<adam_>
then i would be forcing the rest of my team to use those same tools though right
<energizer>
yes
<adam_>
and what about when you are deploying
mbrgm has quit [Ping timeout: 272 seconds]
<adam_>
surely you don't want to ship your linter/language server/type checker
<cole-h>
adam_: The way you originally worded your question made it sound like you wanted to install e.g. `python3Packages.requests` in `environment.systemPackages`, and also use it in a project alongside `python3Packages.click` (which is a project-local dependency)
<adam_>
you just want the smallest subset of packages needed to run your code
<energizer>
right, distinguish between runtime and development dependencies
mbrgm has joined #nixos
<cole-h>
The way you're describing now seems more sane and I would think already works?
<adam_>
when I think of development dependencies I think of something like pytest
<adam_>
which the rest of the team can all agree on, and is used to run tests
<adam_>
but when I think of a tool that is a personal choice, since others may be using a different language server, or an ide, why would I package it with the project?
<{^_^}>
[nixpkgs] @cdepillabout pushed 5 commits to haskell-updates: https://git.io/JUUNA
<adam_>
your pylintrc could for sure be project specific
<energizer>
what about the pylint version
<energizer>
and the mypy version
<energizer>
and the bandit version
<colemickens>
fwiw I install Codium and Rust-Analyzer system-wide and then just make sure to start my session from a nix-direnv shell.
<energizer>
there is such a thing as "development dependencies". linters and test runners are what that phrase refers to
cjpbirkbeck has joined #nixos
<adam_>
okay well then can you have nested nix shells?
<adam_>
because in my home directory I have a init.vim
<adam_>
whenvever I am modifying that I want to have the vim-language-server
<adam_>
but in ~/code/project-a I want a different environment
<dominikh>
he also mentioned things like language servers and other local development tools, energizer. not all development tools are shared among the whole team, or relevant to CI or tests or anything that reproducible. consider some dev tool that can do some refactoring, that just adam_ wants to use locally from his editor.
<adam_>
but that's nested within ~
<dominikh>
his peers won't care about that tool.
<energizer>
dominikh: if peers dont care about a tool then it can be somewhere else
<dominikh>
that "somewhere else" is presumably what he's asking about
<adam_>
^
<energizer>
ok sounds like you've got it figured out then
<colemickens>
adam_: I don't know that you can nest them, but since they're just nix you could compose/combine them.
<colemickens>
One could have a minimal shell.nix and a shell-devenv.nix that includes the opinionated developer tooling too?
<colemickens>
I'm thinking about something like that for Nix-powered Theia devenvs... :) I think there's a neat potential...
alexherbo24 has joined #nixos
alexherbo2 has quit [Ping timeout: 240 seconds]
alexherbo24 is now known as alexherbo2
<azazel>
I'm using that exactly like colemickens writes...I'm using a next-stable dev sub-environment to test upgrades
steell has quit [Remote host closed the connection]
pamplemousse has quit [Ping timeout: 246 seconds]
cmk_zzz_ has joined #nixos
<pie_>
davidak[m]: i guess it just bothers me that idk how to get a device without an ip
cmk_zzz has quit [Ping timeout: 246 seconds]
robin[m]3 has joined #nixos
sigmundv__ has quit [Ping timeout: 240 seconds]
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/JUUA4
jluttine has quit [Ping timeout: 240 seconds]
jluttine has joined #nixos
cole-h has quit [Quit: Goodbye]
cole-h has joined #nixos
pamplemousse has joined #nixos
Rusty1 has joined #nixos
veleiro has joined #nixos
cole-h has quit [Quit: Goodbye]
cole-h has joined #nixos
<daddy_james[m]>
anyone know how to fix lightdm randomly booting to a black screen?
Extends has quit [Remote host closed the connection]
<cole-h>
I don't want to keep my secrets in my main repo tho :(
<colemickens>
I couldn't get git-crypt to work even without submodules because of how the flkae source gets copied to the store, but I think bqv told me a way around that
<pjt_tmp>
did the 20.03 release change anything major about graphics drivers?
<colemickens>
I wound up switching to sops-nix, which doesn't do everything I need quite yet, but also keeps secrets out of the store, so it seems wholistically better
<{^_^}>
[nixpkgs] @cdepillabout pushed 2 commits to haskell-updates: https://git.io/JUTeE
<bqv>
colemickens: i can't remember exactly what that would have been. i think i had big plans for secrets systems but i never got around to implementing them. but yeah mine are currently in $flake/secrets, cause i don't use submodules and i'm not overly worried about stuff being in the store
<bqv>
but the gist of having stuff avoid the store is just encrypting it with a(n ideally per-host) key, and this https://github.com/NixOS/nixpkgs/pull/93659 would be perfect for that
<{^_^}>
#93659 (by ju1m, 4 weeks ago, open): nixos/security.gnupg: provisioning GnuPG-protected secrets through the Nix store
Supersonic112 has joined #nixos
Supersonic has quit [Ping timeout: 244 seconds]
Supersonic112 is now known as Supersonic
stoile has joined #nixos
<Graypup_>
does cabal-jailbreak fail to remove version bounds on flag-gated dependencies?
