GolDDranks has quit [Remote host closed the connection]
<cole-h>
Third time's the charm -- any committers wanna merge a firefox-beta-bin bump so I can stop using the impure overlay (for now)? https://github.com/NixOS/nixpkgs/pull/96447 ❤️
<aranea>
Is there a way to get the nix daemon to perform one particular build in a different place than /tmp?
marcusr has quit [Remote host closed the connection]
marcusr has joined #nixos
<aranea>
My /tmp is a tmpfs, and normally I'm happy to build there, but right now I'm building an exceptionally large package and find myself running out of space.
knerten2 has joined #nixos
dbmikus has joined #nixos
knerten1 has quit [Ping timeout: 240 seconds]
gustavderdrache has quit [Quit: Leaving.]
<roconnor>
man nix-store suggests:
<roconnor>
TMPDIR
<roconnor>
Use the specified directory to store temporary files. In particular, this includes temporary build directories; these can take up substantial amounts of disk space. The default is /tmp.
<roconnor>
man nix-store suggest setting TMPDIR environment variable.
rajivr has joined #nixos
buckley310 has joined #nixos
<aranea>
I already tried setting TMPDIR in nix-build's environment; it indeed placed some temporary files there, but the build directory itself stayed in /tmp.
<aranea>
I'd probably need to pass a TMPDIR variable to nix-daemon itself, but I don't want to make global changes for a temporary workaround.
<evanjs>
Soooo are environment variables automatically propagated to the cargo build phase for buildRustPackage
<evanjs>
?
<evanjs>
I would imagine not _every_ env variable bubbles up automatically. Anyway, trying to grab an env variable during build, and it doesn't seem to be picking it up from my environmeny
<{^_^}>
Found in packages: glib, ibus, grocy, gnulib, polkit, tt-rss, scribus, shaarli, glib.dev, gtk3-x11, appstream, limesurvey, boost-build, appstream-qt, deepin.go-lib, logstash6-oss, logstash7-oss, appstream-glib, gnome_doc_utils, scribusUnstable, gettextWithExpat, libaccounts-glib, ibus-with-plugins, rubyPackages.i18n, gettextWithExpat.doc, rubyPackages_2_5.i18n, rubyPackages_2_7.i18n, python37Packages.sphinx, python38Packages.sphinx, and 4 more
<pjt_tmp>
oooooooh
<pjt_tmp>
thats why >_<
<evanjs>
I typically try `nix-locate -w1` if I need to filter out all the "probably not"s :P
<evanjs>
Also wondering what issue you encountered based on that response tho lol
moet has joined #nixos
SanchayanMaity has joined #nixos
<pjt_tmp>
evanjs: tryna build an armv6l install iso
<pjt_tmp>
and gettext is a PAIN to get to compile
<evanjs>
based on my limited experience with nixos on arm... I'm sorry lol
<pjt_tmp>
eh, it's not bad when it works
<evanjs>
Yeah, depends on your hardware and etc, I guess
<pjt_tmp>
there's this real fun problem with child processes where sometimes one of them will just....hang. Like indefinitely.
<pjt_tmp>
I think it's a bug
<pjt_tmp>
I can *sometimes* 'fix' it
<evanjs>
I was trying to get it working on a wandboard before. And they're pretty flaky when it comes to support re "what driver am I supposed to use" and etc
<pjt_tmp>
by sending the process a sighup
<evanjs>
though it might more be my particular use case (WiFi in master mode), but whatever
<pjt_tmp>
never tried one o' those before
<pjt_tmp>
pi zero and 3b over here
maier has joined #nixos
<pjt_tmp>
evanjs: think I could reliably automate the signal sending part and get any benefitout of it?
<pjt_tmp>
I only gotta do this once, then I'm pinning gettext till the end of linear time
<cole-h>
Without knowing the exact "somethings" involved, only a guess: one was modified locally or is otherwise not cached, while the other is cached.
<Miyu-saki>
cole-h: The specifics are in the gist.
<cole-h>
(Which you posted milliseconds before I responded; looking.)
<Miyu-saki>
(I run a patched Linux, so I know that this is not available on any substituter)
moet has quit [Quit: leaving]
<cole-h>
I have two guesses. 1) You previously built Linux 5.7.14, and thus it's already in the store. 2) anbox-modules isn't a default kernel module and thus still needs to be built.
zuserm has quit [Ping timeout: 244 seconds]
<pjt_tmp>
Miyu-saki: you sure it's exactly the same? hashes match?
<Miyu-saki>
Yeah, you can highlight over the `/nix/store/path.drv` and then find it in the page
<Miyu-saki>
Or even
<cole-h>
Hashes look the same to me (at least the one for `linux-5.7.14`)
<cole-h>
And obviously the anbox-modules hash will be the same because only the drvs are printed (and not the final outpath)
dbmikus has quit [Ping timeout: 258 seconds]
<Miyu-saki>
I'm wondering if my /nix/store's metadata broke?
<Miyu-saki>
I have a feeling if I remove my /run/current-system link, I'll be forced to rebuild linux again
<cole-h>
Definitely don't do that
<Miyu-saki>
And I'd rather not do that, since it takes ~2 hours.
<Miyu-saki>
This one also causes a rebuild of the kernel
<cole-h>
Miyu-saki: So, what's the actual problem? Tbh, that all seems fine to me... If your kernel is customized, you'll obviously need to build kernel modules to match your changes
vv has joined #nixos
<Miyu-saki>
cole-h: The problem is that /nix/store/10ag6plmx44bfdlab72gx62db9wvp3bm-linux-5.7.14.drv is already built, but if I build a package depending on it, /nix/store/10ag6plmx44bfdlab72gx62db9wvp3bm-linux-5.7.14.drv gets rebuilt
<cole-h>
Does it actually get rebuilt, though?
