<fresheyeball>
so I have a reallll weird one folks
<fresheyeball>
I am still playing with my tablet
<fresheyeball>
and krita is doing a segmentation fault
<fresheyeball>
when hitting Open or Save
<fresheyeball>
I was trashing about when I made a discovery
<fresheyeball>
it works fine, IF I lauch krita from termite
<fresheyeball>
and not from the gnome launcher,
<fresheyeball>
or from the gnome terminal
<fresheyeball>
if I lauch from there, seg fault on file open
<fresheyeball>
but if I lauch from termite, it works fine
<fresheyeball>
what the hell
<thefloweringash>
First guess would be differing values of QT_PLUGIN_PATH due to configuration in shell setup which the desktop environment might somehow be skipping. It's a pretty wild guess though.
<fresheyeball>
thefloweringash: lets see if there is a difference
<thefloweringash>
You could dig into the crashes by enabling coredumps and inspecting `coredumpctl`.
<sphalerite>
edef: separately packaged kernel modules like nvidia or kernel modules that are part of linux?
<edef>
sphalerite: out-of-tree, i'm hacking on wireguard and want to override it with my own source/patches
<sphalerite>
edef: I think it'd be a matter of making an overlay that looks something like self: super: {linuxPackagesFor = kernel: let superLinuxPackages = super.linuxPackagesFor kernel; in superLinuxPackages // { wireguard = superLinuxPackages.wireguard.override {…}; }
Dedalo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<edef>
sphalerite: that sounds like an approach worth trying, thanks
<sphalerite>
CMCDragonkai: it should work but is a bit stateful. Alternatively, putting kitty into systemPackages on the server should make the terminfo available too
<__monty__>
CMCDragonkai: With just ssh or with tmux?
<CMCDragonkai>
kitty terminal -> ssh -> NIXOS
<CMCDragonkai>
i don't like the idea of putting kitty terminal into systemPackages
Ariakenom has joined #nixos
<__monty__>
You don't need the entirety of kitty though. You could have an expression that just installs the terminfo files.
<CMCDragonkai>
Actually @sphalerite it doesn't work
<edef>
sphalerite: i guess i can override boot.kernelPackages
verrchu has joined #nixos
<sphalerite>
edef: oh yeah, that should also work
<edef>
sphalerite: i think it's because kernelPackages.extend will discard my changes
<sphalerite>
aaah yeah
<verrchu>
Hi guys. i am experiencing an issue with ssh-agent
<edef>
and that might also be the thing i should invoke to modify it
pointfourone has joined #nixos
pointfourone has quit [Client Quit]
<verrchu>
While trying to do 'ssh-add <somekey>'
<verrchu>
i get 'Could not open a connection to your authentication agent.
<edef>
error: The option value `boot.extraModulePackages.[definition 1-entry 1]' in `/home/src/local/nixpkgs/nixos/modules/services/networking/wireguard.nix' is not of type `package'.
<edef>
that looks working ^_^
<srhb>
verrchu: How did you start the agent?
<verrchu>
doing 'eval 'ssh-agent -s'' doesn't help
goibhniu has joined #nixos
<verrchu>
in mu configuration.nix i have 'programs.ssh.startAgent = true'
<CMCDragonkai>
`infocmp` is not in `ncurses`? I just installed, it still no command
<{^_^}>
[nixpkgs] @hyperfekt opened pull request #52352 → crosvm: init at R72-11316 → https://git.io/fpNam
<__monty__>
No idea. Haven't looked into it yet. I just assumed haddocks are never built.
<srhb>
I believe they are
<srhb>
(eg. nix path-info nixpkgs.haskellPackages.aeson.doc --store https://cache.nixos.org is good)
goibhniu has quit [Ping timeout: 240 seconds]
oldandwise has joined #nixos
<srhb>
__monty__: I do not get a rebuild of aeson if I ask for a ghcWithHoogle with it included.
<oldandwise>
Myrl-saki: i'll dropby Insular by 9pm tonite...
