<dongcarl>
I love it! It's the perfect number for it too! XP
<henry_>
The saga continues! Nix tells me that it was not compiled with libsodium ("error: Nix was not compiled with libsodium, required for signed binary cache support") and yet `./configure` says that Nix was compiled with libsodium ("checking for libsodium... yes").
eoli3n has quit [Ping timeout: 272 seconds]
m0rphism has quit [Ping timeout: 260 seconds]
<henry_>
What do I even do about this?
<infinisil>
:D
eoli3n has joined #nixos
<henry_>
ornxka: "the end result of that is that the entire ecosystem, in unix land and elsewhere, is just this quicksand swiss cheese mire of buggy broken stuff trying to communicate with other buggy broken stuff and nothing works and you spend hours a day having to debug it" It's almost like we need some sort of...official, dare I say government, standard.
<ornxka>
i am all for it
noudle has quit []
<ornxka>
brutal totaliatarian free software open standards oligarchy
<slack1256>
json based
<slack1256>
Well no, nix-based.
<slack1256>
Almost.
<henry_>
One System to Rule Them All / One System to Find Them / One System to Bring Them All / And in the Sunshine Bind Them.
<ornxka>
literally this
<ornxka>
i think in the future such a thing will happen
<henry_>
How could it not?
<henry_>
It's all up from here, baby.
<ornxka>
:D
<henry_>
;)
<henry_>
C is the first target.
<henry_>
There's just no way around it.
<ornxka>
well... its tricky
<ornxka>
its true that its easy to make mistakes in C
<henry_>
Undefined behaviors will be illegal and possession of undefined behaviors will be punished by public flogging.
<ornxka>
haha
<ornxka>
i agree with that
<ornxka>
but i think the solution is to both use c, and also to not have undefined behavior
<gchristensen>
lol
<ornxka>
people think the solution is to use languages that are "correct by construction", where you cant make certain kinds of mistakes
<henry_>
And the Second Commandment will be: Thou shalt not permit out-of-bounds access.
<ornxka>
but i think that just kicks the can further down the road because a lot of the most important properties of programs arent just "has no RCE errors" but are properties specific to the purpose of the program
<ornxka>
you cant make a programming language where you can only implement your cryptographic algorithm correctly without making mistakes, the entire idea of it doesnt make sense
<rooke>
Sure, but it helps to rule out the errors you can rule out
<gchristensen>
languages that prevent certain kind of mistakes don't prevent all kinds of mistakes
<ornxka>
so stuff like rust only gets you halfway to the finish line, the place where things are known to be correct
<gchristensen>
so your statement about "without making mistakes" doesn't really follow
<henry_>
Formal verification of correctness.
<ornxka>
and then, things that are correct can be correct even in languages where it is easy to make mistakes - you can write C programs that dont have any bugs in them and are more secure than badly-written rust programs, javascript RCE poc code, etc
<ornxka>
yes, thats the only thing that gets you to the promised land of things actually working
<ornxka>
you need to win the whole war and not just the RCE battle
jluttine has quit [Ping timeout: 240 seconds]
<henry_>
You know...
<ornxka>
not having remote code execution bugs is really just the simplest possible property a program can have, that it doesnt turn into an assembly language interpreter if you give it the right input
<henry_>
If correct software could be written in C, you would think that people would've figured it out by now.
<ornxka>
big boy programs do stuff like onion routing and complicated cryptographic protocols that have just as much, if not more, of a need to be correct
<ornxka>
henry_: it exists
<gchristensen>
it turns out to not be a simple property
<ornxka>
it just takes an enormous amount of labor ro write
<gchristensen>
anyway, maybe thi sshould go to #nixos-chat
<ornxka>
avionics software is often written in C++ iirc
<ornxka>
oh is that what its called? i joiend #nixos-offtopic and nobody was there..
<henry_>
Imagine using a language where integer overflows are a thing.
<henry_>
I'm sure it's ugly and inelegant, but I would feel uncomfortable if I knew my airplane was running something important in C++.
<henry_>
I simply know too much.
<Dandellion>
so I tried overriding a package and uhh I got this `/nix/store/3g5jkbi1yabcgwm5hbxv8vvyq1p340zd-libaom-1.0.0-g33e9b7fb1/nix/store/3g5jkbi1yabcgwm5hbxv8vvyq1p340zd-libaom-1.0.0-g33e9b7fb1/` as the nix-store path. how does that even happen tbh
<Dandellion>
like I just added a simple overlay like this:
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/JfUyZ
shibboleth has quit [Quit: shibboleth]
<nh2>
Dandellion: that looks like an issue that sometimes apepars with cmake, libaom seems to use cmake
selfsymmetric-mu has joined #nixos
<Dandellion>
do I just... move them back manually
ixxie has quit [Ping timeout: 256 seconds]
<henry_>
Dude, did your Nix install Nix within your Nix?
ixxie has joined #nixos
<Dandellion>
I HEARD YOU LIKED NIX STORES
<henry_>
lol
<ornxka>
henry_: i just know there are some standards for writing c++ code for mission-critical stuff and its full of stuff like "no functions over 100 lines" "no using any c++ features except namespaces and god help you if you try using stuff like exceptions and vectors" etc
morgrimm has joined #nixos
<rotaerk>
I love how every time I update linux, *something* breaks
<{^_^}>
[nixpkgs] @emilazy opened pull request #85589 → Add ACME maintainers team → https://git.io/JfUya
<henry_>
ornxka: Yeah pretty much, lol.
<rotaerk>
troubleshooting time...
<nh2>
Dandellion: I can't quite find the issue, but I think this can be fixed by passing some `-DCMAKE_INSTALL_PATH` explicitly, or osmething like that
<ornxka>
somehow i knew what that was going to be before i clicked on it
morgrimm has quit [Ping timeout: 265 seconds]
hlolli has quit [Ping timeout: 272 seconds]
<henry_>
"OK, so this new language had to divorce itself from Unix, by hiding all the system calls that bound the two together so nicely. This would enable guys who only knew about DOS to earn a decent living too."
<henry_>
On a completely unrelated note, my Nix-installing-Nix stdout is up to ~12k lines of stdout.
<cole-h>
That's good
<cole-h>
Soon it'll fail hard with an error from 10k lines ago >:)
<cole-h>
(I kid)
ixxie has quit [Ping timeout: 260 seconds]
<henry_>
=/
ixxie has joined #nixos
<henry_>
Imagine being able to understand your system.
<MichaelRaskin>
Like, actually understand?
<MichaelRaskin>
I am afraid this requires too much destruction…
<cole-h>
Here is what I understand: 101010 -> IRC -> endorphins when someone pings me
<selfsymmetric-mu>
None of these look like my nix files.
<ornxka>
MichaelRaskin: highly doubtful
<MichaelRaskin>
selfsymmetric-mu: well, the last part tells the problem
<MichaelRaskin>
Your config needs ZODB and the package is marked broken
<selfsymmetric-mu>
MichaelRaskin: Indeed, I've gotten THAT far.
<selfsymmetric-mu>
But I don't know _what_ part of it requires it.
<selfsymmetric-mu>
Oh, I can try doing a `nix store --query`.
<henry_>
I believe in having a system in which my syscalls are the assembly.
<cole-h>
I would start by looking at anything touching python
<MichaelRaskin>
You can instantiate with NIXPKGS_ALLOW_BROKEN=1 environment variable, then query the reference graph
<selfsymmetric-mu>
cole-h: That's what's weird. I have all my Python sandboxed into shells. It might be something built with python.
<cole-h>
Oh hey, it looks like it's from vulnix?
<cole-h>
while evaluating the attribute 'propagatedBuildInputs' of the derivation 'vulnix-1.9.4' at /nix/store/vkalyc5421xsfhxfki6pd1qv8xw3byad-nixpkgs-20.03.1405.a17e021b948/nixpkgs/pkgs/development/interpreters/python/mk-python-derivation.nix:105:3:
<nh2>
Dandellion: great, maybe that fix should go straight into the package?
inkbottle has quit [Ping timeout: 256 seconds]
lmz has quit [Remote host closed the connection]
<Dandellion>
it hasn't had a release yet, I'm just building the master branch, but once the package needs updating itll be needed I guess
<selfsymmetric-mu>
Hey, henry_ what's Nix-installing-Nix? Sounds like dark magic.
Diogo has quit [Ping timeout: 265 seconds]
_viz_ has joined #nixos
reallymemorable has quit [Quit: reallymemorable]
<henry_>
selfsymmetric-mu: It does, and it is.
<henry_>
My compiled Nix is installing Nix's Nix.
wavirc22 has quit [Ping timeout: 260 seconds]
<emily>
bqv: so I thought my flakes setup was working fine, but... my `flake.lock` has an old commit of my nixpkgs fork under `"locked"`, it doesn't update when I rebuild my system or even explicitly `nix flake update`, but new commits of my repo *do* get used when I rebuild. no idea what's going on here
<emily>
bqv: if I do `nix flake update --recreate-lock-file` it seems to recreate it and update it properly. but I'm not sure what the point of the lock file is at all if `nixos-rebuild` is just ignoring it?
<selfsymmetric-mu>
henry_: godspeed on your strange adventure
markus1189 has quit [Ping timeout: 258 seconds]
<henry_>
selfsymmetric-mu: Thanks
<henry_>
MichaelRaskin has suggested that I uninstall the manually compiled Nix ("remove non-Nixpkgs Nix").
<henry_>
Hopefully that will be straightforward.
<MichaelRaskin>
Depending on where you installed it…
<MichaelRaskin>
Hmmm. I now have a crazy suspicion
<henry_>
Hit me.
<MichaelRaskin>
'I guess one can pull Hydra-built Nix for armv7 on a non-armv7 machine via substituters
Diogo has joined #nixos
<MichaelRaskin>
One couls just rsync its closure to the target system…
hugbubby has quit [Ping timeout: 264 seconds]
<MichaelRaskin>
It will not recognise itself as valid, but might be able to pull _and_ register another Nix from cache (preferably a stdenv rebuild away, so there are no conflicting paths)
<MichaelRaskin>
Then the new Nix is a valid Nix-installed Nix…
<henry_>
Radical.
<henry_>
Where were you a week ago?
<henry_>
lol
ixxie has quit [Ping timeout: 256 seconds]
reallymemorable has quit [Quit: reallymemorable]
<MichaelRaskin>
I don't even know what exactly is available in cache for ARMv7!
<henry_>
What is Hydra, what is a closure (in this context), and what is the cache?
<MichaelRaskin>
No, standard install script uses a build of Nix that can be downloaded without Nix
<atemu12[m]>
gchristensen: Yeah, it has a lot of packages that are annoying to build locally like mozjs.
<atemu12[m]>
What's the status on getting hydra to build armv7l successfully again by the way? I can see that a bunch of jobs are queued but none are being built.
<henry_>
MichaelRaskin: Well it will be very cool if your idea works out.
<MichaelRaskin>
Haven't you pretty much succeeded in building Nix anyway?
<MichaelRaskin>
(by now)
<henry_>
I think so, probably. But I like elegant solutions.
reallymemorable has joined #nixos
<henry_>
Especially when I've spent as much time as something (like this) as I have.
<henry_>
*on something
<henry_>
It's been a pain, let me tell you.
<henry_>
The size of my /nix is now larger than my entire uncompressed Raspbian image.
<MichaelRaskin>
That is normal…
<cole-h>
I was wondering why I only had 50G free
<cole-h>
75106 store paths deleted, 85528.04 MiB freed
<cole-h>
That'll do it
<ornxka>
error: build of '/nix/store/k4m1j5pmd8rkc4k5cj1caj43fq5w736f-compiler-rt-9.0.1.drv' on 'ssh://builder' failed: builder for '/nix/store/k4m1j5pmd8rkc4k5cj1caj43fq5w736f-compiler-rt-9.0.1.drv' failed to produce output path '/nix/store/3xwnxx4hxim49w2h97xgnkcqsy4gcjxy-compiler-rt-9.0.1'
<ornxka>
what does this error mean
<cole-h>
From that snippet, it looks like it failed to build. Can you share more logs?
wavirc22 has joined #nixos
proofofkeags has joined #nixos
<atemu12[m]>
cole-h: What the hell did you have your store that took 80G?
hio has quit [Quit: Connection closed for inactivity]
klntsky has joined #nixos
fusion809_ has joined #nixos
<henry_>
I'm sure it is normal, MichaelRaskin, but isn't it also kind of absurd?
fusion809 has quit [Ping timeout: 246 seconds]
h0m1 has quit [Ping timeout: 272 seconds]
<emily>
is there a way to get "nix build" to print all the packages it's planning to build up-front, a la "nix-build"?
