magnetophon has quit [Remote host closed the connection]
lopsided98 has quit [Quit: Disconnected]
lopsided98 has joined #nixos
jperras has joined #nixos
<colemickens>
If there's anyone involved with the Meson PR / Gnome 3.30 update, I'd be curious if there's anyway I can help out. I need the newer Meson and am eager for it to be available.
<samueldr>
j t o j n a r is working on the gnome 3.30 update, (spacing the name so this message doesn't ping 'em)
<boomshroom>
I just got Opus Magnum on Steam from the Humble Bundle. When I tried running it on my NixOS machine, it wouldn't start and when I tried running Steam from the terminal, I got the following error:
<boomshroom>
`./Lightning.bin.x86_64: error while loading shared libraries: libGL.so.1: cannot open shared object file: No such file or directory`
<jasongrossman>
samueldr: "helpees" is definitely a word.
<samueldr>
haha, wanted to make this sentence funny, no need to be stuffy in the comment :)
<slack1256>
mdash: It would answer questions like "What nixpkgs generation I was using on user generation X"
<slack1256>
Or correlate different system-generations with user-generations
Synthetica has quit [Quit: Connection closed for inactivity]
<jasongrossman>
samueldr: IMO there's NEVER a need to be stuffy. Unless you're competing in a stuffiness competition for big prize money.
<samueldr>
:) I'll at least keep professional and using known good words in the official manuals
orivej has quit [Ping timeout: 252 seconds]
<mdash>
slack1256: mmm, there's no correlation between system and user generations though
<slack1256>
Exactly, yet you know sometime around 2 weeks ago you had a good system + user + channel generation and you don't remember which values they were.
<mdash>
it's the saddest thing
<boomshroom>
Hmm. I just ran into a Steam Play (Windows) game that failed because the chroot doesn't provide Python 3.
<boomshroom>
Not the error I would expect from a game meant for Windows.
<boomshroom>
`/usr/bin/env: ‘python3’: No such file or directory`
<samueldr>
it's proton itself that needs python 3
<samueldr>
IIRC there was an update in nixpkgs related
<teto>
we have an alias linux_latest for the kernel, do we have an alias linux_lts ?
<teto>
can't find it but that would be practical
mayhewluke has quit [Ping timeout: 252 seconds]
mayhewluke has joined #nixos
<samueldr>
hm, it would have to be "latest_lts"
<samueldr>
soon we'll have three!
<samueldr>
oh, and we *do* have an alias for the latest LTS already
<samueldr>
it's the default, isn't it?
<samueldr>
oh, soon we'll have four LTS kernels in nixpkgs
* samueldr
didn't know 4.4 was still LTS and updated
<teto>
samueldr: "we *do* have an alias for the latest LTS already" what is that ? seems like 4.14 is last lts
<samueldr>
yes, and `linuxPackages = linuxPackages_4_14;`
<samueldr>
the default is the LTS
<samueldr>
I expect it will be switched to 4.19, either at 4.19's release, or near the following
<samueldr>
(though I have no sources)
<teto>
ok so I can use linuxPackages.kernel to get the latest lts kernel
<samueldr>
I'm not curious about when the switch to 4.19 will happen, but yeah, I am using `linuxPackages` and related attributes personally as "latest LTS"
<{^_^}>
[nixpkgs] @peterhoeg merged pull request #47482 → nixos on hyperv: load proper modules and make installer use the module → https://git.io/fxveA
<boomshroom>
I suppose I should mention that the newer version I installed doesn't break everything, so that's nice.
<samueldr>
boomshroom: updating made proton work?
<{^_^}>
[nixpkgs] @peterhoeg opened pull request #47699 → nixos on hyperv: load proper modules and make installer use the module → https://git.io/fxIRl
<boomshroom>
No. I meant a previous NixOS update made my system unbootable.
<boomshroom>
I'm no longer running 2 generations behind.
<samueldr>
oh, progress I guess
<boomshroom>
Looks like Dwarf Fortress failed to update because it couldn't rm something.
<sphalerit>
colemickens: I don't have an example but it should be as simple as setting boot.kernelPackages = pkgs.linuxPackages_4_18; boot.kernelPatches = [{name = "magic-trackpad-2"; patch = ./magic-trackpad.patch; }];
<sphalerit>
Or something like that
<sphalerit>
(Typed on my phone from memory so it might well be a bit wrong
Ariakenom has joined #nixos
Mr_Keyser_Soze has joined #nixos
Itkovian has quit [Read error: Connection reset by peer]
jtojnar has quit [Remote host closed the connection]
xorkle_ has joined #nixos
iyzsong has quit [Ping timeout: 245 seconds]
ryantm_ has joined #nixos
mog- has joined #nixos
arahael1 has joined #nixos
carlosdagos has quit [*.net *.split]
tertl3 has quit [*.net *.split]
charleshd has quit [*.net *.split]
_deepfire has quit [*.net *.split]
spacefrogg has quit [*.net *.split]
{^_^} has quit [*.net *.split]
r5d has quit [*.net *.split]
globin has quit [*.net *.split]
Arahael has quit [*.net *.split]
v0latil3 has quit [*.net *.split]
mog has quit [*.net *.split]
demize has quit [*.net *.split]
colemickens has quit [*.net *.split]
adelbertc_ has quit [*.net *.split]
xorkle has quit [*.net *.split]
canndrew has quit [*.net *.split]
Hotkeys has quit [*.net *.split]
joepie91 has quit [*.net *.split]
woffs has quit [*.net *.split]
rihardsk[m] has quit [*.net *.split]
pasukon has quit [*.net *.split]
yastero has quit [*.net *.split]
ryantm has quit [*.net *.split]
typetetris has quit [*.net *.split]
Aexoden has quit [*.net *.split]
kwork has quit [*.net *.split]
phreedom_ has quit [*.net *.split]
klntsky has quit [*.net *.split]
mkaito has quit [*.net *.split]
mog- is now known as mog
civodul has joined #nixos
<teto>
the error when you try to access a set member starting by a number is less than stellar "attempt to call something which is not a function but a set" (triggered by a = { "56" = 5; b = 4;} a.56)
patrl has joined #nixos
Mr_Keyser_Soze has quit [Ping timeout: 252 seconds]
mvnetbiz has joined #nixos
<mvnetbiz>
Can someone help me with cmake? cmake manual says all install directories are prepended with CMAKE_INSTALL_PREFIX, but this does not seem to be the case in the nix builder somehow?
selfsymmetric-pa has quit [Ping timeout: 252 seconds]
LogicAside is now known as ThatOtherGuy
endformationage has quit [Quit: WeeChat 1.9.1]
<mvnetbiz>
jackdk, yeah I did read through that. I thought it was some nix idiosyncrasy but I believe I just misunderstood cmake manual, it appears only relative paths get that appended or something, because an install command for destination "bin" works but it is the install for destination "/usr/lib" that fails
Sigyn has quit [Quit: People always have such a hard time believing that robots could do bad things.]