<Graypup_>
asking for a friend
<Graypup_>
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
<{^_^}>
[nix] @bqv opened pull request #3961 → Add a settings option for log-format → https://git.io/JUTey
<{^_^}>
[nixpkgs] @cdepillabout pushed 2 commits to haskell-updates: https://git.io/JUTep
<cole-h>
Ah crap, I have a lot of impure fetchTarball stuff due to niv
<bqv>
oh no, my vim ruined the whitespace
<cole-h>
It's really quite annoying how you can't do `nix build`, get an error, and `nix build --show-trace` because of caching
<bqv>
this is one reason i love emacs, i can just always have show-trace, and reliably scroll up
<bqv>
there is --rebuild incoming, fwiw
<bqv>
and --impure usually busts cache too
<cole-h>
bqv: Any ideas why I'm getting infinite recursion when adding `inputs` to my configuration.nix's top-level thing?
<bqv>
self
<bqv>
probably
<cole-h>
bqv: Oh, I'm using `"${inputs.home-manager}/nixos"` in my imports
<cole-h>
Seems like it doesn't like that?
<bqv>
huh.
<bqv>
cole-h: can i interest you in inputs.home-manager.nixosModules.home-manager?
<cole-h>
I have my home-manager config in my configuration.nix
proofofk_ has quit [Remote host closed the connection]
<bqv>
i imagine that's what you're after, anyway
<bqv>
my flake.nix, line 409
<bqv>
*249
<bqv>
sorry, my local changes are massive
<cole-h>
bqv: This whole modulesFor thing makes my brain hurt lol
<bqv>
heh, yeah.. i just kinda separate my config into vague segments, and inject the various modules from external flakes
<bqv>
evalModules can take a list of modules, as well as just a module
<bqv>
the recImport part is definitely a clusterfuck but it ain't broke so i don't feel the massive need to fix it
<cole-h>
bqv: So, how exactly do I use `inputs.home-manager.nixosModules.home-manager` and the `home-manager.users` stuff?
<cole-h>
I imagine I'd have to add the nixosModules.home-manager thing to my `mkSystem`'s modules list
<bqv>
it's a module, if you want, you could just import it like that string above
<cole-h>
But what next?
<bqv>
it should be equivalent to ${home-manager}/nixos, i think
<bqv>
but probably work better with flakes
<bqv>
for unspecified reasons
<bqv>
but yeah, you can also shove it directly into nixosSystem with your config, and it'll just create that config.home-manager tree
<bqv>
no other change needed
<cole-h>
bqv: So, I've tried to set `configuration = { imports = nixos.modules; };` copied from your flake.nix, but nix complains that the attribute `modules` doesn't exist?
<xd1le>
I found a flake with the line: outputs = { nixpkgs, nix, self, ... }@inputs: { ... }
<xd1le>
What is the @ doing here? Search engines don't like the @ :/
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
la-jesystani has joined #nixos
<cole-h>
@ binds the `...` of `{nixpkgs, nix, self, ... }` to the variable `inputs`
<la-jesystani>
howdy, im failing to build the haskell package beam-core, its saying its missing dependencies, but im unsure how to go about resolving this, whats the usual way?
<cole-h>
So you could use `nixpkgs`, or `inputs.nixpkgs`
<xd1le>
cole-h: thank you!
<cole-h>
bqv: I finally understand what spicy stuff is
<bqv>
lmao
<tsujp>
People here are using nixos as their desktop/laptop distro right?
<tsujp>
As in, while not specifically designed for that purpose it is useable as such yeah?
<la-jesystani>
i am yeah tsujp
<tsujp>
Perfecto, looking to swap to it from arch
<tsujp>
Probably gonna start messing around with in a VM to start off with, replicate my setup yada yada yada
<cole-h>
also bqv: is there a reason `let nixos = ....${name}.modules;` ends in the modules attribute?
<tsujp>
Any insights onto non-systemd and libre-kernel la-jesystani ?
<la-jesystani>
its generally pretty good until one of your haskell libraries wont build and then you have to sit in irc waiting for someone smarter than you ;)
<la-jesystani>
never tried that myself, i think systemd is pretty baked in iirc
<tsujp>
i know 0 haskell so this is probably an opportunity to learn for me hohoho
<dminuoso>
la-jesystani: Alright. So in nixpkgs, you dont have the wide variety of hackage.
