<elvishjerricco>
reivilibre: But more seriously, you just have to wait for the unstable channel to update, which happens once a certain set of tests pass. I think this is normally on the order of a day or two?
<reivilibre>
However, thanks for letting me know 'as soon as it's written'; it's really only just bumping the version of a Python package which has been blessed by upstream as stable and more performant etc etc
cyounkins has quit [Ping timeout: 268 seconds]
<ottidmes>
reivilibre: that depends on a lot of matters, the quickest way is to use nixos-unstable, but that still can take some time depending on whether the channel as a whole passes certain tests, so if it is about a package, I would just use an overlay for the time being, and if it is for a module, you could copy it over and use disabledModules to overwrite it with your own copy of the module
<doyougnu>
I saw in the nixos manual there is a section on customizing the kernel. Is it appropriate to put these things in linux.blah.blah.extraConfig?
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/fhJe6
silver has quit [Read error: Connection reset by peer]
hedning has quit [Quit: hedning]
ddellacosta has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @bendlas pushed to master « clojure: add $out/bin to wrapper PATH »: https://git.io/fhJvv
<{^_^}>
[nixpkgs] @bendlas pushed to master « emacsPackages: update elpa and org generated »: https://git.io/fhJvY
rfold has quit [Quit: leaving]
<reivilibre>
OK I'm sorry, I still don't understand how I can install matrix-synapse 0.34.0 as a NixOS service when it's on master but not yet arrived on nixos-unstable..
doyougnu has quit [Ping timeout: 268 seconds]
<gchristensen>
probably best to just wait
<reivilibre>
Is it possible to add it as a second source like you might do in Debian or something? I don't want to be a pita but if I can't even upgrade someone else's package then I don't think I have a chance of writing my own here..
<dmj`>
gchristensen: is there any way, possible, to cache /nix/store on Travis CI
<dmj`>
gchristensen: I have a cache in s3, but fetching from it is hella slows
<gchristensen>
don't know :/
<gchristensen>
maybe something with nix-store --export and --import?
Lisanna has quit [Ping timeout: 246 seconds]
freeman42x[nix] has quit [Ping timeout: 246 seconds]
Mr_Keyser_Soze has quit [Remote host closed the connection]
erictapen has quit [Ping timeout: 250 seconds]
pie__ has quit [Remote host closed the connection]
thc202 has quit [Ping timeout: 272 seconds]
pie___ has joined #nixos
init_6 has quit []
<jomik>
w00t! I think I managed to make a bunch of smaller nix expressions for downloading android sdk stuff, and it seems to work, with accepting license and all that jazz! :D
jasongrossman has quit [Ping timeout: 268 seconds]
palo has quit [Ping timeout: 240 seconds]
palo1 is now known as palo
<infinisil>
Yeah, I've also just upgraded to Human 2.0, it's been great so far
init_6 has joined #nixos
<{^_^}>
[nixpkgs] @worldofpeace opened pull request #52660 → nixos/displayManager: only install wayland sessions if they exist in … → https://git.io/fhJk8
Kim has quit [Ping timeout: 250 seconds]
kisonecat has quit [Ping timeout: 240 seconds]
Kim has joined #nixos
cyounkins has quit [Remote host closed the connection]
endformationage has quit [Quit: WeeChat 2.3]
hamishmack has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<juhe>
monokrome: You can install libvirt and use virsh directly or use virt-manager to do configuration via UI. I use nixops for deploying to nixos based machines.
<dgarzon>
Good morning, I am using dockerTools.buildImage to create an image with some pre-cached derivations. I can build the image, but when loading it to Docker, I get the following: “Error processing tar file(duplicates of file paths not supported):”. Has anyone ran into a similar issue?
rawreraw has joined #nixos
<dgarzon>
It is worth noting that my contents list is big, at around 100 base derivations!
<{^_^}>
[nixpkgs] @timokau pushed commit from @r-ryantm to master « whipper: 0.7.2 -> 0.7.3 (#52662) »: https://git.io/fhJq9
Dagger2 has joined #nixos
Guanin has joined #nixos
iyzsong-x has joined #nixos
oida has joined #nixos
pie___ has quit [Remote host closed the connection]
pie___ has joined #nixos
<ottidmes>
angerman: if you have the path to the derivation in $store_path then this should give you that: nix-store --query --outputs "$store_path" > >(tail -1)
dermetfan has quit [Quit: WeeChat 2.3]
<Myrl-saki>
dgarzon: I have a feeling I know the answer to this one.
<Myrl-saki>
dgarzon: I guess full disclaimer, i don't use Docker.
<Myrl-saki>
And seems like I don't know the answer to that one. :C
rawreraw has quit [Quit: WeeChat 2.3]
Lears has joined #nixos
<dgarzon>
No worries Myrl-saki, hopefully someone else does! I can build the image with dockerTools.buildLayeredImage, but it is too slow, and the resulting file system looks really weird!
[Leary] has quit [Read error: Connection reset by peer]
Ariakenom has quit [Quit: Leaving]
<Myrl-saki>
dgarzon: I tried colliding binaries (python2 and python3) but got nothing from that.
<hyper_ch>
how can I set for systemCronJobs = [ .... ] that it should use utf-8?
<Myrl-saki>
hyper_ch: Why not systemd and timers, FWIW?
<dgarzon>
ottidmes I am using that, but still getting “Error processing tar file(duplicates of file paths not supported):” not sure it works for what I need. I think my problem is I am including for example two `bash` with different paths, but when they get added to /bin/bash they conflict or something :(
<Myrl-saki>
Nix doesn't seem to have a uniqBy though.
<Myrl-saki>
dgarzon: Nix doesn't concern with that, although it can.
