<danderson>
pie_: I mean, ideally nixos would have a way of handling keys without making them world readable ;). But absent that, I'm doing what I can :/
<mdash>
danderson: i'm working on a similar problem
<danderson>
mdash: what's your approach? Got a link?
<mdash>
except i'm using bitwarden instead of pass and i haven't worked my way up to nixops yet
<danderson>
hrm. I explicitly bailed on nixops because the other aspects it enforces are jarring for managing standalone machines (the whole state thing, and the difficulties of managing the remote machine's nix channels)
qubasa_ has joined #nixos
<mdash>
danderson: Right. Looking at morph instead is next on my list
<elvishjerricco>
One day I'll find a suitable replacement for NixOps... It does some things that are important to me, but it's quite a frustrating system.
<danderson>
but I'm ending up with something similar to nixops's implementation of key management
<infinisil>
Didn't even have to look up the issue number :)
<danderson>
basically, populate /etc/keys and adjust their permissions with an activation script snippet (so that users/groups exist when we do the chown)
<danderson>
and I store the secrets with git-crypt in the repo. So there's no crypto dance during push
<elvishjerricco>
pie_: Hm?
<danderson>
but "advanced secrets access" is not part of my design goals, I don't need stuff like "only the target machine can decrypt the secret"
<pie_>
elvishjerricco sorry bad highloght
<danderson>
and it ends up with the same problems as nixops's key management, which is that the dependencies aren't visible to nixos, so it doesn't know when to restart services etc.
<danderson>
at least with the thing I have, pushing a correct config to a fresh machine _should_ work, because it'll install the secrets before activation reaches "start systemd services"
<danderson>
but it's still veeeeery loose coupling :(
<elvishjerricco>
I'd really like to modify NixOS to separate configuration from hardware and from data (secrets being specific data). Rolling back my config shouldn't have issues with changed hardware or data. Revoking a secret shouldn't break a config. I'd like all three of these things to be separate generational systems.
<danderson>
yeah, that would be nice. But all three are intertwined, and need to take action on each others's changes
<danderson>
well, at least configuration needs to update when hw or secrets change.
<aminechikhaoui>
danderson what was the problem with nixops when you used it ?
<danderson>
aminechikhaoui: the whole state management is unnecessary and cumbersome for my use (which is managing bare metal machines); and nixops updates the machines based on the pushing client's channel configuration.
<danderson>
the latter you can work around with awkward pinning, but it's very not built into the tooling, and still feels very easy to accidentally push my dev box's unstable into prod.
<aminechikhaoui>
anyway, danderson channel management doesn't seem like a problem to me or maybe I didn't get your use case very well
tsrt^ has quit [Ping timeout: 258 seconds]
<aminechikhaoui>
for state, yeah I guess that's probably not ideal, but the none backend doesn't really keep that much state
<danderson>
evanjs: morph does look like what I'm building now, secrets aside
hpfr[m] has joined #nixos
<hpfr[m]>
What would be a good way to symlink dmenu to rofi?
tsrt^ has joined #nixos
singleton has joined #nixos
<Shados>
Is there a tool for converting .drv files to JSON or some other common format?
<aminechikhaoui>
Shados nix show-derivation
<evanjs>
Shados: there’s nix show-derivation if you just need to view it
<evanjs>
Dangit haha
<Shados>
yeah I literally just ran into it >.>
<Shados>
guess my initial search had the wrong terms
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/fjpQ1
<Shados>
was thinking too much like a programmer and had 'parsing' rather than 'converting' originally
<evanjs>
Shados: never been there before /s sometimes I’ll be a whole complex thing I want to do and when it comes down to it I just need a setting or something stupid and easy
boxscape has quit [Remote host closed the connection]
<evanjs>
mdash: like... the default systemPackages, or..?