<dminuoso>
Think of nixpkgs as having only one version per package. Sometimes, we publish multiple meta packages for parallel versions, but this does not tie into cabal.
sangoma has joined #nixos
<la-jesystani>
is there a better way to be getting my packages?
<dminuoso>
What cabal is telling you "Im trying to solve for hashable >=1.1 && <1.3, vector-sized >=0.5 && <1.3, but I cant find those (given the version constraints!)"
<la-jesystani>
i like being able to pin to specific versions, and be able to open a shell with the packages ready to be used in ghci
<dminuoso>
la-jesystani: Is it possible that you're on some unstable nixpkgs channel?
<dminuoso>
On 20.03 I reckon this should build
alp has joined #nixos
<la-jesystani>
yeahp im on unstable
<dminuoso>
That's the problem then. beam-core needs its bounds relaxed
<la-jesystani>
is there a way to ask it to build with a specific version? hold on ill pastebin the stuff im using
<dminuoso>
Either fix beam-core, and wait for the update to be released to hackage, or use 20.03 for nixpkgs
<cole-h>
bqv: btw, h-m + flakes is super easy since that change
<dminuoso>
la-jesystani: Do you understand the fundamental issue here?
<bqv>
cole-h: i'm unhappy of a state of affairs where home-manager isn't a flake
<bqv>
if anything should be a flake, that should be
<dminuoso>
bqv: By the way! Turns out I should be using overrides instead of extend after all.
<la-jesystani>
I believe so? Im just not sure how to ask this derivation to use something different than what my system is on
<bqv>
dminuoso: figured :p
<cole-h>
bqv: Well, good news is that your (global, not just you) efforts have made it so easy that all I needed to do was copy over the `flake.nix`, remove the `nixpkgs` stuff from the imports, and boom!
<bqv>
:D
<cole-h>
bqv: Maybe with more people being turned on to flakes, ryc*e will be comfortable merging support, even if they don't use it themselves (yet)
<dminuoso>
bqv: Turns out, that .extend and .override at the same time doesn't work, and .extend somehow causes missing attribute errors for .override
<dminuoso>
Which makes me wonder how to do staged overrides
<dminuoso>
*shrugs*
<dminuoso>
la-jesystani: I see! So the usual way is roughly this:
<bqv>
you can override multiple times. that's why i said it "works best"
<dminuoso>
heh
<bqv>
cole-h: does that mean a flake.nix is trivial now, though?
<la-jesystani>
dminuoso: thank you :) ill have a read
<bqv>
cole-h: oh, wonderful
<bqv>
cole-h: given you've already done it, you're free to do the PR if you like, but otherwise I can do it now :D
<cole-h>
All I did was copy the flake.nix from bqv-flakes ;)
<bqv>
good god how is my server *still* deploying
spudly has quit [Ping timeout: 258 seconds]
spudly- is now known as spudly
work_ has quit [Quit: Connection closed for inactivity]
noudle has joined #nixos
<cole-h>
bqv: I do have a question with the nixpkgs-mozilla overlay: I want to add the `latest` attribute to my overlays. I previously had this by doing `inherit (import sources.nixpkgs-mozilla final prev) latest;`, but this doesn't appear to work when adding it to my flakes-based overlays
<cole-h>
colemickens: Even if I use your nixpkgs-mozilla fork, I get `undefined variable 'latest'`. Am I doing it wrong?
<colemickens>
cole-h: sounds like it?
<colemickens>
or I misunderstood
<colemickens>
do you mean my actual nixpkgs-mozilla fork or flake-firefox-nightly?
<cole-h>
actual nixpkgs-mozilla
<colemickens>
my change to nixpkgs-mozilla is only relevant to flake-firefox-nightly.
<cole-h>
o
sangoma has quit [Ping timeout: 240 seconds]
<colemickens>
you won't be able to use 'latest" from nixpkgs-wayland with flakes' pure eval
<colemickens>
meaning every system build will be --impure
<colemickens>
I can extend flake-firefox-nightly to expose the other variants that are exposed in "latest" and update them as well but I haven't had time to do it.
<colemickens>
(for more bg: flake-firefox-nightly does an impure eval every hour to get the version info, "pins" it, and exports the nightly build as a flake output)
<bqv>
oh that's cool
<bqv>
wait, no it's not
<bqv>
does that mean the builds aren't reproducible?
<bqv>
colemickens: it's not quite about it working in the context of nixos, more that if the home-manager module is consumed outside nixosSystem but the pkgs overlays are set inside, they won't propogate to h-m
<bqv>
that's why i just cut all losses and fully construct my pkgs outside nixosSystem so there's no room for error
<colemickens>
how does one consume the h-m module outside nixosSystem?
<colemickens>
or why
<colemickens>
bqv: I feel like it's a lost cause if I can't *enforce* that it happens outside nixosSystem. I'm actually surprised I'm still allowed to do what I'm doing, a bit.