Lears has quit [Read error: No route to host]
<dgarzon>
So I think uniqBy would be more appropriate because I can use the keys instead of values
<hyper_ch>
still, same issue
<Myrl-saki>
dgarzon: You'd have much better luck processing that in Bash.
<hyper_ch>
I even prefixed the actual cron entry with export LC_ALL=en_US.UTF-8; export LANG=en_US.UTF-8; export LANGUAGE=en_US.UTF-8; export PYTHONIOENCODING=utf-8;
<ottidmes>
hyper_ch: another solution, use something like makeWrapper or create your own wrapper that does those exports, or did you mean that you already tried that? Other than that I would suggest just to just systemd timers instead
<dgarzon>
Found the issue: “tar -tf layer.tar | sort | uniq -c | grep -v '^ *1' = 2 ./nix/“
<dgarzon>
So it means something is creating two ./nix directories
<hyper_ch>
ottidmes: I have no idea what you mean
<Myrl-saki>
dgarzon: Ah. Hm.
<hyper_ch>
ottidmes: how to use systemd timers?
<dgarzon>
Or at least the produced layer.tar contains repeated paths
<ottidmes>
hyper_ch: the first point would be to use something like: * * * * * eelco ${runCommand "..." '' ... makeWrapper ...''}
<Myrl-saki>
dgarzon: Try binary searching it, I guess?
<Myrl-saki>
That'd actually be a cool function. Binary search in Nix.
<hyper_ch>
ottidmes: I created now a bash script where I first export all the utf-8 stuff and then run the actual python program..... and link the cron job to that bash script
<hyper_ch>
ottidmes: so systemd timers require like a gazillion lines instead of just 1?
<Myrl-saki>
dgarzon: Btw, do you use the fromImage?
<dgarzon>
Nope, I am creating my own base image now :)
<ottidmes>
hyper_ch: would depend on your use case, and lines of config is not really a good metric
<hyper_ch>
it is :)
<hyper_ch>
didn't you know? there's a world shortage of new lines ;)
<hyper_ch>
ottidmes: and how do you configure the systemd timers as which user it should be run?
<ottidmes>
hyper_ch: if you are code golving, I rather use 10 lines of something that works well, than 1 line that I have to put a lot of workarounds in to make behave, if you count your workaround exports too, its not too different anyway
<ottidmes>
hyper_ch: standard systemd stuff, you would define that in your service file
<hyper_ch>
I shouldn't even be required to use workarounds.... if I run it manually from the according user, it works.... if I tell cron to run it, it gives error about utf8/ascii stuff
<hyper_ch>
ottidmes: not everyone is on intimate terms with systemd..... why would I even need a service files?
<ottidmes>
hyper_ch: different environments, so not that surprising
<hyper_ch>
so much for workaround.... I need a systemd service file in order to use a systemd timer
<hyper_ch>
I tried to crate systemd service files before and they don't really run as they should
<ottidmes>
hyper_ch: this is the last that I will say about it, because why bother if you are not interested in the solution. You need the service because the timer is just about the timing aspect, and the service is the thing you normally would run, its more flexible that way, keeping those two things separate
Ariakenom has quit [Read error: Connection reset by peer]
<juhe>
Myrl-saki: for 18.09, though that is cross compilation of hello package only. I'd like to get the sd image. When I run the command specified in nixos/modules/installer/cd-dvd/sd-image-aarch64.nix, I get assertion that it can be built only on Aarch64/ARM64 platform :-/.
<Myrl-saki>
Ah.
<juhe>
The idea behind is to try to add another target, similar to sheeva plug, based on kirkwood, though would like to bootstrap it somehow...
<Myrl-saki>
juhe: The new nixos-rebuild accepts --build-host.
<Myrl-saki>
You just have to override your nix configuration file using -I or using an environment variable.
<Myrl-saki>
juhe: Or --target-host.
<juhe>
Myrl-saki: interesting, I'll definitely investigate this option!
<psy3497>
Hello! I'm tyring out NixOS and NixOps by first provisioning a simple NixOS configuration with VirtualBox.
<psy3497>
However As soon as I deploy and it gets an IP, it tries to connect using it and fails saying:
<psy3497>
ssh: connect to host 169.254.135.102 port 22: No route to host
<psy3497>
amadeus> could not connect to ‘root@169.254.135.102’, retrying in 1 seconds...
<psy3497>
I wonder, before first connecting to the server, does it allrleady have my configuration? Or is the image baked with an initial configuration
jomik has joined #nixos
<psy3497>
whih is then replaced with mine?
<psy3497>
In other words, is failing to SSH a fault of my simple configuration? Or is it unrelated since my configuration wasn't even provisioned yet?
civodul has joined #nixos
jtojnar has quit [Ping timeout: 246 seconds]
<juhe>
psy3497: the 169.254.x.y is IPv4LL, are you sure the IP address was assigned from DHCP server?
<psy3497>
I compared the ip that is in VirtualBox with the IP that is in NixOps state and they match.
<psy3497>
But looking at VirtualBox's host-network settings, dhcp is disabled. Should I enable it?
<psy3497>
Is IPv4LL not good?
<juhe>
Hm, it's worth of shot, though if IPs are assigned and nixops found the one used by VBox, it shouldn't matter.
<juhe>
IPv4LL is OK.
<juhe>
Can you nmap 169.254.135.102 to see if ssh (or what services) are running there?
<psy3497>
Also, do you know if my NixOS configuration is baked into the VM image from the start?
<juhe>
I'd say config should be there, though I didn't use vbox with nixops before.
jtojnar has joined #nixos
<juhe>
Can you actually ping the vbox host? The error message is "No route to host".
<juhe>
So I'd say that none of your local network interfaces (on machine you deploy from) doesn't have any IPv4LL address assigned and system doesn't know how to reach that host.
<juhe>
Enabling DHCP might help.
thblt has joined #nixos
<psy3497>
I see.