<teto>
how can I force the merging of a specific configuration entry before nixos or the module naturally does it ? for instance I would like to merge { "8139TOO_PIO" = yes; } and { "8139TOO_PIO" = no; } as { "8139TOO_PIO" = yes; } . These are defined as submodule items and I can merge merge them via evalModules. Problem is I need to merge them before passing them to the module ?
<teto>
s/?/!
<joko>
teto: can't you use lib.mkForce in yes?
<joko>
So in the respective submodule, { "8139TOO_PIO" = lib.mkForce yes; }
<clever>
teto: have you seen mkMerge?
<teto>
the reason is I want to do kernel.override { structuredConfig= merge (config1 config2) } with config 1 { "8139TOO_PIO" = yes; } and config 2 { "8139TOO_PIO" = no; }
ThatDocsLady has joined #nixos
<clever>
teto: ah, then i would just use //
LysergicDreams has quit [Ping timeout: 252 seconds]
<mvnetbiz>
mkBefore?
<teto>
I can make it work with kernel.override { structuredConfig=[ config1, config2]} but this breaks backwards compatibility as structuredConfig expects a set.
<teto>
clever: the problem with // is that it won't respect the merging rules
tg has joined #nixos
<teto>
last set will override all others
<clever>
teto: nothing in .override can respect merging rules
<teto>
problem is mkMerge will return a list (I already got the merging working with a list, I want to get rid of the list syntax), what I want is ` structuredConfig = ENFORCE_MERGE (lib.mkMerge [ config1 config2 ])`. Is that possible ?
<clever>
teto: mkMerge returns a special set, that the module framework will merge as its processing all modules
Xal has joined #nixos
<clever>
teto: so you would need to treat structuredConfig as a module, and then look at the final config tree after running the modules
<teto>
yeah right, let me try sthg I had in mind then forgot
patrl has joined #nixos
phreedom has joined #nixos
charleshd has joined #nixos
spacefrogg has joined #nixos
pasukon has joined #nixos
demize has joined #nixos
r5d has joined #nixos
colemickens has joined #nixos
mkaito has joined #nixos
typetetris has joined #nixos
Aexoden has joined #nixos
Hotkeys has joined #nixos
{^_^} has joined #nixos
rihardsk[m] has joined #nixos
woffs has joined #nixos
adelbertc_ has joined #nixos
yastero has joined #nixos
joepie91 has joined #nixos
globin has joined #nixos
<teto>
wooooowww the mkMerge succeeds. I don't get it I had tried it already and it was failing. I must have messed up.
<teto>
it should soon become easier to compose your kernel config o/
<teto>
ok firefox shows me "We have just installed an update in the background. Click Restart Firefox to finish applying it.". lol
* sphalerite
only just realised the parallel between NIX_PATH <lookup> syntax and C includes
<teto>
clever++
<{^_^}>
clever's karma got increased to 28
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « haskell-free and base-orphans: use latest versions to fix build with ghc-8.6.x »: https://git.io/fxIyL
<{^_^}>
[nixos-weekly] @garbas pushed commit from NixOS Weekly Robot to gh-pages « Preview of '1be343da8f3527dda0e35bf3b49b4968219e9a3d' commit built by Travis-CI \#436541732 »: https://git.io/fxIS1
<{^_^}>
[nixpkgs] @binarin opened pull request #47728 → rabbitmq: make all dependencies explicit → https://git.io/fxISM
domenkozar has joined #nixos
<roberth>
domenkozar and I are working on a brand new Continuous Integration service for Nix users. Check out https://hercules-ci.com!
<sphalerite>
roberth: ooh! Is this related to the previous work on a hydra successor?