<bqv>
...yeah that was kinda awful phrasing tbh. i mean, if you nixosSystem a list of modules [ ./configuration.nix home-manager ], where that configuration.nix has some overlays inside it, the pkgs argument passed to the home-manager module will not contain those overlays
<{^_^}>
[nixpkgs] @FRidh pushed to master « extra-utils: build a full lvm2 without udev support, fixes #96197 »: https://git.io/JUTT1
mallox has joined #nixos
<bqv>
whereas with my setup, where i manually set pkgs for nixosSystem, that's not the case
<colemickens>
oh ok I get it
<{^_^}>
[nixpkgs] @FRidh closed pull request #96290 → nixos boot extra-utils: ensure we use working lvm2 → https://git.io/JUUoR
<bqv>
basically, with my route, you have one unique and coherent pkgs, and with every other route, you may end up with several incoherent versions
<colemickens>
bqv: but nothing keeps you from forgetting and throwing nixpkgs.overlays = [] in some random file at the edge though
<{^_^}>
[nixpkgs] @FRidh pushed 86 commits to staging-next: https://git.io/JUTTM
<colemickens>
afaict, it's just that you chose to do that diligently.
<colemickens>
(just checking my understanding)
<bqv>
actually, that wouldn't make a difference
<bqv>
because as a part of my thing, i then set nixpkgs.pkgs to the pkgs in the module arguments
<bqv>
which then overrides every other nixpkgs.* option
<bqv>
i'm still salty about eelco nuking my nix store by forgetting a semicolon
<cole-h>
Haha. What happened?
<{^_^}>
[nixpkgs] @danieldk opened pull request #96339 → fossil: install man page and bash shell completions, use system sqlite → https://git.io/JUTkF
GolDDranks has quit [Remote host closed the connection]
lsix has joined #nixos
<bqv>
hm, can't find the commit now. but i updated nix at just the wrong time and ended up with a version that recursively created derivations and invalidated a lot of store paths
<bqv>
it'll probably be in the irc logs somewhere, i was quite upset that night
<bqv>
but i reckon the commit (titled "whoops") got lost in the flakes merge
<{^_^}>
[nixpkgs] @expipiplus1 opened pull request #96355 → <!-- To help with the large amounts of pull requests, we would appreciate your reviews of other pull requests, especiall… → https://git.io/JUTsy
<pemeunier>
Hi! What's the current story with NixOps? If I want to use a new provider, should I consider terraform or writing a plugin (or something else)?
stu_[m] has left #nixos ["User left"]
orivej has quit [Ping timeout: 258 seconds]
<{^_^}>
[nixpkgs] @zimbatm opened pull request #96357 → nix-build-uncached: remove the nix wrap → https://git.io/JUTZs
Dr8128 has quit [Read error: Connection reset by peer]
<superherointj>
I did something stupid, using nix-env installed `busybox.lspci` in my system and I need to reverse it now. Because the usual commands are not working. But I not really sure how I should be reversing this. I already tried to remove busybox. And `lspci` is already removed.
<rizary>
when I build my project using `nix build` I got an error like, `failed with exit code 1; last 10 log lines:`. the bad part is the `last 10 log lines` doesn't give me enough information. Any idea how to get the full log? I'm using nix flake btw.
<sphalerite>
so, one of my nixops systems is failing to boot because one of the filesystems can't be mounted. I can't start sshd in order to redeploy it, (I think) because it depends on mounts.target. Anybody have any great ideas for how I can get out of this pickle?
<edef>
hmm, does the test runner have a convenient way of jumping into a test machine?
<edef>
my current test wrapping is `rm -rf -- $XDG_RUNTIME_DIR/vm-state-* && env -u DISPLAY tests='exec(os.environ["testScript"])' (nix-build test.nix -A driver)/bin/nixos-test-driver`
<edef>
based on how the normal runner does it
<edef>
(but i need internet connectivity for the current incarnation of my test, so i wanted it outside a drv anyway)
<sphalerite>
edef: as in getting a shell on the test machine?
<sphalerite>
edef: if so, if you don't unset DISPLAY, you can log in on the consoles, root has an empty password
<edef>
ah, that's a good idea
<edef>
thanks!
acarrico has joined #nixos
sangoma has quit [Ping timeout: 240 seconds]
kreyren has quit [Ping timeout: 240 seconds]
Wizek has joined #nixos
CptCaptain has joined #nixos
turlando has quit [Ping timeout: 258 seconds]
alexherbo2 has joined #nixos
fendor has quit [Remote host closed the connection]
<infinisil>
Oh true, I'll try to see if it's always the same ones
alexherbo23 has joined #nixos
<q3k[m]>
i wouldn't expect a broken fastly node, but this could take a different internet route
<q3k[m]>
since both of these addressses are in different blocks announced differently
<infinisil>
Hm nope, both with 151.101.2.217 and 151.101.194.217 I can get failures and successes for both
<q3k[m]>
if you can pin down failure to be dependent on one address vs the other, that would be an useful datapoint
<q3k[m]>
alright, that's also good to know
alexherbo2 has quit [Ping timeout: 260 seconds]
alexherbo23 is now known as alexherbo2
<q3k[m]>
so yeah, i don't have any better idea than to look at a packet capture now
<infinisil>
Same with 151.101.66.217
<gustavderdrache>
wouldn't this be the kind of thing a traceroute could help with?