<psy3497>
Just a second I'll try things in order
<juhe>
Sure, take your time.
<psy3497>
on `ip addr show`
<psy3497>
I see I have an interface for `vboxnet0`
<psy3497>
169.254.136.1/24
<juhe>
that looks good
<psy3497>
trying to ping, it failed.
<psy3497>
with host unreachable
<psy3497>
although VM is running right now
<thblt>
Hey! I've just switched from XMonad to Gnome in my NixOS box, and I have a small issue with gpg-agent. I use it for SSH connections along a Yubikey, and I can't get it work on Gnome.
<thblt>
(ssh SERVER fails with permission denied: publickey)
<thblt>
the yubikey works for regular PGP operations
<psy3497>
I wonder if the initial nixos configuration that the image starts with doesn't have ssh nor listening for ping?
<psy3497>
In that case it may be my configuration's fault.
<juhe>
psy3497: ping should work more-or-less always.
<juhe>
psy3497: maybe a firewall on your host is preventing that?
<psy3497>
I didn't enable anytning explicitly
<juhe>
psy3497: what is the output of "ip route"?
<psy3497>
I think I'll try destoying and recreating the VM with a cconfiguration that explicitly enables ssh and opens the ports.
<psy3497>
Although I do not understand if my configuration was already applied from the start, or if it isn't even applied yet because NixOps can't get a connection.
<juhe>
that will surely help, but probably won't solve the "no route to host" part.
<psy3497>
Really?
<psy3497>
The meaning of "no route to host" is something that is outside of the VM?
<juhe>
Can't say for sure, just gut feeling.
<psy3497>
I see. If that's the case then I'll try also playing with VirtualBox's settings.
<psy3497>
Thanks for trying to help!
<psy3497>
Perhaps I should just try NixOS with virtualbox without using NixOps?
<juhe>
No problem. I'd check the route settings to see if you have route for 269.254.x.y set locally.
<juhe>
There should be default route and net routes, I'd expect you see something like 269.254.0.0/16 in output of your "ip r" command
<psy3497>
169.254.136.0/24 dev vboxnet0 proto kernel scope link src 169.254.136.1 linkdown
<psy3497>
I already destroyed the VM, if that's related to the "linkdown"
<{^_^}>
[nixpkgs] @risicle opened pull request #52678 → [r18.09 backport] cloudfoundry-cli: fix build on multiple platforms, notably darwin, to produce correct binary for targe… → https://git.io/fhJsL
<juhe>
Probably yes.
aristid1 is now known as aristid
<juhe>
it should be up and then ping should work
<psy3497>
Is there a way to bring it up?
<psy3497>
Is it a matter of running an `ip` command?
<juhe>
ip link set dev vboxnet0 up
<juhe>
but I think the problem is network mask, your route is set for 169.254.136.0/24, IPv4LL is, AFAIR, 169.254.0.0/16
<psy3497>
tried running that and it's still `linkdown`
<psy3497>
I see! But that was VirtualBox's default so I doubt it is set wrong.
<psy3497>
You'll have to excuse me as I gotta go now.
<psy3497>
I'll keep researching this tomorrow morning.
<juhe>
ok no problem
<psy3497>
Thanks for the attempt to help!
psy3497 has quit [Quit: WeeChat 1.9.1]
<jomik>
Hey guys - how do I merge my own package expressions into `pkgs`?
<ottidmes>
jomik: overlays? or what do you mean?
<ottidmes>
,overlays jomik
<{^_^}>
ottidmes: Did you mean overlay?
<{^_^}>
jomik: Overlays look like `self: super: { foo = ...; }`. Use the self argument to get dependencies, super for overriding things and library functions (including callPackage). More info: https://nixos.org/nixpkgs/manual/#sec-overlays-install
<jomik>
ottidmes: I tried that..! I have a `mypkgs = pkgs.callPackage ./mypkgs {}`. But how do I actually merge "mypkgs" into pkgs, so that I can just refer pkgs.<my package drv>
<ottidmes>
jomik: that looks confusing, you either define a package, or a set of packages, but not a package set as a package
<ottidmes>
jomik: normally mypkgs would be the set defined in your overlay
<jomik>
so I can do `self: super: { inherit mypkgs; }` ?
<ottidmes>
jomik: yeah change that to self: super: instead of { callPackage, ... }: and use super.callPackage, and you have yourself an overlay
<ottidmes>
jomik: you basically are defining your own overlay format, if you want to stick with what you have (which I would not recommend), you could do this: self: super: super.callPackage ./mypkgs {}
boogiewoogie has quit [Quit: Leaving]
<ottidmes>
that call to super.callPackage would then return the attrset that represents what you normally would have returned directly by: { android = super.callPackage ./android {}; }, but again, its probably better to stick to conventions
<jomik>
Okay, so now I changed that and have `import ./mypkgs`.
<jomik>
If I have a derivation in my store, can I open that in nix-shell?
<jomik>
The build failed.
cyounkins has joined #nixos
<ottidmes>
jomik: if it is trying to do network stuff at build time outside the fetcher it will fail to do so by default due to sandboxing
<jomik>
Huh
arjen-jonathan has quit [Ping timeout: 252 seconds]
agander has quit [Read error: Connection timed out]
<jomik>
It worked in nix-shell, guess they differ there.
agander has joined #nixos
<ottidmes>
jomik: in nix-shell you are working outside the sandbox
<jomik>
That makes sense then. Though, how do I figure what is trying to do network stuff?
<jomik>
Can I somehow "cheat" it, by fetching it, and just rerouting their network request?