<Miyu-saki>
Yeah, what I sent recently spits out a huge log
<Miyu-saki>
The (mkShell one)
<cole-h>
I more meant the `nix-build /nix/store/....-anbox-modules....drv` command
<Miyu-saki>
Yeah, that should be the same thing.
mariatsji has quit [Ping timeout: 240 seconds]
<Miyu-saki>
Actually, I know something which causes it to rebuild.
<Miyu-saki>
If I change my /etc/nixos/configuration.nix to *anything*
<Miyu-saki>
It will cause a rebuild
<cole-h>
what
<Miyu-saki>
Of the kernel
<cole-h>
Sounds like you have some impurity then
<cole-h>
That shouldn't happen, I don't think.
<Miyu-saki>
Hm
zuserm has joined #nixos
<Miyu-saki>
I can't exactly `nix-store --delete --force` the kernel either lol
<Miyu-saki>
This is
<Miyu-saki>
a bit confusing
<clever>
Miyu-saki: --force has major side-effects, avoid it whenever possible
<Miyu-saki>
Yeah
<Miyu-saki>
I once accidentally fubar'd my system, luckily it's easy to recover with the livecd.
<dsal>
What's the easiest way to run a really dumb web server? I feel like I might need to write one. :/
<clever>
dsal: services.nginx.enable = true;
<Miyu-saki>
dsal: There's a python runnable module to just host static files
<Miyu-saki>
python -m http.server IIRC?
<Miyu-saki>
clever: Any thought on my gist? Linux should already be built. If I use the kernel as a buildInput, same thing.
<pjt_tmp>
dsal: pshs?
<dsal>
Miyu-saki: Yeah, I do that on the commandline sometimes. I just need a dumb service.
<clever>
Miyu-saki: what was the error?
night has quit [Ping timeout: 240 seconds]
<Miyu-saki>
clever: No error. linux.drv has already been built. If I build something which directly depends on linux.drv, linux.drv is forced to rebuild.
<Miyu-saki>
I don't think this is __impure related either, because linux.drv would be rebuilt if I call `nix-build` on it
<clever>
Miyu-saki: was linux.drv based on an override?
<Miyu-saki>
I haven't tried letting the full build process go, because it'd take ~2 hours.
<Miyu-saki>
clever: I used ` kernelPatches = [{name = "Add fan control for the L440"; patch = ./L440-fan-control.patch;}]; `
<Miyu-saki>
boot.kernelPatches
<clever>
Miyu-saki: you probably garbage collected one of the outputs of the kernel
<clever>
and the new drv needs that output
<Miyu-saki>
Ah
<clever>
and the cache doesnt have it
<Miyu-saki>
Yeah, that makes sense.
<clever>
so nix has to rebuild all outputs
zuserm has quit [Remote host closed the connection]
<Miyu-saki>
Since I just recently garbage collected
<Miyu-saki>
Yeah, thanks. That's 100% it.
<Miyu-saki>
`nix-build` only builds the first output, right?
<clever>
it builds all i believe
<Miyu-saki>
Hm
NightA has joined #nixos
sangoma has joined #nixos
spudly- has joined #nixos
spudly- has joined #nixos
spudly- has quit [Changing host]
<{^_^}>
[nixos-homepage] @github-actions[bot] pushed commit from GitHub Actions to master « Update flake.lock and blogs.xml [ci skip] »: https://git.io/JUmlv
<{^_^}>
[nixpkgs] @zowoq merged pull request #96891 → ipfs-cluster: pin to go 1.14 → https://git.io/JUmBQ
<{^_^}>
[nixpkgs] @zowoq pushed commit from @AluisioASG to master « ipfs-cluster: pin to go 1.14 »: https://git.io/JUm0G
<{^_^}>
[nixpkgs] @zowoq merged pull request #96837 → nixos/cri-o: unset hooks dir to avoid dir creation on startup → https://git.io/JUqXS
<{^_^}>
[nixpkgs] @zowoq pushed commit from @saschagrunert to master « nixos/cri-o: unset hooks dir to avoid dir creation on startup »: https://git.io/JUm0C
cfricke has joined #nixos
sgrunert has quit [Remote host closed the connection]
lopsided98 has quit [Quit: No Ping reply in 180 seconds.]
<{^_^}>
[nixpkgs] @vbgl pushed commit from @marsam to master « ocamlPackages.merlin: 3.3.6 -> 3.3.8 »: https://git.io/JUmuT
ris has quit [Ping timeout: 246 seconds]
lopsided98 has joined #nixos
<Reventlov>
Hey there.
n1tsu has quit [Ping timeout: 256 seconds]
mariatsji has joined #nixos
<Reventlov>
What would be the best way to give access to a nginx directory to another user? Right now, I add the other user into the nginx group, and do a preStart with a chmod g+rwx on the directory (which is not too deep, thanksfully). Is there a cleaner method ?
<{^_^}>
[nixpkgs] @peti pushed 2 commits to haskell-updates: https://git.io/JUmzf
<teto>
Upon running 'nix shell poetry#poetry', I get cannot write modified lock file of flake 'flake:poetry' (use '--no-write-lock-file' to ignore) . Adding --no-write-lock-file does solve the issue but why does nix want to modify the lock file ?
<superherointj>
Hello. How do I configure a set up like this? LUKS -> LVM -> LUKS. Basically I need to open a LUKS encrypted disk but to not mount it and keep order of actions.
fendor has quit [Remote host closed the connection]
<Mic92>
superherointj: noauto as mount option should do the trick.