<emily>
or at least print just the package names (and nothing else) as it builds them?
<emily>
I kinda rely on that to tell when the channel is having issue and I probably shouldn't build Chromium...
fusion809_ has quit [Remote host closed the connection]
h0m1 has joined #nixos
<cole-h>
Would `--dry-run` work?
fusion809_ has joined #nixos
<cole-h>
Hmm, doesn't seem like it...
<emily>
unexpected downsides of switching my system config over to flakes >.>
<atemu12[m]>
<emily "or at least print just the packa"> Does nix not have the --no-build-output flag?
<emily>
it doesn't print output by default
<henry_>
emily: With software I never try to paint outside the lines.
<emily>
but it only shows currently building packages in a live-updated status line
<emily>
rather than keeping them on-screen in a log, or printing them all upfront like nix-build
<henry_>
Trying to use an Unofficial Config is invariably a total waste of my time.
<henry_>
Just saying.
drakonis has quit [Quit: WeeChat 2.8]
slack1256 has quit [Remote host closed the connection]
<emily>
henry_: not sure what you're talking about, sorry
<cole-h>
emily: You might be able to try `nix log`
<henry_>
emily: You know the old joke, Linux is free if your time is worth nothing? It's like that.
<selfsymmetric-mu>
nix-top doesn't help here, right?
<emily>
henry_: what do you mean by unofficial config?
<emily>
selfsymmetric-mu: I don't think so, since it'll only notice stuff once it actually starts building, right?
<henry_>
"unexpected downsides of switching my system config over to flakes >.>"
<emily>
I admit this specific set of requirements is kind of niche-sounding... but I often glanced at the stuff to build when updating my profile to know tell when there's channel sync issues
<cole-h>
henry_: flakes is official. It's just experimental. Depends on your definition of official, I guess.
<emily>
henry_: ah. well, flakes is an official feature that's going to replace the existing channels machinery. it's true that it's currently experimental, but given that I run nixos-unstable and regularly contribute PRs I'm probably one of the better-suited people to try being an early adopter
<ornxka>
...what is the point of the boot.zfs.enableUnstable option when all of the linuxPackages.*.zfsUnstable packages are the same version as linuxPackages.*.zfs package
<emily>
in this case it's just because the flakes-enabled nixos-rebuild uses the fancy newer nix 2 build UI
<henry_>
I have no doubt you're both far more expert than me, lol.
<emily>
which is generally nicer, but lacks a few specific features like this
<cole-h>
emily: But I agree, if there isn't tooling that gives the same (or similar) output as `nix-build`, it should. Maybe time to file an issue? None of the usual experts appear to be here right now.
<emily>
ornxka: sometimes they change
<emily>
ornxka: (if an unstable zfs release supports the latest kernel but stable doesn't, say)
<emily>
generally speaking you only want to turn enableUnstable on when your system breaks because zfs doesn't work :P
<ornxka>
i see
<henry_>
And sure, I'm here after spending twenty-ish hours chasing down and corralling obscure Linux-isms in order to manually compile and configure Nix for an architecture it doesn't easily support.
<ornxka>
zfs is indeed not working
<emily>
cole-h: mhm, I might open an issue or something on the discourse
<ornxka>
i guess ill try unstable kernel
<emily>
is it a build issue? might be that a kernel update has pulled zfs out of sync
<ornxka>
it locks up under load
<cole-h>
ornxka: Do you have swap in a zvol, by chance?
<ornxka>
nope
<emily>
ah, that's more likely to be configuration, hardware, or upstream openzfs issue than something relating to nixos kernel versions
<ornxka>
i dont use swap generally
<ornxka>
i might enable it later if i can get everything working though since only 2GB of ram
Maxdamantus has quit [Ping timeout: 258 seconds]
<cole-h>
Oh, I was asking because it's a "Known issue" of NixOS on ZFS: "You shouldn't use a ZVol as a swap device, as it can deadlock under memory pressure"
<emily>
it's not quite the RAM hog it has a reputation for being unless you enable dedup but that still sounds like skirting the limits a little
<ornxka>
cole-h: ah yeah
wavirc22 has quit [Quit: ZNC 1.7.2+deb3 - https://znc.in]
<ornxka>
i dont really mind if its slow i just want it to work
<ornxka>
where "work" precludes 15 minute lockups
Maxdamantus has joined #nixos
<emily>
right. well, 15 minute lockups sound like the kind of thing that could happen when the memory manager gets really unhappy, theoretically
<emily>
but i'm no expert, so it could be something entirely different
<emily>
one thing to remember is that zfs memory management is kinda detached from the rest of the kernel because of the solaris compatibility layer / heritage
<emily>
which can cause some oddities
wavirc22 has joined #nixos
<ornxka>
it had memory free when it locked up is the weird thing but it might still be the problem, who knows
<freeman42x[m]>
how do I change NixOS to use en_US instead of en_IE for everything? getting this error: https://i.imgur.com/Sxisz74.png, and I'd like to switch to en_US
<samueldr>
freeman42x[m]: should be under i18n.*
<freeman42x[m]>
samueldr: do you know what the setting to change is called and what would be a valid value? it is not in my file yet
selfsymmetric-mu has left #nixos ["gone to the land of dead hiccups and extinguished light bulbs"]
<cole-h>
Probably looking for `i18n.defaultLocale`
<qy[m]1>
<emily "unexpected downsides of switchin"> I call it an upside :p I just use `nix log` more
<qy[m]1>
<emily "bqv: if I do `nix flake update -"> I thought that was intentional, for the purpose of reproducible builds. Update does add new inputs, etc
<freeman42x[m]>
so useful, didn't know there was site also for searching configuration.nix options, I was using `man configuraiton.nix`
_viz_ has quit [Remote host closed the connection]
<emily>
qy: how are you meant to update to a new version of dependency if not "nix flake update"?
gentauro has quit [Ping timeout: 264 seconds]
<emily>
"nix flake update --recreate-lock-file" seems like a very weird thing to have to write every time you want your project to not be on a months-old nixpkgs
<emily>
maybe I just don't understand the model, but I was expecting "nix flake update" to work like "niv update"
wavirc22 has joined #nixos
brent__ has joined #nixos
<bqv>
emily: I think you can force update a specific input, but I normally just recreate
ixxie has quit [Ping timeout: 256 seconds]
<emily>
re upside: "nix log" can't tell me what the hundreds of things that are queued to build after networkmanager and flatpak and wfupd are
<emily>
wfupd → fwupd
<emily>
just, those packages are core enough that I start sweating whenever I see them getting built and scared for the Chromium or kernel compile that's coming next
CMCDragonkai_ has joined #nixos
<qy[m]1>
Heh, fair enough.
fusion809_ has quit [Ping timeout: 240 seconds]
<qy[m]1>
dry-build could, if you grew a habit of using it, but yeah it's less straightforward
<CMCDragonkai_>
I have a Python derivation that exports an executable script that is meant to call another CLI executable in one of its dependencies.
<CMCDragonkai_>
Putting it in propagatedNativeBuildInputs or nativeBuildInputs does not expose that CLI executable to my script when the whole derivation is built.
<CMCDragonkai_>
Is the right way to deal with this is `wrapProgram`?
codygman has quit [Read error: Connection reset by peer]
gentauro has joined #nixos
codygman has joined #nixos
wavirc22 has quit [Ping timeout: 250 seconds]
<{^_^}>
[nixpkgs] @delroth opened pull request #85591 → pythonPackages.pyopengl: avoid dragging in 500M of -dev packages → https://git.io/JfUHo
<cole-h>
What about `propagatedBuildInputs` (no native)? Just spitballing, but that's what is used in the fish-shell derivation because it needs access to python and gettext and things.
codygman has quit [Read error: Connection reset by peer]
codygman has joined #nixos
morgrimm has joined #nixos
aw has quit [Quit: Quitting.]
spacefrogg has quit [Quit: Gone.]
aw has joined #nixos
spacefrogg has joined #nixos
ixxie has quit [Ping timeout: 265 seconds]
brent__ has quit [Ping timeout: 250 seconds]
<CMCDragonkai_>
cole-h: no that does not work
<delroth>
CMCDragonkai_: the usual way of doing this would be to not require on PATH to find your dependency's executable, but instead "hardcoding" the path
<delroth>
I don't know about Python, in many native programs you would have e.g. a ./configure option to point to the executable's location
morgrimm has quit [Ping timeout: 256 seconds]
<cole-h>
You could `wrapProgram`, or you could replace the path in the source
<delroth>
but you could just sed -i in patchPhase
cortexman has left #nixos [#nixos]
brent__ has joined #nixos
<delroth>
e.g. patchPhase = "substituteInPlace foo.py --replace myprog ${pkgs.myprog}/bin/myprog"
<cole-h>
(Usually, that's done in postPatch, so that patchPhase still takes effect)
<delroth>
good point
<CMCDragonkai_>
delroth: I reckon using wrapProgram is better
<CMCDragonkai_>
delroth: That way you don't need to know every file where its being used
zupo has joined #nixos
<CMCDragonkai_>
delroth: Plus you can use makeWrapperArgs too
<emily>
qy: hm, howso? --dry-run just spits out the system .drv and bails
<CMCDragonkai_>
also i put the wrapProgram in `postFixup`
butterthebuddha has quit [Ping timeout: 272 seconds]
HeN has quit [Quit: Connection closed for inactivity]
<qy[m]1>
emily: ah, yeah i think i had that happen recently too. strangely it worked fine once i started the system build and cancelled it as soon as it was querying/fetching/building
<CMCDragonkai_>
I wonder if there's a difference between using setup.py scripts or console scripts?
_viz_ has joined #nixos
<qy[m]1>
(i usually use it retroactively once i realise it's trying to build 3 thousand things and i spit out my coffee
<qy[m]1>
* (i usually use it retroactively once i realise it's trying to build 3 thousand things and i spit out my coffee)
<emily>
ah interesting
<emily>
thanks for the tip, that's better than nothing at least
<emily>
it does indeed seem to work then
<emily>
ok, something is seriously up, because it wants to rebuild gcc and bzip2 and coreutils
zupo has quit [Ping timeout: 256 seconds]
brent__ is now known as fusion809
<emily>
I wonder what on earth is up with my nixpkgs
<qy[m]1>
`nix-diff` will save you hours there, i think
stevenxl has joined #nixos
<emily>
sigh, so much exciting opportunity to use these fancy tools :P
<qy[m]1>
:D instantiate coreutils from a plain version of whatever channel you're using, and then nix-diff will probably give you an idea of what the root difference is if you diff that drv with the one nix is trying to build
ixxie has joined #nixos
<emily>
mhm
wavirc22 has joined #nixos
<qy[m]1>
i did the same when i accidentally modified systemd...
<emily>
unfortunately everything seems cached when I just do "nix run -f . whatever" in ~/src/nixpkgs
<emily>
it's only when building my system config, pinned to the exact same revision, that stuff starts breaking somehow
<emily>
so I'll probably have to nix-diff the systems themselves
<qy[m]1>
well if you can sleuth one of the drvs you're building from the nix store with bash magic, you can just fetch the corresponding drv from hydra
<qy[m]1>
nix-instantiate channel:nixos-unstable -A coreutils
<qy[m]1>
* `nix-instantiate channel:nixos-unstable -A coreutils`
<emily>
mhm
felixfoertsch23 has joined #nixos
<emily>
is there a way to go from a (not yet existing) output store path to the .drv path that's meant to produce it?
