joined #nixos
fendor has quit [Remote host closed the connection]
<lexos>
hello, I would like to install an additional font in my nixos, and set it in VScode. font package name is noto-fonts-emoji-git-2016-03-17 . Does nix-env -i noto-fonts-emoji-git-2016-03-17 will make this font available for VsCode? in which path VsCode look for fonts?
nick_l has joined #nixos
<nick_l>
How do I disable iptables temporarily on NixOS?
<nick_l>
(without doing a rebuilding of the configuration)
<aminechikhaoui>
any idea what can I look at to investigate the issue with microphone
<srhb>
Like, traffic from inside is natted to appear like traffic from outside
<Dezgeg>
I wonder if there's some sort of X-Y problem here, because I wouldn't expect most apps to care about the network interface
<clever>
nick_l: line 30 says it will accept any icmp with type 8, which should be ping-request
<nick_l>
Dezgeg: I am running keepalived, have a program to send UDP packets and I just want ipvs to forward those packets to some machines. At this point, I'd be happy with wireshark actually seeing forwarded packets on the keepalived machine.
<Dezgeg>
so you want your machine to _forward_ packets from other machines, not send them itself?
<aminechikhaoui>
oh the github issue is probably not related, as I see the same output from before the switch to 18.03
<nick_l>
Dezgeg: forwarding/sending doesn't matter for me.
pkill9 has joined #nixos
<nick_l>
Dezgeg: I just want to see packets moving at this point.
<Dezgeg>
well, try nc -l -u -p 1234 to have an udp server up on port 1234
<Dezgeg>
and nc 10.0.0.10 -u 1234 to have a udp client connect to that port
<nick_l>
Dezgeg: .10 should be the virtual ip of interest or the private ip?>
<nick_l>
Dezgeg: running on the virtual ip of interest.
<Dezgeg>
doesn't any IP do?
<Dezgeg>
any IP of the server machine?
<nick_l>
Dezgeg: echo hi came to the other end.
<Dezgeg>
sounds like you have packets moving then
<nick_l>
Dezgeg: yes, but it is keepalived which should be moving them.
<nick_l>
Dezgeg: so, a packet should be sent to the virtual_server of keepalived, and then it should forward those packets to available backend servers.
<nick_l>
Dezgeg: my configuration works in libvirtd. It's just AWS that doesn't work.
<Dezgeg>
I don't know what keepalived is or what it does
<nick_l>
And the only difference is that ping -I eth1 <ENI IP>
<nick_l>
Dezgeg: it is a configuration system for ipvs.
<Dezgeg>
but if you need to have your system actually forward packets on the IP level you need /proc/sys/net/ipv4/ip_forward set
<nick_l>
Dezgeg: I have that set to 1.
<Dezgeg>
among probably other stuff I can't remember
<srhb>
I don't remember having to set up anything special for keepalived.
<nick_l>
srhb: on AWS?
<srhb>
Uh, no, just NixOS :)
<srhb>
I don't know how the AWS stack works.
<nick_l>
srhb: no, on NixOS it works fine. It's the combination that doesn't work.
<srhb>
And they're not using some out-of-band firewalling between your machines?
<srhb>
That would be yucky, so I guess not...
<nick_l>
srhb: all I am testing is still on a single machine.
<nick_l>
srhb: I.e. wireshark with packets coming in and out.
<srhb>
Ah
<nick_l>
srhb: since this is NixOS, can I add single printf statement in the kernel code for ipvs?
<nick_l>
srhb: if I can do that, then I'd hope that I can just get a message output like "we have received a packet".
<nick_l>
ipvs can also be configured with debugging, I think. Not sure how.
<srhb>
I mean sure (though probably printk), but that's some big cannons. :P
<srhb>
I'm inclined to believe we're missing something obvious...
logzet has quit [Remote host closed the connection]
<abbradar>
I'm investigating a strange Nix "daemon out of memory" error on one of my machines. It seems strange that nix-daemon retains all the memory it allocated during evaluation when the build happens
<abbradar>
Say, I run "nixos-rebuild switch", the daemon eats 350mb of memory and then build starts but all the memory is still uesd -- shouldn't at this point just ready derivations be in memory?
<abbradar>
(well, maybe it is the case but then it's strange they use so much space)
<AMDmi3>
infinisil: yes, that would be nice; although for repology it's only needed for a link to .nix file correctly, not sure if it's enough to justify it
<nick_l>
srhb: Let's say I really, really badly wanted to patch that kernel module, how do I tell NixOS that?
<srhb>
nick_l: boot.kernelPatches
<nick_l>
srhb: I am basically out of other ideas.
<nick_l>
srhb: I learned more about networking then I wanted to know too.
<srhb>
nick_l: Heh, I know that feeling. :P
<srhb>
Diagnosing network issues is always like voodoo.
<srhb>
And, in my personal experience, mostly because I don't know what's going on.
peacememories has joined #nixos
<srhb>
Which is always fun!
<Dezgeg>
FWIW I can reproduce on an arch linux machine that 'ping -I wlp3s0 10.112.22.133' doesn't give pongs either
John882 has joined #nixos
<ekleog>
Hmm… Anyone knows whether we have a mdadm module that would run mdadm --monitor, so as to get emails when disks fail? I can't find any, but am pretty sure it must exist, so before writing one…
WhittlesJr has joined #nixos
<cransom>
if you are doing IP level weirdness on AWS, if packets are leaving your machine that don't have it's source address, you need to remove source/dest packet checks
<WhittlesJr>
So... is it "Nix Oh Es," "Nix Ahs," "Nix Ohs," or something else?
<nick_l>
cransom: the ENI is already created not to do that.
<johnw>
I say Nix Oh Es
<shlevy>
Whittlesjr: I do the first
simukis has quit [Ping timeout: 240 seconds]
<gchristensen>
I used to say "nix ahs" but then trained myself to say "Nix oh es" before going to Nix Con to avoid the embarrassment
<shlevy>
Though in formal context it is of course "Nicks-oh-ess, the purely functional Linux distribution TM"
<srhb>
shlevy: *A* purely functional... :P
dan_b has quit [Ping timeout: 240 seconds]
<cransom>
nick_l: are there multiple interfaces involved? you confirmed with tcpdump that packets are leaving the interface?