<infinisil>
Yeah I'll try tcpdump next
metreo has quit [Quit: Leaving.]
<infinisil>
Or traceroute might be a good idea yeah
<q3k[m]>
i'm going to guess that this is indeed an MTU issue
<cransom>
traceroute will be far less useful than a tcpdump.
<q3k[m]>
because after this, you receive a pretty large set of data
<q3k[m]>
which might get immediately coalesced into a large packet
dm9 has joined #nixos
<infinisil>
Ahh
<q3k[m]>
which is dropped somewhere because MTU, and a return ICMP 'fragmentation needed' packet is blocked because shitty firewall
<q3k[m]>
are you doing this from a home link or a corporate network?
<q3k[m]>
do you have any weird network virtualization stuff on your machine?
<q3k[m]>
and yeah a traceroute would be pretty much useless here, unless, i don't know, you see like >1% drops after some point
<infinisil>
Home network, but I do route everything through wireguard to another server
<q3k[m]>
(which would mean a broken internet connection)
<q3k[m]>
yep it's the wireguard
<infinisil>
I think I tried turning it wireguard already though
<infinisil>
I'll try again though
<q3k[m]>
try it without the wireguard if possible, see if that fixes the issue
<q3k[m]>
if so, we can fix your wireguard tunnel next :P
<{^_^}>
[nixpkgs] @Emantor closed pull request #96382 → gnum4: include fix without fetchurl, update comment → https://git.io/JUkOm
<teto>
adisbladis: I was advised to bump my poetry version to 1.1.0b2 to get better help debug about a poetry issue. I tried to bump it locally but I got an issue with 'No module named poetry" (as a follow up of other errors). Do you have a custom working branch ?
<q3k[m]>
as a test you can also bump the inner WG MTU up to 1500 (which will cause fragmented UDP packets on the outside, which will makes thing shitty but at least also make it work)
<infinisil>
Oh yeah I get less failures at least when turning off wireguard
<infinisil>
And the failures I do still get are for a different reason, probably DNS, as I get "* Resolving timed out after 2000 milliseconds"
<q3k[m]>
arianvp: nothing's really bad, just typical MTU shenanigans
<arianvp>
Btw q3k the logs on our concourse still don't load due to MTU shenanigans
<arianvp>
:)
<lordcirth_>
Finally installed NixOS on my main PC - got a weird bug. Every boot, it stops in emergency mode, but pressing Enter resumes and everything works fine?? Root is on ZFS, and EFI boot.
<q3k[m]>
teto: oh, you mean this nix installer from nixos.org? duh.
gustavderdrache has joined #nixos
<simpson>
arianvp++ Huh, is there one of those magic aliases for it? I don't see it on the online package search. TIL, thanks.
<{^_^}>
arianvp's karma got increased to 12
<{^_^}>
[nixpkgs] @vcunat pushed to master « nixos installer tests: add a missing package »: https://git.io/JUkcP
<teto>
q3k[m]: yeah, I know hte difference between nixos/nix. I am trying to infect my company with nix (and maybe nixos afterwards) but the online install script had some issues on ubuntu and I wondered if I could replace with a .de
sigmundv__ has quit [Read error: Connection reset by peer]
sigmundv__ has joined #nixos
<teto>
actually I was not there when the install messed up, something about incorrect permissions on a folder
<teto>
I thought the nix .deb had reached ubuntu but not yet in fact :(
<q3k[m]>
lack of sudo to chown /nix?
<q3k[m]>
i hit that fairly often when installing nix on other distributions
<q3k[m]>
although i'm not sure if it's a lack of sudo or it just bailing for some other reason and asking to fix with sudo
<teto>
something in the .../profiles/per-user . We have a thin wrapper around the nix-installer to upgrade it when already present. It's a bit adhoc so I hope I can improve the situation a bit
<infinisil>
superbaloo: fetchGit (note the capital G) works for that since it just uses the `git` in your PATH
marsh has quit [Ping timeout: 244 seconds]
<superbaloo>
infinisil: I was wondering about how to inject secrets in a build
<superbaloo>
to be quite fair :)
<{^_^}>
[nixpkgs] @NeQuissimus pushed 6 commits to release-20.03: https://git.io/JUk4W
<infinisil>
superbaloo: Got an example?