<li_matrix>
I dont think you can override-remove
<mdash>
evanjs: like, in my configuration.nix I want to replace something added to environment.systemPackages in another file
<li_matrix>
sounds like you wanna write an option
<li_matrix>
without a module to configure it, youll have to edit the file in question
vika_nezrimaya has joined #nixos
<mdash>
I cheated and used ".overrideDerivation (o: { priority = 1; })"
<evanjs>
*screams*
<mdash>
it's fine, it's fine, if this all works out I'll turn it into a pull request and i can take out the hackery
<mdash>
nah it's fancier than that, maybe "hacquerie"
jluttine has quit [Ping timeout: 252 seconds]
<li_matrix>
well done.
alex``` has quit [Ping timeout: 246 seconds]
<li_matrix>
has anyone ever moved a build job halfway done, to another host
pareidolia has quit [Ping timeout: 268 seconds]
<li_matrix>
Id use that a lot
<li_matrix>
might be an interesting project
alex``` has joined #nixos
alj[m] has joined #nixos
<alj[m]>
Can somebody help me override rust.packages.stable.rustc ? I can't figure it out.
pareidolia has joined #nixos
<evanjs>
alj[m]: version, plugins, etc, what do you want to override?
<alj[m]>
Im on nixos 19 and I need to override the nix-expression creating rustc inparticular
drakonis_ has joined #nixos
ekleog has quit [Quit: WeeChat 2.4]
ekleog has joined #nixos
<alj[m]>
I don't know why but I have a rustc.nix file that does not correspond to the version on github
drakonis has quit [Ping timeout: 276 seconds]
<evanjs>
alj[m]: the master branch doesn't necessarily match the packages on your computer. there are multiple channels for packages in nix, and depending on which channel you're using, some things might not match what is on the master branch on GitHub
<alj[m]>
evanjs: So when i try to use the nightlies I'm missing rustPlatform (and/or rustPackages) to build the actual binary i want to build with rust
<alj[m]>
evanjs: It's past 3am on my side of the earth and I basically tried everything i could figure out on my own. After finally figuring out the exact location of the error I am now at a loss on how to "patch" it...
MmeQuignon has quit [Ping timeout: 245 seconds]
<alj[m]>
It all boils to this, i think: How do I download and substitute a nix-expression for my rust?
<alj[m]>
I should be able to use a nix-expression that doesnt have the error, if i can manage to override rust entirely but i dont nkow how
drakonis has joined #nixos
<evanjs>
alj[m]: Right right. Sort of odd the nix expression doesn't work for the platform you're running on, though
<alj[m]>
or would my binary still not use the proper expression then?
ajirx[m] has joined #nixos
timclassic has joined #nixos
roadt[m] has joined #nixos
pingveno[m] has joined #nixos
rihardsk[m] has joined #nixos
Diamaths[m] has joined #nixos
jonreeve[m] has joined #nixos
bpye has joined #nixos
regnat[m] has joined #nixos
ArtemVorotnikov[ has joined #nixos
TravisRt2botio[m has joined #nixos
otini has joined #nixos
Notkea[m] has joined #nixos
mariusz-kowalski has joined #nixos
Seb[m] has joined #nixos
Smith[m] has joined #nixos
ronny has joined #nixos
wrunt[m] has joined #nixos
haslersn has joined #nixos
midi[m] has joined #nixos
yoho[m]1 has joined #nixos
Dan[m]8 has joined #nixos
GerdFlaig[m] has joined #nixos
gudea[m] has joined #nixos
steshaw[m] has joined #nixos
wangoe[m] has joined #nixos
white_bluff[m] has joined #nixos
vdot0x23[m] has joined #nixos
zmlww[m]1 has joined #nixos
pachacuti[m] has joined #nixos
rycee has joined #nixos
NfixEstrada[m] has joined #nixos
KarelWDingeldey[ has joined #nixos
zmlww[m] has joined #nixos
yangm has joined #nixos
tyx2q[m] has joined #nixos
tokudan[m] has joined #nixos
gspia[m] has joined #nixos
sauyon[m] has joined #nixos
ptival[m] has joined #nixos
sm[m] has joined #nixos
kxlljoyy[m] has joined #nixos
jwaksbaum[m] has joined #nixos
kyleondy[m] has joined #nixos
groggy[m] has joined #nixos
grim_reaper[m] has joined #nixos
infinisil-m has joined #nixos
grahamc[m] has joined #nixos
rnhmjoj has joined #nixos
klbrn[m] has joined #nixos
sputny[m] has joined #nixos
smolboye[m] has joined #nixos
info_supplicant[ has joined #nixos
Gopal[m] has joined #nixos
orospakr[m] has joined #nixos
peterhoeg[m] has joined #nixos
ZD[m] has joined #nixos
Woland[m] has joined #nixos
vxzero[m] has joined #nixos
Yakulu[m] has joined #nixos
syntaxtree[m] has joined #nixos
tindvik[m] has joined #nixos
tonic0x539[m] has joined #nixos
tobim[m] has joined #nixos
Shyim[m] has joined #nixos
y0x3y[m] has joined #nixos
pop[m]1 has joined #nixos
alex[m]16 has joined #nixos
xbreak[m] has joined #nixos
schmittlauch[m]1 has joined #nixos
<alj[m]>
evanjs: Like My binary is build by an expression so i need nix to know the right rust
reallymemorable has joined #nixos
<alj[m]>
oh and with stupid i mean that I'm tired and what ive written is too simple to just work
alex``` has quit [Ping timeout: 250 seconds]
jluttine has joined #nixos
alex``` has joined #nixos
<evanjs>
alj[m]: the weirdest part is that as far as I know, that l shouldn't really matter, at least in all the cases I've worked with arm stuff. What's the actual error you're getting, anyway?