<johnw>
NixOSisNotGuix
<nick_l>
cransom: yes, two interfaces, eth{0,1}
<WhittlesJr>
shlevy: hmm, not quite enough syllables... shouldn't it really be "GNU/Linux?"
<cransom>
nick_l: the routing table has updated to get the outgoing traffic out the correct interface?
<nick_l>
cransom: packets are sent from 172.31.50.6 to 172.31.50.6 (so, their source address is also weird)
<nick_l>
cransom: I had expected .7 as the source address (which is the primary private IP)
<nick_l>
cransom: I did something in that area:
<nick_l>
cransom: ip rule add {to,from} 172.31.50.6/32 table out
<cransom>
nick_l: ips wouldn't change unless you are doing masquerading or source translation
<nick_l>
cransom: but the important part is that I don't have the impression that the packets ever end up at some kernel part that actually creates new UDP packets again.
txp284 has quit [Ping timeout: 276 seconds]
<nick_l>
cransom: I have actually seen it shortly perform the right job (i.e. create packets to its backend servers)
<nick_l>
cransom: but that was last Friday before I stopped the servers. After a restart, it was broken.
<WhittlesJr>
oh, speaking of Guix... can I just admit that one of the biggest barriers to me using that OS is actually its name? "Hey man you should switch to GEEKS ES DEE it's sweet"
<cransom>
nick_l: and this `out` table exists after reboot? there's a whole bunch of things that could be it between tables, rules, if RPfilter is kicking in
<nick_l>
cransom: out table didn't exist last Friday.
<cransom>
are your two interfaces on the same subnet?
<srhb>
Bloody hell, did tarsum break _again_? Q_Q
sanscoeur has joined #nixos
AMDmi3 has quit [Ping timeout: 256 seconds]
* srhb
flails
AMDmi3 has joined #nixos
tertle||eltret has joined #nixos
peacememories has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<zybell>
Everybody new to networking makes that mistake, why do you think nick_l hasn't. cransom be realistic.
peacememories has joined #nixos
<nick_l>
cransom: yes
<johnw>
Whittlesjr: I have trouble taking pijul seriously because of its name too
<nick_l>
cransom: is that not allowed?!
<cransom>
nick_l: do you plan on pushing multiple gigabits through that machine?
<nick_l>
cransom: no
<nick_l>
cransom: at least, not per second.
pkill9 has left #nixos ["WeeChat 2.0.1"]
<nick_l>
zybell: can you please elaborate?
<cransom>
nick_l: it's allowed, but its also hard... and weird because 1 interface is enough.
<nick_l>
cransom: I need one interface to do unicast communication with the failover.
<nick_l>
cransom: or at least, I think so.
<cransom>
in aws, everything is unicast
<nick_l>
cransom: but keepalived needs one IP to listen on and one IP to send VRRP announcements over.
<cransom>
and plumbing in another interface isn't going to make that any more reliable when you are just a vm in a cloud.
<nick_l>
to*
bkchr has quit [Ping timeout: 260 seconds]
<fearlessKim[m]>
johnw: u better take me seriously or I nuke you :p
stanibanani has joined #nixos
<nick_l>
cransom: what should I do?
<cransom>
actually, lets step back here.... keepalived? is it going to make ec2 api calls to move ips between the HA machines?
<nick_l>
cransom: yes
<nick_l>
cransom: but that's not relevant right now.
<zybell>
nick_l: As I have learned networks, every network cable gets its own IP. A math formula describes the relationship between IP of card and IP of cable.
<nick_l>
zybell: well, they have a different IP, just the same subet.
<nick_l>
zybell: well, they have a different IP, just the same subnet.
<nick_l>
cransom: what is relevant is getting a packet to ipvs and from ipvs to somewhere else. That's it.
<nick_l>
cransom: and perhaps this whole ping issue.
<nick_l>
cransom: anything more advanced is a step too far at this point.
<zybell>
If cables have the same subnet, its nearly impossible that they have different IP.
<cransom>
nick_l: ok, well. good luck. i've gotta go. a lot of the complexity would go away if you used multiple subnets, or even better, a load balancer that aws will gladly provide
<nick_l>
cransom: aws doesn't do UDP loadbalancing.
<nick_l>
cransom: I can easily add another subnet.
<nick_l>
cransom: so, you are suggesting that one machine will have two subnets (one per interface)?
<nick_l>
cransom: why would that make things simpler?
<zybell>
because simpler routetables
dan_b has joined #nixos
<ekleog>
hmm so I guess there's no mdadm module for running mdadm --monitor, then :/
<nick_l>
zybell: can you explain why ipvs doesn't ever seem to appear to receive a packet?
<nick_l>
zybell: or at least try to send one?
<nick_l>
zybell: the--stats option suggests that lots of packets come in, btw.
<nick_l>
zybell: but nothing ever comes out (in those same stats) which actually is normal when using the DR (direct routing option)
<nick_l>
zybell: so, it's not clear for me how to actually test what is and what isn't working.
<nick_l>
zybell: I would also appreciate it if you could be more specific.
<WhittlesJr>
johnw: Eek. Nothing quite beats "Guh-noo" though imho
<gchristensen>
is that "noo" like "new" or an elongated "no"
<TweyII>
Yod-dropped ‘new’ of course
<rycee>
fearlessKim: What infinisil said :-)
<infinisil>
huh?
<zybell>
if you don't have a route there, you can't send pkgs there. And if can't send there the 'martian'(yeah, is so called)-filter will remove all pkt that can't be send from recv.
<WhittlesJr>
gchristensen: HAHA I was definitely going for "guh-new" but the thought of "Guh-NO" tickled me. "Guh-NO thank you."
__Sander__ has quit [Quit: Konversation terminated!]
<infinisil>
rycee: what did i say?
jrolfs has joined #nixos
jensens has quit [Ping timeout: 240 seconds]
<fearlessKim[m]>
infinisil: u loved icecream (how to import nixos config)
<zybell>
Try first to configure only one interface. Ping it from inard. Ping it from outer. Check TCP. From in. From Out. Check UDP. From in. From Out. In this order. Only then configure second IP. On different subnet.