<Ke>
so reducing your nixos problem to a systemd problem is trivial, you just need to read up on systemd dependencies
<chiiba>
To any Hashicorp Vault gurus here: Can I create a policy to allow reading a specific secret engine path? Or do I have to create new secret engine and seperate secrets between engines to mangage ACLs?
<Ke>
like requires, after, wants, before
<superherointj>
Mic92, what is the full option name?
<superherointj>
Tried option `noauto` and did not decrypt the volume. Then tried to mount it as test errored as `mount: /mnt/crypt_aux_ssd: unknown filesystem type 'crypto_LUKS'` (makes sense). What I might be looking for is the equivalent of crypttab in NixOS.
D_ has joined #nixos
cosimone_ has joined #nixos
cosimone has quit [Read error: Connection reset by peer]
<Raito_Bezarius>
(well not that broken because it did deploy a node on my Proxmox cluster \o/)
stteevveen has joined #nixos
<Raito_Bezarius>
Thanks to everyone I pestered with my questions
<arianvp>
sgrunert: I have a question about cri-o
<arianvp>
Why is conmon a thing? Why do podman and cri-o use it to double fork and monitor the cgroup; instead of just running "systemd-run <command here>" to hook it into systemd's cgroup hierarchy directly?
cosimone_ has quit [Remote host closed the connection]
<arianvp>
It seems like a reinvention of something that is already shipped in systemd anyway. and then we don't need to jump through hoops when running podman in a systemd service because we can just use Type=simple no?
cosimone_ has joined #nixos
<arianvp>
i've been trying to look for an answer for this; or a motivation; but i cant find it
<stteevveen>
Hi! Not really a Nix question, although I need this for trying to package an app. Do you know how to access the list of archives exposed by github ? Either by browser and/or commandline ? I succeed doing wget -O - https://github.com/project/user/tarball/master | tar xz so as to get the latest, but I have not found how to list everything in
<sgrunert>
arianvp, I'm not that familiar with systemd-run. I know the initial plan for conmon was to keep it very light, and only keep track of the running process.
<arianvp>
systemd-run basically creates a temporary systemd service; and then systemd monitors the cgroup and the process
<sgrunert>
I'm not sure if it was just a decision about having the flexiblity to change it in an easy way, or if there were other technical aspects. I'll raise the question to the Red Hat friends in the Kubernetes slack.
<arianvp>
(But yeh; it does require you to run systemd; so I guess that'd be a good argument to not use it in some cases; but I think having more tight systemd integration is very useful from a system-administrator perspective)
D_ has quit [Ping timeout: 240 seconds]
<sgrunert>
arianvp, OTOH we recommend to use the systemd cgroup driver (instead of cgroupfs), the same applies to the Kubelet. So having "no need for systemd" is not that valid argument any more.
dsg has quit [Ping timeout: 260 seconds]
<arianvp>
yeh that was my exact thought. if you're requiring the systemd cgroup driver anyway; why not let systemd monitor the process directly; instead of conmon. cut out the middleman
<arianvp>
I might give a shot at comming up with a PR for CRI-O to see if it's a feasible idea
<sgrunert>
Sounds like a plan, thank you arianvp. If we look at the sources of conmon we have to admit that the complexity increased over time…
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
zupo has joined #nixos
drgbr has quit [Ping timeout: 260 seconds]
<stteevveen>
Self-answering my question, in anycase someone is interested in how to do this: https://api.github.com/repos/project/user/releases will return a JSON with all available releases. Consequently It can be retrieved with curl, or wget on the command line. Adding a jq filter may help to get just the needed information.
kreyren_ has quit [Remote host closed the connection]
kreyren has joined #nixos
criptonauta_ has joined #nixos
zupo has joined #nixos
cript0nauta has quit [Read error: Connection reset by peer]
Blackraider has joined #nixos
Thra11 has quit [Ping timeout: 240 seconds]
Blackraider has quit [Remote host closed the connection]
D_ has quit [Ping timeout: 240 seconds]
<gentauro>
is it possible to `hot swap` the Linux kernel in NixOS (in other words, swap kernel without rebooting)?
<symphorien[m]>
short answer: no. Long answer: there is kexec, but it is not much better than rebooting.
D_ has joined #nixos
<gentauro>
symphorien[m]: roger that. thx.
dsg has joined #nixos
Extends has joined #nixos
<Ke>
in general nixos is quite the same as other distros in this regard, there are some hotpatch solutions that create loadable modules or kprobes or whatever that change kernel behaviour on the fly to fix problems or similar
mniip has quit [Quit: This page is intentionally left blank.]
cosimone has joined #nixos
<{^_^}>
[nixpkgs] @kevincox pushed 2 commits to kevincox-phodav-udev: https://git.io/JUmP3
<georgyo[m]>
ingenieroariel: You may want to look at https://github.com/catern/nix-utils specifically the tarball. It creates a self-extracting nix install.
vykook has joined #nixos
<georgyo[m]>
catern: you really need to write a blog post (or some docs) about your magical install anywhere nix installer.
stteevveen has quit [Remote host closed the connection]
<c4droid>
Hi, I added gnome-tweaks to my system packages, when I execute `nixos-rebuild switch' give the undefined variable 'gnome-tweaks' error. `nix search' can search that package, just in reconfigure system report that.