<emily>
presumably it's possible since nix can look it up to do the build but I don't know what the nix-store command is... I should probably try and memorize its manpage
felixfoertsch has quit [Ping timeout: 256 seconds]
felixfoertsch23 is now known as felixfoertsch
<emily>
hehe, thanks (and sorry, I hit that part of the manpage right as you said that)
<qy[m]1>
:D
<emily>
I feel like after quite a while using NixOS I know 10% of the commands, and in 5 years I will have learned twice as much stuff as I currently know and still somehow only know 10% of the commands
<qy[m]1>
also, i just checked, and i have 19 coreutils drvs on my system
<qy[m]1>
..
<qy[m]1>
to be fair, some of the incantations are straight up arcane
phreedom has quit [Remote host closed the connection]
<emily>
it, um, actually wants to build coreutils 3 times for the one system
phreedom has joined #nixos
<qy[m]1>
well that explains that.
jarlg has joined #nixos
<emily>
it also wants to build 5 curls??
codygman has quit [Ping timeout: 240 seconds]
<emily>
it's possible nix build --dry-run --no-link -L .#nixosConfigurations.renko.config.system.build.toplevel is not quite the right incantation
<qy[m]1>
can't speak for curl but coreutils might be part of the gcc bootstrap process
hlolli_ has quit [Ping timeout: 272 seconds]
<emily>
oh, --deriver seems to only work for a prebuilt derivation
<emily>
(at least it complains error: path '/nix/store/psmplh6x0fja24yai0r4law2ap8dvcnh-fish-3.1.0' is not valid)
<qy[m]1>
ah, right, yeah
codygman has joined #nixos
<qy[m]1>
i imagine if gcc is changing, most of your system is. might be easier to diff a higher level drv
<qy[m]1>
that's only compiling once
ixxie has quit [Ping timeout: 258 seconds]
<emily>
yeah, fish was my choice of derivation -- just need to figure out how to, uh, fish the .drv path out of my system
ixxie has joined #nixos
<qy[m]1>
hail mary: gc your store, eval the system, then find the drv with no corresponding output :D
<emily>
ahaha
<emily>
ooh maybe I can pass the target path to nix build and strace it to see the .drv it opens
<emily>
oh, I can't since it uses the daemon :')
<qy[m]1>
ahah
<energizer>
i'm using a program that has a bug in it and i want to report. the program doesn't have a --version command, how can i find out?
evertedsphere has joined #nixos
<cole-h>
`readlink $(which <binaryname>)` might have some useful information
<energizer>
it doesn't tbh
<cole-h>
Does it not point to a path in your store?
<cole-h>
(Or, I'll just remember the command I wanted to use in the first place... :P)
<bqv>
now you're going to type realpath -f someday :)
fusion809 has quit [Ping timeout: 272 seconds]
<cole-h>
Does `nix build` have an analogue to `nix-build -E 'with import <nixpkgs> {}; callPackage somedrv.nix {}'`?
butterthebuddha has joined #nixos
<infinisil>
cole-h: nix build '(<some expression>)' should do it
<energizer>
how do you install your dotfiles?
<cole-h>
infinisil: Could you expand on how that '<some expression>' would translate to a `callPackage`? (e.g. I want to build a "test" derivation not in nixpkgs)
<infinisil>
some expression = with import <nixpkgs> {}; callPackage somedrv.nix {}
<emily>
(or a function from settings to nixpkgs, depending on how you count)
<energizer>
then the `pkgs.` in that line is unncessary, right?
morgrimm has quit [Ping timeout: 256 seconds]
growpotkin has joined #nixos
<bqv>
yes, as it happens
<bqv>
just habit :p
<cole-h>
How does one determine closure size from the `result` symlink again? I remember asking before, but it's been a while since I actually checked...
<bqv>
huh, here's a thought. if a source repo has a default.nix, i can add it as a non-flake input to my config, and update it without having to do that silly tofu dance
<bdju>
is there a quick way to find out when a package was last updated? not for me, but in general
<bdju>
so far I'm searching nixpkgs issues on github
<srhb>
bdju: git log if the commit message is sane, otherwis I guess git blame.
MYMYRSRS has quit [Quit: leaving]
<cole-h>
mkg20001: When you think your PR is mostly ready (if it isn't already), trigger an eval yourself (commend `@ofborg eval`, if you didn't know). GitHub's being a bit whacky right now, so there are no automagic evals
<cole-h>
s/commend/comment/ lol
<bdju>
srhb: I'm not a heavy user of git, would I have to do a pull of all nixpkgs for that to work? I'm just in a browser atm
<mkg20001>
cole-h: `@GrahamcOfBorg eval` or `@ofborg eval` ?
<srhb>
bdju: Yes. But otherwise looking through the commits in the github sort of works too, or visiting the file that defines the package and viewing history, though the search is probably worse than the git cli.
<srhb>
in the github UI*
<cole-h>
Both work, but gchristensen has been wanting to decouple the bot from his name, which is why I suggested `@ofborg` (which won't autocomplete -- don't worry) :P
morgrimm has quit [Ping timeout: 256 seconds]
<mkg20001>
cole-h: ok
ddellacosta has quit [Ping timeout: 265 seconds]
<mkg20001>
cole-h: pr is ready now btw, just a simple jetbrains update. shouldn't set things on fire :P. tho wondering why none of the editors have an updateScript assigned...
<cole-h>
np, I just saw you pushed something else so I didn't want to eval something that wasn't finished
chagra has joined #nixos
<{^_^}>
[nixpkgs] @KAction opened pull request #85595 → laminar: init at 0.8 → https://git.io/JfUdV
chagra_ has quit [Ping timeout: 260 seconds]
<bqv>
by the way, if anyone gets buildMachines working with flakes, hmu
<bqv>
i have an issue but i'm not sure flakes are the cause
<delroth>
hmmm... anyone been getting /tmp chmod issues when activating a new configuration on their system recently?
<delroth>
[/etc/tmpfiles.d/tmp.conf:11] Duplicate line for path "/tmp", ignoring.
<cole-h>
danderson: I'll give you the same advice I have earlier: `@ofborg eval` when your PR is ready to be eval'd. GitHub's webhooks seem to be screwy right now.
<cole-h>
s/have/gave
<danderson>
thx
<cole-h>
np. Nothing obvious is showing up in t he logs, so I can't do much more than warn people :P
asheshambasta has joined #nixos
palo1 has joined #nixos
<{^_^}>
[nixpkgs] @danderson opened pull request #85598 → nixos/tailscale: set a CacheDir in the systemd unit. → https://git.io/JfUFu
palo has quit [Ping timeout: 260 seconds]
palo1 is now known as palo
Jackneill has joined #nixos
zupo has joined #nixos
<{^_^}>
[nixpkgs] @peterhoeg opened pull request #85599 → nixos/https-dns-proxy: system level support for DNS over HTTPS [WIP] → https://git.io/JfUFo
<sb0_>
I switched to 20.03 from 19.09, and I have the hydra-migration package installed
<sb0_>
however hydra-backfill-ids does not work:
<sb0_>
main::backfillJobsJobsetId(): DBI Exception: DBD::SQLite::db prepare failed: no such column: jobset_id [for Statement "SELECT COUNT(*) FROM jobs WHERE jobset_id IS NULL
Baughn has quit [Read error: Connection timed out]
Baughn has joined #nixos
thongpv871 has joined #nixos
<{^_^}>
[nixpkgs] @FRidh merged pull request #85591 → pythonPackages.pyopengl: avoid dragging in 500M of -dev packages → https://git.io/JfUHo
<{^_^}>
[nixpkgs] @FRidh pushed commit from @delroth to master « pythonPackages.pyopengl: avoid dragging in 500M of -dev packages »: https://git.io/JfUbi
<qy[m]1>
jluttine: your best bet is home-manager. There was a pr somewhere for adding a gtk/qt module to nixpkgs but its not going to get merged in our lifetimes
<jluttine>
qy[m]1: hmm.. why doesn't just plain installation work? installing gnome3.adwaita-icon-theme adds icons under /run/current-system/sw/share/icons/Adwaita. however, my apps don't seem to find those, but that might be a separate issue - or not.
<jluttine>
i guess i'd need to understand how apps are trying discover icons..
<jluttine>
i thought they just look under $XDG_DATA_DIRS/icons
chagra has quit [Ping timeout: 256 seconds]
ixxie has quit [Ping timeout: 260 seconds]
lsix has joined #nixos
ixxie has joined #nixos
snupples[m] has joined #nixos
<{^_^}>
[nixpkgs] @parthy opened pull request #85603 → edk2: Support build on macOS → https://git.io/JfUpe
smatting has joined #nixos
maddo has joined #nixos
hmpffff_ has quit [Ping timeout: 265 seconds]
hlolli_ has joined #nixos
hmpffff has joined #nixos
fusion809 has quit [Read error: Connection reset by peer]
<turion>
Iterate over all entries, generate the new attributes you want to add, merge the attrsets
rauno has quit [Remote host closed the connection]
<turion>
Wow, the standard library reference now has Haskellish type signatures... was that always there and I didn't notice?
snicket has joined #nixos
Vikingman has quit [Ping timeout: 256 seconds]
cfricke has joined #nixos
broccoli has joined #nixos
<turion>
In tests, is there something like `machine.writefile` or so? Or do I need to do `machine.succeed("cat '....\n...\n...' > myfile.txt")?
fusion809 has joined #nixos
<jluttine>
is there some way to check if an icon theme is installed correctly? or how do applications search for icons? if i say, use icon foobar, then how is searched? is there some standard tool to find the icon from the system?
<turion>
peelz: I think lib.attrsets.mapAttrs is the easiest
<peelz>
turion: I don't think I can iterate over `users.users` because it would result in infinite recursion (e.g. `users.users = map (...) config.users.users`)
<peelz>
It needs to be something that gets evaluated when merging values
<turion>
peelz: How do you currently configure your users? Do you just plain write users.users = .... in your configuration.nix?
<peelz>
turion: yes but I'm trying to create modules and I don't want to hardcode users
<turion>
peelz: Ok that's fine. But as long as you have control over where they're defined, you can just
<turion>
* import the result in your configuration.nix
<turion>
* apply mapAttrs (_username: attrs: attrs // { extraGroups = ["somegroup"]; }) to them
<turion>
* define your own attrset myUsersBeforeAddingGroup = { userfoo = {...}; ...};
<turion>
In that example, myUsersBeforeAddingGroup is where you do user definition
<turion>
Or did I misunderstand and you want to imperatively declare users? (i.e. by executing useradd)
<peelz>
turion: yeah I think I'll be building my users from another module so I can just reference those usernames when building `config.users.users.<name>.extraGroups`
<peelz>
turion: no, all declarative
<turion>
Ok, then this should work, I think
<peelz>
yep thanks a bunch! turion++
<{^_^}>
turion's karma got increased to 1
<turion>
Feel free to send a gist if it doesn't quite work yet :)
<turion>
srk: When I grep for it, I get swamped by the thousands of uses :/
<turion>
Ah ok
<srk>
turion: jump into lib folder
<turion>
Fair enough :)
<turion>
With other things like writeFileDir I'm always clueless how to find them :/ I don't really want to start memorizing the nixpkgs directory structure
<srk>
turion: yeah, it's not ideal - lib and pkgs/build-support/trivial-builders.nix are quite common
<srk>
hmm, I might be able to do such function lookup..
<turion>
I start to understand why guix is a thing... nix is far from a "proper" programming language, missing a lot of tooling
<srk>
turion: I somewhat agree but due to different reasons. what are you missing btw?
<peelz>
looks like functions aren't comparable lmao
<manveru>
anyway, the __ naming is a side-effect of how the parser works, afaik...
<peelz>
oh is it? :O
<manveru>
they're exactly the same as the builtins, just shorter to write :)
<peelz>
ahh interesting
<peelz>
is it bad practice to use the __ prefix?