<how>
Hi, I'm reading nixos wiki on moving /nix to a a separate lvm volume but I'm unsure on some of the steps
<gchristensen>
srhb: mind if I PM? :)
<srhb>
gchristensen: You're always welcome :)
<how>
e.g. how to bind /mnt to /nix
<how>
and bind mount / to /old_root but there's no /old_root
davidlt_ has joined #nixos
<clever>
how: id recomend moving it while the system is not running, just rsync the data to the new filesystem, delete the old, and convince it to boot with the new fs mounted at the right place, then fix the nixos config
<how>
clever: what do u mean when the system is not running?
<clever>
how: boot from another system, like the installer iso or a recovery image
<clever>
how: for rescue-boot, you add imports = [ ./rescue_boot.nix ]; to your configuration.nix, and it will generate a new grub option that boots into a rescue OS
kai_w has quit [Quit: kai_w]
<clever>
how: ive used that to do exactly what your doing, move my laptop /nix into its own zfs dataset
<how>
clever: so I just add that import to my configuration.nix, then reboot into the rescue os, add filesystems."/nix" = ... and rsync, then reboot and delete old /nix?
<clever>
how: you can add filesystems."/nix" = ... and `nixos-rebuild boot` before you move it, to save some trouble
<clever>
then youll have a new generation in grub that expects it to be moved, and an old that doesnt
<how>
clever: then what am I doing in the rescue-boot?
<clever>
make the new LV, format it, and move all of /nix over
<how>
hmm i have already done that should i do it allover again?
<clever>
if the data is already moved over then you just need to update configuration.nix and reboot i believe
<how>
besides the last one is copy instead of move obviously
ndrei has quit [Remote host closed the connection]
<how>
so I don't even need to rescue-boot?
<clever>
yeah, if the data has already been moved, you dont need it
ndrei has joined #nixos
Arcaelyx has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<how>
ok thanks, last question: what configuration options do I need in filesystem."/nix" = ...? I guess just device and neededForBoot?
<clever>
neededForBoot is automatic
<clever>
it will need device and fsType
<how>
clever: thanks so much!
<clever>
yep
blankhart has quit [Ping timeout: 240 seconds]
anixos has joined #nixos
<anixos>
hello, I'm having problems logging in with gdm. My user was created with configuration.nix (hashed password and all that), but gdm prompts me to enter my username and password, and when I do that it doesn't let me in (says something went wrong)
<anixos>
when I add an user with useradd homever, everything works fine and gdm detects my user automatically. What could be the cause?
fresheyeball has joined #nixos
llhart has joined #nixos
<clever>
anixos: are you using hashedPassword or initialHashedPassword?
coconnor has joined #nixos
<fresheyeball>
hello out there
<coconnor>
morning all.
<fresheyeball>
last week I was trying to get my hakyll blog working again
<fresheyeball>
and made a ton of progress with help from this most glorious IRC community
<anixos>
clever: hashedPassword
<fresheyeball>
but did not succeed
<fresheyeball>
the unit tests for hlibsass do not pass (a haskell package)
<clever>
anixos: i think you need to set mutableUsers = false; for that to work
<anixos>
okay thanks, I'll come back and report if it works
<fresheyeball>
even when using an overlay and a super.haskell.lib.dontCheck
xeji has joined #nixos
<fresheyeball>
I am at a loss for how to get the unit tests to stop being run in my build
<clever>
fresheyeball: can you gist the overlay?
<fresheyeball>
clever: sec
<coconnor>
I have an odd one today - the ninja build tool when used by qmake is ignoring NIX_BUILD_CORES
ndrei has quit [Remote host closed the connection]
blankhart has joined #nixos
<clever>
fresheyeball: // overwrites the hlibsass atttribute of the resulting haskellPackages, but it does not make other haskell packages use the new version
<coconnor>
which, on this system, causes it to hit swap building qtwebengine
<llhart>
Hello everyone! Is there any known problems with radeon driver currently? It was working fine ~2months ago but not anymore... :/
<clever>
fresheyeball: you must use super.haskellPackages.override { overrides = { hsself: hssuper: { ... }; }; }; to write a haskellPackages overlay
ndrei has joined #nixos
anixos has quit [Remote host closed the connection]
daGrevis is now known as daGrevisdaGrevis
ndrei has left #nixos [#nixos]
<coconnor>
llhart: which driver specifically? paste your xserver config and kernel config?
daGrevisdaGrevis is now known as daGrevis
<fresheyeball>
clever: can you show me specifically how you would do that?
<anixos>
clever: unfortunately, GDM still asks for username and password, and when I type it in still says something went wrong
<lejonet>
Hmm, using environment.etc."thing".source, is that the recommended way of symlinking a dir into etc? (I'm going to try and symlink in the etc/OpenCL dir from nixos.amdsdkappFull, hoping that is enough to get opencl functional)
<anixos>
perhaps theres some special file useradd modifies? or some group?
<clever>
anixos: double-check the /etc/shadow file, and confirm if the hashed pw is actually in there
<anixos>
clever: it is indeed
<anixos>
I can login just fine through any of the ttys
<clever>
anixos: and do you have isNormalUser set on your user?
<srhb>
Remind me, is there a nicer way to backport than cherrying and redoing the entire PR? It's so tedious...
<anixos>
clever: yes
<srhb>
There should be a backport bot...
<coconnor>
anixos: are you in the video group?
<anixos>
huh... nope, going to try this one now
ndrei has quit [Remote host closed the connection]
anixos has quit [Remote host closed the connection]
ndrei has joined #nixos
<coconnor>
EG: the /dev/fb0 device is root:video. I didn't think it mattered (yet) for other "video" things but...
<clever>
coconnor: X is already running and using fb0 by the time the login prompt is open
anixos has joined #nixos
<coconnor>
AH
<anixos>
that still hasn't worked, I can post my configuration.nix if that helps
how has quit [Quit: leaving]
<coconnor>
although... the login X is a different user
<coconnor>
anixos: yes
<anixos>
I'd like to note that this isn't a regular GDM screen where it shows your username and only prompts for a password, it shows an username field
<coconnor>
enter user name then password? LIght colored with nixos logo?
reinzelmann has joined #nixos
<coconnor>
I think that's sddm (the default) not gdm
<anixos>
no coconnor, GDM login screen
<clever>
anixos: check "ps -eH x" and see what process is near X in the tree
<clever>
fresheyeball: you need to tell cabal to -lsass, because the static library lacks that metadata
ryanartecona has joined #nixos
<anixos>
yes coconnor, indeed I can
<{^_^}>
[nixpkgs] @dtzWill closed pull request #37810 → udisks2: 2.1.6 -> 2.1.8, extend workaround to include musl → https://git.io/vx0LU
zybell has quit [Ping timeout: 268 seconds]
<sasoiliev>
Hi all, I'm running 17.09.3204.d16ffa15090 and I'm trying to install Viber, but the viber.deb artifact has been updated and the hash in 17.09 does not match.