<steveeJ>
c4droid: you probably missed the "gnome3". it's in <nixpkgs>.gnome3.gnome-tweaks
<Extends>
the package is located in gnome3.gnome-tweaks
<truby>
I'm having trouble with steam, it starts fine but can't connect to the steam network. anyone else seeing this issue? Steam from flatpak works fine weirdly
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<{^_^}>
[nix] @edolstra pushed to master « github: Use access token when calling .../commits API »: https://git.io/JUmMa
<pheoxy>
so I'm trying to build a raspberry pi image and I keep getting "error: Package ‘raspberrypi-firmware-1.20190925’ in /nix/store/5q1ss9jz0b10kvgmcpf75ahbsabncvjq-nixos-20.03.2868.ff6a070b4ef/nixos/pkgs/os-specific/linux/firmware/raspberrypi/default.nix:20 is not supported on ‘x86_64-linux’, refusing to evaluate."
<pheoxy>
I have "nixpkgs.config.allowUnsupportedSystem = true;
<pheoxy>
boot.binfmt.emulatedSystems = [ "aarch64-linux" ];" set in configuration.nix
<pheoxy>
I'm running "nix-build '<nixpkgs/nixos>' -A config.system.build.sdImage -I nixos-config=./sd-image.nix"
<truby>
hmm. Steam has let me on now after a few hours of trying. No idea what is going on or if it will fail again with the same error :')
<ingenieroariel>
pheoxy it looks like this line would allow your system to cross-compile, which is good: boot.binfmt.emulatedSystems = [ "aarch64-linux" ];" s
<ingenieroariel>
but I am not sure this line "nix-build '<nixpkgs/nixos>' -A config.system.build.sdImage -I nixos-config=./sd-image.nix" is telling nix-build to cross-compile
<pheoxy>
ingenieroariel: do I need to define the emu system somewhere?
zupo has joined #nixos
<ingenieroariel>
do you have this line: <nixpkgs/nixos/modules/installer/cd-dvd/sd-image-aarch64.nix> in your configuration.nix ?
<ingenieroariel>
err - sd-image.nix
<pheoxy>
no
<pheoxy>
got it
<pheoxy>
missing arg
<ingenieroariel>
ok, next I was going to suggest was --argstr system aarch64-linux
<pheoxy>
"--argstr system aarch64-linux"
<ingenieroariel>
ha
<pheoxy>
thanks ingenieroariel
<pheoxy>
nice
dbmikus has joined #nixos
c4droid has left #nixos ["ERC (IRC client for Emacs 27.1)"]
<ingenieroariel>
as you add packages, notice that some will not cross compile correctly, even though they will build fine on native compilation
<ingenieroariel>
so try to use established branches to make it more likely that you can just download the artifacts for most aarch64 packages
<ingenieroariel>
if you go crazy with a custom nixpkgs, then try to just compile the minimum and finish on the target system
orivej has quit [Ping timeout: 260 seconds]
<pheoxy>
weird, why's that
zupo has quit [Ping timeout: 240 seconds]
<evanjs>
Anybody have an idea as to why nixUnstable is printing some items twice?
<evanjs>
This is basically the command I'm using: `nix run nixpkgs.nixUnstable --command nix-build -E "(with import <nixpkgs> { overlays = [ (import /home/<path>/nixos/overlay/overlay.nix ) ]; }; callPackage ./"${nix_file}" {$args} )" --show-trace`
zupo has joined #nixos
<evanjs>
For some reason, various commands are printed twice
dbmikus has quit [Ping timeout: 265 seconds]
<evanjs>
Well, messages, whatever
kayg04 has joined #nixos
fnlaai has joined #nixos
<evanjs>
e.g. "Compiling zip v0.5.6" (in the context of `cargo build`)
<allgreed>
howdy, I'm trying to use nix in an unprivilaged containerized CI environemnt and have some trouble with the generated docker image (via `dockerTools.buildImage`) - I cannot simply do `docker load` since the container doesn't have required privilages (namely cannot create mount namespaces, but guess that's just the latest error and if I solved that
<allgreed>
there would be more). But I can do `docker build`. Is there an easy way for nix to generate a Dockerfile instead of producing the image? Or maybe you use nix in a similar environment and have a workaround for the `load` problem ;)
lawr3nce has quit [Remote host closed the connection]
lawr3nce has joined #nixos
gustavderdrache has joined #nixos
<steveeJ>
allgreed: are you sure the mount namespace is the issue? you can verify that with `unshare -m`
<allgreed>
steveeJ: I got this error message "Error processing tar file(exit status 1): Error creating mount namespace before pivot: operation not permitted" when attempting to load, so pretty sure
criptonauta_ has quit [Ping timeout: 240 seconds]
<allgreed>
and as I've mentioned it's not just the mount ns - I'm atemtping to run docker in docker without the `--privilaged` flag or mounting the socket and afaik that's a no-go
cript0nauta has joined #nixos
<steveeJ>
I've not tested any of that. could you try podman instead?
jumper149 has joined #nixos
lawr3nce has quit [Ping timeout: 240 seconds]
zuserm has joined #nixos
<zuserm>
does anyone know where I can get the command "sftp-server"? It's expected by a fork of proton I'm trying to build, and google is unhelpful
erasmas has joined #nixos
<infinisil>
,locate bin sftp-server
<{^_^}>
Found in packages: lsh
<zuserm>
thanks!
<infinisil>
> :p lsh.meta
<{^_^}>
{ available = true; description = "GPL'd implementation of the SSH protocol"; homepage = "http://www.lysator.liu.se/~nisse/lsh/"; license = { fullName = "GNU General Public License v2.0 or later"; sho...
<infinisil>
Hmm
<maralorn>
Is there a a time of day when the branch-off will happen on friday?