<roberth>
sphalerite: it's a new service using the old project's name
<sphalerite>
oh that looks really cool
rembo10 has quit [Remote host closed the connection]
v0|d has quit [Remote host closed the connection]
Xal has joined #nixos
rembo10 has joined #nixos
kai_w has quit [Ping timeout: 252 seconds]
v0|d has joined #nixos
simukis has joined #nixos
<{^_^}>
[nixos-weekly] @domenkozar pushed to master « suggestions »: https://git.io/fxIHO
<{^_^}>
[nixos-weekly] @garbas pushed commit from NixOS Weekly Robot to gh-pages « Preview of 'fe3c28fbe859405e99ce5b4e3628c5dd8c18053c' commit built by Travis-CI \#436548860 »: https://git.io/fxIH8
asymmetric has joined #nixos
<{^_^}>
[nixos-weekly] @domenkozar pushed to master « English »: https://git.io/fxIH2
<{^_^}>
[nixos-weekly] @garbas pushed commit from NixOS Weekly Robot to gh-pages « Preview of '2fa902804cca868b993209f632a608a944d064f9' commit built by Travis-CI \#436549698 »: https://git.io/fxIHX
iyzsong has joined #nixos
legojames has joined #nixos
<legojames>
Hello, anyone here able to help me customise a package from nix-pkgs
<{^_^}>
[nixos-weekly] @domenkozar pushed 5 commits to production: https://git.io/fxIQO
<{^_^}>
[nixos-weekly] @domenkozar pushed to master « Call for Content: 2018/11 »: https://git.io/fxIQn
<{^_^}>
[nixos-weekly] @domenkozar opened pull request #66 → Call for Content: 2018/11 → https://git.io/fxIQW
<{^_^}>
[nixos-weekly] @garbas pushed commit from NixOS Weekly Robot to gh-pages « Release of (commit '2d4ee0200c4272481ef67401e3864d460728b23a') built by Travis-CI \#436552401 »: https://git.io/fxIQ8
<legojames>
I want to install a package (in this case Minecraft) and override a variable set in the default.nix file. I want to change the exec param on line 8 here:https://github.com/NixOS/nixpkgs/blob/a88b8da55436e4d74ff5fa50c5c20308966a34cb/pkgs/games/minecraft/default.nix#L8
<legojames>
I've read about package overides and overlays even tried to make them work but with no success
<clever>
legojames: variables in a let block cant really be overridden
<clever>
legojames: the simplest option is to copy the whole default.nix, edit the copy, and just have a minecraft = pkgs.callPackage ./the_copy.nix {}; in your overrides
<{^_^}>
[nixpkgs] @xeji merged pull request #47695 → nixos/activation: fix systemd-user daemon-reload in auto-upgrade service → https://git.io/fxIZB
<{^_^}>
[nixpkgs] @xeji pushed commit from @Ma27 to master « nixos/activation: fix systemd-user daemon-reload in auto-upgrade service (#47695) »: https://git.io/fxIbX
<{^_^}>
[nixpkgs] @Mic92 pushed commit from @r-ryantm to release-18.09 « wireguard-tools: 0.0.20180918 -> 0.0.20180925 »: https://git.io/fxINm
freeman42x has quit [Ping timeout: 252 seconds]
sb0 has joined #nixos
[Leary] has quit [Remote host closed the connection]
Lears has joined #nixos
<{^_^}>
[nixpkgs] @Mic92 pushed 0 commits to nixos-tests-debug: https://git.io/fxINz
<{^_^}>
[nixos-weekly] @lheckemann opened pull request #67 → Fix typo in title → https://git.io/fxINX
<{^_^}>
[nixpkgs] @xeji pushed commit from @Ma27 to release-18.09 « nixos/activation: fix systemd-user daemon-reload in auto-upgrade service (#47695) »: https://git.io/fxIN5
patrl has quit [Ping timeout: 252 seconds]
teser has joined #nixos
<teser>
hi, does anybody know if firejail can be run under normal user, without sudo? A related question is: how can a program define that it wants to run under a different username, or possibly just create the username and run under that? I know how this can be done, I'm interested in how nixos handled that when a program requires it?
Ridout has joined #nixos
<Yaniel>
the suid call and flag are for that
lassulus has quit [Ping timeout: 268 seconds]
<sphalerite>
emily: re your suggestion about wpa_supplicant, networkmanager is pretty huge and I'm not sure we want to pull that into the installer. There's always the graphical installer if you want networkmanager
<srhb>
teser: Usually through security.wrappers
<sphalerite>
emily: what might be nice though is shipping iwd with it
<{^_^}>
[nixos-weekly] @domenkozar pushed 2 commits to production: https://git.io/fxIA7
<{^_^}>
[nixos-weekly] @domenkozar merged pull request #67 → Fix typo in title → https://git.io/fxINX
<sphalerite>
teser: for the case of firejail, we have the `programs.firejail.enable` option to set it up automatically
<{^_^}>
[nixos-weekly] @garbas pushed commit from NixOS Weekly Robot to gh-pages « Release of (commit 'a51422af7634912320ecad04cda91764a2fd26b0') built by Travis-CI \#436573656 »: https://git.io/fxIxv
<sphalerite>
teser: ah, it seems this was added after the 18.03 release
<sphalerite>
teser: you could try the 18.09 beta (release should be pretty soon anyway), or you could just copy the module into /etc/nixos and put imports = [ ./firejail.nix ]; in your configuration.nix to make it available
<teser>
sphalerite, is there the https://nixos.org/nixos/options.html website for that release - I was always wondering why the options are not up-to-date on the website, it would be great if we could select the release version on there?
<sphalerite>
teser: there isn't, but you can use `man configuration.nix` which contains all the same information as it applies to your current system
<sphalerite>
emily: I'm just trying iwd, it seems to work quite nicely and has a fairly convenient command-line interface
<sphalerite>
meh, its closure isn't tiny either. Still 3x smaller though
lassulus has joined #nixos
<teser>
sphalerite, thanks, I didn't know about the man page
<teser>
will come in handy
<sphalerite>
:) I use it all the time, except when I want to show somebody an option (which is when I fall back to options.html)
<teto>
sphalerite: do you know if it works for broadcom too or it's intel specific ?
<teser>
sphalerite, one more question: awesome 3.5 was removed a month ago https://github.com/NixOS/nixpkgs/compare/ff449e176a60...8f6d3a3c3300, but I require that in order for my modules to work (I have a module that's not compatible with the 4.x version). It is possible to bring that back and add an option into the configuration.nix to re-enable this?
<MrAngel>
Has anyone managed to install powerline via nix itself and get powerline.zsh in the right place?
<MrAngel>
It very obediently copies the python binaries to my path but not the shell script that makes the magic happen
<{^_^}>
[nixpkgs] @thefloweringash opened pull request #47734 → thunderbird: fix build with rustc that supports `aarch64-fuschia` → https://git.io/fxIpo
<sphalerite>
MrAngel: stuff installed via nix won't go into widely shared namespaces, only places like the system profile (for nixos) and ~/.nix-profile (for everywhere nix-env is used without specifying a differnet profile)
<MrAngel>
The system profile should be fine thought, right? It's in the path
<{^_^}>
[nixpkgs] @srhb pushed commit from @periklis to release-18.09 « kubectx: 0.5.1 -> 0.6.1 »: https://git.io/fxIpy
<MrAngel>
I mean, heck, even my user profile should be fine as that's in my user's path
<sphalerite>
MrAngel: so what you probably want is to source ~/.nix-profile/share/zsh/site-contrib/powerline.zsh in your zsh profile or similar
<MrAngel>
... I failed to find that file for some reason
<sphalerite>
or /run/current-system/sw/share/…
<MrAngel>
Package manager is doing its job fine, PEBKAC
<sphalerite>
:D
<sphalerite>
teser: there are various ways to do this. You can probably add it back by grabbing the package definition from a verison that had 3.5, and doing something like nixpkgs.overlays = [(self: super: {awesome = super.callPackage ./awesome.nix {};})]; in your configuration.nix to have awesome replaced with awesome 3.5
<sphalerite>
the overlay might not really be necessary if the awesome module allows setting the package, maybe there's a programs.awesome.package option or something
<MrAngel>
Aaaaand powerline for tmux still isn't behaving itself -_-
<MrAngel>
I must have just missed an actual install instruction
<sphalerite>
in which case you can just do programs.awesome.package = pkgs.callPackage ./awesome.nix {}; or something
<sphalerite>
MrAngel: how so?