<ottidmes>
jomik: you can, for example some fetchers, like fetchFromGitHub have an attribute called `extraPostFetch` that you can use to do additional network stuff, I use it for a game that needs to do a few curl requests to fetch additional source files
cyounkins has quit [Ping timeout: 240 seconds]
<jomik>
Does that allow them through or?
b has joined #nixos
<ottidmes>
jomik: if you just have some Java executable you call to build your thing and that happens to include those network calls, then it will not work
<ottidmes>
jomik: I have to go now, best of luck!
jonaswouters has joined #nixos
<{^_^}>
[nixpkgs] @Mic92 merged pull request #52678 → [r18.09 backport] cloudfoundry-cli: fix build on multiple platforms, notably darwin, to produce correct binary for targe… → https://git.io/fhJsL
<{^_^}>
[nixpkgs] @Mic92 pushed 2 commits to release-18.09: https://git.io/fhJsb
<thblt>
Is there a way to disable the Gnome keyring when using Gnome, or at least to prevent it from breaking GPG-agent in SSH-agent emulation mode? Thanks!
<jonaswouters>
Is there a reason why fetchTarball in an import statement does not work during nixos-install?
jomik has quit [Ping timeout: 268 seconds]
balsoft has joined #nixos
<balsoft>
Hello everyone!
<balsoft>
Can someone help me with nixos on arm?
<balsoft>
I am trying to run NixOS on OrangePI PC2 (which is aarch64, allwinner H5 SOC microcomputer). I have confirmed that it works by uploading official Ubuntu on it.
<balsoft>
leotaku: Yes, you are right, but you generally should use it "embedded" in some other construct, like list or NullOr
<balsoft>
jomik: Do you know of any good way to do imports in home-manager when using it as NixOS submodule?
<balsoft>
I would like to organise the mess that is my configuration into modules, but I would like to keep it as a NixOS submodule at the same time.
<jomik>
balsoft: I don't have home-manager in my /etc/configuration.nix. I use it only in my home.
<jomik>
You can use the imports attribute though.
<balsoft>
Yes, I understand.
<balsoft>
jomik: No, it's not available when you use it as NixOS submodule.
<balsoft>
Before I used it as you do, in .config/nixpkgs/home.nix, and had everything as a module
<balsoft>
But I don't know of any good way to do that now, with home-manager as submodule/
<jomik>
Why do you use it in the system config? :P
<balsoft>
jomik: because I have 3 machines that I constantly work on and about 7 of them in total
<balsoft>
And it is very nice to have everything inside a single config
<jomik>
But, I am pretty certain you can do imports = [ ./myhome-manager ]; in configuration.nix. Can I see your configuration.nix?
<jomik>
I use dotfiles-sh for that stuff :P
<jomik>
I have /etc/configuration.nix for system specific stuff, I think. Mainly because I don't want to sudo :P
<balsoft>
jomik: Yes, you can import stuff from configuration.nix, but it's not trivial to import home-manager modules because of the way they are build
<balsoft>
That's the way I split my config logically into system configuration and home configuration, but I would like to split home configuration further into modules
<ottidmes>
balsoft: I have only experimented with home-manager shortly, but I dont see the problem, why can't you split it into modules? you cant just import them like any other module?
nD5Xjz has joined #nixos
pie__ has joined #nixos
<ottidmes>
balsoft: if imports is not an option, I know of a workaround that would get you the same behavior
pie___ has quit [Remote host closed the connection]
<balsoft>
ottidmes: the problem is that you can't imports = [] from home.nix because it's not an option for using home-manager as nixos submodule, and you cannot imports = [] from configuration.nix because it doesn't merge correctly
<balsoft>
ottidmes: what's the workaround?
<ottidmes>
balsoft: ah, I see, you mean that you would have to namespace everything with home-manager.users.balsoft and supply the device argument with e.g. a custom option if you were to want to use the normal module system
<ottidmes>
balsoft: well the simplest solution would be to do something like defining a list that defines your home.nix attrset, that list would be a list of file paths and you just merge the attrsets, but this is not so nice, because it would not leverage the module system so the merging logic of the types would not be applied
dgarzon has quit [Quit: dgarzon]
<ottidmes>
balsoft: so instead I would map the list of files to prefix the attributes with home-manager.users.balsoft and then import those, shall I make an example?
<balsoft>
Perhaps yes, but that would be quite hacky
jomik has quit [Ping timeout: 272 seconds]
<balsoft>
ottidmes: Because that would only work for home-manager.balsoft
<balsoft>
home-manager.users.balsoft I mean
<ottidmes>
no?
<ottidmes>
I could make that variable
<infinisil>
balsoft: You should be able to do `{ home-manager.users.balsoft = import ./some-home-manager-module; }`
<balsoft>
infinisil: and that would merge correctly?
<balsoft>
And I can define options in one file and use them in another?
<infinisil>
Or maybe even `{ home-manager.users.balsoft = mkMerge [ ./foo.nix ./bar.nix ]; }`
<balsoft>
I'll try that right now
<infinisil>
Hmm, not sure about option definitions
<balsoft>
Ah, that's what I meant.
<infinisil>
Yeah I think those don't work
<infinisil>
That's a problem I've been wanting to solve for a while, but it's non-trivial
<balsoft>
Because It's pretty trivial to just import some stuff, I already do that, but that often requires calling import ./foo.nix {arg1 = "foo"; arg2="bar;} and that's awful
<ottidmes>
balsoft: I would have said, imports = map (f: { home-manager.users.balsoft = import f; }) [ ./file1.nix ./file2.nix ], but infinisil suggestion is prettier
<balsoft>
ottidmes: Yeah, but that's not "imports" per se as it doesn't allow for option definitions
<ottidmes>
balsoft: wut? it would be like any module, no? I guess I leave it to infinisil, I don't see the problem, but maybe because I am not using home manager anymore
dontdieych has quit [Ping timeout: 272 seconds]
<infinisil>
You can only define options from imports in top-level modules
<infinisil>
(currently)
<balsoft>
ottidmes: Thanks for help anyways, but that wouldn't be proper "imports" replacement as it wouldn't allow for option definintion
<balsoft>
infinisil: I have heard that one could override options inside submodules
<infinisil>
Yeah, but only from top-level modules too
<infinisil>
I typed this up before you showed your code: `config.networking.allowedTCPPorts = mapAttrs (cfg: cfg.custom-port-option) (lib.attrValues config.home-manager.users);`
<balsoft>
infinisil: thank you!