<{^_^}>
[nixpkgs] @atlaua opened pull request #96912 → nixos/qemu-vm: Fix and update system.requiredKernelConfig entries → https://git.io/JUm73
ingenieroariel has joined #nixos
sangoma has quit [Ping timeout: 256 seconds]
<numkem>
Are systemd units unable to reach the nix store? I want to do a `docker load` on start through a oneshot service and I'm getting a `no such file or directory` while the path in the nix-store does exists
hiro99 has joined #nixos
jonringer has joined #nixos
dbmikus has joined #nixos
<makefu>
numkem: could you run `ls -la` instead of `docker load` for debugging? normally /nix/store is available to all services so my guess would be that what you think is an image is actually a directory or that the systemd service is somewhat restricted to access the store
<gleber>
Found an interesting research project: https://sci-hub.tw/10.1007/978-3-030-00244-2_16 . Using Nix there would have saved them a lot of effort, but their idea of using whole-program optimization for libs/bins placed into a container.
hiro99 has quit [Remote host closed the connection]
hiro99 has joined #nixos
evelyn has joined #nixos
<evelyn>
Hi
<evelyn>
I'm having a really hard time trying to work out how I can get sox to build with LAME support on my system.
<evelyn>
I have sox installed as a system package. I would like to install it with MP3 support; there's enableLame in the buildInputs in the expression, but I am not sure how to get this into my system's configuration.nix
<evelyn>
The cheat sheet isn't very helpful in this respect. Does anybody know?
<{^_^}>
[nix] @domenkozar opened pull request #3981 → Allow HTTP binary cache to request absolute uris → https://git.io/JUm5C
<simpson>
evelyn: Currently seeing if `sox.override (_: { enableLame = true; })` will work.
civodul has quit [Ping timeout: 244 seconds]
<simpson>
(There does appear to be one example of using `override` in https://nixos.wiki/wiki/Cheatsheet but it's buried deep and not obvious.
<simpson>
Ugh, stupid OCD.)
<ToxicFrog>
I think .override just takes a map, unlike .overrideAttrs which takes a function? So something like
<zuserm>
so, I noticed that nixpkgs has 2.2k pull reqs, whic makes me wonder how long it usually takes to get a pull req merged?
superherointj has joined #nixos
alp has quit [Ping timeout: 246 seconds]
<Raito_Bezarius>
Is there a recent docs on how to set up remote building on non-NixOS machines?
sgrunert has quit [Remote host closed the connection]
<maralorn>
zuserm: That depends a lot on the PR. If it's complex and touches a part that no one feels really responsible for, it might take quite a while. If it is simple and or touches a part that people that have time care about it can go very quickly. On average the high number of PRs is more caused by a lot of PRs, most of them get dealt with quickly.
quinn has joined #nixos
jonringer has joined #nixos
<evanjs>
jonringer: finally we can bug you for PRs here lol
o1lo01ol1o has joined #nixos
<ingenieroariel>
zuserm as an outsider I think nixpkgs is a lot easier to get things into than other distros, here are some of my experiences:
<ingenieroariel>
tegola: a simple golang app with no other dependencies, accepted without changes, very fast (days turnaround)
<simpson>
zuserm: I have a graph that I like to show folks for this. Imagine that the number of open PRs scales with the number of packages that we have. We have more packages than other ports trees: https://repology.org/graph/map_repo_size_fresh.svg
dbmikus has joined #nixos
<ingenieroariel>
hydra (an OAUTH library with massive dependencies in golang) was rejected becasue it was too big and of limited use (in a few days too)
<zuserm>
nice
<ingenieroariel>
apache airflow (took a couple of weeks and a ton of work, then went in and broke several times over the next months causing several people to step in, because it has a massive number of python dependencies)
<ingenieroariel>
so zuserm i'd say if you are asking because you have a specific package, not all PRs are created equal zuserm and what simpson said
MtotheM1 has joined #nixos
<zuserm>
I don't have a specific package, I'm just getting into nixos, and I've been trying to convince a friend to join me, so I was curious.
waleee-cl has quit [Quit: Connection closed for inactivity]
alp has joined #nixos
ingenieroariel has quit [Ping timeout: 245 seconds]
<evanjs>
any rust peoples here: is there some weird behavior when building projects with include_bytes!() ? It's almost like it's trying to parse the file, as if I had used include!() or something
werner291 has quit [Ping timeout: 246 seconds]
mariatsji has quit [Remote host closed the connection]
<evanjs>
not found"e.g. I'll get messages like: "/nix/store/cnyylxxm43dajck44dagpaiba89vj6sp-rust-1.46.0-2020-08-24-04488afe3/bin/cargo: 1: ELF
MtotheM1 has quit [Quit: Leaving]
<evanjs>
omg weechat lol
mariatsji has joined #nixos
mariatsj_ has joined #nixos
mariatsji has quit [Remote host closed the connection]
leibniz[m] has joined #nixos
bqv has quit [Quit: WeeChat 2.9]
<leibniz[m]>
Can I use NixOS as an unbreakable source-based distro? (So I could have the latest versions of the few apps I use?)
<jumper149>
leibniz[m]: I guess yes, but what exactly do you mean with latest version?
bqv has joined #nixos
sascha has joined #nixos
sascha is now known as Guest48841
<leibniz[m]>
jumper149: In source-based distros like Gentoo, I can just compile the app myself. But in Ubuntus, people say that doing this would mess with OS libraries
Guest48841 has quit [Client Quit]
<jumper149>
Because nixpkgs-unstable is a rolling release channel and updates are quick.
orivej has joined #nixos
<evanjs>
mmkay, note to self, consider rust via nix on WSL1 unstable lol
<leibniz[m]>
jumper149: So I would still need to wait for nixpkgs-unstable to provide packages?