<peelz>
or does it not matter?
sigmundv__ has joined #nixos
morgrimm has joined #nixos
magnetophon has quit [Ping timeout: 256 seconds]
<manveru>
i use it all the time... but not sure if that'll ever change at this point
<peelz>
is it considered bad form for nixpkgs code?
<manveru>
probably :)
<peelz>
aight :P
<manveru>
the weird thing is that not all stuff in `builtins` is also available as prefixed version
<peelz>
HUH
<manveru>
and there are some that aren't prefixed
<peelz>
like map?
<manveru>
yeah
<peelz>
yeah I guess I'd rather just use `builtins.` lol
<manveru>
just open a nix repl and hit tab :)
<peelz>
true
<peelz>
huh what's dirOf?
<manveru>
gives you the parent directory
<peelz>
ah
<manveru>
> dirOf ./.
<peelz>
interesting
<{^_^}>
/var/lib/nixbot/state
<peelz>
oh yeah, what's placeholder?
<manveru>
placeholder is magic :)
<peelz>
I saw it being recommended for referring to "out" but it seems so trange
<peelz>
uh oh
<peelz>
I don't like magic xD
<manveru>
afaik it only works inside a derivation, and you can use it to get the computed out location in nix without having to go through the build scripts
<kazimazi>
hi, how do I build and run ARM (32 bit) Assembly in nix-shell?
ramses_ has joined #nixos
vidbina_ has quit [Ping timeout: 260 seconds]
<ramses_>
Hi guys, I am trying to package a small python web-app with nix. I succeeded to get it built and the python file which I metionned in scripts in setup.py is present in the packages bin directory. But I cannot figure out how to handle static html and css files
ashesham` has joined #nixos
aveltras has joined #nixos
<ramses_>
I added the files in package_data in setup.py, but I cannot see them anywhere in the resulting nix package
sb0_ has quit [Quit: Leaving]
cr4y1_ has joined #nixos
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<ornxka>
it looks like its deleting the store path because its empty
<ornxka>
and then complaining that it failed to produce the output path
<ornxka>
despite actually compiling successfully
<ekleog>
bqv: could you open a PR so it's easier to review? :)
o1lo01ol1o has quit [Remote host closed the connection]
o1lo01ol1o has joined #nixos
<turion>
srk: In terms of tooling & ecosystem, I miss: Typechecking. A module and package system. A central package repository. Autogenerated reference docs. Searchable reference docs.
emilsp has joined #nixos
Baughn has quit [Quit: ZNC 1.6.2+deb1 - http://znc.in]
OmnipotentEntity has quit [Ping timeout: 256 seconds]
<turion>
By package system I mean that I don't have to collect my favourite nix tools on github and import them via fetchgit.
<turion>
srk: Further tooling: Some kind of effect system that allows some limited IO & metaprogramming to get rid of all the hacky helper scripts that semi-automatically update stuff like hackage lists.
<turion>
srk: I've heard about Dhall, but it doesn't do what nix does, right? It's great for creating config, but it doesn't do anything like nix does with derivations?
thibm has joined #nixos
evertedsphere has quit [Ping timeout: 256 seconds]
<peelz>
oops, I guess I've been staring at this for too long -.-
<turion>
I see build systems and metaprogramming as two similar principles. You have some notion of stages across which you want to preserve stuff (in nix, the output of the derivations).
<turion>
My test kills the VM with an OOM error as soon as my program (agda in this case) collects more than 260 MB memory, no matter to what value I set virtualisation.memorySize
<srk>
turion: yes, that's correct regarding Dhall, it offers a lot of interop libraries tho, for example dhall-nix
<wedens[m]>
can I just remove symlinks in /nix/var/nix/profiles/ and do gc?
<turion>
wedens[m]: nix-collect-garbage --delete-older-than I guess
<wedens[m]>
I want to remove specific generations, not just something older than X
<peelz>
How can I make the attributes in userModule.config (line 29-32) merge into the parent config? https://0x0.st/i1fg.txt
<turion>
srk: Ah cool, I didn't know about. "This package targets people who wish Nix had a type system." I confess guilty :D
piezoid has joined #nixos
cosimone has joined #nixos
<srk>
turion: lol!
<turion>
Even if it were a simple type system like Hindley-Milner + row-polymorphic records, it would already help so much, because then we could have something like hool
<turion>
srk: Another tooling point: If there was a decent package system, we could separate nixpkgs.lib from nixpkgs. And then people could reuse it for their own purposes
<{^_^}>
[nixpkgs] @petabyteboy pushed 2 commits to staging: https://git.io/JfTk6
brightone has joined #nixos
<Baughn>
While I think I know what the answer will be, uh...
<Baughn>
Is it still possible to run KDE 3 or 4?
<brightone>
Hello, can someone help me with Hydra configuration? I'm trying to set up GithubPulls plugin, and it just doesn't pick up my tokens in config. If this is offtopic, please let me know wher I could get this help. Thanks!
<infinisil>
Baughn: If it's in an older version of nixpkgs the answer might be yes
<Baughn>
I don't particularly want to run an older version of nixpkgs, though. :P
<infinisil>
Then the answer might be no :P
<Baughn>
Well, I knew it wasn't going to work out. It's just that unfortunately none of the desktop environments that are available actually work for me.
<Baughn>
They all have some critical bug or other.
aveltras has quit [Quit: Connection closed for inactivity]
<infinisil>
Go DE-less, there's dozens of us!
<Baughn>
I used to run xmonad!
<Baughn>
But DE-less doesn't work either.
<infinisil>
Oh I do that now :o
<Baughn>
I'm on a triple-monitor HiDPI setup with *different* DPIs for each monitor.
<infinisil>
Ohh, gchristensen will tell you about how sway can handle such cases really well
<Baughn>
Isn't sway Wayland-specific?
<gchristensen>
yeah, Wayland is required
<Baughn>
No thanks, then. Never had a good experience with wayland.
<gchristensen>
fair enough
<Baughn>
It seems to leak GPU memory every time I try to run an opengl application.
<infinisil>
Baughn: Was that recently or a while ago?
<gchristensen>
wayland does, or sway does?
<gchristensen>
(since Wayland is just a protocol, the memory leaks would be up to the implementation)
<Baughn>
GNOME on Wayland, to be specific.
<etu>
Gnome tend to use lots of memory :)
<Baughn>
But surely that's a server issue, not up to the window manager?
<gchristensen>
wayland doesn't have "a server"
<Baughn>
Last time I tried it was just after I got my Radeon VII, so... two months ago?
<Baughn>
By accident.
<Baughn>
I found that after crashing Minecraft four or five times running under IDEA, the entire DE stopped working.
<gchristensen>
Baughn: indeed, the window manager is responsible for all that jazz
<{^_^}>
[nixpkgs] @worldofpeace pushed 2 commits to master: https://git.io/JfTIn
<Baughn>
gchristensen: Doesn't this make it really difficult to code a custom window manager?
<Baughn>
...sheesh.
<seku>
would have loved to give sway a try too, but im still running nvidia :)
<Baughn>
This does not make me think better of wayland, but ok. I'll try it, if you can offer a config template.
<Baughn>
I'm desperate to get off of Windows. >_>
<gchristensen>
Baughn: I have a lot of reasons to not think very highly of Sway or most of the wayland ecosystem. the only reason I use it is I have multiple screens of different DPIs and sway handles it nicely, and hotplug too
<Baughn>
Hotplug is crucial... two of the screens are connected through a KVM sooo
<Baughn>
(A very dumb KVM that does electrical passthrough of the DP signal)
teto has joined #nixos
ramses_ has joined #nixos
nschoe has joined #nixos
iqubic` has joined #nixos
leotaku has quit [Ping timeout: 256 seconds]
andresilva has joined #nixos
evertedsphere has joined #nixos
leotaku has joined #nixos
ggp0647 has quit [Remote host closed the connection]
ggp0647 has joined #nixos
iqubic has quit [Ping timeout: 272 seconds]
romildo has joined #nixos
<andresilva>
what is the scope that a module gets when called with `(import ./module.nix)`? in particular the config variable inside `module.nix` are different than on the module doing the call
<infinisil>
andresilva: Not entirely getting the question, can you be more concrete?
<{^_^}>
[nixpkgs] @Xe opened pull request #85610 → moonplus: init at 0.3.8 → https://git.io/JfTLL
morgrimm has quit [Ping timeout: 258 seconds]
<andresilva>
infinisil maybe it's easier if I show my actual issue as I don't know Nix enough to properly explain myself. I added an option to my global configuration, unfortunately I cannot see this attribute in the config here (https://github.com/andresilva/nixos-config/blob/master/home/home.nix), which I believe is because of the way this file is imported
cr4y1_ has quit [Read error: Connection reset by peer]
cr4y1 has joined #nixos
teto has joined #nixos
ramses_ has joined #nixos
<ramses_>
Hey guys, is there any way to load a default.nix from a remote git repo into my nixos config and call it with callPackage? Or should I add the default.nix itself to my nixos config repo? I would like to keep the default.nix with the actual code
morgrimm has quit [Ping timeout: 265 seconds]
kenran has quit [Quit: leaving]
<hlolli>
ramses_ that hasn't been a problem, you will need to clone the whole repo, I'll show you one example I have in my config.
kenran has joined #nixos
<hlolli>
https://pastebin.com/wWsfpHSu as you see, calling import on the repo works without any specific configuration because default.nix exists there at the root dir.
<alexarice[m]>
I'm trying to write a nixos test, is there a way of creating a temporary mutable directory and passing this into a derivation, (This should simulate the home folder on the computer)?
<infinisil>
alexarice[m]: If it's a NixOS test, you can just configure a user with a home directory as you would on NixOS
<alexarice[m]>
infinisil: ah great, will give that a try
<{^_^}>
[nixpkgs] @pbogdan opened pull request #85616 → haskellPackages.text-format: don't mark as broken → https://git.io/JfTmz
proofofkeags has joined #nixos
miscellaneousbus has joined #nixos
miscellaneousbus has left #nixos [#nixos]
relaxmax has quit [Ping timeout: 240 seconds]
romildo has quit [Quit: Leaving]
<hlolli>
ramses_ you can use this library to do that which my demo happens to showcase, or if you're lazy like me, you write wrong hash and nix will correct you and tell you which one is the right one
<ramses_>
hlolli: it seems that it doesn't redownload when I change the revision (commit)..
<hlolli>
yes, that's a common pitfall
henry_ has joined #nixos
<hlolli>
if you don't change the sha, nix will think nothing changed, sha is everything, I sometimes have false positive builds because of it (which I find out eventually when ghamofborg fails to build it)
andresilva has quit [Remote host closed the connection]
kenran has quit [Quit: leaving]
<hlolli>
just think about it, how should nix know if two commits don't contain the exact same hash, it's like telling nix it's the exactly same data, in which case, no need to redownload
<symphorien>
ornxka: replace the overlay by something which does not evaluate :)
<ornxka>
lol i literally just tried that
<ornxka>
30 seconds ago
<ornxka>
it didnt work and then i realized i needed llvmPackages_10 and not llvmPackages
<ornxka>
i think there might also be a way to print it with nix repl?
<ornxka>
but i wasnt able to get that working
<ramses_>
hlolli: I see, makes sense
opthomasprime has joined #nixos
<ramses_>
A bit annoying though, given that it's also very strict in it's verification of the hash value, so not easy to get it to give me the new hash
waleee-cl has joined #nixos
opthomasprime has quit [Remote host closed the connection]
<hlolli>
make a typeo in the shasum, then nix will be forced to recheck
<hlolli>
I often make an e a f or 0 a 1, sometimes it doesn't fit the sha algorithm, then I just change some other value
opthomasprime has joined #nixos
ddellacosta has joined #nixos
<{^_^}>
[nixpkgs] @andys8 opened pull request #85617 → Haskell: apply-refact version has to match ghc → https://git.io/JfTmA
<ramses_>
hlolli: now it seems that my statement is not of type package. default.nix returns the result of calling pythonPkgs.callPackage. I'm sorry for all the questions, I'm pretty new at this
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
evhan has quit [Quit: De IRC non curat rex...]