<infinisil>
Something like that, not tested though
<jomik>
The issue is that this jar file fetches something if it isn't cached, so I want to create the cache manually :P
<ottidmes>
jomik: if the package is part of a service or could be made to be, you could just write a systemd service that creates this folder for you
<jomik>
It's a lar
<balsoft>
jomik: I don't think this is a good thing to fetch something from a nix build
<balsoft>
This is impure
<balsoft>
Or are you doing something outside of the build before?
<balsoft>
before the build
<balsoft>
infinisil: this kind of works, but what if I need to say use another home-manager option in my implementation? Like I would like to write some stuff to xdgConfig
<infinisil>
balsoft: Yeah you can't write such recursive things, you need to always use a specific option on the left to avoid recursion
<gmarmstrong>
I'm working on packaging an application from npm. I'm using the instructions in the nixpkgs manual for Node.js packages (https://nixos.org/nixpkgs/manual/#node.js-packages). Running ./generate.sh takes a very long time, and it's changing thousands of lines in pkgs/development/node-packages/
<infinisil>
Ah but yeah, that's a good issue
jomik has quit [Ping timeout: 268 seconds]
<gmarmstrong>
The application works, I'm just hesitant to make such a large PR.
<gmarmstrong>
Most of the changes are incrementing version numbers for other node packages (dependencies of the application, I think).
<gmarmstrong>
Okay, I see from https://github.com/NixOS/nixpkgs/pull/43587 that generate.sh just needs to be run occasionally. I'll make a PR with the updates and then another with my package.
<hexa->
except it sounds like it's more for microcode and other bootable stuff
<ottidmes>
hexa-: there is also boot.initrd.secrets, but that does not work out of the box with grub, but if you want I could give you a snippet that gives grub support for it
<hexa->
I'm past grub actually
orivej has joined #nixos
<hexa->
grub openes the it up just fine, but there is no way to pass the secret to the initramfs, so the easiest way to not enter the secret twice was to add a keyfile to the initrd
<balsoft>
As it seems there is more activity now, I will repeat my question: can somebody help me with NixOS on ARM
<balsoft>
?
<ottidmes>
hexa-: I am using secret files in my initrd with grub, so I have to disagree with "no way"
<jomik>
Alternative seems to be to just hardcode the hash that is input when accepting licenses with sdkmanager... But, that is pretty flawed, imo.
<simpson>
jomik: You can run a Java program during the build step. You can't get the program impurely, though, and networking needs to be done through Nix or with the sandbox disabled.
<jomik>
simpson: Issue is networking part.
<jomik>
The sdkmanager for android sdks wants to fetch a lot of things, for its cache.
<ottidmes>
jomik: AFAIK it is all or nothing with sandboxing, I do not believe you can disable sandboxing per package. I have no experience with them, but I believe if you can be certain of the output being the same every time, then you should be able to use fixed output derivations, who are allowed to do networking
<jomik>
ottidmes: fixed output derivations? And I can't really, cuz I don't control the cache that they fetch :P
<jomik>
ottidmes: doing that allows me to use java and such to fetch stuff?
<ottidmes>
jomik: they allow for impurities (which is what you want), while still ensuring purity of the output (it always has to result in the same output) so that it can safely be used further down the line
<jomik>
That makes sense. Okay. I'll have to ponder that :D
<jomik>
Tho, in that case, I guess I may as well update the license acceptance has and completely avoid fetching these things.
<samueldr>
you can also massage the output of a fixed output derivation
<samueldr>
e.g. if it outputs a file with the current time and a desired static file, you can rm the file you don't want :)
<samueldr>
from memory, the fixed output derivations for maven do a lot of massaging, e.g. removes a lot of files which are not needed, and do not have a fixed output
<samueldr>
well, not for maven, but using maven
<jomik>
Oh, that could maybe work.
<jomik>
But, it'll take the hash of the whole thing, right? Including symlinks and such, which may vary depending on the user's setup.
jperras has joined #nixos
<samueldr>
then if it varies due to its inputs it's not a fixed output derivation?
<ottidmes>
luckily I have not yet needed it, but I can see how that is useful, so far for me failing sha's have been actual failures of the source being changed or removed
<samueldr>
I'm not sure what you're doing, but maybe you can make multiple fixed-output derivations and merge them later according to the user's configs?
<samueldr>
and the fixed output derivations are not necessarily the end-results; you could also have a fixed output derivation which is a "template file" which can be manipulated by another derivation, which that one isn't fixed
<samueldr>
it's, after all, what most builds do: fetch* are fixed output derivations, and the build processes use them to do more things
<samueldr>
let me try on my zero plus 2, using the cross-built u-boo
<balsoft>
I have spent about two days trying to figure it out myself before asking here
<balsoft>
Oh, that would be great
<balsoft>
Building it doesn't take too long
<samueldr>
yeah, shouldn't using the manual steps I linked previously takes ~10 mins max, and some of the time is due to human input :)
<samueldr>
yeah, it shouldn't; *
<samueldr>
when you say "the thing doesn't boot", do you mean u-boot doesn't show up neither on serial nor hdmi?