<MrAngel>
I have a green bar at the bottom (the default tmux display) rather than Powerline
<MrAngel>
Despite the fact that I'm now sourcing the correct .zsh file and have everything present in my path
<MrAngel>
So I can only conclude there's something you need to do to enable powerline for tmux specifically that I've just plain missed
<sphalerite>
I don't think that would affect tmux's status bar?
<sphalerite>
yeah
<MrAngel>
Powerline has a plugin that's supposed to override tmux's status bar
<sphalerite>
you probably want to put `source ~/.nix-profile/share/tmux/powerline.conf` in your tmux.conf :)
<MrAngel>
I mean, that sounds like a very plausible suggestion, I sort of assumed just sourcing it in my shell would be enough since the docs don't mention this
<MrAngel>
Ah, it's hidden under configuration
<sphalerite>
sourcing it in your shell will do for using powerline for your shell prompt ;)
oida has joined #nixos
<MrAngel>
I agree with you that a distribution model relying on sourcing a file in one's shell to affect tmux would be a bit bonkers
<MrAngel>
But I read the installation instructions a cool five times and didn't see anything about doing it
<legojames>
clever: I've added a local copy of the minecraft package and amended some details, but not being picked up: Is this right in configuration.nix:
<sphalerite>
teto: emily: my earlier statements about iwd working pretty well may be disregarded. It works well at first, but then seems to spontaneously disconnect, or dhcpcd suddenly gives up on the addresses and doens't get them back, or something…
<sphalerite>
meanwhile wpa_supplicant is fine on the same network
<d1rewolf>
tilpner: ok, thanks. i'll give it a shot
<teto>
sphalerite: interesting ty
<d1rewolf>
tilpner: wow...you're not kidding. it just repetitively launches the login screen (launched 20-ish before I killed it)
<{^_^}>
[nixpkgs] @andir pushed 2 commits to release-18.09: https://git.io/fxLIw
<Avaq>
Hello! I have a short question regarding gnupg. My 'programs.gnupg.agent.enable' is set to false (as verified by nixos-option), and GPG_AGENT_INFO is empty, yet when I unlock my gpg key (through 'pass'), it remains unlocked for 15 minutes. What could be doing this?
<sphalerite>
Avaq: nowadays gnupg will not run without an agent, and automatically spawn one if needed
<tilpner>
d1rewolf - I suspect this behaviour is not specific to NixOS. If this is important to you, I suggest you replicate this behaviour on a non-NixOS machine and report it to whoever made that AppImage
<sphalerite>
Avaq: I think putting max-cache-ttl 0 in your gpg-agent.conf should stop it from caching credentials
<Avaq>
sphalerite: Ah, that makes sense. It is actually the behaviour I want, so I have no need to disable it. I just to clear up how 'programs.gnupg.agent.enable' affects the gpg behaviour. I guess it makes no difference. Is it a legacy thing?
<sphalerite>
Avaq: I'm guessing from the option's documentation that it uses systemd socket activation — i.e. rather than the client noticing that no agent is running and spawning one itself, it connects to the socket which prompts systemd to start the agent.
<d1rewolf>
is there any effort to make nixops work with vmware? we use it primarily as our local virtualization solution, so it'd be nice to use nixops for provisioning
erasmas has joined #nixos
obadz has quit [Ping timeout: 252 seconds]
<Myrl-saki>
d1rewolf: You can also nixops to a nixos machine.
<Myrl-saki>
d1rewolf: You just have to manage the starting and stopping of vmware outside of nixops.
<d1rewolf>
Myrl-saki: yeah...was hoping for a way to instantiate remotely, but if that's the best option
<{^_^}>
[nixpkgs] @rnhmjoj opened pull request #47764 → nixos/xserver: add option to install custom xkb layouts → https://git.io/fxLns
obadz has joined #nixos
Itkovian has joined #nixos
<pmade>
Updating nixos-18.03 wants patches in the form of https://git.gnome.org/browse/vte/patch/?id=HASH which redirects to a GitLab login page. I think the URL should be `commit' and not `patch'.
<typetetris>
Myrl-saki: you are right, there isn't a patch yet.
<Myrl-saki>
typetetris: Maybe just set meta.broken
<Myrl-saki>
But I'm not sure what meta.broken means.
<sphalerite>
pmade: right… it should be extra-substituters. People seem to run into this quite a bit, I'm not really sure why (if cachix does this wrong or something..?)
<sphalerite>
pmade: anyway, you should be able to rebuild normally using nixos-rebuild switch --option substituters https://cache.nixos.org/
<{^_^}>
[nixpkgs] @peti pushed 8 commits to haskell-updates: https://git.io/fxLw8
<d1rewolf>
tilpner: so in this case, since it's a very old version, if we wanted to try phabricator on nixos, would you recommend forking and updating this proejct
<yurb>
adisbladis: thanks, and which nix command should this be an argument for?
<adisbladis>
yurb: You put that in your configuration, for example: environment.systemPackages = [ (mumble.override { jackSupport = true; }) ]
<yurb>
adisbladis: I see. Thanks!
<samueldr>
d1rewolf: last year I tried to update phabricator, it went good, but didn't end up using it
<{^_^}>
[nixpkgs] @florianjacob opened pull request #47775 → nixos/munin: move from /var/run to /run → https://git.io/fxLr5
<samueldr>
phabricator event did its upgrade successfully from the 2014 revision to its 2017 revision!