<hlolli>
so you it's solved ramses_ ?
evhan has joined #nixos
<ramses_>
hlolli: no... My default.nix returns the result of pythonPkgs.callPackage, but when putting the statement in systemPackages, nix tells me that it's not of type package
<hlolli>
are you importing this or callPackage-ing it?
<ramses_>
Seems like I'm still missing something
<ramses_>
I tried both
<hlolli>
and did you try with and without function call (eg. import ./default.nix; import ./default.nix {};) ofc better if I saw the derivation
<ramses_>
hlolli: I think I figured it out, callPackage needed an extra argument and the error message was not very clear
<ramses_>
Nix does remain a very opaque beast to me
<hlolli>
true, the error messages can often help, can take time to get used to them, and even if you get used to them, they aren't always clear.
justanotheruser has quit [Ping timeout: 265 seconds]
<hlolli>
I'd nix handles poorly beginner mistakes, I wold love to see "this is not type x, did you mean y?"
<nschoe>
ramses_: "Nix does remain a very opaque beast to me" -> I can relate and understand. What greatly helped me is reading the Nix Pills, have you read them?
<ramses_>
I skimmed through the nix pills, but didn't read them front to back
opthomasprime has left #nixos [#nixos]
<ornxka>
based and nix pilled
opthomasprime has joined #nixos
<evanjs>
what's the easiest way to get the current texlive package set version via e.g. `nix eval`?
<ramses_>
Next challenge: how does nix deal with the python module search path? What should I do to have a module be detectable in a systemd service? Somehow the site-packages dir in my nix store need to be scanned by python, I guess
zupo has joined #nixos
pamplemousse has joined #nixos
<adisbladis>
ramses_: Use `python.withPackages` and create a python environment
<adisbladis>
Writing a package manager is a good undertaking if you don't value your own sanity
<adisbladis>
Especially a Python one
<ldlework>
adisbladis: I have done this.
<ldlework>
I am not smart enough to finish it.
<hlolli>
adisbladis I wasn't too serious about it, I was more curious to see how much money nixers are willing to pay me not to write a new python package manager, I know many color hair strains have turned gray because of this.
<adisbladis>
Writing poetry2nix made me really, really appreciate just how complex python packaging actually is
<simpson>
It doesn't have to be complex. The Python community chooses for it to be complex.
<adisbladis>
simpson: It doesn't _have_ to be, but it is :)
<simpson>
adisbladis: Things don't have to be the way that they are. Especially with code; the way that code is is merely the way that it currently happens to be, and it's infinitely plastic.
opthomasprime has left #nixos [#nixos]
knupfer has quit [Remote host closed the connection]
knupfer has joined #nixos
o1lo01ol1o has joined #nixos
proofofkeags has quit [Remote host closed the connection]
o1lo01ol1o has quit [Remote host closed the connection]
reallymemorable has quit [Quit: reallymemorable]
nixuser has quit [Read error: Connection reset by peer]
<ar>
hi. i have a strange problem with my nixos install recently. trying to boot normally takes ages; trying to do nixos-enter from a nixos livecd also takes ages, and it looked like update-users-groups.pl script hanged. the few times it finished successfully, it took a few GiB of ram, and once it got OOMkilled
rardiol has joined #nixos
inkbottle has joined #nixos
zebrag has quit [Ping timeout: 265 seconds]
reallymemorable has joined #nixos
<ar>
the /nix/store/...-users-groups.json it apparently tries to operate on is a bit over 16k
<lewo>
ar: this could happen if you read a big directory since it is loaded in RAM. Are you sure your are not doing something such as `src = /` somewhere?
<ar>
lewo: i don't think so
<{^_^}>
[nixpkgs] @misuzu opened pull request #85625 → nix-simple-deploy: init at 0.1.1 → https://git.io/JfT3I
sarcasticadmin has joined #nixos
erasmas has joined #nixos
jarlg has quit [Ping timeout: 272 seconds]
henry_ has quit [Quit: henry_]
arjen-jonathan has quit [Ping timeout: 272 seconds]
<evanjs>
hopefully nobody else is trying to update texlive packages atm, going to try doing so :D
butterthebuddha has quit [Remote host closed the connection]
<rail_>
infinisil: hey, sorry for an extra ping here. Looks like you are the last person touched znapzend for reals. There is a pending PR at https://github.com/NixOS/nixpkgs/pull/83257. No rush, when you have time, can you either look or redirect to someone else please. Thank you in advance!
<{^_^}>
[nixpkgs] @worldofpeace pushed 5 commits to master: https://git.io/JfT3D
tex has joined #nixos
<{^_^}>
[nixpkgs] @worldofpeace pushed 4 commits to release-20.03: https://git.io/JfT39
<tex>
Hello everyone, sorry for interrupting everyone, I ask you for a minute. My family is experiencing difficulties due to Covid-19, I kindly ask you if you can help me by activating my Tik Tok invitation code so that we can receive the necessary help at this inopportune time. My Code 251313031
<{^_^}>
[nixpkgs] @worldofpeace pushed 0 commits to refs/tags/20.03: https://git.io/JfT35
<evanjs>
any tex maintainers here or others know if bumping to an "unstable" mirror is okay if it's not part of a release?
relaxmax has joined #nixos
<evanjs>
Still not totally familiar with nixpkgs releases
<evanjs>
**nixos, I guess
<evanjs>
ahhh wait the unstable releases don't have `archive` folders, so I'm unsure how that would even work
fenedor is now known as fendor
amanjeev_ is now known as amanjeev
mallox has quit [Ping timeout: 250 seconds]
relaxmax has quit [Quit: Leaving]
ilikeheaps has quit [Ping timeout: 265 seconds]
mallox has joined #nixos
ilikeheaps has joined #nixos
Izorkin has quit [Ping timeout: 256 seconds]
<jared-w>
gchristensen: Your stateless zfs setup is awesome. I'm really considering it (cause #yolo). Do you run full-disk encryption (sans bootloader) with it?
<jared-w>
Taneb: something like this: https://github.com/shlevy/nix-plugins seems really cool, but it's complete inability to "one-click-install-and-just-work" for other people kinda kills the appeal for me
<{^_^}>
#72473 (by urbas, 24 weeks ago, closed): Rust compiler does not build on Raspbian Linux (Raspberry Pi 4)
<henry_>
How is this accomplished?
turion has quit [Ping timeout: 265 seconds]
Henson has joined #nixos
tpanum- has joined #nixos
tpanum has quit [Ping timeout: 240 seconds]
magnetophon has quit [Ping timeout: 260 seconds]
<{^_^}>
[nixpkgs] @druimalban opened pull request #85627 → Fix mg building on darwin and update to most recent release (20180927) → https://git.io/JfTZN
<Raito_Bezarius>
then mkPoetryPackages could take some globalPreferWheels, etc.
<{^_^}>
[nixpkgs] @xaverdh opened pull request #85628 → treewide: add bool type to enable options, or make use of mkEnableOption → https://git.io/JfTco
mallox has quit [Ping timeout: 260 seconds]
<hexa->
is anybody using home-assistant on nixos and considers it fun? :<
mallox has joined #nixos
<Raito_Bezarius>
hexa-: modulo Python modules… :D
<simpson>
hexa-: I found that the base setup worked, but lots of plugins were broken. I've basically committed to writing my own, with Prometheus and duct-tape.
<hexa->
case in point
<Raito_Bezarius>
I had HA on a NixOS RPi
shibboleth has joined #nixos
shibboleth has quit [Remote host closed the connection]
<hexa->
very sad, I have it running on Debian still, want to convert the rest of the machine to NixOS, but HA is a showstopper imo
<Raito_Bezarius>
I don't see what you're afraid of
<hexa->
guess i'll give it a shot
<cole-h>
srid: How would a builtin install from nixpkgs? Is there any precedence for this? I don't think it's unreasonable to expect a user to add git to their path if they expect to be able to use git...
<Raito_Bezarius>
do a VM
<Raito_Bezarius>
and play around with NixOS + HA
<Raito_Bezarius>
that should give enough insights
<Raito_Bezarius>
and once you're done, you can just copy the config and reuse it
<hexa->
true, thank you
<srid>
cole-h: That makes sense. But is there way to avoid builtins in this expression? nix-env -iE '_: let src = builtins.fetchGit { url = "https://github.com/srid/neuron"; ref = "master"; }; in import src.outPath { gitRev = src.shortRev; }'
<srid>
xy problem: Just install from default.nix, but mandatorily pass the git revision to it.
<cole-h>
How do you expect to get the git revision if git isn't on the path?
opthomasprime has joined #nixos
<{^_^}>
[nix] @mkenigs opened pull request #3520 → Try to substitute tree for all uses of fetchTree instead of just in flakes → https://git.io/JfTCJ
<{^_^}>
[nixops] @grahamc closed pull request #1238 → Even more mypy type checking → https://git.io/Jv2Ws
<srid>
Well, I suppose I could wrap the whole thing, which is already an annoyingly longish CLI, with `nix-shell -p git --run ...` ...
proofofkeags has joined #nixos
<cole-h>
srid: You could recommend `fetchFromGitHub`, which uses the tarball at the specified `rev`
<cole-h>
or fetchTarball itself, I guess
<cole-h>
because fetchFromGitHub mandates a checksum, while I don't think fetchTarball does.
<srid>
yea, checksum is no go
<srid>
Hmm. fetchTarball might do the trick
hopheynananey has left #nixos ["WeeChat 2.8"]
proofofkeags has quit [Remote host closed the connection]
proofofkeags has joined #nixos
sigmundv_ has quit [Read error: Connection reset by peer]
<Ilya_G>
srid: Ive used lib.fakeSha256 to fetch and trigger error that yields the expected hash, the extract proper value from output and use that. Unless I am missing the point here
<Ilya_G>
*then extract
<srid>
Wait, then I won't have access to `shortRev`. Ugh
<cole-h>
I don't understand why you need access to the shortRev, if you want to support users without `git` in their path?
bbarker has joined #nixos
<srid>
It is used internally to write generated sources, which would be imported to compute the full version of the program (which includes a semblance of `git describe ...`). cf https://github.com/srid/neuron/pull/55
<srid>
Although that means you can't install from master branch. You always have to specify the revision.
Ilya_G has quit [Remote host closed the connection]
jgeerds_ has quit [Ping timeout: 256 seconds]
<pie_[bnc]>
evanjs: i have not messed with texlive in a while, and i havent tried to do anythin with that yet
<pie_[bnc]>
evanjs: would be cool if you throw some notes together :P
<pie_[bnc]>
unless its actually trivial/obvjous
turion has joined #nixos
<morgrimm>
Does anyone know if HFS+ is supported natively in NixOS?
scott has quit [Ping timeout: 245 seconds]
ardumont has quit [Ping timeout: 272 seconds]
scott has joined #nixos
<iclanzan>
I am having trouble with NixOS on an Asus T102H tablet: battery not detected, power button freezes system, plugging an SD card in freezes the system. I tried a couple of versions of PopOS and none of these issues are present there. Tried kernel versions from 4.19 up to 5.6. Anyone know what is going on?
Intensity has joined #nixos
<Raito_Bezarius>
adisbladis: is it expected that "make test" fails on generate.py?
zupo has joined #nixos
seku has quit [Ping timeout: 256 seconds]
Acou_Bass has quit [Ping timeout: 264 seconds]
ardumont has joined #nixos
koleesch__ has quit [Remote host closed the connection]
Acou_Bass has joined #nixos
<energizer>
nixos-generate-config is not adding any `fileSystems` attributes. why is this happening and how can i fix it?
<Raito_Bezarius>
Also, what's a simple way to "test" a poetry2nix fork? If it was standard soft, I could just nix-env install it, but as I want to test its functions, I'm not sure how to do that
<Raito_Bezarius>
I could use a nix-shell I suppose (?)
<cole-h>
Kyndig: It's because the homepage updated the flake it was using to be 20.03, even though 20.03 has yet to be released.
<Kyndig>
is that common practice to update the websites before releasing?