<{^_^}>
oldandwise: 6 days, 17 hours ago <nh2> Sorry, I had to step out of a moment. This is how to patch the bluetoothd derivation: https://gist.github.com/nh2/94049ce3d3cc8ab82a814ded9f7b5277 You can bind it to a variable and use that in `hardware.bluetooth.package`
<rardiol>
anyone knows why both SDL-1.2.15-dev/include/SDL/SDL_opengl.h and libGLU-combined/include/GL/glext.h have conflicting definitions for GL_GLEXT_VERSION ? Which one is 'correct' ?
<__monty__>
srhb: I don't have docs for megaparsec for example, even though I do have it installed with ghcWithHoogle.
<srhb>
__monty__: Maybe you've altered mkDerivation or something?
<__monty__>
Maybe the docs aren't built for nixpkgs-unstable?
rfold has quit [Ping timeout: 250 seconds]
<srhb>
__monty__: They are.
<__monty__>
Not on purpose.
<srhb>
__monty__: Sure, but I don't think it cares about intent. :D
oldandwise has quit [Client Quit]
<__monty__>
Maybe myEnvFun alters mkDerivation?
<srhb>
__monty__: I doubt it. Something else is going on.
* __monty__
puts on his deerstalker
<srhb>
__monty__: The cache definitely has the doc output of megaparsec too. :) And it should for everything that doesn't explicitly have haddocks disabled
<etu>
Izorkin: It seems to build at least
<etu>
Izorkin: I haven't tested to execute all of the resulting files
robstr has joined #nixos
<srhb>
__monty__: Just to make sure, what does nix eval nixpkgs.haskellPackages.megaparsec.doc.outPath give you?
<srhb>
__monty__: It should be vsfa2drzpgw08qmj5diwk0x210bdimaa on current nixpkgs-unstable
<__monty__>
srhb: Hmm, for me it says it doesn't, for 6.5.0. Also, I have to use -f '<nixpkgs>' otherwise I get attribute nixpkgs not found.
<srhb>
__monty__: Ah, maybe yours is called nixos?
<robstr>
Hi, currently facing strange error with might not be related to nixos (works on my macbook). https://gist.github.com/rsoeldner/2fa9d905534c2dafa338ab6788a495d5 trying to run a executable - zsh tell me no such file or directory. `file` and `ls -al` output appended. some ideas ? :-/
<symphorien>
if you can build this executable from source, you will probably save you a lot of troubles
<robstr>
hm they get shipped my hyperledger fabric
<{^_^}>
[nixos-org-configurations] @zimbatm pushed to zimbatm/route53-zone « terraform: describe the nixos.org records in route53 »: https://git.io/fpNi9
<{^_^}>
[nixos-org-configurations] @zimbatm opened pull request #68 → terraform: describe the nixos.org records in route53 → https://git.io/fpNiQ
<{^_^}>
[nixpkgs] @globin merged pull request #52380 → [18.09] nixos/nextcloud: add basic module documentation and warn about current upgrading issues → https://git.io/fpNi4
<{^_^}>
[nixpkgs] @globin pushed commit from @Ma27 to release-18.09 « nixos/nextcloud: add basic module documentation and warn about current upgrading issues »: https://git.io/fpNPg
<{^_^}>
[nixos-org-configurations] @zimbatm pushed to zimbatm/route53-zone « terraform: describe the nixos.org records in route53 »: https://git.io/fpN1Q
agander has quit [Ping timeout: 244 seconds]
ckauhaus has quit [Quit: WeeChat 2.2]
<schopp0r>
hi. I was trying to build coq and proofgeneral (from the unstable channel). It aparently has autoconf as dependency, and autoconf fails to build because "GNU M4 1.4.6 or later is required; 1.4.13 is recommended"
<schopp0r>
is this a bug?
<symphorien>
is the sandbox enabled ?
<schopp0r>
symphorien: I guess not.