<samueldr>
even did*
<samueldr>
(sorry, I don't have the branch around to share)
<d1rewolf>
samueldr: ah...so there's some builtin upgrade capability within it? I'm completely new to it, but it looks like it could be a nice replacement for our hodgepodge of tools
<samueldr>
d1rewolf: I meant that once nixpkgs's revision is updated, phabricator handles migrating its database, and the upgrade from that old to a newer revision worked in my case
<d1rewolf>
tilpner: so what exactly is the difference between the file I referenced and that one (phd.nix)? why are there two phabricator-focused derivations?
<samueldr>
it looked like a solid contender for using, but at $work we didn't have time to invest into porting our workflow into phabricator (yet)
<d1rewolf>
samueldr: I see. k. out of curiosity, what did you end up going with?
charleshd has quit [Remote host closed the connection]
<samueldr>
status quo: not changing a thing, staying on gitlab.com :/
fendor has quit [Ping timeout: 252 seconds]
charleshd has joined #nixos
<d1rewolf>
this raises a general nixos question for me. as a user, i would expect to search for phabricator, and, if found, would probably expect it to be a fairly recent version. is there something in the nixos workflow which tracks aging packages? are they eventually marked for removal?
<samueldr>
and as far as your question: <nixos/modules/services/web-servers/apache-httpd/phabricator.nix> is a module that autoconfigures a phabricator instance for a nixos configured apache, <pkgs/misc/phabricator/default.nix> is the actual derivation with the software
<samueldr>
ah, and phd.nix seems like the service for the other misc. daemons phabricator would run
<d1rewolf>
interesting. that seems...a little messy
<samueldr>
not sure it is, the httpd module isn't privvy to the service modules :)
<d1rewolf>
and nix search phabricator only shows two packages (arcanist and phabricator)
<samueldr>
yeah, nix search only searches derivations, it doesn't know about nixos modules
<samueldr>
the phd.nix services module uses `pkgs.phabricator`, as does the httpd module
<d1rewolf>
samueldr: modules are new to me. reading about it now. thanks very much. the nixos footprint is wide at times....i still feel like a newb a couple months in.
<d1rewolf>
that said, I haven't *fully* read the nixos and nix manuals, so my fault prob
<samueldr>
it is wide, nixpkgs is all about derivations, packages, software; nixos modules intend to use those to build a complete system, with some configuration implementation details taken care for you, so it definitely is a big thing
<thoughtpolice>
d1rewolf: I installed Phab on Nix myself, and I just wrote automation around it to let phab automatically update itself
<thoughtpolice>
Yes, I am, but not on NixOS. (I migrated my Phabricator to a different provider that didn't have it)
<d1rewolf>
thoughtpolice: I see. mind if i dm you to get your thoughts on it? off-topic here
<d1rewolf>
samueldr: given just a nix file, how would one tell if you're looking at a derivation or a module?
<thoughtpolice>
Sure, though for now I can't really guarantee a timely reply (just sort of popping in and out)
<samueldr>
d1rewolf, thoughtpolice, unless you want to be 100% private, you could always share thoughts on phabricator on #nixos-chat (I still am interested in one day looking again at phabricator)
<samueldr>
d1rewolf: hmmm, esperience, and derivations also most often have `mkDerivation`, as a function, and nixos modules most often look like a function returning a set options, and configurations
<d1rewolf>
samueldr: thoughtpolice: that'd be fine as well. I have no need to be private
<d1rewolf>
samueldr: ok, great. thank you
slyfox has joined #nixos
patrl has quit [Ping timeout: 264 seconds]
asymmetric_ has quit [Ping timeout: 252 seconds]
<d1rewolf>
so if i wanted to test the older phabricator which is already there in a nixos-container, would i use phd.nix, the nix derivation returned by "nix search phabricator", or the module?
<samueldr>
you would enable the phabricator httpd module, and the phd module for the daemons, which in turn both are implicitly using the phabricator derivation
<nschoe>
has anybody tried to build ROs package "joystick_drivers"? It says it's missing "diagnostic_updaterConfig.cmake". And I have no idea why this .cmake file is missing :/
<nschoe>
Any help is welcomed (or any suggestion) :)
<samueldr>
d1rewolf: do note that the networking options are maybe not the best, they are probably there only ro simplify the test
<thoughtpolice>
In general I'm really not sure we should package Phabricator tbh
<d1rewolf>
samueldr: sounds like it might be easier to just test in an debian vm or something
<thoughtpolice>
Upstream sort of explicitly advises against it and recommends you update very regularly.
<thoughtpolice>
Like, Debian does this and there's already been multiple times with GHC where people tried installing the 'arcanist' package and it didn't work
<d1rewolf>
thoughtpolice: could you run it on nix without packaging it? I haven't tried that yet
<thoughtpolice>
I guess if you scripted it to do bi-monthly updates it wouldn't be bad
<thoughtpolice>
Yes, that's mostly what my module does
<d1rewolf>
k thx
<thoughtpolice>
All the module does is really install a script that runs commands for you conveniently
<thoughtpolice>
So it doesn't use systemd services, or anything like that
<samueldr>
thoughtpolice: if there was a maintainer for the package, it probably wouldn't be an issue :/, having the system configuration in nixpkgs itself, especially since it's tested, imho is good
<thoughtpolice>
It just installs PHP and whatnot and sets up all the configuration files
<thoughtpolice>
Well, yeah. If we went the way my module does it, I think it would be fine
<thoughtpolice>
But packaging it explicitly is maybe a bad idea
<thoughtpolice>
It needs the network to work that way, however
<d1rewolf>
thoughtpolice: so to use your phabricator.nix, i'd just download and nix-env -i ./phabricator.nix?
<samueldr>
dunno, I'm not knowledgeable enough with phabricator and related to know for sure
<thoughtpolice>
No, it needs to be installed as a NixOS module
<thoughtpolice>
I haven't used it in a while, it's probably busted now
<d1rewolf>
thoughtpolice: ok. going to read up on those. thanks
<thoughtpolice>
The version I use on Ubuntu is kinda close
asymmetric_ has joined #nixos
Ariakenom has quit [Read error: Connection reset by peer]
<catern>
i'm packaging "latrace". It depends on an old version of flex, which is already in nixpkgs. But there's a small patch to latrace that makes it work with modern versions of flex
<catern>
Should I provide it the old version of flex, or should I use the patch?