<superbaloo>
Ideally I would like a fetchCompanyGitlab using credentials injected from "somewhere"
<superbaloo>
my question is, where the "somewhere" should be
<superbaloo>
pretty much :D
<superbaloo>
should that be a parameter of the derivation, or could that be injected any other way
<superbaloo>
ideally I'd like the credential not to be in the derivation repository
<superbaloo>
and be able to inject a different credential when using build server vs on my workstation
<superbaloo>
if that makes sense
<{^_^}>
[nixpkgs] @matthewbauer merged pull request #95956 → runInLinuxVM, test-driver: use -cpu max instead of -cpu host → https://git.io/JUeLb
<{^_^}>
[nixpkgs] @matthewbauer pushed 2 commits to master: https://git.io/JUk44
spudly- has joined #nixos
spudly- has joined #nixos
spudly- has quit [Changing host]
<wpcarro>
Is there a version of writeShellScriptBin that doesn't nest the shell script in result/bin? Where result itself is the executable...
<superbaloo>
pkgs.writeScript ?
<ToxicFrog>
superbaloo: off the top of my head, I'd say: it's a parameter and when building it you inject it as something like `pkgs.mypackage.override { server = foo; user = bar; pass = baz; }`
marsh has joined #nixos
<superbaloo>
ToxicFrog: yeah I dont think I have to many choices
cole-h has joined #nixos
<cole-h>
bqv: Just found a shortcoming of flakes...
spudly has quit [Ping timeout: 260 seconds]
spudly- is now known as spudly
<wpcarro>
writeShellScript and writeScript... Hmmm what's the difference
<cole-h>
Can't use h-m's `config.lib.file.mkOutOfStoreSymlink` :(
<cole-h>
Note: toString ./weechat-conf also doesn't work in flakes (points to the store path), so I need to use `${config.xdg.configHome}/..../weechat-conf` :')
<LnL>
can't you make a drv that points to a path outside the store?
<clever>
cole-h: i think its because nix copied the flake into the store before parsing the nix
<clever>
cole-h: so ./. is already in the store to begin with
<cole-h>
Yeah :(
<LnL>
yeah ./. won't work but absolute paths should
davidv7_ has joined #nixos
viev[m] is now known as Viev[m]
kreyren has joined #nixos
<bqv>
cole-h: oh? You can use ln in scripts, though, right?
<cole-h>
clever++ LnL++ Just needed to change from `./config` to `${config.xdg.configHome}/nixpkgs/modules/weechat/config` and mkOutOfStoreSymlink works again :-D
<{^_^}>
clever's karma got increased to 497
<{^_^}>
LnL's karma got increased to 77
justanotheruser has joined #nixos
<LnL>
yeah that or make config itself a symlink to somewhere and add it to the flake
dani9292 has quit [Remote host closed the connection]
domogled has joined #nixos
<{^_^}>
[nixpkgs] @flokli pushed to remove-perl-test-driver « nixos/lib/test*: remove perl test driver »: https://git.io/JUkRA
<{^_^}>
[nixpkgs] @flokli pushed 0 commits to remove-perl-test-driver: https://git.io/JUkRp
<bqv>
No, I missed the point again, but what you were saying makes sense, heh
<bqv>
cole-h: fwiw I think you can get the same effect by just specifying a mode
<bqv>
Cause obviously modes don't exist in the store, so that forces a copy
<cole-h>
I don't want a copy :(
<bqv>
Ah
<cole-h>
I want a symlink because that can be mutated
<cole-h>
(The contents)
<bqv>
Right, mutated in the secrets vault
<{^_^}>
[nixpkgs] @flokli opened pull request #96396 → nixos/lib/test*: remove perl test driver → https://git.io/JUkRj
cosimone has joined #nixos
butcher_ has joined #nixos
<LnL>
I abuse symlinks for even more weird stuff sometimes tho :)
fendor has quit [Remote host closed the connection]
<cole-h>
"string '...' doesn't represent an absolute path" :(
fendor has joined #nixos
<lordcirth>
Finally installed NixOS on my main PC - got a weird bug. Every boot, it stops in emergency mode, but pressing Enter resumes and everything works fine?? Root is on ZFS, and EFI boot.
<cole-h>
Problem solved, s@~@/home/user@ lol
<LnL>
:D
<cole-h>
Now it's ever-so-slightly less declarative (because I'm hardcoding my username)
<cole-h>
but meh
alp has quit [Ping timeout: 260 seconds]
tarruda2 has quit [Ping timeout: 258 seconds]
<LnL>
lordcirth: I think I also ran into that when I migrated my pool, but can't remember what the problem was
<bqv>
I think I'd be upset if my secrets weren't in the repo direct, it makes deploying from scratch slightly nontrivial cause you gotta bootstrap the secrets
<Extends>
just wondering, what's the purpose of flake-compat ? I understood it provides a shellNix and defaultNix that can be used, but what do this options provide ? Many people have been using this with no additional options that importing the tarball and add .shellNix at the end or .defaultNix, what does this do ?