<anixos>
I'd like to clarify that it is indeed the GDM screen, but instead of showing the username and only prompting for password it prompts for both, and when the correct are typed it fails to login. I'd like to add that it works when using useradd(1).
<fresheyeball>
clever: not sure what you mean, I have never seen that flag
<clever>
how: oh, and that copy lacks the new nixos you just made with `nixos-rebuild boot`, thats why it cant boot
<clever>
you have to make that copy after running `nixos-rebuild boot`
<how>
oh so i just need to nixos-rebuild boot now, rsync again, and rebooting should work?
<clever>
yeah
<how>
clever: I'll try that thanks!
anixos has joined #nixos
<infinisil>
Hmm, wouldn't it be nice if there was 'pkgs.channels.unstable' reflecting all unstable packages? Would be super easy to use specific stuff from a different channel
<anixos>
just confirmed useradd(1) works perfectly and GDM lists the user correctly. Tried setting GECOS field for my own user to no luck.
zybell has joined #nixos
<srhb>
infinisil: That sounds like a can of worms!
<srhb>
infinisil: Which unstable commit does a stable commit correspond to etc :P
<infinisil>
srhb: ah no! I mean using IFD and fetch the nixos-unstable branch, not pinned
<infinisil>
Would be really convenient when you just wanna use a different branch but don't want to deal with nix-channel, overlays, git cloning or whatever. Everybody could use this easily
<srhb>
I never understood quite how it works wrt caching.
Lisanna has joined #nixos
<infinisil>
1 hour it gets cached
<srhb>
I see.
<Lisanna>
...did the 4.9 kernel get removed from nixpkgs? ):
<srhb>
Well, I certainly see the convenience, but it does seem a little tricky.
<infinisil>
srhb: what tricky part is there to it?
<srhb>
Lisanna: Nope.
<srhb>
infinisil: Well, the whole "what are you currently running" shenanigans.
<srhb>
It's hard enough with channels in general.
<Lisanna>
srhb oh good, I just can't find the package for it
<srhb>
Which should just die already. :-P
<coconnor>
llhart: what is failing for you? is X server failing to start?
<srhb>
Lisanna: linux_4_9 ? Or linuxPackages_4_9 ?
<Lisanna>
srhb no results for either
<anixos>
clever: coconnor: jorunalctl says "gkr-pam: no password is available for this user"
<srhb>
And it uses NIX_PATH, which terrifies me :P
<srhb>
(Or relies on, rather)
michas_ has joined #nixos
<infinisil>
srhb: it does!
<infinisil>
?*
<srhb>
Doesn't it?
ndrei has joined #nixos
<srhb>
I thought that's where it got the names from for the channel in question.
<srhb>
Maybe not!
<infinisil>
Channels are just mappings from name -> commit afaik
<srhb>
ocharles: According to the short help, https is not supported.
<infinisil>
Ah yeah it does if you use it like import <unstable> {}
<infinisil>
But that's a nix feature, has nothing to do with channels directly
<srhb>
Right, it's mostly how it's used that bothers me.
<srhb>
I'd prefer a canonical nixpkgs.json tool over nix-channel --update and company.
<srhb>
Or whatever. Something simpler :)
<srhb>
And closer tied to the code.
<llhart>
coconnor: Thank you. My system is working but when using xrandr --listproviders, I got modesetting. (My radeon is a discrete GPU so I think I'm falling back on my intel GPU)
<infinisil>
srhb: hmm.. not sure what you're thinking of
<srhb>
infinisil: I want the version to embedded near, say, configuration.nix
blankhart has quit [Read error: Connection reset by peer]
blankhart has joined #nixos
<coconnor>
llhart: try pasting the output of "journalctl -b -u display-manager"
<coconnor>
llhart: that is the X server (the display manager in this case) startup
<coconnor>
llhart: will have details on what devices were probed and what drivers worked
<infinisil>
srhb: you know what would be nice: numbered channel updates
<anixos>
coconnor: any ideas? nothing seems to be working.
chisui has quit [Ping timeout: 260 seconds]
<infinisil>
The first version of 18.03 would have number 1, the next update 2, etc.
<srhb>
infinisil: Sure. It's an indirection that isn't necessary, I feel, but it's better than just "some point at channel" :-)
<anixos>
coconnor: seems to be an issue with the password, at least from my useradd(1) debugging. using merely useradd <NAME> doesnt make GNOME list it, but when you passwd it does
ThatPako has joined #nixos
<coconnor>
llhart: huh! indeed. see if "journalctl -ru display-manager" has anything more
<infinisil>
srhb: ah i see what you meant: to have a json file per channel that gives the hash and url of it, then have some nix code that imports that file and uses it to e.g. set pkgs.channels.foo
<coconnor>
llhart: the "-b" was "since last boot" or something.
steveeJ has joined #nixos
<Lisanna>
srhb okay, verified that it's still there with nix repl, funny that it doesn't show up in the web search... thanks
anixos has quit [Remote host closed the connection]
<srhb>
Lisanna: You're welcome! I find the cli tools much more accurate :)
<srhb>
infinisil: Yes exactly. Sorry, I wasn't clear. :)
<srhb>
infinisil: "Less Magic"^tm
anixos has joined #nixos
<dahirsch>
clever: are you there?
<clever>
dahirsch: yes
<anixos>
seems to not work with configuraiton.nix created user and passwd
<infinisil>
srhb: ah and the main thing to make this happen is to adjust the nix-channel tool
<{^_^}>
→ c9ea1ff5 by @Baughn: mpv: Add openal support
<{^_^}>
→ 47205202 by @Mic92: Merge pull request #37745 from Baughn/mpv-openal
humanoyd has quit [Quit: WeeChat 2.1]
pop has quit [Quit: WeeChat 1.0.1]
<dahirsch>
clever: i experimented a bit with the uefi issue. i can boot with 'uefi only' via systemd-boot. but grub only works then i use 'legacy support', does not matter if i choose uefi or legacy first then. any ideas why grub does not work?