<thoughtpolice>
Use the patch.
Ariakenom has quit [Quit: Leaving]
<thoughtpolice>
Well, depends on what you mean by 'old'. We have a few versions. Generally I try to keep things off old versions if possible, but we do have a few uses of flex_2_5
<mpickering>
Any ideas how to fix "<command line>: can't load .so/.DLL for: libz.so" when building a haskell package using cabal? I have added zlib to the environment
<thoughtpolice>
Actually, a better criteria is if the patch for latrace has been accepted upstream. If it has, I would definitely use the patch, personally.
<thoughtpolice>
This is all just my opinion, of course.
<catern>
The patch hasn't been accepted upstream
<catern>
latrace is unmaintained AFAICT
jtojnar has quit [Ping timeout: 252 seconds]
Avaq has quit [Quit: Leaving.]
<catern>
(but it's packaged in many other distros, and generally useful)
<thoughtpolice>
I see, did some other distro write the patch then (or whoever?)
Ariakenom has joined #nixos
jtojnar has joined #nixos
<thoughtpolice>
In any case, I stand by my recommendation that keeping things off old versions is preferable (less overrides/special cases in all-packages.nix, or un-obvious dependency inputs like flex_2_5_35 in the expression inputs, makes it easier to ween ourselves off eventually) if, as the maintainer, you're willing to test it and say it works.
asymmetric_ has quit [Ping timeout: 252 seconds]
<thoughtpolice>
Which I assume you obviously are!
<{^_^}>
[nixpkgs] @thoughtpolice pushed commit from @r-ryantm to master « libressl_2_8: 2.8.0 -> 2.8.1 »: https://git.io/fxLMZ
<gchristensen>
the cycle time is so slow and you can't know if it'll work on a fresh host unless you get a fresh host. for that reason, I always created new servers instead of applying chef to an existing node.
<d1rewolf>
gchristensen: i'm looking of a nix-ish like experience (declaring a configuration and then go do it), but across our infrastructure, which includes different flavors of linux and windows. chef seems to have good support for both
<d1rewolf>
gchristensen: oof. so there's no guarantee that a configuration is accurately applied?
<gchristensen>
chef doesn't provide a nix-ish experence. it pretends to be declarative, but it is actually not and the declarative feeling falls away almost immediately
<gchristensen>
well it isn't declarative under the hood, so it is trying to take a declarative spec of just a tiny subset of the entire system, and convert it in to an imperative setof commands to run
<d1rewolf>
gchristensen: ugh. so I couldn't, say, keep my 6 personal machines all in sync with the same software/config with it?
<gchristensen>
you _could_ but it is a lot of hard work
<d1rewolf>
gchristensen: out of curiosity, and not to belabor the point, but could you give an example of why keeping 6 machines in sync would be a lot of hard work?
<gchristensen>
well chef / puppet only do what you tell it, so it is basically as hard as it would be if you wrote a bash script to do it... except with a pretending-to-be-declarative language on top.
<mdash>
d1rewolf: the chef/puppet model is "change a bunch of stuff" instead of "build it all from scratch"
<d1rewolf>
ah...k. i see
<gchristensen>
right, Nix's model is build everything from scratch every time, so you know _exactly_ what you have. Chef / puppet's model is, indeed, just change stuff so it matches the spec... but your spec only covers a teeny tiny portion of all the system and you have a massive glacier of stuff undescribed.
sigmundv__ has quit [Ping timeout: 272 seconds]
patrl has joined #nixos
<d1rewolf>
that makes sense. so all the cruft which would be installed from a default debian or ubuntu iso
<gchristensen>
yeah
<gchristensen>
or that one time you installed another package just b/c you're experimenting
<sphalerite>
one of the key points is that apt for instance *does not provide any way* to go back to a previous state — if you install a package and want to restore everything to the way it was before, theo nly way to do so is to restore a backup
<d1rewolf>
sphalerite: right. that is indeed ugly.
<sphalerite>
I'm not sure if rpm/dnf/whatever else is better nowadays
<elvishjerricco>
sphalerite: apt doesn't even log the changes it makes in a way that can be automatically reversed? That's insane
<d1rewolf>
someone commented in the chef channel that you could approach uninstall ability with zfs. not sure what they meant...maybe snapshotting of the entire system and then rolling back to that snapshot
<hodapp>
d1rewolf: yes, don't some folks do almost exactly that on FreeBSD?
<d1rewolf>
hodapp: not sure
<hodapp>
it's a normal ZFS-on-root application I think
<elvishjerricco>
Ah. Well in that vein, NixOS isn't even really safe since services can mutate state
<elvishjerricco>
But I guess it's much better than that
<sphalerite>
sure, but that's nixos modules. You still have the guarantee that if you add something to systemPackages or install it with nix-env, you can go right back to where you were before
<sphalerite>
why, you can even do it instantaneously with --rollback :D
mayhewluke has quit [Ping timeout: 268 seconds]
<sphalerite>
of course software likes to create its wee "config" (often mixed config/state) files in ~ or /etc or /var but at least you can install it without immediately having that side effect
<sphalerite>
gchristensen: the horizontal rule after the "let's consider the 200-store-path case" paragraph feels out of place..?
<ikwildrpepper>
then you can use nixpkgs, such as : pkgs = import nixpkgs {};
<bigvalen>
What's the differenice between nixos-rebuild & nixos-install ? (I seem to have an install that doesn't boot properly - hangs starting networking, and a few other services, for some reason)
<bigvalen>
I managed to boot into an installer. nixos-enter & nixos-switch moan that it's running in chroot, and when I reboot, I still get the same 7 or so services that aren't starting, maybe a reinstall will sort it)
<samueldr>
put simply, nixos-install is used to do the housekeeping required to nixos-rebuild the target system while into the installer
romildo has joined #nixos
<bigvalen>
So, probably can run that safely, with the existing system under /mnt, and it'll try heal my /boot too
<Unode>
ikwildrpepper: so fetching a specific commit. Got it. Thanks
<romildo>
Is there in nix a function to build a string of paths separated by : from a list of paths?