<{^_^}>
[nixpkgs] @zimbatm pushed to master « nix-build-uncached: remove the nix wrap (#96357) »: https://git.io/JUkE4
arjen-jo1 has quit [Ping timeout: 260 seconds]
<cole-h>
flake-compat is for non-flakes-enabled Nix
<{^_^}>
[nixos-search] @garbas closed pull request #159 → add support for `nix develop` → https://git.io/JUk8T
orivej has quit [Ping timeout: 240 seconds]
<cole-h>
So if you were to use Nix 2.2, you can still have access to a project that has since switched to the flakes infrastructure
<Extends>
ah so using a system that doesn't allow flakes with flake-compat allow me to use the outputs in a shell.nix for instance ?
<{^_^}>
[nixos-search] @garbas reopened pull request #159 → add support for `nix develop` → https://git.io/JUk8T
<cole-h>
e.g. `nix-build` will still work as expected
<lordcirth>
,tofu
<{^_^}>
To get a sha256 hash of a new source, you can use the Trust On First Use model: use probably-wrong hash (for example: 0000000000000000000000000000000000000000000000000000), then replace it with the correct hash Nix expected. For inserting 52 0's in vim: <esc>52i0<esc>
<elvishjerricco>
How do I get ZFS to be available in `runInLinuxVM`?
<{^_^}>
[nixpkgs] @chiroptical opened pull request #96399 → pythonPackages.jupyterlab-git: init at 0.20.0 → https://git.io/JUkEE
<lordcirth>
I'm trying to tofu a new package src, using nixUnstable. But the error has the hash in a different format than the nix file??
jbox has joined #nixos
ishaan has joined #nixos
<lordcirth>
Using fetchFromGitHub
<Extends>
what do you mean it has a different format?
<cole-h>
Yes, flakes-enabled Nix now uses SRI hashes by default
<cole-h>
They'll look like `sha256-OrrKUffbhLLhvO36DwVkT+TTPyw+kP1vJK+BYq9En1I=`, for example
<lordcirth>
cole-h, yeah, that's what I see. So what's the easiest way to tofu?
<cole-h>
You can still use that 000000 string as the sha
<lordcirth>
yeah, I did, but then the build error, which I would normally copy in, is SRI
<cole-h>
So, copy and paste it into the sha256 field :)
<cole-h>
Nix is smart enough to be able to use both
<lordcirth>
oh, including the "sha-256" header?
<lordcirth>
Ahah. I assumed the header would bork it
<lordcirth>
Thanks
<cole-h>
However, if you REALLY don't want it to be base64, you can run `nix-hash --type <type> --to-base32 '<hash>'`, which will convert it to the hash format you're used to seeing.
<lordcirth>
thanks!
<lordcirth>
Is there a recommendation for a (master branch) nixpkgs PR?
<cole-h>
I don't think we have a standard, so whatever makes you most comfortable I guess.
waschtl[m] has joined #nixos
<lordcirth>
Ok thanks
<cole-h>
The lowest support Nix version for Nixpkgs is 2.2, which is the first version to support SRI hashes (IIRC), so there should be no problems either way
<LnL>
yeah the minimum version was bumped so all formats are ok now
rajivr has quit [Quit: Connection closed for inactivity]
<colemickens>
so say ZFS is telling me some store paths are corrupt, how can I clear and then re-realize them without deleting all the other paths up (down?) the tree
<flokli>
bqv: sorry to hear that. There were some follow-up fixes after this, which should have trickled down to unstbal
<flokli>
s/unstbal/unstable/
<flokli>
How did it break?
<bqv>
hmm, i'm not entirely sure. i did just update most of my nixpkgs inputs, so i should have any followups, but evidently something in my flake still uses the old fontconfig then
Jackneilll has joined #nixos
<flokli>
I can't really help with that, sorry.
__Sander__ has quit [Quit: Konversation terminated!]
<flokli>
but I fear the /applications/ need to have this code. So if you run applications coming from a flake that still uses another fontconfig that behaves differently…
<flokli>
you migt need to backport these changes into your flake, if that is a thing
<q3k[m]>
<lordcirth "Ok, two bugs: 1) When I enable v"> hm, are you sure this is not a first-vs-subsequent-startups thing?