<symphorien>
retry with
<schopp0r>
symphorien: afair, the kernel doesn't support sandboxing or something
<fyuuri>
Hi. After updating one of my systems to 18.09 booting fails with the message "Failted to start File System Check". Does anyone know what the problem is?
<symphorien>
something related to fsck ?
<symphorien>
maybe run it by hand if a previous generation can boot
<fyuuri>
I tried. It did not help :(.
<symphorien>
journalctl should have details
graphene has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @bignaux opened pull request #52396 → ps2client: init at 2018-10-18 → https://git.io/fpNDx
<{^_^}>
[nixpkgs] @globin pushed to master « notmuch: use python 3 »: https://git.io/fpNye
<fyuuri>
journalctl says I should run fsck manually
kreisys has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
kreisys has joined #nixos
<{^_^}>
[nixpkgs] @globin pushed to master « Revert "notmuch: use python 3" »: https://git.io/fpNy4
<{^_^}>
[nixos-org-configurations] @zimbatm pushed to zimbatm/route53-zone « terraform: describe the nixos.org records in route53 »: https://git.io/fpNyB
<makefu>
i stumbled upon yet another dead source url in nixpkgs, are there any ambition currently to mirror them via some kind of service? right now i am essentially manually uploading sources to archive.org for the stuff i really would like not to worry about.
<jonaswouters>
after initial installation, why does it take so long to do nix-env -i vim?
<__monty__>
jonaswouters: Probably because they don't want you to install things imperatively on NixOS ; )
<__monty__>
jonaswouters: No, when you use a name like that nix needs to evaluate the whole of nixpkgs basically. That's why everyone recommends doing it by attribute `nix-env -iA vim`.
<samueldr>
,-iA
<samueldr>
hmm
<jonaswouters>
oh ok, I thought the A was to evaluate the whole nixpkgs, I'll try that thanks
<__monty__>
jonaswouters: You should normally be doing declarative package management on NixOS though.
<__monty__>
,declarative
<{^_^}>
There are multiple ways of managing declarative profiles. 1) Attrset, compatible with imperative use of nix-env https://git.io/fAQHW ; 2) buildEnv, providing more control over the paths that are linked into the profile https://git.io/fp0aU ; 3) home-manager, providing nixos-like config for your ~ https://github.com/rycee/home-manager
<__monty__>
Ah, nice.
<jonaswouters>
__monty__: that is what I', doing, I just setup a new system and trying stuff out, and I want vim! :D
<samueldr>
I often cheat and instead of (temporarily) installing to the profile I use `nix-shell -p vim` or whatever I need to try things :)
<__monty__>
Ah, if it's only temporary I'd use a nix-shell -p vim
<jonaswouters>
I actually linked you my git repo with the config ;)
<makefu>
jonaswouters: probably because drakonis i checked out softwareheritage but it seems like they are only mirroring a couple of large sites, not stuff like the VirtualBox sources or really anthing non-standard
<samueldr>
it feels like there's echo around here :)
chimay has joined #nixos
<jonaswouters>
ah haven't looked into nix-shell yet. Will add that to the todo
<LnL>
if the gcc build depends on stuff like bootstrap-tools it's part of that list, but you'll probably never have that on your system since you (hopefully) don't have to build that yourself
<LnL>
but you could filter out everything in nix-store -qR $(nix-instantiate '<nixpkgs>' -A stdenv) that's not part of nix-store -qR $(nix-build '<nixpkgs>' -A stdenv --no-out-link) to catch most
<nschoe>
okay.
<{^_^}>
[nixpkgs] @rycee pushed commit from @r-ryantm to master « jhead: 3.00 -> 3.02 »: https://git.io/fpNQR
<nschoe>
Can you point me to a direction to solve my issue: I have written a cabal file and turned it into a default.nix with cabal2nix. Then I have written a release.nix file that calls the default.nix to build it. This failed because _someone_ (and I can't find which deps) required doctest==0.13.*. But the doctest version is more recent. So I followed a how-to in order to override the doctest version. Now it complains because some other dep
<nschoe>
(still don't know which :/) now want doctest > 0.15. I'm wondering what's the correct way to go to build haskell packages ^^
<nschoe>
This is why I asked how to get the dependencies of a derivation. But unfortunately, what I get with nix-store -q --tree doesn't correspond to what hackage lists for the package, which is something that I don't understand either.