<jumper149>
leibniz[m]: You don't need to use only one channel. You can use multiple channels and use whatever you want from anywhere.
<jumper149>
Nix will install different versions of dependencies.
mariatsj_ has quit [Remote host closed the connection]
<simpson>
leibniz[m]: Libraries are only in different directories. It's no slower than when one has libraries in both /usr/lib and /usr/local/lib.
xcmw has joined #nixos
<V>
It's probably faster in a bunch of cases since there's more hardcoding going on here
mariatsji has joined #nixos
<jumper149>
In fact as long as you stick to a single channel you will have dependencies shared across all applications.
<leibniz[m]>
Can I have Nix Package manager in Ubuntu?
<jumper149>
Yes.
<cole-h>
Yep.
<simpson>
V: Uh? How would that work? The libraries are still dynamically linked, and the actual loading and runtime linking of code objects is what is slow, not checking (cached) inodes.
<V>
simpson: a lot of places use hardcoded paths instead of looking up stuff in $PATH
<leibniz[m]>
jumper149: So why everyone is using snaps, appimage, etc?
kuznero has joined #nixos
<leibniz[m]>
I mean what's the catch? Is it hard to set up?
<simpson>
V: I'd have to see measurements to be convinced, sorry. But it's a fun idea for designing an experiment.
<cole-h>
The opposite. Snaps is probably easier to setup than Nix or NixOS.
<jumper149>
leibniz[m]: They just got popular. Why are so many people using Windows?
<V>
simpson: I doubt it makes much of a difference in practice
orivej has quit [Ping timeout: 240 seconds]
mariatsji has quit [Read error: Connection reset by peer]
mariatsji has joined #nixos
<kuznero>
Hi all! Where can I check why latest vscode 1.48.2 is not getting landed into stable? It is about 6 days passed since it was merged. Normally things go slightly faster...
<{^_^}>
#96479 (by doronbehar, 5 days ago, open): nixos-unstable is blocked because of failing luks-format1 test
<kuznero>
I see, thanks
bennofs__ has quit [Ping timeout: 260 seconds]
<immae>
Is there a "zfs mount -a" or equivalent executed automatically during boot? I heard about such thing but I cannot see anything that confirms it in stage 1 or stage 2 scripts
maier has joined #nixos
<immae>
(I’m thinking of "non-legacy" mounts here, legacy mounts are correctly declared in the fstab)
smyds has quit [Remote host closed the connection]
<kuznero>
But if it is blocked, does it make sense to not to upgrade that luks-format1 thingy?
<V>
kuznero: that's already in master, there's a proposed PR to revert the offending commit
<kuznero>
Super! Thanks!
<V>
hydra builds happen after merges here .-.
<V>
so master can be broken
<V>
having master broken for up to ~2 weeks is not uncommon
maier has quit [Ping timeout: 240 seconds]
<ToxicFrog>
immae: it's meant to, but as a systemd unit, not as part of the boot scripts
<{^_^}>
rust-lang/rust#75075 (by Shnatsel, 4 weeks ago, open): There is no cross-platform way to use `include_bytes!`
<ToxicFrog>
in my experience this happens too late, so I have a snippet in `boot.postBootCommands` that deletes all the stuff the activation script created in not-yet-mounted mountpoints, then does `zpool import` and `zfs mount`
<ToxicFrog>
(I wrote this a while ago; it's possible that the activation script stuff that creates directories wrong has been fixed by now)
<immae>
ToxicFrog: Yes I saw that kind of warning in the wiki, I’m just trying to get a grasp about how things are organized for now rather than trying to make things work correctly ;)
rgrau has joined #nixos
<cole-h>
If you add `neededForBoot = true;` to the filesystem you want mounted at boot in your hardware-configuration.nix, it will indeed be mounted at boot
<Extends>
with the new flake system in home-manager, how does one rebuild the configuration ?
<ToxicFrog>
cole-h: the problem is that if you have a lot of zfs datasets you need to manually set each one to legacy mounting and configure it there rather than trusting `zfs mount -a` to do the right thing
<cole-h>
¯\_(ツ)_/¯
<cole-h>
I don't see a problem with that (because I had NixOS in mind when I set up ZFS, and thus created all my datasets with legacy mounting).
<immae>
ToxicFrog: my end goal is to put a RequiresMount dependency on the service that needs its dataset mounted
hiro99 has quit [Quit: hiro99]
<immae>
(still not exactly clear if it’s workable or not so I’m doing tests)
<immae>
so I’d only have "critical" ones in fstab and let the other ones get mounted later by zfs mount -a
<immae>
(Also I don’t know how smart systemd will be about that RequiresMount dependency, since he wouldn’t know from the top that "later" a mount will appear under that tree)
<buffet>
hey, i have an attrsOf option somewhere, and im trying to generate a bunch of systemd services from it. how can change the attribute name (systemd.services.???) without resorting to something like fromJSON?
<cole-h>
infinisil: Nah, second one doesn't change `is deprecated` to `has been removed` for ease of applying the uncontroversial (IMO) part of the suggestion.
evelyn has left #nixos ["Guidheam ort droch bhreitheanas nuair thig ort latha na fìrinn."]
cr4y1_ has quit [Remote host closed the connection]
cr4y1_ has joined #nixos
sangoma has joined #nixos
<{^_^}>
[nixpkgs] @mohe2015 opened pull request #96923 → Bugfix in apache httpd → https://git.io/JUYk0
shibboleth has joined #nixos
xensky has quit [Quit: No Ping reply in 180 seconds.]
xensky has joined #nixos
shibboleth has quit [Quit: shibboleth]
<jared-w_>
Extends: wouldn't setting args.enableParallelBuilding explicitly to false not trigger the `or true`? The `?` bit only happens if the property doesn't exist in args, iirc.