<evanjs>
In terms of syntax, you can override things like `pkgs.rust.packages.stable.rustc.overrideAttrs (oldAttrs: rec { stdenv.targetPlatform.system = "armv7-unknown"; } )`, but this example will _not_ work for what you're trying to do, as setting the stdenv is a little more complicated
<alj[m]>
evanjs: My error is akin to " Failed to load specificatoin for target "armv7l-unknown-linux-gnueabihf""
<evanjs>
jtojnar: yeah what’s confusing me is the top part where they define PYTHONPATH using a command but then they redefine it later. “PYTHONPATH =.*” didn’t seem to match.
<jtojnar>
evanjs: I do not see it redefined
felixfoertsch23 has joined #nixos
felixfoertsch has quit [Ping timeout: 264 seconds]
das_j has quit [Remote host closed the connection]
<evanjs>
jtojnar: Ah you're right. The way they had the defaults worded was throwing me off. It looks like it's used like they mention in the defaults everywhere else. After realizing that, I guess I just need /lib/python3.7/site-packages etc for PYTHONDIR for it to not act dumb
<evanjs>
:rubberduck:
lambda-11235 has joined #nixos
<growpotkin>
so am I crazy or is our default `ftp` command bunk?
lambda-11235 has quit [Max SendQ exceeded]
lambda-11235 has joined #nixos
lambda-11235 has quit [Client Quit]
<growpotkin>
I cannot use `ftp USER@somehost.com` or even `ftp ftp://somehost.com` without it whining. I feel like it's totally different from the `ftp` command I get on my University (Ubuntu) machines, or Arch, or Kali
phreedom has quit [Remote host closed the connection]
phreedom has joined #nixos
<{^_^}>
[nix] @Infinisil opened pull request #3080 → docs: Note that tryEval doesn't do deep evaluation → https://git.io/fjpdc
growpotkin has quit [Ping timeout: 244 seconds]
rauno has quit [Ping timeout: 264 seconds]
<angerman>
what are my options to debug/profile nix memory consumption?
Necro^Byte has joined #nixos
growpotk- has quit [Ping timeout: 258 seconds]
<Ashy>
angerman: in general "add more ram" or failing that "add more swap" or both
<angerman>
Ashy: I'm easily past 40G at this point.
<angerman>
Ashy: I'd ideally like some tool that let's me understand *why* my nix expression ends up eating so much memory, or even give me a hint.
<elvishjerricco>
angerman: I was happy to find that Nix the package manager builds like any other C++ project; so normal techniques for C++ ought to work
<infinisil>
,profiling angerman
<{^_^}>
angerman: Use NIX_COUNT_CALLS=1 and/or NIX_SHOW_STATS=1 to profile Nix evaluation
<angerman>
elvishjerricco: I did consider looking into compiling nix with xcode and running it through instruments. Was just hoping for something less involved :-)
<infinisil>
(might only shows something if it finished successfully though)
<angerman>
infinisil: yea, this just causes my system to grind to a halt :-(
<angerman>
elvishjerricco: it *does* seem to be within the module logic in haskell.nix though :p
<elvishjerricco>
Yea I won't claim the NixOS module system is efficient :P
<angerman>
specifically within the inheritance logic we now have...