<nschoe>
__monty__, no I don't think it does, it fetches from Hackage.
<__monty__>
nschoe: ghc needs to know about packages to be able to compile them. Because of reasons you can't just specify alternative directories for ghc to look for them. That's why there's a ghcWithPackages that sets up ghc the right way to use nix dependencies.
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/fpN7N
<nschoe>
__monty__, no: when it failed to build because it could not find the doctest==0.13.* (which was the last err message), I use gabriel's method to override the doctest attribute.
chimay has quit [Read error: Connection reset by peer]
<__monty__>
Ah, so the problem is there's a conflict between the versions several dependencies need. One wants ==0.13 the other >0.15, those can't be reconciled. So you'll never fix that by picking a different version. The usual solution is to use `doJailbreak` in the override, which basically removes all version constraints in a package. Something like this: `doctest = doJailbreak (haskellPackages.callPackage ...);`
sigmundv_ has joined #nixos
<__monty__>
Uhm, except you wouldn't jailbreak doctest.
<__monty__>
You'd jailbreak one of the packages that depends on it.
<__monty__>
Or you could use `dontCheck` then it probably wouldn't need doctest. That'll keep tests from being run by nix while building.
<nschoe>
__monty__, I see, but about the doc test version, it does compile and work with cabal new-build. So it _is_ possible to build my package.
<nschoe>
okay, I'll try dontCheck too.
<__monty__>
Yes, cabal new-build only builds your current package, doesn't test anything.
<nschoe>
oh so this is just a test problem? I'll test dontCheck right now.
ambro718 has quit [Quit: Konversation terminated!]
<{^_^}>
[nixpkgs] @globin pushed commit from @r-ryantm to master « bwidget: 1.9.12 -> 1.9.13 »: https://git.io/fpNdg
drakonis1 has quit [Ping timeout: 252 seconds]
<nschoe>
__monty__, hub is my package ^^
<nschoe>
I don't use doctest in it.
<__monty__>
nschoe: You need to `dontCheck` every package that *does* use it.
<__monty__>
It's not transitive.
<nschoe>
Argh
kim0 has joined #nixos
<nschoe>
__monty__, so in my release.nix I have to override every package (like I did with doctest) by using dontCheck?
<eyJhb>
Anyone able to run a build on a newpackage PR?
<__monty__>
nschoe: Well, only the problematic ones.
<nschoe>
yeah sure.
<__monty__>
nschoe: Afaiu this it's because while nixpkgs does use stackage snapshots to base the packages on, it doesn't erase the version constraints, which means that even if packages are compatible "according to stack" the version bounds may still be in the way.
graphene has quit [Remote host closed the connection]
<nschoe>
I am disablign checks one y one, but it takes forever.
graphene has joined #nixos
<__monty__>
Yeah, way I do it is I override the haskellPackages my ghc comes from so I only have to do it once, gradually, rather than per project again and again.
<steveeJ>
samueldr: thanks! any chance there's one built from master?
<samueldr>
let me check, but I think there was an issue with the hydra built ones
<nschoe>
__monty__, at last! It finally managed to build! Cool. Thanks for sticking with me.
<__monty__>
nschoe: No problem, you did the actual work.
tzemanovic has joined #nixos
<nschoe>
__monty__, regarding these versions mismatch, is this bound to get better as time flows? Does it simply mean the nixOS snapshot is out of date?
<steveeJ>
samueldr: very nice, thanks a lot!
<__monty__>
nschoe: No. Do you know about the difference between stack and cabal?