<balsoft>
samueldr: it doesn't
<balsoft>
Unfortunately currently I have no way to check serial
<samueldr>
okay, so it's a case of u-boot seemingly not working; helps to know what I should look for
<balsoft>
Because I am away with just my laptop and this thing
<balsoft>
But connecting it to TV didn't work
gmarmstrong has quit [Ping timeout: 246 seconds]
<jomik>
Weird things happens sometimes... Now I have a package that has the same _name_ as one in nixpkgs. Which appaenrly makes mkDerivation see it as that one, because it starts its derivation.. Why?
<balsoft>
And it doesn't even light up LEDs
<samueldr>
it's *possible* that the nixpkgs-built mainline kernel won't output to HDMI yet; I haven't verified yet with H5-class hardware, but on A64-class there's a missing kernel config needed for HDMI
<balsoft>
The board works, it starts up with Ubuntu
<samueldr>
(but u-boot should work in any cases)
<jomik>
Ooooh, nvm. Overlay didn't run here, my bad.
<balsoft>
samueldr: It should work, as it presumably does on Lite
worldofpeace has joined #nixos
jperras has quit [Ping timeout: 250 seconds]
<samueldr>
balsoft: hdmi output on the mainline kernel? yes it should, if the kernel configuration required for A64 isn't for H5
<worldofpeace>
Can someone with access to building on the darwin platform with ofborg build #52628 ?
<jomik>
I tried going into a nix-shell to print the rpath of the file in the store, and it only has one path, which points to its own store location/lib64
cyounkins has quit [Remote host closed the connection]
cyounkins has joined #nixos
<Laalf>
does anyone have a package for http://www.seamonkey-project.org ? i have no idea about packaging on nixos but id try my luck on it if nobody has a package
<balsoft>
samueldr: sorry for being officious, but should I wait for you to test out cross-compilation on your Lite or am I better off getting some sleep?
<samueldr>
balsoft: don't worry; I need to setup the things still, so it all depends on whether or not it's pressing for you
<jomik>
Are you not allowed to refer to $out in the rpath of your elfs?
<balsoft>
It's not pressing, it's fine.
kisonecat has joined #nixos
<balsoft>
I'm just frustrated for spending so long on a little board
<samueldr>
balsoft: try this in a nixpkgs checkout of nixos-unstable (just in case) nix-shell -I "$PWD" -p 'with (import ./. {}); let plat = pkgsCross.aarch64-multiplatform; in plat.buildUBoot{defconfig = "orangepi_pc2_defconfig"; extraMeta.platforms = ["aarch64-linux"]; BL31 = "${plat.armTrustedFirmwareAllwinner}/bl31.bin"; filesToInstall = ["u-boot-sunxi-with-spl.bin"];}'
<samueldr>
here It's not mixing platforms, and using the same "platform" as the other builds do
<samueldr>
and then go to sleep and I'll notify you of the results of both your previous build and that one
<Dedalo>
Hi guys, I have an X1 Carbon with a WQHD (2560x1440) display and I'm using the i3 window manager. I'm having trouble to scale the dpi. I have added to the configuration.nix service.xserver.monitorSection = '' DisplaySize 338 190 '', but when I reboot I can't login anymore. It says: session cannot start, or something like that. Anyone using the same laptop can share his settings?
<balsoft>
samueldr: thank you very much, gonna do that.
arjen-jonathan has quit [Ping timeout: 252 seconds]
<balsoft>
samueldr: unfortunately, I can't stay connected.
<samueldr>
no worries, I'll tell the bot to tell you, next time you say something
<balsoft>
Ok
<jomik>
So apparently because I create a dir called build-tools, and refer to a dir in that, this audit-tmpdir script reacts..?
<balsoft>
samueldr: by the way, is it fine if I flash sd image of NixOS just once and then flash U-boot multiple times?
<balsoft>
Or do I need to clean out starting sectors before re-flashing?
lokado has quit [Quit: Connection closed for inactivity]
<jomik>
Ah, cuz the audit-tmpdir script is still from back in august.
<samueldr>
two issues I had in mind: 1. there's an eMMC on the zero plus 2, just verified that it's not using that
<samueldr>
2. zeroed out the section it boots from to re-test, in case something else was causing bad tests
<balsoft>
That means that I just have to build in QEMU and it'll work?
<balsoft>
Or did I misunderstand something
<samueldr>
I don't know yet
<samueldr>
and yeah, after (1) zeroing out, and (2) re-flashing the mixed platforms build, that one doesn't work
<samueldr>
now I'm trying (1) zeroing out (2) flashing the non-mixed cross build
<balsoft>
Oh, ok, got it
<balsoft>
I guess you don't need to zero out if it doesn't work :)
<samueldr>
let's be thorough
arjen-jonathan has joined #nixos
<samueldr>
or alternatively: I'm an idiot and I must re-do all steps: I accidentally had `of=/dev/sdg of=/dev/sdX` :/
<nDuff>
If my function is passed /nix/store/<hash>-whatever, I want the derivation it resolves to be named "whatever-myfunc", not "<hash>-whatever-myfunc". I assume there's going to be a way to extract "whatever" (the derivation name) already in nix, or in the nixpkgs standard library -- where should I look for that?
<jomik>
What is the best way to make a derivation configurable ? I need to be able to specify version and a sha256. the sha is used for the fetch together with the version.
<balsoft>
samueldr: It seems like serial is silent
<samueldr>
what's the nixpkgs commit you're using to build?
<samueldr>
I'll try on that one
<balsoft>
My internet is too slow to download nixpkgs
<samueldr>
if you're not specifying from a checkout, (-I with nix-shell) nixos-version should be able to tell
<balsoft>
So I'm using latest nixos-unstable
<balsoft>
Yes, I know, channels are not good
<samueldr>
uh? channels are fine
<samueldr>
you can use `nixos-version`, though, to know which commit it is exactly
<{^_^}>
[nixpkgs] @worldofpeace pushed commit from @r-ryantm to master « syslogng: 3.18.1 -> 3.19.1 »: https://git.io/fhJRW
<samueldr>
thanks balsoft, it's building u-boot for my board with that commit, let's see
<samueldr>
(I've been using a checkout built from master, *maybe* there's been changes since the 14th)
<nDuff>
jomik, ...kinda' depends on context. If you define a function that retuns a derivation, that's one way to parameterize; another is to rely on nix-build --arg to provide default values to be used for automatic function invocation.