<dahirsch>
*when
<srhb>
infinisil: Yeah, it's quite non-invasive.
<srhb>
But channels have generations, which you want for rollbacks
<srhb>
So meeeh
ndrei has quit [Ping timeout: 240 seconds]
<clever>
dahirsch: nope, ive not done much with uefi
<clever>
dahirsch: but you could start by comparing what is in /boot and efibootmgr in each setup
sasoiliev has quit [Ping timeout: 260 seconds]
<{^_^}>
[nixpkgs] @dtzWill reopened pull request #37810 → udisks2: 2.1.6 -> 2.1.8, extend workaround to include musl → https://git.io/vx0LU
<dahirsch>
clever: this would have been my next step. thanks
<infinisil>
srhb: well i think that's acceptable
<llhart>
coconnor: Hum Just got the same output but multiple times :/
<ThatPako>
do I have to provide a hash for `fetchFromGithub`?
<srhb>
ThatPako: Yes.
mounty has quit [Ping timeout: 264 seconds]
<ThatPako>
srhb: How would I obtain that hash?
<srhb>
!tofu
<infinisil>
srhb: you might have to refetch the channel on rollback, but that's okay imo
<{^_^}>
To get a sha256 hash of a new source, you can use the Trust On First Use model: use probably-wrong hash (for example: 0000000000000000000000000000000000000000000000000000) then replace it with the correct hash Nix expected.
<srhb>
ThatPako: ^
<coconnor>
llhart: doh. might be filtered to your user. "sudo journalctl ..."?
<ThatPako>
oh, I see
<ThatPako>
is fetchFromGithub documented somewhere?
<srhb>
ThatPako: Nope.
<ThatPako>
Damn.
<infinisil>
Unfortunately
<clever>
ThatPako: the main args are owner, repo, rev, sha256
<srhb>
ThatPako: It's defined in pkgs/top-level/all-packages.nix if you want to check it out :)
<srhb>
But that's probably less than helpful. You probably just want those 4 arguments and be done with it.
<{^_^}>
[nixpkgs] @grahamc opened pull request #37881 → nixpkgs docs: Use SVGs for callouts → https://git.io/vxEFv
<llhart>
coconnor: Same thing :(
marsam has quit [Read error: Connection reset by peer]
marsam has joined #nixos
anixos has quit [Quit: Lost terminal]
coot has joined #nixos
mounty has joined #nixos
<phry>
can someone recommend a file browser/image previewer that works fine without kde installed? everything I try supports zero image formats. I guess missing plugins/libraries.
<krey>
phry: pcmanfm is pretty good
<krey>
phry: mirage for image preview
<ThatPako>
what should go into the `rev` field for `fetchFromGitHub`?
<clever>
ThatPako: a git revision
<phry>
krey: mirage does not recognize images for me, installing pcmanfm right now
<ThatPako>
clever: how do I get the most recent one? I never heard the word "revision" before :/
<clever>
ThatPako: which repo are you trying to fetch?
<clever>
ThatPako: click on the commits link near the top, then the msg for that top commit
<clever>
ThatPako: the revision is the sha1 in the URL
<ThatPako>
ohhhh, I copied that into the hash, not the rev. that's why it wasn't building. derp
<ThatPako>
Thanks!
<krey>
phry: that's pretty weird
<krey>
phry: what do you mean "it doesn't recognize images"?
<ThatPako>
Okay, now it build but the result folder is empty... Did I do something wrong? :/
<phry>
krey: it just doesn't show image types in the file selection. although I now installed it via nix-env instead of `nix run` and it seems to give better results
<ThatPako>
actually... looks like it needs some non-standard build steps. damn
<clever>
ThatPako: can you gist the nix expression?
<johnw>
with Nix 2.0, is there any way to pass "-k" to "nix build"?
<infinisil>
Need to read the ascii art in reverse
<niksnut>
johnw: --keep-going
<srhb>
johnw: Maybe --option keep-filed true will do it?
<srhb>
oh, keep-going...
<johnw>
ok, nix build --help just doesn't mention it
<ThatPako>
Any idea how to fix `/nix/store/klwxxrlgalq49cjnczvgha3j7l00jixw-binutils-2.28.1/bin/ld: cannot find -lcurses
<ThatPako>
?
<{^_^}>
[nixos-homepage] @edolstra merged pull request #197 → Copy non-standard style and JS DOM nodes from the generated docs for richer docs → https://git.io/vxEp3
<{^_^}>
[nixos-homepage] @edolstra pushed 2 commits to master: https://git.io/vxueA
<{^_^}>
→ 64dceab3 by @grahamc: Copy non-standard styles from the generated docs + JS for richer
<{^_^}>
→ 4584abbc by @edolstra: Merge pull request #197 from grahamc/rich-docs
<infinisil>
which simulates a "standard" linux environment
<srhb>
d-fish: In nix land we usually deal in expressions that are distinct from the package "artifacts"
<srhb>
d-fish: Ideally you'll be able to write a nix expression that in itself is self contained and has the instructions for fetching and installing your package. Be it from source or binary.
<ThatPako>
REEEE `make: *** No rule to make target 'install'. Stop.`
<ThatPako>
I'm so close...
xy2_ has joined #nixos
<srhb>
ThatPako: Check if the binaries are actually all cooked. If they are you probably just need to override installPhase to copy them to $out
<srhb>
Otherwise you'll probably have to dig deeper into whatever the build system needs to produce a neat install :)
Meizikyn has joined #nixos
winem_ has quit [Ping timeout: 256 seconds]
<Meizikyn>
In the nix-expression/Attribute-path "nixos.linuxPackages.kernel.dev", what is the proper name of the element "*.dev" in that string? Is it also considered part of the attribute or is there a specific name for that segment of the attribute path?