<steveeJ>
`passthru.updateScript` is pretty exciting
tzemanovic has quit []
<nschoe>
__monty__, yeah, stack creates snapshot of inter-compatible packages AFAIU
tzemanovic has joined #nixos
<__monty__>
nschoe: Yeah, so cabal is solver based. It uses the constraints to find compatible package versions. Stack ignores the constraints.
agander_ has quit [Quit: Leaving]
<nschoe>
__monty__, but I thought nixOS snapshots were based on stack's?
<__monty__>
haskellPackages is basically the versions from the most recent stackage snapshot + latest versions from hackage for everything not in stackage.
tzemanovic has quit [Read error: Connection reset by peer]
<__monty__>
*But* all of these with their constraints.
Goldsum has joined #nixos
<__monty__>
And it uses cabal-install to do the actual building of packages afaik.
tzemanovic has joined #nixos
acarrico has joined #nixos
<__monty__>
So you can still get conflicts with constraints that can't be solved, even if all the packages are in the latest stackage snapshot.
<nschoe>
Hum... okay. This is weird: why keep the constraints?
<__monty__>
Cabal uses the solver approach so the constraints aren't going anywhere. That means you'll always have to deal with the conflicts in nixpkgs. However, I'm sure there's better technical solutions to come. Maybe pool resources on determining what packages need jailbreaking in the current nixpkgs.
<__monty__>
Because the stackage approach doesn't actually work. It doesn't mean packages are compatible, just that they build together.
<nschoe>
So this is a sort of stand-still? Nothing "really" works well, right?
<simpson>
Kind of. If one zooms out a bit, this kind of situation seems very endemic to Haskell but doesn't necessarily show up elsewhere in nixpkgs.
<simpson>
Stackage requires manual intervention by its maintainers in order to stay afloat. I figure that we do, too.
ambro718 has quit [Quit: Konversation terminated!]
<nschoe>
simpson, what I don't understand is: nixpkgs sort of solved this issue (barring Haskell packages). Why can' we apply the same principles to Haskell packges?
<simpson>
nschoe: Haskell is one of those ecosystems where the package-management language is *tighter* than the actual underlying API.
crmlt has joined #nixos
<immae>
Hey there, is it safe to remove manually old versions of /nix/var/nix/profiles/per-user/<user>/channels-*-link ? If not, how should I do to cleanup? (I couldn’t find anything in nix-store --gc or nix-channel)
<Goldsum>
Hey, so I was having trouble getting samba to work with nix. I am using nix on Gentoo as a package manager, and I am a little bit lost on which files I have to change and work with as oppose to on NixOS
<immae>
gchristensen: hmm it does not remove old versions of channels
<nschoe>
Hum, okay.
<simpson>
nixpkgs works because the APIs that programs usually use to communicate are either ancient and set-in-stone or extremely flexible about versioning. Meanwhile, the typical Haskell package narrowly insists that it will only work on a specific slice of Haskell compilers with the right version of `base` and `containers` and etc.
<__monty__>
simpson: Elm has a form of that too.
<gchristensen>
immae: check out its docs, like -d
<simpson>
__monty__: It's a popular idea! I want Monte to have something similar; the entire idea is that one should be able to *try* a package upgrade, automatically run all of the unit and integration tests, and then decide whether to commit or rollback.
<immae>
Ah right thanks, but I’d like to remove only the channels, not the other things
<__monty__>
Nix doesn't actually solve the problem with haskell dependencies though. It solves the problem of having two packages with incompatible dependency requirements.
<__monty__>
simpson: Wait, why wouldn't that be possible with haskell?
<buttbutter>
Hey, just getting stared with nix. I'm trying to launch polybar along with xmonad, but I'm not sure what the "nix" way of doing this is. I think on Arch I just launched it using .xinitrc.
<simpson>
__monty__: Totally possible but one has to check all of the versions and make them line up. Another way of looking at it is that writing a Haskell app entails writing a Haskell program and also a Cabal program.