<balsoft>
samueldr: I'll try again with Ubuntu, maybe the board was broken in the process of trying to run NixOS on it
doyougnu has joined #nixos
<jomik>
Yeah, I did consider using a function.. Not sure it is as pretty as it could be tho.
<jomik>
I'll just do that :P
<balsoft>
Would be kinda a shame, but then it's like 30$
<samueldr>
the way allwinner boards work, I wouldn't think software would break it, but you never know :/
<samueldr>
though yeah, just tried and using the particular commit your system is at, it works here
<samueldr>
your last `dd`, did you copy from buildInputs or cd in there?
<balsoft>
It doesn't boot!!!!
<balsoft>
With Ubuntu
<balsoft>
It did yesterday
<balsoft>
WHAAAT
<samueldr>
it could be the SD card, also
<balsoft>
It's the same one
<balsoft>
The only thing that changed was power supply
<Mic92>
hexa-: did you got your grub problem sorted out?
<hexa->
Mic92: aye, that overlay worked just fine
<balsoft>
OH WHAT!!!!!
<samueldr>
you could sha256sum the ubuntu image, dd it to the sd card, then sha256sum the appropriately configured dd output (same size as the image) and see if the sd card somehow corrupts the data
<samueldr>
or yeah, power supply could be an issue :/
<hexa->
Mic92: thanks again
<balsoft>
It boots when I plug it into my laptop
dgarzon has quit [Quit: dgarzon]
<balsoft>
But my perfectly good USB charger with much higher power - nooo
sigmundv_ has joined #nixos
<samueldr>
hm, could be how it negotiates for power, with the fancy usb power delivery
cyounkins has quit [Ping timeout: 250 seconds]
<samueldr>
if it fails to negotiate, it could go to 500mA while the laptop maybe provides more
<balsoft>
Is there any easy way to bootstrap NixOS from ubuntu?
<samueldr>
(for aarch64) I don't know
<Mic92>
balsoft: what are using at the moment?
ottidmes has quit [Ping timeout: 245 seconds]
<Mic92>
dd appearently
<balsoft>
Mic92: Yes, I use dd
<balsoft>
Do you think that there is a faulty bit somewhere?
<balsoft>
But I have reflashed multiple times
agander_ has joined #nixos
<samueldr>
now that you know the power supply could be at cause, you could retry using the sd image, flash u-boot using one of the previously made builds and see?
<balsoft>
Yes
<balsoft>
I am doing that right now
<balsoft>
But it's quite a slow process
<samueldr>
yeah :$ I know the feeling
<balsoft>
Because my laptop's internal SD card reader is USB 2.0 for some strange reason
<balsoft>
all of the USB ports are 3.0
jtojnar has quit [Read error: No route to host]
jtojnar has joined #nixos
<samueldr>
in any way, most sd cards won't saturate USB 2.0 speed in constant write, AFAIK
agander has quit [Ping timeout: 240 seconds]
<balsoft>
This is quite a good SD card, it costs more than the board itself :)
<samueldr>
I always test SD cards when I buy them, and the quoted speed if they don't quote write speed, is always read :(
<samueldr>
write is often 10x less than the read speed
<balsoft>
Makes sense to me
<balsoft>
Marketing and all
nDuff has quit [Quit: zzz]
<samueldr>
yeah, just something to keep in mind
<balsoft>
I have learnt not to believe anything that's written by the manufacturer
<monokrome>
Is there, like, a generic video driver that'll work for my server's X server? I don't need fancy graphics or anything, but I have an i9-9900K and "i915" "inteel", or "vesa" doesn't seem to work
<balsoft>
Serial time!
<samueldr>
yeah, give it 5-10 minutes for the first boot, (probably less) since it resizes the ext4 partition
<balsoft>
Oh, it's a 32GB card
<samueldr>
monokrome: are you using the latest kernel or whatever default nixos has?
<zfnmxt>
Is it possible to "extract" the configuration.nix from a certain generation? I made edits to my configuration.nix that broke my system, and I'm not sure what the working config was :D
<balsoft>
zfnmxt: yes,
<balsoft>
let me find the location cause I don't remember it
<zfnmxt>
balsoft: Thanks :D
<balsoft>
Hmm, can't find it
<balsoft>
Pretty sure it was there
<zfnmxt>
balsoft: I found a bunch of configuration.nix's in /nix/store
ilya-fedin has joined #nixos
<ilya-fedin>
Hello. How can I specify in configuration custom configure flags for package and compile it on system rebuild?
<zfnmxt>
But no idea what hash corresponds to what build...
<zfnmxt>
Oh, nevermind. Those don't look like the same thing.
<zfnmxt>
Ah, I'm an idiot. Those are just the man page entries >.>
<ilya-fedin>
For example, I want to add to connman --enable-nmcompat and --with-dns-backend=systemd-resolved
<worldofpeace>
Baughn: They haven't documented it beyond the rfc yet I'm afraid
<monokrome>
wtffff.... If I do `strace lspci -v` it works fine
<monokrome>
If I do `lspci -v` the entire system hangs to the point where dmesg freezes before outputting anything and htop freezed as well
<monokrome>
O_o
<jomik>
Ah, no. It is causing an issue.