<lordcirth>
q3k[m], I have rebooted several times
alexherbo2 has quit [Ping timeout: 258 seconds]
<q3k[m]>
it's not about reboots, it's about what's in your $HOME (wherever xfce keep its user configs)
<q3k[m]>
*keeps
<lordcirth>
q3k[m], it does throw errors about ~/.cache/session/ being empty
<q3k[m]>
would you try making a brand new user and logging in as that, seeing if it's still missing panels?
janneke has quit [Ping timeout: 265 seconds]
<lordcirth>
I guess I could try that
<q3k[m]>
i mean, there's a possiblity videoDrivers influences this, but i find it much more likely something went wrong on your first startup and that caused xfce to save a broken/empty config or something
<lordcirth>
q3k[m], then why does it work perfectly when I revert to without videoDrivers set?
<lordcirth>
Different session saves?
janneke has joined #nixos
<q3k[m]>
oh, if that happens then yeah
<q3k[m]>
that's an interesting interaction.
<lordcirth>
q3k[m], still want me to try another user?
konobi has joined #nixos
<q3k[m]>
i mean, if it wouldn't be too much of an issue, it would be a good (negative) data point to confirm it's not some weird local config thing
<q3k[m]>
s,issue,hassle,
<lordcirth>
No problem. Applying and rebooting
lordcirth has quit [Remote host closed the connection]
<lordcirth>
q3k[m], you might be on to it. user "foo" had default toolbars and icons and such.
<lordcirth>
I don't understand why nvidia would change that, thoug
<q3k[m]>
yeah if your original user has toolbars again when you revert, uh
<lordcirth>
I could move ~/.config/xfce4 anyway and see
<q3k[m]>
maybe it's something like xfce imagining that this is a screen for which it has no config, because technically it sees it as another screen when it's visible via the new drivers?
<lordcirth>
That might make sense...
<q3k[m]>
one stupid thing you can try
<q3k[m]>
is revert to the default drivers
<q3k[m]>
and see if foo loses toolbars
<q3k[m]>
more datapoints more better
<lordcirth>
Ooh, I set my panels to specific monitors...
<q3k[m]>
ah!
lordcirth has quit [Remote host closed the connection]
alexherbo2 has quit [Ping timeout: 264 seconds]
diamondburned has quit [Read error: Connection reset by peer]
ManiacOfMadness has joined #nixos
diamondburned has joined #nixos
lordcirth has joined #nixos
<lordcirth>
q3k[m], sorry, xfce crashed when I opened the panel settings :P
lunik13 has joined #nixos
<lordcirth>
I've rebooted back, I'll set the panel monitors to auto, then boot to nvidia again
<lordcirth>
That's a bad xfce bug, tho
julm has joined #nixos
lordcirth has quit [Remote host closed the connection]
CptCaptain has quit [Quit: WeeChat 2.9]
lordcirth has joined #nixos
lunik1 has quit [Ping timeout: 265 seconds]
lunik13 is now known as lunik1
alexherbo2 has joined #nixos
<lordcirth>
q3k[m], yup, that did it - I have nvidia + working panels now, thanks!
<q3k[m]>
you're more than welcome.
<lordcirth>
crashing the whole desktop when I open settings is pretty nuts, though
tdeo has quit [Ping timeout: 260 seconds]
<Ashy>
god i hate nvidia
<lordcirth>
Seems like more of an XFCE bug to me, tbh
<bqv>
because holy christ i'm lost with our python infra
thc202 has quit [Ping timeout: 240 seconds]
werner291 has quit [Quit: werner291]
alexherbo2 has quit [Ping timeout: 240 seconds]
alexherbo2 has joined #nixos
<zeta_0>
i am finally able to build nixos on my raspberry pi, but for some reason, when i rebooted, it went to a blank screen. maybe has something to do with gnome3/gdb and someone told me it might be because the hdmi display is not working correctly on the pis? here's the repo i used to build nixos image: https://github.com/Robertof/nixos-docker-sd-image-builder , and here's the issue posted on reddit: https://github.com/Ro
<zeta_0>
bertof/nixos-docker-sd-image-builder
ManiacOfMadness has quit [Ping timeout: 272 seconds]
fendor__ has joined #nixos
kenran has joined #nixos
<zeta_0>
do any of you know of anyone that got nixos to correctly build and work on a raspberry pi 4 ?
tdeo has joined #nixos
Zee1nOnly has quit [Remote host closed the connection]
fendor_ has quit [Ping timeout: 256 seconds]
Zee1nOnly has joined #nixos
alexherbo2 has quit [Ping timeout: 258 seconds]
<lordcirth>
zeta_0, ask on #nixos-aarch64
<zeta_0>
also, does anyone here know if emacs27 will be put in the nix unstable channel ? last i checked, it wasn't there.
kenran has quit [Ping timeout: 246 seconds]
<zeta_0>
lordcirth: ok, i'll try that channel, i hope it's not dead though.