<__monty__>
Goldsum: Mine's in ~/.config/nix/nix.conf. I expect it's actually using $XDG_CONFIG_HOME/nix/nix.conf to determine where to put it.
<Goldsum>
Oh, alright, that makes sense, I'll give it a look... thanks!
Lears has quit [Read error: No route to host]
<__monty__>
simpson: Well, with the approach you described you wouldn't have to check the versions though. If you consider your tests passing as the canonical source of truth. The bounds are supposed to be used by the dependency's maintainer to indicate actual, real incompatibilities.
<simpson>
__monty__: Sure. This is a kind of untyped-vs-typed situation, even, maybe!
<__monty__>
There's parallels for sure.
<simpson>
mdash: ^^^
<immae>
gchristensen: After looking at the code of nix-collect-garbage, it looks like `nix-env -p /nix/var/nix/profiles/per-user/<myuser>/channels --list-generations` seems safer to use for the specific task I needed (the other one will delete every generation of every profile/channel of every user, which is a bit too much)
<gchristensen>
great :)
<gchristensen>
if you run it as your own user, it can only delete your stuff
<immae>
Ah ok, thanks
jmeredith has joined #nixos
<nschoe>
What's the correct way to cross compile by .nix file? I tried `nix-build --arg crossSystem aarch64 -A hub release1.nix` but it complains "undefined variable 'aarch64'".
<gchristensen>
aarch64-linux ?
alex`` has quit [Read error: Connection reset by peer]
<nschoe>
gchristensen, same. Alsos the same with aarch64-multiplatform
<nschoe>
What do you mean? The package is in my workspace.
<__monty__>
nschoe: Looks like you need the '{ config = ...}' syntax, just aarch64 doesn't look like it should work.
<nschoe>
__monty__, do you have some examples?
<Unode>
hi all, do any of you know simple cgroups/sandbox software that allows running single binaries in a controlled environment with minimal fuss? Most cgroups interfaces seem overly complicated/cumbersome to use for single commands.
<__monty__>
nschoe: My only examples are at your link : ) I've never cross compiled with nix.
alex`` has joined #nixos
<__monty__>
Unode: I'm assuming nspawn is too cumbersome?
<nschoe>
__monty__, OMG how did I miss this... let me try
<Unode>
Looking for something similar to: `restricted_run --mem X --cpu Y command_to_run`
<Unode>
__monty__: looking... is this a systemd thing?
<__monty__>
Yes, systemd-nspawn, it's for containers.
<Unode>
__monty__: yeah, I don't want to be dependent on systemd
<__monty__>
Unode: Oh, I figured since you're in #nixos : ) Any cross-platform container technology then.
<Unode>
I found firejail but got mixed reviews due to most features being removed from their LTS release.
<Unode>
nsjail also showed up but still seems rather complicated (and unclear on how some things are supposed to work without root/suid)
<__monty__>
Afaik the roughly equivalent linux based solutions to bsd jails are containers.
tzemanovic has quit [Remote host closed the connection]
<Unode>
__monty__: yeah I guess what I'm after is a dumbded down interface to containers...
<Unode>
Also looking at both firejail and nsjail the memory related options seem to only act on virtual memory (which isn't more than 'ulimit -m'). Can containers restrict memory based on RSS instead of VMEM?
<Goldsum>
__monty__: When I make edits to my nix.conf, is there a good test I can put in there to see if it is reading from the file?
rawreraw has quit [Ping timeout: 250 seconds]
<__monty__>
Goldsum: Yes, just put ERROR at the top and run nix-env -q : )
<{^_^}>
[nixpkgs] @Amar1729 opened pull request #52408 → (wip) init infer at v0.15.0 → https://git.io/fpNNF
<Goldsum>
Oh, alright, thanks, I'll give that a shot
vk3wtf has quit [Ping timeout: 252 seconds]
graphene has quit [Remote host closed the connection]
<Goldsum>
Wew
<Goldsum>
Nice
<Goldsum>
Alright, it worked
graphene has joined #nixos
<sphalerite>
Unode: it might be possible, but given my understanding I highly doubt it's desirable, and would result in programs getting SIGKILLed by the OOM killer rather than having failing mallocs which can be handled gracefully.
petersjt014 has joined #nixos
tzemanovic has joined #nixos
<Unode>
sphalerite: the motivation is that some processes (for instance python, haskell, go, ...) tend to reserve much more vmem than they actually need and fail with memory errors quite far from the vmem limit set.