<Extends>
jared-w_: I think that the `or` does that, like, if the value in the set is absent, then the value of the or will be taken, which is why it's kind of weird, as ? true already set enableParallelBuilding to true, and if enableParallelBuilding isn't present, then its value will also be set to true, although not sure about that
noonien has joined #nixos
<noonien>
hello folks!
<noonien>
i'm trying to update a package using an overlay, this is how i'm doing it in the overlay: http://ix.io/2vUq
<noonien>
the problem is that i'm getting this error: http://ix.io/2vUv and i'm unsure how to fix it
<noonien>
i've tried settings verifyCargoDeps to false, and cargoSha256 to unset
<noonien>
but, i'm unsure what the real problem is
<V>
you put verifyCargoDeos, not verifyCargoDeps
<noonien>
argh!
<noonien>
the problem seems to persist
<V>
have you tried doing what the error tells you to
<noonien>
i even tried setting cargoSha256 to all 0s, i get the exact same error
<noonien>
i'm unsure what cargoSha256 has to do with anything, as far as i can tell, the error is triggered by a diff
<noonien>
hmm, is there a way to print the nixpkgs version i'm using for a specific channel?
cr4y1_ has quit [Remote host closed the connection]
lsix has joined #nixos
<jared-w_>
Extends: I think you're right. My inclination is that the ? true is intended to make enableParallelBuilding something that must be explicitly opted out of (by setting it to false) rather than letting it default to missing/false
<noonien>
same issue, there might be something wrong with me setting up the overlay, i'll try to setup a different .nix file and nix-build it, if that works, there's something wrong in my overlay
<noonien>
V: eh, i went without kak-lsp.drvAttrs, since that doesn't really do much on this case. might even do bad
<noonien>
thanks for the help though!
vv has quit [Ping timeout: 245 seconds]
<V>
if you're on macOS it'll make a difference, since that has buildInputs you need to care about
<V>
otherwise it just sets meta for that package
rkallos has quit [Quit: connection reset by purr]
lsix has quit [Quit: WeeChat 2.9]
<V>
this is horribly messy though
<V>
Making a fresh package is probably the simpler thing to do
krasnij has joined #nixos
<noonien>
i'm on nixos. yeah that's what i went with. kak-lsp doesn't seem to have any special arguments to buildRustPackage, and i don't want any stray drv attrs ending up as special buildRustPackage arguments by mistake :D
zettabyte has joined #nixos
<V>
probably wise
<V>
I'm not sure I know of a nicer way of doing this
<{^_^}>
[nixpkgs] @jtojnar pushed commit from @r-ryantm to master « cldr-emoji-annotation: 37.0_13.0_0_1 -> 37.0_13.0_0_2 »: https://git.io/JUYL7
<zettabyte>
is there a way to declaratively specify a service to be installed but disabled by default (service is fully configured but can be manually started with systemctl start <service>)
<{^_^}>
#96928 (by buffet, 49 minutes ago, open): rclone: fusermount not in PATH
<V>
it's a suid binary, so you can't just magic yourself up privilege escalation
<buffet>
> su: must be run from a terminal
<{^_^}>
<LAMBDA>
<buffet>
oops
<buffet>
basically i wanna set the suid thing on something
<V>
take a look at security.wrappers
<buffet>
will do, thanks
<V>
I'm confused though, there's a wrapper for fusermount
o1lo01ol1o has quit [Remote host closed the connection]
<buffet>
any great ways to add that to the path of rclone?
o1lo01ol1o has joined #nixos
<buffet>
i was mosty playing around as i dont really have a solution, so dont expect any great ideas there
erasmas has quit [Quit: leaving]
<buffet>
i tried to symlinkJoin the two packages (fuse3 and rclone), but 1. missing suid, 2. fuse3 only ships fusermount3 not fusermount, and i kinda got sidetracked at that point
o1lo01ol1o has quit [Ping timeout: 246 seconds]
cr4y1 has quit [Remote host closed the connection]
fendor has quit [Read error: Connection reset by peer]
<V>
buffet: what's the easiest way of testing rclone? does it support ssh, etc?
<V>
oh, 'sftp'
<buffet>
my issue specifically is with `rclone mount`
<V>
I can't mount anything unless I configure it
<buffet>
should also support sftp though
dansho has joined #nixos
<buffet>
there is rclone config, which takes you through the steps, so thats pretty quick, but yeah it needs a config
<V>
made a 'local' remote
<V>
get the same error
<V>
buffet: add /run/wrappers/bin to $PATH
<V>
probably using makeWrapper
<V>
(or wrapProgram)
xcmw has joined #nixos
<{^_^}>
[nixpkgs] @bcdarwin opened pull request #96934 → f3d: init at 1.0.1 → https://git.io/JUYtx
noudle has quit []
<buffet>
ok so that seems to work, but i somehow created a folder with ????? owned by ?:? so lemme fix that rq
<buffet>
before i know if it actually worked
<buffet>
ok so root has permissions denied, thats likely bad, and it reoccurs so i blame fuse not working
ManiacOfMadness has quit [Ping timeout: 260 seconds]
ManiacOfMadness has joined #nixos
nealean[m] has joined #nixos
<buffet>
V, yes, works
<V>
whee
<buffet>
is that good or bad?
<V>
presumably it working is good?