<elvishjerricco>
You could disable checking and see if that helps. I think it's an argument to evalModule
<infinisil>
It's a proposal by niksnut for a module system built into nix itself
<angerman>
ahh, that one. I think I've read that a while ago. But it's not usable yet, so ...
<infinisil>
I sure wonder if a faster module system could be implemented in nix itself
<elvishjerricco>
It's also... magical. I dunno; I'm not a huge fan because of how magic it seems
vika_nezrimaya has quit [Ping timeout: 258 seconds]
<angerman>
infinisil: well... that brings us back to profiling tools, doesn't it?
<elvishjerricco>
Yea, it'd be cool if there were a good solution in Nix itself
endformationage has quit [Quit: WeeChat 2.5]
<angerman>
nix and haskell share some of the same pain, writing fast expresions require intimite knowledge in how to write fast expressions.
<infinisil>
At least Nix is interpreted, which makes it a bit easier
<angerman>
I've tangentially discussed that with Bartosz Milewski and a few other during ifcp...
<elvishjerricco>
I was so disappointed when I found out Nix doesn't actually implement maximal laziness, but I would wager that was to avoid being slow as hell :P
<angerman>
infinisil: does it? how so? I must be missing something.
<infinisil>
elvishjerricco: What's not maximally lazy? You mean attribute names?
mac10688 has quit [Ping timeout: 244 seconds]
<infinisil>
angerman: Well with GHC you have the compiler that produces some code which might be good or not, and then it gets run with an RTS which might run well or not
<infinisil>
angerman: With Nix those don't exist, all code just gets executed straight away, and that should make it a lot simpler
<angerman>
infinisil: true, but I have -prof, eventlog, and a host of other inspection tools.
<elvishjerricco>
infinisil: Check out his paper on the subject. He lays out a memory system where data is hashed as it's allocated to deduplicate it. Basically, any references to identical data will point to the same object in a table
<infinisil>
Huh interesting
<elvishjerricco>
So you can have recursive structures that take constant memory by way of luck :P
<elvishjerricco>
Unfortunately Nix doesn't implement this like I thought it did
<angerman>
with check = false, I'm still at 18G and growing although more slowly, ... still not fit for any practical purposes.
<infinisil>
angerman: What are you evaluating?
<elvishjerricco>
angerman: I doubt this is relevant... But I did make a commit or two that assumed we had maximal laziness. The recursive buildPackages or whatever was defined with this assumption. I doubt it's the issue, but it should use more memory than I originally thought
<angerman>
#57477 is really light on details. That's such a shame. Profpatsch's comment is right on the money though.
<angerman>
elvishjerricco: it potentially could. But, this is still bonkers.
<elvishjerricco>
Yea it'd only be an issue if an absolutely ridiculous number of packages had build-time Haskell deps. So it's probably more related to the module system
<angerman>
elvishjerricco: how about *every* package has a build-time haskell dependency?
<elvishjerricco>
angerman: What do you mean? Like you blanket added a buildPackages.foo dep to all packages?
<elvishjerricco>
Wouldn't be a big problem
<angerman>
yea like that.
<elvishjerricco>
The only reason that recursive code would be an issue is that a dep could evaluate a recursive dep which could evaluate a recursive dep and so on
<elvishjerricco>
So just pulling one dep from a package set and giving that to everyone would be constant overhead
freeman[NixOS] has quit [Remote host closed the connection]
freeman42[NixOS] has joined #nixos
freeman42[NixOS] has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @kwohlfahrt opened pull request #67989 → nixos/hardware.deviceTree: Allow use of dtmerge → https://git.io/fjpFh
<bpye>
Hey - something weird is going on my system. I'm on 19.03 (I upgraded from 18.09 a while back) - I keep getting errors that "Warning: the new NixOS configuration has an ‘init’ that is incompatible with the current configuration."... /nix/var/nix/profiles/system/init-interface-version shows 'systemd 2' whilst /run/booted-system/init-interface-version gives 'systemd 3'
<bpye>
I'm not sure how it's ended up getting downgraded?
<bpye>
I do have nixos-unstable in my nix-channel list but it is named 'nixos-unstable' so I don't believe it should be used?
<bpye>
Any ideas?