<sphalerite>
Unode: If they don't actually need that much memory they should try again with a smaller allocation AFAIK.
<Unode>
sphalerite: A quick test with python and large lists, for a vmem of 1.5G I only managed to get up to ~1G RSS before getting MemoryError.
<sphalerite>
but at least you get a MemoryError and not a SIGKILL :)
<Unode>
how many python programs do you know that can recover from a MemoryError exception? ;)
<arianvp>
how can I Figure out "why" machine and gitlab-runner are being rebuilt?
<Unode>
buttbutter: I used to use supervisord, but that's equivalent to systemd --user which is already in the system anyway. So I converted from .xprofile to systemd some months ago. It's a bumpy road, systemd isn't great for jobs that depend on X (and require $DISPLAY)
rardiol has quit [Remote host closed the connection]
random_yanek has joined #nixos
random_yanek has quit [Max SendQ exceeded]
Rusty1 has quit [Quit: Konversation terminated!]
random_yanek has joined #nixos
nD5Xjz has quit [Ping timeout: 268 seconds]
<nschoe>
hum, am I missing something: I have used flag `--arg crossSystem '{ config = "aarch64-linux"; }'`, but 1) I can run the resulting executable on my laptop, and 2) when I use `nix copy` on my raspberry pi, it complains about wrong executable format
orivej_ has quit [Ping timeout: 250 seconds]
<ktor>
hi, I'm new with nix/nixos however I have a need to build gradle project (i'm new to gradle as well) that's using embedded node/npm in one of it's tasks. I have searched a lot for an article about such combo but couldn't find one. Could anyone help me with that?
<infinisil>
nschoe: You should try the new cross thing via the pkgsCross attribute
nD5Xjz has joined #nixos
<infinisil>
nschoe: E.g. `nix-build -A pkgsCross.aarch64-multiplatform.hello`
<nschoe>
infinisil, yes, this works for building "standard" packages, but I have my own `release.nix` file, and I'd like to cross-compile it for rasPi
shapr has joined #nixos
<nschoe>
infinisil, what's the correct way, then? Should I add my package to the pkgsCross.aarch64-multiplatform set?
ottidmes has quit [Ping timeout: 250 seconds]
hedning has joined #nixos
<infinisil>
Hmm not sure, maybe look at how pkgsCross is implemented
<infinisil>
I guess you could do a standard nixpkgs overlay and then use pkgsCross.<your package> too
<run500>
here's an example: nix-shell -p jq --run 'sudo jq'
<nschoe>
run500, ho sorry, I thought you were looking for a way to run the command and --run did not work. I don't know about sudo :/
<nschoe>
run500, does it work when you simply enter the shell (i.e. not use --run or --command) an then from inside use `sudo yourCommand`?
<nekroze>
Trying to write a nix expression for the evelauncher https://pastebin.com/ASFJ03D2 which I have at least gotten to start up earlier versions of it in the past. Now it seems to need libz.so.1 so I have added it to prop build inputs and the library bundle path I am causing the launcher to load but it still cannot load libz.so.1 for some reason. Any suggestions on getting it to find the library?
ntd has quit [Quit: ntd]
<nschoe>
nekroze, I _think_ you need the -dev
<run500>
nschoe: doesn't work in an interactive shell either eithout --run or --command
<run500>
without*
<nschoe>
nekroze, I mean the ".dev".
kvda has joined #nixos
<nschoe>
nekroze, I had a similar issue with Haskell dev env, and I think I had to add both `zlib` and `zlib.dev` in the `buildInputs`. Can you try this?