<sphalerite>
bigvalen: yes,it won't take care of eg. mounting /mnt/boot though, so make sure to do that
<sphalerite>
teser: it seems the option is called virtualisation.virtualbox.host.enableExtensionPack nowadays, based on a cursory search of the configuration.nix manpage
<teser>
sphalerite, my current config is set like so: config = { pkgs }: import nixCustomConfig { inherit pkgs; };
<bigvalen>
sphalerite, Thanks! Made slightly more tricky on a zfs root :)
<clever>
teser: `config = import nixCustomConfig;` will behave identically
<philippD>
Is there a function in lib or something that extracts the name of a nix-store path?
<teser>
clever, sphalerite how can I include the "nixpkgs.config.allowUnfree = true" in there as well?
<clever>
teser: either set it in the nixCustomConfig function, or do `config = (import nixCustomConfig) // { allowUnfree = true; };`
<bigvalen>
This is odd, though. "cannot find a grub drive for /dev/disk/by-path/usb-HP_iLO_Internal_SD-CARD_00002311231231-0:0"
<bigvalen>
"Check your device.map"
<bigvalen>
Is that...generated anywhere ? I would have assumed Grub devices were just pulled from /sys somewhere
<clever>
bigvalen: i tend to just set it to a normal device, like /dev/sdb
<clever>
but with usb devices, that has the risk of being the wrong device
patrl has quit [Ping timeout: 260 seconds]
<mdash>
by-uuid tends to be more stable
slack1256 has quit [Remote host closed the connection]
<clever>
so i change it to just "nodev" after it has installed, and then grub cant update anymore
<clever>
mdash: grub needs the root device, not a partition, and only partitions have uuid's
<clever>
teser: oh right, and since your doing something diff now you have to undo the improvement i gave
<mdash>
clever: huh, how did i forget that
<bigvalen>
clever, Weirdly, /dev/disk/by-uuid only has partitions, not whole disks. So I can see my /dev/sdf1 there (which is /boot) but not sdf. I only see sdf in /dev/disk/{by-id,by-path}
<mdash>
bigvalen: never mind me
<sphalerite>
clever: weird how the GPT disk identifier isn't exposed in /dev/disk/by-*…
<teser>
clever, I would like to use the improvement you proposed.
<clever>
teser: it would also help to know what nixCustomConfig contains, can you just post all the files to a gist?
<teser>
clever, this works now
trevorriles has quit [Client Quit]
trevorriles has joined #nixos
<sphalerite>
teser: why are you doing this importing business? it should be fine to just use the nixCustomConfig file as your configuration.nix AFAIU
jperras has quit [Quit: WeeChat 2.2]
<teser>
sphalerite, that config basically contains "packageOverrides = pkgs: rec { ... }" that I like to keep separate from configuration.nix
<clever>
teser: i would just put the allowUnfree = true; in that file then
<trevorriles>
Has anyone here used AWS cfn-init to update nixos configuration? So when the cloudformation metadata gets updated, it triggers a process that writes the new config then calls "nixos-rebuild switch"
<sphalerite>
teser: then your configuration.nix should probably look roughly like { pkgs, ... }: { nixpkgs.config = import nixCustomConfig { inherit pkgs; }; virtualisation.virtualbox.host.enable = true; virtualisation.virtualbox.host.enableExtensionPack = true; }
<sphalerite>
although you shouldn't need pkgs for any of your nixpkgs.config definitions, otherwise you're likely to end up with infinite recursion
<teser>
sphalerite, I think I'm going to keep the allowunfree in this file, since virtualbox is defined here, so if I remove everything in the nixCustomConfig, everything still works.
<sphalerite>
teser: try not to get too hung up on how you'll split your configuration into pieces before you've got everything working
<sphalerite>
it's an easy trap to fall into, but really, there's nothing wrong with just putting everything in configuration.nix for a start
<sphalerite>
in fact it's probably better to do that then start splitting it once you get a second machine to configure, because that helps you find good "splitting points"
<teser>
I've been using configuration.nix on multiple machines, but I need certain parts only on servers, while others on clients, etc.
<{^_^}>
[nixpkgs] @volth opened pull request #47794 → dhcp: fix build on Debian-aarch64 → https://git.io/fxLAf
<teser>
I know I can improve the current setup that I have, however it's working just the way it is. The only thing I need is actually learn nix at some point.
<d1rewolf>
thoughtpolice: i'm curious...did you find you *had* to install phabricator with a nix file, or is it possible to do it by hand on top of nixos?
<bigvalen>
Gah. Even after a nixos-install the system still boots up, with systemd reporting 14 jobs in "startup".(ZFS pool scrubbing, some cron jobs for snapshotting, kernel auditing, cpu frequency governor setup, samba service daemon, address config for eno1 (shouldn't be used), winbindd
<bigvalen>
I wonder do I just need to wait until the zfs scrubbing is done
teser has quit [Ping timeout: 268 seconds]
matthewbauer has quit [Read error: Connection reset by peer]
jperras has joined #nixos
matthewbauer has joined #nixos
matthewbauer has quit [Remote host closed the connection]
<nico202>
saueldr: yeah garbage collection worked :)
acarrico has joined #nixos
<Friedem>
does `nix-store --export whatever.drv` capture all the binaries needed? or just a description of them?
<sphalerite>
Friedem: it only exports the drv itself along with its metadata (i.e. store path)
<sphalerite>
Friedem: you may be interested in nix-store --export $(nix-store -qR whatever.drv) which exports its entire drv closure (which is enough to build it on a system without access to the binary cache, but still requires internet to download stuff like source tarballs)
<Friedem>
is there some other way to capture my nix environment that I would be able to transfer to another machine without internet access?