<{^_^}>
→ a80856ce by @grahamc: nixpkgs docs: Use SVGs for callouts
<{^_^}>
→ da179745 by @grahamc: Move the overridden CSS to overrides.css which can be applied on nixos.org
<{^_^}>
→ 63406fa1 by @grahamc: Merge pull request #37881 from grahamc/doc/svg-callouts
<{^_^}>
[nixpkgs] @grahamc merged pull request #37881 → nixpkgs docs: Use SVGs for callouts → https://git.io/vxEFv
Arcaelyx has joined #nixos
asuryawa_ has joined #nixos
<{^_^}>
[nixpkgs] @grahamc pushed 2 commits to release-18.03: https://git.io/vxuTm
<{^_^}>
→ c0c6e986 by @grahamc: nixpkgs docs: Use SVGs for callouts
<{^_^}>
→ a50e5538 by @grahamc: Move the overridden CSS to overrides.css which can be applied on nixos.org
asuryawa_ has quit [Remote host closed the connection]
Meizikyn has joined #nixos
asuryawa_ has joined #nixos
asuryawanshi has quit [Ping timeout: 276 seconds]
<ThatPako>
just found this in a file called `install.sh` `binaries=${1:-"/usr/local/bin"}`. What should I `sed` this to?
Intensity has joined #nixos
ixxie has quit [Ping timeout: 256 seconds]
<ThatPako>
also, can I make it install the standard packages there to?
<makefu>
ThatPako: how do the lines with $binaries look like?
<srhb>
ThatPako: You can't/shouldn't use either of those directories.
<ThatPako>
yea
<makefu>
if they look like this: $binaries/bla then you could sed away $binaries/
<makefu>
and make the package work by assuming stuff is in PATH
<srhb>
I don't think that's exactly what you want though. Try looking for defaultPackagesDirectory for instance
<ThatPako>
okay, so... the emojicode compiler can be found at /build/Compiler/emojicodec. should I copy that file to `${out}/bin`?
Xal has joined #nixos
<ThatPako>
still not sure where the packages should go though :/
<Xal>
I'm trying to package a qt5 application and I'm getting a "contains a forbidden reference" error
<srhb>
ThatPako: Probably somewhere like $out/EmojiCode
<srhb>
Er, $out/EmojicodePackages
<Xal>
strangely, when I build the application with qt4 it works fine
<ThatPako>
is it possible to "wrap" the compiler somehow then? it has to know about this place after all
<srhb>
ThatPako: It's actually compiled into the binary where it looks by default.
asuryawa_ has quit [Read error: Connection reset by peer]
<srhb>
ThatPako: So you'll want to find the configuration option for that.
coot has joined #nixos
asuryawanshi has joined #nixos
coconnor has quit [Remote host closed the connection]
<ThatPako>
oh, true. gotta figure out how to change that then
<srhb>
ThatPako: Look for defaultPackagesDirectory
<ThatPako>
oh, yea
how has quit [Ping timeout: 276 seconds]
<{^_^}>
[nixpkgs] @shlevy pushed to release-18.03 « haskellPackages.lenz-template: Fix against lenz 3 »: https://git.io/vxukD
<{^_^}>
[nixpkgs] @shlevy pushed to master « haskellPackages.lenz-template: Fix against lenz 3 »: https://git.io/vxukS
Neo-- has quit [Ping timeout: 256 seconds]
<d-fish>
How would I add an sddm theme?
pie_ has joined #nixos
asuryawanshi has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @coreyoconnor opened pull request #37889 → qtwebengine: correct how NIX_BUILD_CORES is propagated to the gn ninja → https://git.io/vxuIL
Ariakenom has joined #nixos
asuryawanshi has joined #nixos
orivej has joined #nixos
coconnor has joined #nixos
<ThatPako>
why is `out` considered an undefined variable in `PatchPhase`?
ryanartecona has joined #nixos
<coconnor>
build systems all the way down
<{^_^}>
[nixpkgs] @tobimpub opened pull request #37890 → cquery: init at 2018-03-25 → https://git.io/vxuIy
hamishmack has joined #nixos
jensens has quit [Ping timeout: 256 seconds]
<srhb>
ThatPako: It's not.
<srhb>
ThatPako: It's a shell variable though, not a nix variable.
<Meizikyn>
Found something odd, derivation attribute-path for "nixos.linuxPackages.kernel.dev" , "*.dev" is called an "output". looking at its meta section (in xml) it shows <meta name="outputsToInstall" type="strings">\n\t<string value="out" /> is there another way via. a cli-tool to show possible outputs of a derivation?
<LnL>
ThatPako: it is, what's the problem you are running into?
<ThatPako>
I tried `sed -i 's,/usr/local/EmojicodePackages,${out}/EmojicodePackages' Compiler/CL
<coconnor>
I still haven't found the nix 2.0 equivalent for dry-run
<srhb>
LnL: come again?
<ThatPako>
do I have to mkdir `$out` too?
d-fish has quit [Quit: Konversation terminated!]
<srhb>
coconnor: --dry-run ?
<srhb>
Hmm, no, not quite the same.
<srhb>
ThatPako: In essense, you have to do everything in order to populate $out. Normally, the build system will have some sort of install command that, when fed a prefix, does all that for you.
<srhb>
ThatPako: That does not appear to be the case here.
<srhb>
ThatPako: So it's your job. :)
<LnL>
srhb: hmm?
<srhb>
LnL: I didn't understand $(out) is make syntax not bash
<ThatPako>
I tried it with cmakeFlags but how do I get my `out` in there?
<LnL>
it gets executed as make 'FOO=$(out)'
<LnL>
make interprets the out not bash
<coconnor>
srhb: ... oh
<coconnor>
hahaha
<coconnor>
excellent!
<srhb>
ThatPako: That was why I asked about the quoting there.
<ThatPako>
about what quoting?
<srhb>
In cmakeFlags
<srhb>
Because I was unsure how to feed cmake $out
<ThatPako>
I still don't understand it :/
<srhb>
Which I think is what LnL is explaining, and I don't understand it either :)
pie_ has quit [Read error: Connection reset by peer]
<LnL>
:/
<ThatPako>
so it should be `cmakeFlags = [ "-DdefaultPackagesDirectory=$(out)" ];`?
pie_ has joined #nixos
<LnL>
yes, that should work
<srhb>
Aha!