<Guest14627>
the closest i got is pkgs.lib.overrideDerivation (pkgs.lua53Packages.lua) (old: { NIX_CFLAGS_COMPILE = "-x c++"; }) -- but this results in "warning: '-x c++' after last input file has no effect" during the build and has no effect.
<Guest14627>
i assume there is a better way than just copying the derivation and applying the change i want?
Jackneill has joined #nixos
<boxdot1>
HI — I cannot find any docs on what builtins are. Is it a module somewhere? Can I read source code of it?
<Guest14627>
teto, i thought about that, but following the usual examples of doing that still results in CC being a C compiler. i feel like it should be possible to accomplish this with overrideCC somehow, but i don't see how.
schjetne has joined #nixos
<teto>
Guest14627: the configurePhase sets CC in some specific case, maybe you hit those ?
o1lo01ol1o has quit [Remote host closed the connection]
dansho has joined #nixos
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Okinan has quit [Quit: Konversation terminated!]
cfricke has joined #nixos
thc202 has joined #nixos
<alj[m]>
Hey! Is it possible to cross compile a whole system (in this case armv7)? I'm thinking about just renting a strong server for a bit and letting it compile, instead of losing (another) 3 days waiting for my little board to compile.
vika_nezrimaya has quit [Ping timeout: 245 seconds]
joshuagl has joined #nixos
schjetne has quit [Ping timeout: 246 seconds]
<talyz>
eyJhb: yeah, kinda.. my problem was that virtualbox wasn't working at all, but I never got around to fixing it - I just found the offending commit..
<talyz>
I tried getting the libvirtd backend in nixops up and running as a replacement for virtualbox, but I'm getting some pretty strange issues: mainly nix store corruption. Maybe the base image is broken..?
schjetne has joined #nixos
<teto>
talyz: do you mean files are owned by "nobody" ? there are some solutions on nixops tracker
<{^_^}>
[nixpkgs] @mmahut opened pull request #67994 → prometheus-aws-s3-exporter: init at 0.3.0 → https://git.io/fjpNC
<{^_^}>
[nixpkgs] @roberth opened pull request #67996 → lib.cleanSourceWith: Allow name to be set, optional filter, doc → https://git.io/fjpNz
<sphalerite>
Is there a way to enable both grub and systemd-boot? I want to be able to use systemctl kexec, which wants /boot/loader/loader.conf (the systemd-boot config) to exist, but actually boot with grub usually
<hatefulmoron>
hi, does somebody mind helping me use 'nix-shell' for building my project?
<__monty__>
hatefulmoron: Best way to get your questions answered is to ask them.
<hatefulmoron>
i've created a shell.nix file -> https://pastebin.com/BsDAKvVj, and I'm using `nix-shell --pure --run "make"` to run my makefile. The issue is that my script complains that it can't find "libglib-2.0.so", which should be provided by the `glib` package.
freeman42[NixOS] has joined #nixos
<hatefulmoron>
I've done a search through my filesystem, and it reports that the file is present in the store folder for the `glib` package elsewhere, but it's not being provided to the environment described in shell.nix
<hatefulmoron>
as in, it's in the path `/nix/store/1ax2nivc09afxky6ggpz7b9sqghhlvic-glib-2.58.2/lib/libglib-2.0.so` but it seems to not be symlinked in the shell.
<tilpner>
hatefulmoron: I'm not sure which output is selected here, but I suspect it's the first output, which does not contain the .so
<tilpner>
> glib.outputs
<{^_^}>
[ "bin" "out" "dev" "devdoc" ]
<tilpner>
So try glib.out maybe
<hatefulmoron>
sorry do you mind expanding a bit
<hatefulmoron>
when you say try glib.out, what should I change about my script?
<tilpner>
Replace glib with glib.out and see if that helps
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<jschievink>
I have a proprietary software package that contains both 32-bit and 64-bit exectables, how do I get that to work with NixOS?