<cole-h>
The next release branch is created about a month before the month in which it would ideally be released -- I believe we branched off in February for 20.03.
snicket has quit [Ping timeout: 240 seconds]
Mateon2 has joined #nixos
dermetfan has quit [Ping timeout: 265 seconds]
<Raito_Bezarius>
theorically, 19.09 is EOL around April 2020
<Raito_Bezarius>
IIRC
chagra has quit [Quit: WeeChat 2.7.1]
<cole-h>
Kyndig: It was a mistake, because the manuals are built from the nixpkgs flake, which was (as I said) recently updated to 20.03. See https://github.com/NixOS/nixos-homepage/issues/400 for more nifo
<{^_^}>
nixos-homepage#400 (by samueldr, 2 days ago, open): Decouple the nixos/nixpkgs manuals from the nixpkgs flake input
<cole-h>
s/nifo/info
nschoe has quit [Ping timeout: 265 seconds]
<Kyndig>
so building from the nixpkgs was a mistake
<Kyndig>
ok
<Kyndig>
thx
pamplemousse has quit [Ping timeout: 256 seconds]
shibboleth has quit [Ping timeout: 240 seconds]
<Raito_Bezarius>
> End of support is planned for end of April 2020, handing over to 20.03.
<{^_^}>
error: syntax error, unexpected ',', expecting ')', at (string):304:48
<samueldr>
that's not what's to understand from that issue, the "nixpkgs" identifier here could have been anything
<Raito_Bezarius>
oops
<samueldr>
Raito_Bezarius: "planned" :) it's always a full month after release + best effort if anything serious soon after that month
<srhb>
Raito_Bezarius: Plans are fickle things.
<Raito_Bezarius>
Definitely
<cole-h>
samueldr: Oh, sorry for putting words in your mouth.
<Raito_Bezarius>
But I guess it's not that a good idea to stay on 19.09 as 20.03 might go out shortly
<Raito_Bezarius>
(especially for stuff like NextCloud, etc.)
<srhb>
Depends on your use case, I'd say.
<Raito_Bezarius>
Right
<Raito_Bezarius>
wait, what is nixos-option, has it been introduced in 20.03 ?
<samueldr>
if you upgrade now, and face a serious issue (unlikely) at least it'll be possible to raise the issue before the release
<srhb>
Raito_Bezarius: No, but it got better recently.
<samueldr>
nah, been there for a good while
Mateon2 has quit [Ping timeout: 250 seconds]
<Raito_Bezarius>
Wow
<Raito_Bezarius>
couldn't this be used for, let's say, auto-complete in your own configuration.nix ?
<Raito_Bezarius>
or does it only work on existing declared options?
luna has joined #nixos
Mateon1 has joined #nixos
chloekek has joined #nixos
<yorick>
why does grub-config depend on btrfs-progs?!
<yorick>
why does system-path depend on bcache-super-show? :/
<cole-h>
Because it has btrfs support?
<yorick>
well, can I disable that for closure reasons?
<yorick>
I would like for my minimal image not to depend on the entire userspace for nfs, nilfs, f2fs, zfs, btrfs, bcachefs
<Raito_Bezarius>
you could override it I suppose
<srhb>
Raito_Bezarius: It works on whatever option is in your config, from upstream or defined locally.
<negaduck>
How after after defining `d = derivation { name = "myname"; builder = "${coreutils}/bin/true"; system = builtins.currentSystem; }`? I mean, it's only interpolated into a string and is not passed anywhere otherwise
<cole-h>
Yep. But you'll see a lot of broken things if you just try to run `haskellPackages`.
<rednaZ>
srhb: Thank you. That make things easier.
<negaduck>
how does it refer to coreutils?
<{^_^}>
[nixpkgs] @ryneeverett opened pull request #85632 → [20.03] pythonPackages.stem: remove marked as broken → https://git.io/JfT8M
<srhb>
negaduck: Strings can be magical in nix in that they can have dependencies
<Henson>
the nix package search website used to include links to the derivations. About a month ago those links have been replaced with "Not specified". Is there a reason for that?
<srhb>
negaduck: That's much of how automatic dependency tracking happens.
pareidolia has joined #nixos
<lordcirth>
negaduck, ${coreutils} means "insert the path to coreutils here"
<rednaZ>
cole-h: I see them. Why are they broken?
henry_ has quit [Quit: henry_]
<lordcirth>
That automatically depends on building coreutils.
<cole-h>
rednaZ: Because... they're broken?
<cole-h>
I don't really know how to answer that lol
morr[m] has joined #nixos
<rednaZ>
They are marked as broken, are they not?
<negaduck>
yeah, I mean "${coreutils}" — it's a string, right? Or it's something else?
<cole-h>
Yes, they're marked as broken. Therefore, the packages are broken.
<srhb>
negaduck: It's a string, which is the path to coreutils.
<lordcirth>
negaduck, it's a string, which is then processed to evaluate any uses of ${}
<lordcirth>
Just like "$HOME" is a string in bash, but when you run it, it's going to evaluate it
<negaduck>
so in nix I can do something like "${coreutils}".arguments?
<lordcirth>
negaduck, A file path doesn't have a .arguments
<negaduck>
After defining `d = derivation { name = "myname"; builder = "${coreutils}/bin/true"; system = builtins.currentSystem; }` how does it know we used the coreutils var here?
<cole-h>
So, the simplest answer is: it knows about coreutils because it is currently in scope.
<Raito_Bezarius>
fail
hlolli_ has joined #nixos
<srhb>
negaduck: What is "it"? What do you mean "know we used" it? I'm not sure exactly which part of this is causing confusion, perhaps you can try to rephrase your question a bit?
cbarrett has quit [Ping timeout: 265 seconds]
nschoe has joined #nixos
<cole-h>
srhb: I think they are unclear on where `coreutils` is actually coming from.
<cole-h>
In this case, it's in scope because of `nix repl '<nixpkgs>'`
cbarrett has joined #nixos
hlolli has quit [Ping timeout: 265 seconds]
<negaduck>
Raito_Bezarius: this is what I'm asking for, thanks. Is it documented anywhere?
<Raito_Bezarius>
negaduck: that's more of an impl detail, beyond some docs, you'll have to read the source code or ask experts here
<srhb>
negaduck: I think you may actually be looking for Eelcos thesis.
<Raito_Bezarius>
or this ^ indeed
<srhb>
negaduck: The hand-wavy story is "we can recognize any nix store path and say it's a kind of dependency"
<cole-h>
281 pages of PDF. Better get comfortable.
<negaduck>
haha
<negaduck>
thanks
<srhb>
It's actually really well-written and easy to digest, fwiw
<{^_^}>
[nixos-homepage] @samueldr pushed to fix/released-manual « Decouple build nixpkgs from released nixpkgs »: https://git.io/JfT4O
<srhb>
negaduck: But: You're also coming up on some chapters in the nix pills that detail runtime dependencies, which may make the idea clearer to you
reallymemorable has quit [Quit: reallymemorable]
<{^_^}>
[nixos-homepage] @samueldr opened pull request #401 → Decouple build nixpkgs from released nixpkgs → https://git.io/JfT4n
<cole-h>
Hehe, to get something fixed faster, just misrepresent what someone was talking about! (cc samueldr hehe)
<cole-h>
✨ samueldr
<{^_^}>
samueldr's karma got increased to 212
<srhb>
If you can accept that "nix takes care of this" for now, until it crystallizes, at leaast :)
<samueldr>
cole-h: I would have fixed it earlier, but this week-end I told myself "no nixpkgs, no nixos, no mobile-nixos" and I did that
shibboleth has joined #nixos
<MichaelRaskin>
samueldr++ samueldr++
<{^_^}>
samueldr's karma got increased to 213
dermetfan has joined #nixos
<cole-h>
samueldr: Just taking some time to decomrpess? Hope you enjoyed your time off :)
<gchristensen>
samueldr++
<{^_^}>
samueldr's karma got increased to 214
<negaduck>
srhb: I mean, in the docs it's not clear for me how it's inferred that coreutils is a dependency. `derivation = { ... builder = "${coreutils}/bin/true" ...}`, than the coreutils is suddenly an input in .drv. So, it's a hidden feature of the language apparently
<negaduck>
in the pills
<samueldr>
I had an horrible time converting a wifi led light bulb to tasmota → #nixos-chat for the rest of the story
henry_ has quit [Quit: henry_]
<cole-h>
;^)
<srhb>
negaduck: I would consider it a "hidden feature of the language" for now and keep going :)
<srhb>
negaduck: It becomes clearer.
<MichaelRaskin>
negaduck: yeah, for each string that got constructed using an output path, the dependencies are tracked by the interpreter
magnetophon has quit [Ping timeout: 256 seconds]
<negaduck>
cool. Thank you guys
<MichaelRaskin>
But that part just works and you will only need to understand it a few years later
<{^_^}>
[nixos-homepage] @garbas pushed to plasma5-rename « adding some stats »: https://git.io/JfT4a
<cole-h>
samueldr: Though I'm a little curious why you point to `edolstra/nixpkgs` for released instead of `NixOS/nixpkgs`
reallymemorable has joined #nixos
<samueldr>
cole-h: good question, and the answer is surprisingly simple: that's how the flake is registered in the registry, and that's the flake that was used prior to the breaking change
<cole-h>
"breaking change" being publishing the 20.03 manual before release?
chagra has quit [Ping timeout: 258 seconds]
turona has joined #nixos
<cole-h>
samueldr: Though I suppose my confusion stems from the fact that the `nixpkgs` flake (sans `-released`) comes from `NixOS/nixpkgs`. Since NixOS appears to be available, why not use that?
gagbo has quit [Quit: I'm out !]
Ilya_G has joined #nixos
<{^_^}>
[nixos-homepage] @garbas pushed 6 commits to plasma5-rename: https://git.io/JfT4d
<samueldr>
cole-h: 19.09 official doesn't have a flake.nix, so it's not a flake
<cole-h>
Ohhhhhhhhhhhhhhh
<samueldr>
that's why, I figure, eelco did it that way
<cole-h>
Thanks, the explains it :D
<cole-h>
s/the/that
<cole-h>
As you can tell, I'm not well-versed (or versed at all, really) in flakes. It makes sense that it requires a flake.nix
<qy[m]1>
i quite intentionally don't use the online registry
<cole-h>
bqv: Someday (after I move to NixOS), I'll set up flakes. When that day comes, I'll be pestering you for a while ;)
<bqv>
:p
<{^_^}>
[nixops] @gilligan opened pull request #1309 → Add .mypy_cache to .gitignore → https://git.io/JfTBe
henry_ has joined #nixos
nschoe has quit [Ping timeout: 272 seconds]
ixxie_ has quit [Quit: leaving]
<energizer>
i'm trying to set up a machine. nixos-generate-config doesn't create any `fileSystems` entries. what's wrong?
ixxie_ has joined #nixos
ixxie_ has quit [Client Quit]
<energizer>
will it fail to boot if i try?
<cole-h>
Maybe? Did you setup your filesystems before you generated the config?
<energizer>
yes i have it set up, the machine is running nixos. but there's no `fileSystems`
<finnwww[m]>
I'm having a hard time installing bspwm on nix os. does anyone know a tutorial?
<qy[m]1>
oh you made it
<qy[m]1>
i remember you from the guix room :p
<cole-h>
energizer: Where are you checking for `fileSystems` entries?
<finnwww[m]>
oh yeah you too ☺️
<energizer>
finnwww[m]: it's probably just `services.xserver.windowManager.bspwm.enable = true;`
magnetophon has joined #nixos
<Ankhers>
Does anyone have an example shell/default.nix file that I could look at in order to build a simple gtk application?
<{^_^}>
[nixops] @gilligan opened pull request #1310 → Extract ci actions to ./ci/ scripts → https://git.io/JfTBL
<finnwww[m]>
energizer: i have that. none of the shortcuts are working...
<Ilya_G>
So I am trying to create a derivation that would allow me to build a systemd unit configuration outside of nixos-generators, can anyone point me in the right direction.