<{^_^}>
[nixpkgs] @dtzWill opened pull request #37892 → virglrenderer: fix missing include → https://git.io/vxutO
<LnL>
but just $out or ''${out} wouldn't
<infinisil>
Whoa, I just found the best systemd manpage of them all
<infinisil>
man systemd.directives
<infinisil>
Because it shows you in which man pages you can find stuff
<srhb>
kerrhau: Embarrasingly I can't read from that whether you actually have a channel named nixos which is the one you should be using
<tobiasBora>
Hello,
<{^_^}>
[nixpkgs] @dtzWill opened pull request #37894 → grub2: extend platforms to include all linux → https://git.io/vxuYs
<srhb>
kerrhau: The only other thing I can think of is forgetting to allowUnfree in your nixpkgs config, but people always look at me like I'm superstitious when I claim that the error message isn't always clear on that :-)
<tobiasBora>
I'd like to know, why when I install nix using a tarball made by nixpkgs/nixos/lib/make-system-tarball.nix, nix is not in the user profile?
d-fish has joined #nixos
<kerrhau>
srhb: I do have allowUnfree set
Meizikyn has quit [Ping timeout: 260 seconds]
orivej_ has joined #nixos
orivej has quit [Ping timeout: 268 seconds]
coot has quit [Quit: coot]
<MichaelRaskin>
I think the gotacha about allowUnfree is when you _don't_ get any error message.
<srhb>
Ah, right.
<kerrhau>
now I'm getting `warning: name collision in input Nix expressions, skipping '/home/blob/.nix-defexpr/channels_root/nixpkgs'`
<d-fish>
I'm confused on how to change my sddm theme in nixos. In /run/current-system/sw/share/sddm/ I see symlinks to nix store packages, and I've got a theme in my home-folder I want to use a theme
<srhb>
kerrhau: well, after sudo nix-channel --update, either of nixpkgs.soulseekqt or nixos.soulseekqt ought to work. :/
<srhb>
kerrhau: Assuming no other channels are shadowing them somehow.
<pie_>
d-fish, maybe sddm has an environment variable you can set or a dotfile or something
<srhb>
kerrhau: I'd get rid of them aaaaall, but I also find channels very confusing.
<pie_>
nevermind "# libsdtc++.so.6 is `special`, it resides in stdenv.cc.cc.lib (see other packages)"
<ThatPako>
btw, how would I write a function allowing me to build stuff with emojicode next? I have to pass all the packages it depends on to the compiler after all
Myrl-saki has joined #nixos
<Myrl-saki>
How long does `download-from-binary-cache.pl --query` run for?
d-fish has quit [Quit: Konversation terminated!]
<kerrhau>
srhb: `error: selector 'nixpkgs.soulseekqt' matches no derivations`
<kerrhau>
same for nixos
<kerrhau>
-iA just returns no output
<Myrl-saki>
Or is there supposed to be some instructions for updating from 17.09 to 18.03?
<kerrhau>
and I have `nixpkgs.config.allowUnfree = true;`
<ThatPako>
Any idea why I'm getting `sh: c++: command not found` when compiling emojicode code?
<Myrl-saki>
It seems to be stuck at `download-from-binary-cache.pl --query`
<ottidmes>
Myrl-saki: You can safely ignore the warning about signed-binary-caches being unknown
<ottidmes>
Myrl-saki: Were you already using Nix 2.x?
<Myrl-saki>
ottidmes: Nope.
<ThatPako>
`propagatedBuildInputs` are depenencies needed at runtime, right?
<infinisil>
ThatPako: Nah, it's complicated and i don't fully understand it either
<ThatPako>
is there any way I can define runtime dependencies then?
<ottidmes>
Myrl-saki: The problems are caused by the conversion from Nix 1.11 to 2.0
<infinisil>
ThatPako: If you need runtime dependencies you probably want to wrap the executables with a different PATH or patch them
<ThatPako>
because the compiler won't work without a linker avaiable
<ottidmes>
Myrl-saki: I think you will have to revert back to your 17.09 checkout you had. Update only nix.package = pkgs.nixStable2;
<Myrl-saki>
ottidmes: "The problems" being "nixos-18.03 not installing" or "signed-binary-caches"?
<ottidmes>
Myrl-saki: Do a reboot, or at least make sure that nix-daemon is restarted properly and reports being version 2
<infinisil>
ThatPako: Best thing to do would be to patch it, so that the source uses `${pkgs.whatever}/bin/ldd` instead of `ldd`
<ottidmes>
Myrl-saki: The former, signed-binary-caches come from it being removed from nix 2.0, but the new alternative, require-sigs, is true by default
<Myrl-saki>
offlinehacker[m]: Should I use hte script provided in the NixOS site?
<Myrl-saki>
ottidmes: Or do the `nix.package` one.
<ottidmes>
Myrl-saki: And only when you have version 2.0 running of Nix, would I try to update to 18.03
<ottidmes>
Myrl-saki: Definitely the nix.package
<{^_^}>
[nixpkgs] @dtzWill merged pull request #37894 → grub2: extend platforms to include all linux → https://git.io/vxuYs
<{^_^}>
→ e7890021 by @dtzWill: grub2: extend platforms to include all linux
<Myrl-saki>
ottidmes: Yeah, I read `The quickest way to install it on Linux and *macOS* is to run the following in a shell (as a user other than root):`
<Myrl-saki>
ottidmes: As `... NixOS ...`
<ottidmes>
Myrl-saki: Did you already do that, because you will first need to undo that, or else you run into troubles later (I've seen a few reports)
WhittlesJr has quit [Quit: Page closed]
<Myrl-saki>
ottidmes: Luckily not.
<Myrl-saki>
Well, I C-c'd it.
<Myrl-saki>
myrl@myrl:/e/nixo/nixpkgs[heads/nixos-18.03]$ which nix │
<Myrl-saki>
/run/current-system/sw/bin/nix
<Myrl-saki>
I think I'm safe.
<ottidmes>
Myrl-saki: yep, but I see you have nix already, which means you already have Nix 2.0
<Myrl-saki>
ottidmes: What's nixos-rebuild switch in the new nix wrapper?
krey has quit [Ping timeout: 240 seconds]
<Myrl-saki>
ottidmes: Yeah, I did the change you said.
<Myrl-saki>
Something like copying path '/nix/store/w9fpn2ggscfg4m9l6b2bw3741j30ga0d-glibc-2.26-131' from 'https://cache.nixos.org'... (0/5.0MiB)
<ThatPako>
after building and trying to compile some example emojicode stuff I get linker errors
<bennofs[m]>
ThatPako: is this on linux or mac?