<jomik>
openat(AT_FDCWD, "/nix/store/7gx4kiv5m0i7d7qkixq2cwzbr10lvxwc-glibc-2.27/share/locale/en/LC_MESSAGES/libc.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
<Baughn>
Okay, so I assume that the library does reference glibc. (In the nix-store tree structure, I mean; i.e. glibc won't be GC'd while the library is still there.)
<jomik>
Yeah, I guess zlib references it.
kisonecat has quit [Ping timeout: 246 seconds]
<Baughn>
...let me start over.
<Baughn>
Of course zlib references glibc. Practically everything references glibc.
<Baughn>
What makes you think there is an issue?
<Baughn>
It's common for glibc routines to probe the system in that manner, causing a lot of failed opens.
* monokrome
wonders if jomik and I are having the same issue, because my hang is happening during openat on /sys/bus/pci/devices/0000:05:00.0/config - which seems similar?
<jomik>
The issue is the one I pasted in. It doesn't find the library it expects, because of some locale stuff.
<jomik>
monokrome: No, luckily this is just my android sdk environment :P
<monokrome>
ah
<Baughn>
jomik: Can you rebuild the application you're running?
<jomik>
aapt wont run, because it can not find the library it needs I guess.
<jomik>
Baughn: I have done that multiple times :P
<Baughn>
Have you installed it with nix-env?
<monokrome>
My openat succeeds, but reading hangs
<Baughn>
monokrome: That's no kind of libc bug. It's the /sys/ filesystem, which is backed by kernel drivers instead of on-disk files.
<Baughn>
They can do basically whatever.
<Baughn>
Not sure what reading config is supposed to do, but perhaps it pokes a hanging PCI device, and hangs as a result.
<monokrome>
hmm
<monokrome>
It's trying to read a device using vfio driver
<monokrome>
I think
<Baughn>
jomik: 'aapt' doesn't seem to be in nixpkgs.
<Baughn>
Looks like a patched binary, not something we're building from source code.
<jomik>
Yup
<Baughn>
There were some workarounds in the bug you linked. Have you tried those?
<Baughn>
...but the cause is probably that it was originally built against an older libc than the one it ends up running with.
<jomik>
Nah, want to try that last really :D I am just gonna try to patch everything like the nixpkgs version does.
simukis has quit [Quit: simukis]
<dgarzon>
Hi, I am trying to build a Docker image with dockerTools. I can build and load my image with dockerTool.buildImage, but when I use dockerTools.buildImageWithNixDb and then docker load < result, I am getting the following error: “68ae500b7aa7: Loading layer [=================================================> ] 15.12GB/15.13GB Error processing tar file(duplicates of file paths not supported):” I have ran “tar -tf result | sort | uniq
<dgarzon>
grep -v '^ *1’” and “tar -tf layer.tar | sort | uniq -c | grep -v '^ *1’” and I cannot find any duplicate paths. Has anyone ran into this issue before?
<jomik>
Well, the issue seems to be a bit different :D
simukis has joined #nixos
beaky has joined #nixos
<Baughn>
dgarzon: At a pure guess, there might be multiple docker layers and it's aliasing something from a previous layer?
<jomik>
Baughn: I got hat issue with aapt solved. I suppose I was missing a patch on some lib file, and it was then stripped.
<jomik>
So I didn't actually get to see where aapt was looking for the file :D
<monokrome>
So, apparently my video capture card powers down when it's not in use
<monokrome>
If I use vfio_pci with it, vfio tries to read from it while powered down
* monokrome
laughs
<dgarzon>
I am trying something else and now dockerTools.buildImageWithNixDb fails with “error: creating directory '/tmp/nix-build-docker-layer-my-image.drv-0/layer/nix/store': Permission denied” I am inspecting the source code and see it happen after: “Generating the nix database..." Any ideas?
jomik has quit [Quit: WeeChat 2.2]
__monty__ has quit [Quit: leaving]
orivej has joined #nixos
<{^_^}>
[nixpkgs] @veprbl opened pull request #52707 → bzflag: enable on darwin → https://git.io/fhJzd
jtojnar has quit [Read error: Connection reset by peer]
hamishmack has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Dedalo has joined #nixos
[Leary] has joined #nixos
Lears has quit [Ping timeout: 240 seconds]
jtojnar has joined #nixos
Mateon1 has quit [Ping timeout: 245 seconds]
Mateon3 has joined #nixos
hamishmack has joined #nixos
<dgarzon>
Has anyone seen this error when building a docker image with buildImageWithNixDb? “error: build of '/nix/store/z18vv8s5zfsx6lkv9w59wmswy5ypdqv5-nix-registration.drv' on 'ssh://dev-env-nix-docker' failed: invalid character '/' in name '/nix/store/cb0fi0yvrgrdj8kf6mvghj2b51hv0lcw-python2.7-ansible-2.3.1.0’”
Mateon3 is now known as Mateon1
<ilya-fedin>
In continuation of yesterday's theme about the autostart of ModemManager.
<ilya-fedin>
And /etc/systemd/system/modem-manager.service has appeared, but its contents is broken. It turned out that it links to /nix/store/v821wa5w90p4hsvz6xihw9rwpdvyvxq0-unit-modem-manager.service/modem-manager.service, to the package unit-modem-manager.service, in which only this file. However, the real unit file is located in
<ilya-fedin>
I was advised to do this: `systemd.services.modem-manager.enable = true;`
<ilya-fedin>
/nix/store/q4szg0r20859126lmkpa9p6nx09dbalv-modem-manager-1.7.990/etc/systemd/system/modem-manager.service. How to include the real unit file?
<ilya-fedin>
*enable the real unit file
<{^_^}>
[nixpkgs] @veprbl merged pull request #52707 → bzflag: enable on darwin → https://git.io/fhJzd
<{^_^}>
[nixpkgs] @veprbl pushed to master « bzflag: enable on darwin (#52707) »: https://git.io/fhJgK