<cole-h>
Not bored enough to change your handle... :P
<DigitalKiwi>
but not bored enough to make it good
morgrimm has quit [Ping timeout: 256 seconds]
<energizer>
finnwww[m]: qtile is a window manager that offers a bunch of different layouts. i would recommend it if you know python. if you dont, maybe easier to use something else http://docs.qtile.org/en/latest/manual/ref/layouts.html
<DigitalKiwi>
cole-h: but if i change my handle it ruins the joke
<finnwww[m]>
ohhhhh I'll try it
Miyu-saki has quit [Ping timeout: 260 seconds]
jasom has quit [Ping timeout: 260 seconds]
Miyu-saki has joined #nixos
jasom has joined #nixos
morgrimm has joined #nixos
vidbina_ has joined #nixos
Miyu-saki is now known as Guest68777
misuzu has quit [Quit: leaving]
opthomasprime has quit [Remote host closed the connection]
<{^_^}>
#56430 (by pukkamustard, 1 year ago, open): nixos/guix: init at 1.0.0
<bqv>
i don't know why i'm even surprised, it's nixpkgs after all
<Ilya_G>
I can't even sort out a starting point at the moment.
<cole-h>
Well, that one's a year old now... ;^)
<bqv>
yeah
<{^_^}>
[nixops] @gilligan opened pull request #1311 → Drop `dev-shell` and remove `--exclude tarball` → https://git.io/JfTRG
<{^_^}>
[nixos-homepage] @samueldr pushed to plasma5-rename « Decouple build nixpkgs from released nixpkgs »: https://git.io/JfTR4
turona has joined #nixos
mallox has quit [Quit: WeeChat 2.8]
<henry_>
bqv: "i don't know why i'm even surprised, it's nixpkgs after all" What do you mean?
<drakonis>
bqv: its kind of uh
<drakonis>
out of date and still doesnt do guile 3?
<drakonis>
i'd kill for guile 3
<energizer>
do yall enable passwordless `sudo nixos-rebuild switch`?
<drakonis>
that's passwordless sudo you mean?
<bqv>
henry_: let's just not go there
<drakonis>
nixpkgs has all the things
<energizer>
drakonis: specifically for that command, yes
<bqv>
drakonis: i mean you could try guile 3 for mine
<bqv>
i haven't though
<cole-h>
You can allow passwordless sudo for specific commands (and args IIRC)
<drakonis>
there's no guile 3 on nixpkgs yet
<samueldr>
energizer: just saying, that would end up allowing an attacker to configure it so all sudo is passwordless :)
<bqv>
cause guix will just rebuild itself from the guix store properly
<henry_>
bqv: Yeah but I need to know tbh.
<drakonis>
but guix will use guile 3 once you pull it
pamplemousse has joined #nixos
jmatrix1 has joined #nixos
<DigitalKiwi>
henry_: always check github before you start working on something because frequently someone has already done it
<energizer>
samueldr: haha i suppose so
<bqv>
at least it says "Instead of using the Guix binary tarball it would be much cooler to bootstrap Guix from a smaller bootstrap binaries." which is exactly what i've done, so that's a win
<Ilya_G>
DigitalKiwi: How can I update systemd derivation in a running NixOS instance?
vidbina_ has quit [Ping timeout: 260 seconds]
<henry_>
DigitalKiwi: That's fair.
<energizer>
when using `morph deploy`, how do you get your sudo password into the remote?
<henry_>
I'll probably do it actually.
<manveru>
energizer: sudo password?
<manveru>
why wouldn't you just connect to root user?
<{^_^}>
[nixops] @grahamc merged pull request #1311 → Drop `dev-shell` and remove `--exclude tarball` → https://git.io/JfTRG
<Ilya_G>
DigitalKiwi: I am trying to figure out a workflow by which I can run system updates on my server and I am starting to pull my hair while the deadline is rapidly approaching.
<henry_>
bqv: That could work if you're into that sort of thing. I just want Nix to build me packages occasionally. I'm running Raspbian because I like old software with no "surprises".
<henry_>
I've been burned too many times. Eventually you stop touching the stove.
<bqv>
ah ok
<energizer>
manveru: do you `su` before executing `nixos-rebuild switch` on your home machine?
seku has quit [Quit: Connection closed]
<henry_>
Just recently in order to get compiled Nix to install nixpkgs.Nix, I've had to manually disable the `m_hugefile` test of e2fsprogs.
<henry_>
It's a grotesque hack and I'm sure it'll break at some point in the future.
<DigitalKiwi>
bqv: there's a caveat that nix-channel --update updates *all* channels and nixos-rebuild --upgrade only updates one (unless you're me, because i have a wrapper...) (someone made a PR that added an option to update all...)
<henry_>
C/C++ dependency chains make me contemplate the merits of a mountain hermitage.
<bqv>
i really ought to modify all my scripts to use nix directly rather than all these nix shell scripts
<bqv>
henry_: real
alexherbo2 has quit [Ping timeout: 256 seconds]
alexherbo24 is now known as alexherbo2
<Ilya_G>
DigitalKiwi: So the deal is as follows. I have my own Python app. I am able to install it and configure service that runs this app as part of creation of virtual hard drive image.
domogled has quit [Remote host closed the connection]
<{^_^}>
[nixos-homepage] @davidak pushed to Community « Rework community page »: https://git.io/JfT0G
domogled has joined #nixos
<DigitalKiwi>
yeah uh i do my best to avoid python
<Ilya_G>
DigitalKiwi: I update Python App by building nix-store exporting in CI and nix-store importing it on the live system
<Ilya_G>
Now all I need is to point service configuration to a newly installed application
<Ilya_G>
it's not about Python it's really about deployment.
<Ilya_G>
* deployment strategy
<Ilya_G>
I have the python package building down
<Ilya_G>
service configuration and execution down
<Ilya_G>
it is changing that configuration on live system to point to a newly deployed package where I get screwed
<{^_^}>
[nixpkgs] @matthewbauer opened pull request #85636 → BLAS/LAPACK fix fallout from #83888 → https://git.io/JfT0R
<kurnevsky>
No lightdm here. But nix-store -q --referrers says that dbus somehow depends on lightdm... Weird.
<{^_^}>
error: syntax error, unexpected $undefined, expecting ')', at (string):304:47
ixxie has quit [Quit: Lost terminal]
luna has joined #nixos
mgdm has joined #nixos
<worldofpeace>
gchristensen: do you know if I have to use nix.package to use nixFlakes? or can I just use nixFlakes in a shell?
<gchristensen>
I think in a shell should be fine
<gchristensen>
I htink all the goodness is evaluator side
lunik1 has joined #nixos
<worldofpeace>
gchristensen: seem you have to allow the experimental features
morgrimm has joined #nixos
<b42>
kurnevsky: there's also "nix why-depends" which shows the dependency chain
otulp has joined #nixos
<lunik1>
to get ssh in initrd to work do you need boot.initrd.network.ssh.enable = true, boot.initrd.network.enable = true, and to provide a boot.initrd.network.ssh.host*Key?
wip_ has joined #nixos
<{^_^}>
[nixos-homepage] @garbas pushed to update-nixpkgs-on-cron « skip ci commits »: https://git.io/JfTEA
<gchristensen>
worldofpeace: yup, for sure :)
maddo has quit [Quit: See ya]
<{^_^}>
[nixos-homepage] @garbas opened pull request #403 → Update flake.lock in cronjob → https://git.io/JfTEp
lsix has quit [Quit: WeeChat 2.8]
morgrimm has quit [Ping timeout: 265 seconds]
gustavderdrache has quit [Quit: Leaving.]
<{^_^}>
[nixos-homepage] @davidak pushed to Community « Revert NixCon team page change »: https://git.io/JfTuZ
<cole-h>
I was kinda hoping to see "dQw4w9WgXcQ" in there, somewhere
<bqv>
energizer: lmao
<samueldr>
helvetica.ttc? :|
<gchristensen>
samueldr: yesss? :)
knupfer has quit [Ping timeout: 252 seconds]
<energizer>
gchristensen: cool. i'm noticing there's no `users.users.root.openssh.authorizedKeys.keys = [...]`. how do you manage ssh auth?
<kfoley>
I've just started using nix and am trying to install notmuch with emacs support, so I ran `nix-env -iA nixpkgs.notmuch --arg withEmacs true`. Is that right?
<Acou_Bass>
hey everyone, got a bt of a Q - i use bitlbee, got a line of services.bitlbee.plugins = [ pkgs.bitlbee-facebook *others ]; and i want to include a package in that line from unstable... ive got a line in configuration.nix with a nixpkgs override to allow unstable packages so usually id just add them to my package line as unstable.packagename or whatever, but i cant figure out how to add an unstable
<Acou_Bass>
pkg to my bitlbee plugins?
<gchristensen>
energizer: this is my laptop's secrets.nix
<Acou_Bass>
uhh, wasnt expecting that to be so long but there :P
jgeerds_ has joined #nixos
corpix has quit [Quit: Leaving]
<gchristensen>
energizer: for my servers ... that one will take some time to redact :)
<cole-h>
heh
<gchristensen>
but I don't really consider SSH keys to be secret
<gchristensen>
samueldr: now there is some good stuff
<bqv>
i really want a nice solution to ssh too. i thought about doing a yubikey thing, but that's a pain. then i thought about resurrecting my SSH CA, but that's never been ergonomic. i think i might just have keys for every user@host combination and spray those all over my config
<energizer>
gchristensen: but you do set users.users.root.openssh.authorizedKeys.keys in it?
<bqv>
gchristensen: you are a true heathen
<gchristensen>
bqv: vault has nice SSH-CA support
<gchristensen>
energizer: oh, no, not for root
<bqv>
vault as in the hashicorp thing?
<gchristensen>
bqv: yea
<energizer>
gchristensen: how do you remotely install configs on the servers if not with root?
<bqv>
hmm, yeah i think i came across that, can't remember why i decided against
<gchristensen>
energizer: nixops creates and manages its own root SSH key
Kyndig has quit [Remote host closed the connection]
ozychhi has quit [Ping timeout: 252 seconds]
<henry_>
So compiled-Nix has successfully installed nixpkgs.Nix.
<energizer>
gchristensen: so your local nixops process manages a local database generates an ssh key, and then *somehow* it installs that root ssh key onto the servers
<henry_>
MichaelRaskin: You suggested uninstalling compiled-Nix, right?
ozychhi has joined #nixos
<MichaelRaskin>
Once I have checked nixpkgs.nix works, I usually remove compiled-Nix, yes
<henry_>
Okay, so (roughly speaking) nix-build offers the ability to install a package and all of its dependencies at a specific version in a reproducible way, and nix-env is using the capabilities of nix-build to do all of that plus environment specific configuration such as path resolution?
eyenx has joined #nixos
<infinisil>
cole-h: withTests tests fish = flip const tests fish = const fish tests = fish
<infinisil>
cole-h: Super weird, no idea why that's there
<infinisil>
Essentially tests just gets ignored
<cole-h>
Yeah, I'm really confused. Even more confused because it's been hanging around for ~3 years
eyenx has quit [Changing host]
eyenx has joined #nixos
<cole-h>
It seems like Pr*fpatsch was working on some sort of testing mechanism, but stubbed it out because it wasn't merged/ready?
<MichaelRaskin>
henry_: nix-env combines nix-build stuff, buildEnv stuff, and generations-of-build management stuff… then it also has imperative management of what's installed, no -eA matching -iA, and the default profile is usually in $PATH
<infinisil>
cole-h: Ah yeah, that could be it
<infinisil>
Well nowadays it's possible by using passthru
<lovesegfault>
cole-h: Hum? I'm on nixos-unstable and it's on 3.24
<cole-h>
Oh, maybe not on nixos-unstable yet?