<ThatPako>
linux, NixOS
<ottidmes>
Myrl-saki: Not sure, but if you want to read more about what changed, like how binary caches are now unified with the concept of substituters, you can read about it here: https://nixos.org/nix/manual/#ssec-relnotes-2.0
<ThatPako>
bennofs[m]: getting the same linking errors
<Myrl-saki>
Wait wtf. How does configuration.nix work?
<Myrl-saki>
Is there a .drv for it?
obadz has joined #nixos
leat has quit [Ping timeout: 268 seconds]
<bennofs[m]>
ThatPako: ok. what if you do nix-shell -p clangStdenv glibc --pure --command '-/result/bin/emojicodec test.emojic'` (added glibc). if that doesn't work then I don't know, sorry
<ottidmes>
Myrl-saki: Well, building your configuration.nix results in a system deriviation, you could see that as the .drv of the configuration.nix
the-kenny-w has joined #nixos
the-kenny has quit [Quit: WeeChat 2.0]
<ThatPako>
bennofs[m]: downloaded glibc, then I got the same linking stuff
<ThatPako>
damn
jrolfs has quit [Ping timeout: 276 seconds]
Ariakenom has quit [Quit: Leaving]
<Myrl-saki>
ottidmes: How do I get that drv?
<Myrl-saki>
ottidmes: I want to use the why-depends to know what's making my the download so large.
<Myrl-saki>
Oh wait. Found it. It's at the bottommost.
<bennofs[m]>
ThatPako: alright, so I checked /nix/store/klwxxrlgalq49cjnczvgha3j7l00jixw-binutils-2.28.1/bin/ld and it appears to be the real ld. Usually, in Nix builds, ld is a wrapper script that takes extra options from `NIX_LDFLAGS`, and those options include additional linker search paths for all the dependencies specified by buildInputs / nix-shell arguments
<bennofs[m]>
so you might want to find out why it is using the wrong ld, and if you can get emojicodec to use the wrapper script for its linking
jrolfs has joined #nixos
<Myrl-saki>
Though, nix-depends seems to require the out files to be downloaded.
<Myrl-saki>
nix why-dpeends*
<Myrl-saki>
When I'm exactly trying to use it to decrease my download times.
<ThatPako>
bennofs[m]: where's the wrapped ld?
<ottidmes>
Myrl-saki: Yeah, it will need to download additional metadata
<{^_^}>
[nixpkgs] @marsam opened pull request #37895 → Enable gambit/gerbil on darwin → https://git.io/vxuC0
<Myrl-saki>
ottidmes: Does command-not-found for `git clone`'d nixpkgs now?
<ottidmes>
Myrl-saki: I have not actualy used it before, but it is the same idea behind when you want to use nix-diff and you need to get a hold unto the .drv file, while you normally just download the result paths
<bennofs[m]>
ThatPako: nix-store -q --references $(nix-build --no-out-link -A clangStdenv '<nixpkgs>') | grep clang should give you the path to the clang-wrapper and that has a wrapped ld in its bin directory
<ThatPako>
ohhh, ld is part of clang
asuryawanshi has quit [Ping timeout: 260 seconds]
<bennofs[m]>
or, simpler: nix-build '<nixpkgs>' -A clangStdenv.cc
<bennofs[m]>
no, not really. its a symlink to a binutils-wrapper package :D
<ottidmes>
Myrl-saki: That would be part of 18.03, not Nix 2.0, right? I am still on 17.09 with Nix 2.0, so no idea
<Myrl-saki>
ottidmes: Ah, okay, thanks.
blankhart has quit [Ping timeout: 240 seconds]
kerrhau- has joined #nixos
abathur has joined #nixos
jrolfs has quit [Ping timeout: 256 seconds]
kerrhau has quit [Ping timeout: 265 seconds]
<catern>
why do the foolish writers of language package managers never just do things the hard way, and admit that you need a system-integrator between package creators and package consumers
<ottidmes>
ThatPako: I might be completely off mark, but are you sure its a 64 build, I had a similiar error when building grub4dos and it was because I had to use: grub4dos = callPackage ./grub4dos { stdenv = stdenv_32bit; }; i.e. supply it with the 32bit stdenv
<hodapp>
blugh, is 'nix-shell -p python36Packages.mxnet' failing on nixpkgs master for anybody else?
<ThatPako>
ottidmes: I had all of this work on Arch Linux before so I doubt it's a problem with the architecture. Thanks for pointing that out though
<nineteeneightd>
With the latest minimal ISO, I can get to the end of Stage 1, but it fails trying to exec Stage 2 init.
lexos has quit [Ping timeout: 260 seconds]
Myrl-saki has quit [Ping timeout: 268 seconds]
spear2 has quit [Ping timeout: 276 seconds]
ThatPako has quit [Ping timeout: 240 seconds]
Myrl-saki has joined #nixos
davidlt_ has quit [Ping timeout: 240 seconds]
<Myrl-saki>
I get an OOME with `nix copy`... wow?
alexteves has quit [Ping timeout: 240 seconds]
jensens has quit [Ping timeout: 256 seconds]
ThatPako_ has quit [Quit: Lost terminal]
xcmw has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
chessai has quit [Ping timeout: 276 seconds]
<clever>
Myrl-saki: it keeps the entire storepath its copying as a std::string
dahirsch has quit [Remote host closed the connection]
<infinisil>
Ew
<Dezgeg>
nix master has some improvements on that, I think
<infinisil>
Ah
<infinisil>
clever: Wait you're saying it keeps the store*path* as a string? what's bad about that?
ma27 has quit [Ping timeout: 240 seconds]
newbie|6 has quit [Ping timeout: 240 seconds]
<clever>
infinisil: that includes every file within the path
<infinisil>
Ah, so a single string for everything in $(find $out)?
<clever>
for the contents of everything
<clever>
after serializing it into a NAR
<infinisil>
Oh.. okay yeah that's ugly
ottidmes has quit [Ping timeout: 240 seconds]
<nineteeneightd>
If anyone sees anything obvious or can point me in the right direction regarding the boot process failing to pass Stage 1, I'd greatly appreciate it. Here's the log: https://gist.github.com/198d/fb566ad7485343f5111567033771aa93
<Dezgeg>
why are the boot args like that?
<nineteeneightd>
I'm attempting to run NixOS under hyperkit (xhyve).
<clever>
nineteeneightd: was the iso modified in any way?