<nschoe>
run500, okay, then something more fundamental must be going on. For instance `nix-shell -p hello --command "sudo hello"` work on my machine
nD5Xjz has joined #nixos
orivej has joined #nixos
<run500>
nschoe: i see, just did some more testing and it works on a nixos machine but not with nix installed on gentoo or ubuntu
<nschoe>
run500, I have to go, but a last thing you might try:
<nekroze>
nschoe: oh is that a different room here on freenode?
rprije has joined #nixos
<nschoe>
run500, as a normal user, in Ubuntu you probably had to add some command to source the nix.sh or smth similar
jasongrossman has joined #nixos
<nschoe>
run500, maybe you can try to add this into root's .bashrc too. This way _maybe_ when you run the sudo command it also has the nix.sh sourced and can figure out path?
<nschoe>
run500, other than that, I'm clueless :/
<Mic92>
nschoe: sudo is resetting your PATH
<nschoe>
Mic92, ah right.
<Mic92>
you have to prepend ~/.nix-profile/bin
<nschoe>
Mic92, also I _think_ there's a way not to
<nschoe>
Isn't that with -E or smth ?
<Mic92>
but sometimes sudo does not allow -E
<nschoe>
Mic92, yeah... well in this case, I don't know :/
<Mic92>
you can also do someting like nix-shell -p jq --run "sudo env PATH=$PATH jq"
<qyliss>
Can I change the -p value for a patch applied to a derivation?
<Mic92>
qyliss: patchFlags
<qyliss>
cheers
orivej has quit [Ping timeout: 250 seconds]
<run500>
thanks for the ideas
<nschoe>
Anyway, I have to go, run500 look at Mic92's answers, they are more educated than mine :)
<nschoe>
Cheers.
nschoe has quit [Quit: Program. Terminated.]
<nekroze>
nschoe: ah I see, a package suffix, thanks!
<qyliss>
Mic92: oh, but that applies to all patches, right? I can't just set it for a single patch?
orivej has joined #nixos
<Mic92>
qyliss: in that case you could also run patch manually in postPatch.
<qyliss>
Cool, I'll do that. Thanks.
<Mic92>
or fix the patch if it is local
Ariakenom_ has quit [Quit: Leaving]
<Mic92>
just found out that fetchpatch also allows that
<Mic92>
There is a stripLen argument.
<qyliss>
Oh, nice! That's exactly what I need
hamishmack has joined #nixos
<shapr>
clever: any idea why ubuntu installer would boot but the latest nixos graphical installer hangs?
<samueldr>
one of the issue one will face with implementing another init with nixos options is how many options are simply pass-throughs to systemd concepts
<samueldr>
(AFAIK)
<Goldsum>
So did you build it with a NixOS base, then transfer it over to s6?
<qyliss>
Yeah
<Goldsum>
That's really cool :)
<samueldr>
qyliss: it must have helped that systemd isn't in stage-1, right?
<qyliss>
samueldr: I'm content to just implement services myself
<qyliss>
samueldr: sorta. It actually _is_ in stage1
<__monty__>
samueldr: Isn't policykit a freedesktop project too? That's all systemd afaic.
<qyliss>
Because stage1 needs udev
<Goldsum>
I have been using Gentoo, and then my plan was to manage many of the packages with Nix
<samueldr>
it's a security problem that affects systemd, but not systemd's own issue; here I hate going tangentially there, but it's always FUD with systemd, either it implements everything or everything external it uses is now its problem right?
<qyliss>
eg I have elogind, and can run Wayland (I think - didn't get round to actually starting it last time I was working on it)
<qyliss>
but it's been a little while since I touched it
<__monty__>
samueldr: Isn't all of the freedesktop project developed by the same set of maintainers for the most part? Systemd is just metonymy it's not FUD.
<samueldr>
qyliss: if it was only about including another config at boot in the menu, you may be interested in #45345