<jtojnar>
cole-h GNOME 3.36 is on unstable, 3.24 on 20.03, gtk versions are different
<jtojnar>
s/3.24/3.34/
<cole-h>
x)
justanotheruser has joined #nixos
<jtojnar>
lovesegfault I thought we backported a patch for that issue
<cole-h>
Maybe I should stop conjecturing about these kinds of things... x)
<jtojnar>
it was true in the past but gtk 3 is frozen at 3.2 no
<lovesegfault>
jtojnar: maybe it hasn't made it's way into the channel yet
<jtojnar>
s/3.2/3.24/
<jtojnar>
now, since they are focusing on GTK 4
rotaerk has quit [Ping timeout: 240 seconds]
<jtojnar>
lovesegfault I think the patch was part of the upgrade PR, but maybe I am confusing it with similar issue in mutter
<jtojnar>
hedning would know
<lovesegfault>
let's see if 3.24.18 fixes it
<lovesegfault>
building...
<hlolli_>
what do other nixers think about this situation. I'm bumping graalvm8, and it's been a journey, as it usually is when I touch graalvm compilations. Since november, jvmci8 a openjdk8 hotspot for graalvm8 needs to be compiled statically. (the official policy of oracle also changed from transparent builds to "use our builds and we wont help you compile graalvm from sources"). Long story short, static openjdk would be a new change to
<hlolli_>
nixpkgs, and thus far I've been succesfull with most of the jdk's in the jdk8 suite. But my dilemma is if I should be spending time getting the parts of openjdk8 to work which I don't need for graalvm8 to compile, or should I make a complete support in 1 big PR?
Thra11 has quit [Quit: WeeChat 2.8]
<hlolli_>
Because for that I may need to introduce new library like https://github.com/openjdk/portola which are set of makefiles made for static-musl-jdk compilation, that I would defenitely need for hopspot build (which I dont need for jvmci8 because jvmci8 is itself its own hotspot). As an effect of the infamous rubber duck, I think I'll first get that static build support before proceeding with graalvm bump, and do a complete support.
fendor has quit [Read error: Connection reset by peer]
Mrmaxmeier has quit [Quit: Ping timeout (120 seconds)]
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<cole-h>
propagatedBuildInputs is for binaries that should be found in the PATH of a package, right?
Mrmaxmeier has joined #nixos
o1lo01ol1o has quit [Ping timeout: 240 seconds]
incognito9999 has quit [Quit: ZNC 1.7.2+deb3~bpo9+1 - https://znc.in]
incognito9999 has joined #nixos
<luna>
What is the name of that package that handles dot files in nixos?
<slack1256>
home-manager ?
<thequux[m]>
home-manager?
<thequux[m]>
Too slow!
<luna>
Yeah, thanks.
<slack1256>
thequux[m]: heeeeeey
<thequux[m]>
Err, I was too slow :-D
<slack1256>
ping > 400ms
<hlolli_>
cole-h it's a bit of mystery to me, buildInputs behave this way you discribed already, I took it for inheritance, since it was the recommended way to grab 1 python environment and use it in another with pypi2nix, but that's the only use case I had for it.
<{^_^}>
[nixos-homepage] @worldofpeace pushed commit from @samueldr to plasma5-rename « Decouple build nixpkgs from released nixpkgs »: https://git.io/JfTg3
<cole-h>
hlolli_: buildInputs is for build-time dependencies like libraries. nativeBuildInputs is for build-time binary dependencies (like cmake, etc). propagatedBuildInputs is for runtime (binary?) dependencies, AFAICT
<hlolli_>
true tue, I misunderstood
<thequux[m]>
So, I'm trying to build a cross compiler for cortex m4, and performance is going to be critical for my application. Is there any way to make nixpkgs use some special cflags for target binaries?
<hlolli_>
so I guess when an app calls for example "steam" on runtime, it would need steam decleared as propagatedBuildInputs, which makes sense perhaps in the python env's case, as it's a shell with these packages as runtime dependencies. But I'd love to understand this better.
<thequux[m]>
I specifically want to override `-march` and `-mtune`
rotaerk has joined #nixos
incognito9999 has quit [Quit: ZNC 1.7.2+deb3~bpo9+1 - https://znc.in]
incognito9999 has joined #nixos
qycs[m] has joined #nixos
gustavderdrache has joined #nixos
<qy[m]1>
oh yeah, that qy is me and bqv too btw
<ajs124>
can anyone explain aggregate hydra jobsets to me or point me to any documentation (hah)
<ajs124>
specifically, about #85577. I really don't get what was going on there.
<duckfullstop>
dumb beginner's question that is going to have an equally dumb answer I can already tell: I need to use netcat from libressl instead of netcat-gnu (which I think is installed by a dependency), what's the best way to override that on a user profile level?
<ldlework>
i started off by solving my dotfiles situation with nix on OSX
<energizer>
luna: alternatively you could do it with nix's activationScripts but that won't clean up afterwards
<ldlework>
before switching to nixos
<energizer>
ldlework: i use computers that i dont have root on
<cole-h>
worldofpeace: What do you mean I tweak? >:( (✨)
<worldofpeace>
cole-h: adjustments = tweaks
codygman has quit [Ping timeout: 240 seconds]
<luna>
energizer: What wouldn't clean up?
<ldlework>
energizer: i am almost certain that both nix and guix can run without root
<worldofpeace>
not like, ooh you're "tweaking"
<cole-h>
Yeah, I know hehe
<evelyn>
what's to stop you havving them in git for computers you don't control and then nix for everything else
<ldlework>
energizer: and i definitely am not using multiple non-root machines where i need my own dotfiles so sorry you have that constraint
<cole-h>
✨ worldofpeace It's so close I can smell it. And it smells sparkly ✨
<{^_^}>
worldofpeace's karma got increased to 125
<ldlework>
i'll take being able to rice my own actual daily workstation with things like icon locations and so on and have 100% reproducability for that
<energizer>
evelyn: that's one of the options i'm considering. i wrote something yesterday to have hm install my dotfiles repo, but i'm still undecided about it
<ldlework>
I guess I'd put it this way. If I was doing so much stuff on servers, that I felt like I needed my bashrc and emacs config or whatever, I'd just maintain mostly static secondary configs that let me get work done. But this would have literally no bearing on the fact that my primary personal forever workstation is totally managed with Nix and reproducable.
<drakonis>
who wants to do it?
incognito9999 has quit [Quit: ZNC 1.7.2+deb3~bpo9+1 - https://znc.in]
<energizer>
Ashy: check out the 'mountains' episode of Planet Earth
<ldlework>
And not weirdly limited by being the lowest common denominator with what I can also put on work servers.
<drakonis>
Ashy: those are some really badass goats
<qy[m]>
<ldlework "I guess I'd put it this way. If "> this is exactly the reason i refuse to ever have a vimrc over like 10 lines. that way, no matter where i go, vim is ALWAYS the same modulo a few initial commands
<cole-h>
worldofpeace: Is there an RFC for how RMs are chosen? Or any somewhat-formal document about it?
<cole-h>
Curious on the process
<ldlework>
seems like such a devestation to the potential of the tool just for that zealous one property
<energizer>
ldlework: i'm not limited in the way that qy[m] is advocating. i can use my ~/.less on more than one computer by just pulling in my minimal-dotfiles repo as a dependency of my main nix config
shibboleth has quit [Quit: shibboleth]
<ldlework>
energizer: this is actually what i do for emacs
<ldlework>
it's just a sub-module of my main nix repo
<energizer>
ldlework: exactly
<worldofpeace>
cole-h: there's a section in the nixpkgs manual releases
<ldlework>
and nix basically just moves my dotfile to the right place
<worldofpeace>
cole-h: I'll get you the rfc link
<qy[m]>
tbh i don't ever feel the need to modify vim anyway. like i always say, i treat it as a keyboard protocol, not an "environment"
<energizer>
ldlework: except i didnt want to do it for emacs because i have way too much config there
<energizer>
ldlework: but i think i'm ok doing it with vim
<ldlework>
energizer: vim as an portable minimum server editor seems totally rational to me
<ldlework>
but don't forget about tramp
<energizer>
ldlework: yeah i do use tramp as well
<ldlework>
maybe it feels too inconvinent which just means you need a hydra or something
<ldlework>
:)
<qy[m]>
but what if you have to ssh into a random server you don't own? or use someone else's machine for a while?
<qy[m]>
do you then spend half an hour setting up your dotfiles..?
<qy[m]>
cause i just open vim and get going
<qy[m]>
way simpler :p
<ldlework>
yeah i have literally never had this mythical experience of having to literally being expected to do something productive on someone else's machine
<energizer>
qy[m]: not half an hour, it's literally `git clone git@github.com:me/dotfiles.git .`
<{^_^}>
[nixos-homepage] @worldofpeace pushed 2 commits to master: https://git.io/JfT2y
<ldlework>
but people do seem to care a lot about this use case
<ldlework>
i ignore it completely
<cole-h>
worldofpeace++ Thanks, that's cool stuff. So, you're RM for 20.09 as well, with some other yet-to-be-determined victim^W helper?
<{^_^}>
worldofpeace's karma got increased to 131
reallymemorable has joined #nixos
<qy[m]>
i mean, it's easy for me to both ignore and not ignore it, since i use vim and emacs. i use a highly customized emacs, but on any other system (including my servers) i just use vim
<qy[m]>
it's simple really
<ldlework>
seems like that was what we said above
<energizer>
qy[m]: right. but i can have my miniature vim customizations, my .less, and so on
<ldlework>
a minimal static portable config
<qy[m]>
when i say "just use vim" i mean completely without configuration
<energizer>
qy[m]: i dont see the advantage of that
<qy[m]>
so it's absolutely identical on any system i go to
<energizer>
so is mine, because i clone my dotfiles
reallymemorable has quit [Client Quit]
<qy[m]>
it seems like a hassle to me, i just avoid all the new keybinds, all the flashy gimmicks and new features all over the shop, and just get better at using vim for what it is
<qy[m]>
that way i get the same productivity even if the system i'm on doesn't even have internet!
<energizer>
luna: activationScripts won't delete files that have been deleted from your config. home.activation will
corpix has quit [Quit: corpix]
<luna>
qy[m]: but that also means you never will get more productive than what the default allows...
<qy[m]>
but see i don't believe any of these addons do higher the skill ceiling
<luna>
energizer: So basically the only way to delete a file with that is to overwrite it with something else?
<qy[m]>
i think all they do is lower the skill floor
<ldlework>
no just remove it from the config
<qy[m]>
but i understand if you disagree with that
<energizer>
luna: yes, or use home-manager (or one of the other non-nix alternatives)
<ldlework>
oh
<luna>
got it, okay. Thanks, energizer!
worldofpeace_ has joined #nixos
<luna>
qy[m]: Not necessarily needing a skill ceiling, though. If I have 30 lines to comment out for a test, id rather use a plugin to type 3 or 4 keys than however many to comment and uncomment the 30 lines.
<luna>
There isn't much skill in commenting out lines, and the keys don't change the originals. That does increase productivity, though.
<qy[m]>
i see no reason why that needs a plugin, i could achieve that in pretty few keys, and if i wanted it fewer, i can create a macro in seconds to speed up the process
<{^_^}>
[nixos-homepage] @worldofpeace pushed 0 commits to davidak-patch-1: https://git.io/JfT2b
<qy[m]>
it's little gain for lots of pain, from my point of view
drewc has quit [Ping timeout: 240 seconds]
<drakonis>
aight i'm harvesting that sweet sweet karma
corpix has joined #nixos
bhipple has joined #nixos
<kazimazi>
hi I've been tring to build my first derivation of off https://github.com/rafaelrinaldi/pure, I've just given up on it so here I'm asking how you would go about it.
<drakonis>
submitted to hn and lobsters
incognito9999 has quit [Quit: ZNC 1.7.2+deb3~bpo9+1 - https://znc.in]
incognito9999 has joined #nixos
bhipple has quit [Ping timeout: 256 seconds]
<bbarker>
how does one go about getting libpcre.so available? the given derivations only seem to install pcre binaries
bhipple has joined #nixos
<gchristensen>
worldofpeace: by all means :)
<qy[m]>
> pkgs.pcre.lib
<{^_^}>
attribute 'lib' missing, at (string):304:1