<sphalerite>
you could export *everything* in the store in principle I guess
<sphalerite>
not sure how you'd do that without running into command-line lenght limitations though :/
<{^_^}>
[nixpkgs] @volth opened pull request #47809 → remove support for pre-5.26 Perl (2) → https://git.io/fxtkU
<{^_^}>
[nixpkgs] @volth closed pull request #47809 → remove support for pre-5.26 Perl (2) → https://git.io/fxtkU
<{^_^}>
[nixpkgs] @volth opened pull request #47810 → remove support for pre-5.26 Perl → https://git.io/fxtks
<sphalerite>
you could use nix copy to put all the paths in another store, e.g. nix copy --all --to /mnt/usbstick/
__monty__ has quit [Quit: leaving]
asymmetric_ has joined #nixos
<sphalerite>
Friedem: or if you want to take the really gung-ho approach, rsync -a /nix/ /mnt/usb/nix/ :D
<sphalerite>
but I suppose what it really comes down to is what do you want to do on the destination machine?
<dhess>
Anyone around who has built OpenWRT using Nix/Nixpkgs?
<gchristensen>
catern: You work with Shangpeng!!
<dhess>
it'd be great to generate OpenWRT images deterministically
<dhess>
gchristensen: cool blog post on that DAG Docker tool.
<gchristensen>
dhess: thank you :D
<clever>
dhess: #nixos-on-your-router may be of interest to you
<Friedem>
essentially, I want to end up with a directory structure, that if I were to chroot into it, would behave as though i'm in 'nix-shell --pure whatever.nix'
<clever>
Friedem: nix copy is your answer then, one min
<dhess>
clever: ok I'll check that out, although I'm already running NixOS on my router :) Here I'm specifically interested in something I can run on WAPs with very small amounts of flash and RAM
<clever>
Friedem: with this, i now have an example/nix/store and it expects you to chroot into example before using it
<sphalerite>
dhess: the closest to that is telent's work on NixWRT
<symphorien>
Friedem: do you need a nix-shell env (with build tools) or just things in $out/bin ?
<dhess>
gchristensen: Are you running these Docker images on NixOS machines, or k8s?
<gchristensen>
dhess: k8s
<clever>
dhess: ah, thats more in the territory of not-os
<clever>
dhess: not-os compiles down to a 40mb squashfs
<dhess>
sphalerite: yeah I know about NixWRT. It sounds promising but I need something that works now, which is why I'm looking into how to build OpenWRT using the Nix ecosystem.
<dhess>
clever: yeah if I were doing something from scratch I think I'd start with not-os
<symphorien>
If you just need the executables and nix is not on the target just tar nix-store -qR $path
<Mic92>
you can update the revision tested if it works
<dhess>
ok will do
<clever>
symphorien: and yeah, if you already have nix on the target, you would change the --to into --from, to copy from the unpacked tar, to the real store, or just nix copy over ssh
<clever>
symphorien: the main benefit i can see to using --to and --from like this, is getting updates over an airgap
<Mic92>
dhess: if it does not work your best bet is to patch out all the 2>/dev/null redirection they have in there horrible build system
<clever>
symphorien: so you could put wpa_supplicant and some wifi drivers onto a usb, and smuggle them into a laptop with broken wifi
<Mic92>
or you sysdig -c stderr
<dhess>
I need to figure out how to create a config as well. I want to bake images with the whole thing ready to go: flash and it's all configured and everything.
<dhess>
Mic92: ugh :(
<Mic92>
they have a kernel-like configuration system
<dhess>
It *is* tempting to try to bake my own. I just think that's probably a months'-long project
jperras has quit [Quit: WeeChat 2.2]
<Mic92>
I usually edit the .config files rather then click through the menu
<{^_^}>
[nixpkgs] @primeos closed pull request #47669 → nixos/networking: Add an option whether to use a local DNS resolver → https://git.io/fxkgf
<dhess>
Mic92: thanks, that's all useful.
Ariakenom has quit [Read error: Connection reset by peer]
<Mic92>
nixwrt could become very useful there, you can be much worse then buildroot/openwrt.
<dhess>
Mic92: were you cross-compiling these OpenWRT images?
<Mic92>
dhess: yes, always.
<Mic92>
this is what they at least support
<Mic92>
I have also packaged some simple stuff
helloworld has joined #nixos
<dhess>
Mic92: hmm, I haven't done any cross stuff with Nix yet, but I'm surprised that works with your gist. I don't see any cross suff in there, isn't the gcc in the fhsenv native gcc?
helloworld has left #nixos [#nixos]
<Mic92>
dhess: they build there own toolchain, the host compiler is only used to compile their stuff
<dhess>
holy crap, domenkozar is doing the Hercules-as-a-service thing for real?
shabius has quit [Quit: Leaving]
<freepotion>
Guys, in the meantime, I read this old issue https://github.com/NixOS/nixpkgs/issues/24346 But is there any news on this topic? No one is going to add another init (runit, OpenRC) to NixOS?
<{^_^}>
#24346 (by crocket, 1 year ago, closed): Can I replace systemd with OpenRC or runit on NixOS?
<freepotion>
Reading the source code, I came to the conclusion that this is almost an impossible task, because everything is hardcoded for systemd. But still?
<kalbasit[m]>
I don't understand the error, so I'm not sure how to fix it. `warning: path '/nix/store/bp7540dm0hyy5cq8v2ffqidzy4kc4r4g-bazel-watcher-0.5.0-deps' claims to be content-addressed but isn't error: unexpected end-of-file`
<gchristensen>
catern: shangpeng is great! I miss working with him.
<srhb>
kalbasit[m]: Yeah, I'm not keen on that without knowing that it actually builds properly on Hydra. I also don't see why it wouldn't, but then again I didn't try running the ofborg method and understand why it would choke on it.
<{^_^}>
samueldr/ofborg-viewer#22 (by grahamc, 6 seconds ago, open): On a network with webockets blocked you can't view logs of completed builds (or any builds)
<gchristensen>
thanks for the triage :)
<LnL>
oh and same at work (but that could be the domain)
<samueldr>
hmm, it'll have to not fail that noisily in that case
* samueldr
remembers the time there was no state at all
<roconnor>
does optionals evaluate its argument when the flag is false? I.e. is it safe to dereference attributes that are null when the condition is false?