<buffet>
aight
<buffet>
"whee" is just hard to tell whether its a "woohoo" or a harsh sigh
<buffet>
thanks for the help
<V>
Ah, I was contemplating putting \o/ instead but I wasn't in a good enough mood :p
civodul has quit [Quit: ERC (IRC client for Emacs 26.3)]
Maxdamantus has quit [Ping timeout: 265 seconds]
Reventlov has quit [Ping timeout: 244 seconds]
Maxdamantus has joined #nixos
cole-h has quit [Quit: Goodbye]
cole-h has joined #nixos
jumper149 has quit [Quit: WeeChat 2.9]
Church- has quit [Ping timeout: 265 seconds]
werner291 has quit [Quit: werner291]
lurpahi has joined #nixos
zuserm has left #nixos [#nixos]
<lurpahi>
I switched out 19.03 for 20.03 and `nixos-rebuild switch`d and now GRUB is giving me problems. 🤔
andreas303 has quit [Remote host closed the connection]
andreas303 has joined #nixos
<infinisil>
,dnw lurpahi
<{^_^}>
lurpahi: Provide as much info as possible on problems. What specifically doesn't work? What command did you run? What's in the relevant files? What's the error? This makes debugging easier and helps getting faster responses
philr has joined #nixos
gustavderdrache has joined #nixos
<lurpahi>
I meant to do that at that moment, but I got pulled away after I sent that message.
<lurpahi>
I have had this problem before—as evidenced by browser history—but I'm not sure what I did to fix it before. It's not unlikely that I just gave up and undid whatever made it start happening.
<samueldr>
lurpahi: not that this points to a solution, but some assumptions: you are booting using legacy bios boot, rather than uefi, right?
<lurpahi>
Probably. I'm not completely certain.
<lurpahi>
I haven't really gone into anything in years, so I don't remember my own setup too well.
<samueldr>
if you have /sys/firmware/efi/efivars and files under there you're booted as UEFI
<lurpahi>
There is no `/sys/firmware/efi`.
<samueldr>
then that's as I assumes because of the output; knowing this is helpful
<samueldr>
next: "installing the GRUB 2 boot loader on …..." what is it that you ellipsized?
<samueldr>
a partition or a block device?
<lurpahi>
I don't think I ever really figured out the difference between UEFI and whatever else.
<lurpahi>
`/dev/disk/by-label/nixossd`
<samueldr>
right, then this means it's the partition
<lurpahi>
There's just one partition on that drive.
<samueldr>
the difference between legacy bios boot and UEFI is how the "bios" of your machine decides what to boot; the legacy boot option starts whatever program is embedded into the MBR, while UEFI has some more configuration options and start a program that is found as a file in a partition
<lurpahi>
For some reason, `/dev/sdc1` is my root drive.
<samueldr>
having only one partition is likely fine
<samueldr>
usually, grub is installed to the MBR, rather than being installed to a partition, do you know if there is a reason which made you install it to the partition?
<samueldr>
it could be because of dual (or more) boot
<lurpahi>
I've abandoned every other operating system I've installed on any of three drives presently in this computer; this instance of NixOS is all I've got now.
<lurpahi>
I don't know if it being on an SSD—as implied by the name `nixossd`—is relevant. 🤔
<samueldr>
shouldn't matter
<samueldr>
instead of configuring the installation to install to the partition you should be able to switch it to install to the disk
<lurpahi>
In `configuration.nix`, I do seem to have configured `boot.loader.grub.device` to be `/dev/disk/by-label/nixossd`.
<samueldr>
though you might want to have a usb drive with the nixos installer handy just in case
<samueldr>
you can identify a "stable" path name for the hard drive
<lurpahi>
I don't have a flash drive handy that I could do that with; I gave one that I had with NixOS on it to my brother last night, knowing his intent to format it if he used it.
<lurpahi>
And he's asleep now.
<lurpahi>
I have too many non-functioning flash drives.
<samueldr>
though I assume the risk is slim, anything touching the boot configuration like that has intrinsic risks
<lurpahi>
So, should I see what happens when I replace `/dev/disk/by-label/…` with `/dev/disk/by-id/wwn-…`? 🤔
<lurpahi>
If it doesn't work, I'll call it time for a fresh installation of whatever Linux distribution I feel like trying. Probably NixOS again but possibly something else with Nix on top.
<samueldr>
if you want the explanation, you'll have to read through the grub manual, or ressources, about the difference when you install to a disk compared to a partition
<lurpahi>
It intuitively 'makes sense' to a degree, but I feel like installing it to a partition just shouldn't work at all, in the general case.
<samueldr>
actually, that's the annoying part, you likely had a valid mbr program that knew to look at partition to continue booting, which hid part of the situation
<samueldr>
and grub grew since 19.03 just enough to not fit in whatever scheme it uses
mbrgm has quit [Ping timeout: 260 seconds]
mbrgm_ is now known as mbrgm
<hexa->
> This survey is anonymous.
<{^_^}>
error: syntax error, unexpected ')', expecting ID or OR_KW or DOLLAR_CURLY or '"', at (string):322:1
<infinisil>
hexa-: It's optional, but yeah I got confused too
<hexa->
"This survey is anonymized. Please do not enter personally identifiable or private information into this survey. "
<hexa->
:D
<infinisil>
nbathum: Can you still change it/redo it to not even ask for a name? Or ideally not even an email?
<nbathum>
infinisil: yes possibly. is registering an account on limesurvey.org considered not ideal?
<nbathum>
I agree it is confusing
<infinisil>
It can hardly be anonymous when you have to give an email (and name optionally), even if you yourself don't get that data (I feel like limesurvey could get it if they wanted)