<jschievink>
It seems like I can either get it to use the x86-64 dynamic linker the the one for 32-bit x86
<jschievink>
(for all executables though)
<allan>
I’m on macOS: What are the steps to setup bash completion for commands? I have both ~/.nix-profile/share/bash-completion and ~/.nix-profile/etc/bash_completion.d (both directories) and ~/.nix-profile/etc/profile.d/bash_completion.sh which I would expect to automatically be sourced, but even if explicitly sourcing it, there is no completion for the commands that comes with completions (e.g. git)
<Ashy>
been meaning to clean it up and send a PR but that does actually work
drakonis has joined #nixos
iyzsong has joined #nixos
selfsymmetric-pa has quit [Remote host closed the connection]
turion has joined #nixos
<hatefulmoron>
tilpner sorry to ping you again, but the environment variable you mentioned seems to work well, and the little hack I ended up with can be seen here: https://pastebin.com/kP5J0QDu
<hatefulmoron>
thanks for your help
<{^_^}>
[nixpkgs] @johanot opened pull request #68001 → python3Packages.cherrypy: fix tests after upgrade to pytest5 → https://git.io/fjpxo
jtojnar has quit [Quit: jtojnar]
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
jtojnar has joined #nixos
mudri has quit [Quit: Connection closed for inactivity]
<alex```>
warning: The option `services.redshift.longitude' defined in `/etc/nixos/configuration.nix' has been changed to `location.longitude' that has a different type. Please read `location.longitude' documentation and update your configuration accordingly.
<alex```>
where I can find the documentation for the new option?
<vyorkin>
lets say I have the following derivation for the ppx_deriving_protobuf pkg http://ix.io/1U9d but it defines another "library": https://github.com/ocaml-ppx/ppx_deriving_protobuf/blob/v2.7/src/dune#L10 (named `ppx_deriving_protobuf.runtime`). Then if I do the nix-shell and `ocamlfind query ppx_deriving_protobuf` -- the `ocamlfind` is able find it. But if I try to "ocamlfind" the ppx_deriving_protobuf.runtime "library" with "ocamlfind
<vyorkin>
query ppx_deriving_protobuf.runtime" I'd get "ocamlfind: Package `ppx_deriving_protobuf.runtime' not found". I suppose that I should create a separate derivation for it somehow
<evanjs>
alex```: didn’t see anybody mention it, but for option documentation, you can use `nixos-option services.redshift.location.longitude` nixos-option also supports tab completion
<evanjs>
alex```: the documentation from nixos-option should be in sync with the channel you’re using
<etu>
asymmetric: We have a binary cache for aarch64 and images
<infinisil>
etu: man pages are only updated once you can do a successful rebuild
<etu>
infinisil: But so is the system?
<etu>
ah
<etu>
I see what you mean, the man page is synced with the system you managed to build.
asheshambasta has quit [Ping timeout: 245 seconds]
<etu>
But not with the channel you have for the system you may be trying to build
<infinisil>
Yea
<asymmetric>
etu: thanks! i'm looking to install nixos on this thing https://wiki.kobol.io/ - let's see if it's supported
freeman42[NixOS] has quit [Ping timeout: 245 seconds]
<evanjs>
Wait there’s a configuration.nix manpage? 😯 still need to figure out how the get nixos-options explorer working locally without building it from the repo
<dminuoso>
evanjs: Yes. `man configuration.nix` is probably one of the most underused things.
<etu>
evanjs: yes
<dminuoso>
evanjs: If you ever find yourself "googling for nixos options", you should have just typed `man configuration.nix` every time.
<dminuoso>
Its faster and more correct.
<etu>
That manpage is great, quick to search compared to any website
<dminuoso>
And it always matches your exact system.
asheshambasta has joined #nixos
<dminuoso>
Ah wait, you covered that already
_zincy has joined #nixos
<evanjs>
I mean I use nixos-option mostly but that seems easier depending on the use case
<evanjs>
Though, nixos-option also lets you do -I nixos-config=... not sure if the configuration.nix manpage supports that
mjisme has quit [Ping timeout: 244 seconds]
drakonis_ has joined #nixos
<etu>
The manpage is quite static in that regard since it's only updated with a successful system rebuild
<infinisil>
Hm I think you could craft a command to display a nixpkgs' manpage
<infinisil>
Lemme see
<etu>
well, it's probably a derivation? So you could probably just build it and read it somehow? :)
<infinisil>
gchristensen: alex```: Btw the redshift location thing doesn't need to be filed as an issue. Unless you want to file the manpage problem
<gchristensen>
I think it needs changelog?
<infinisil>
gchristensen: Why so?
<infinisil>
The error message explains it well
<infinisil>
s/error/warning
<gchristensen>
it breaks the user's config and the error message isn't very specific about how to fix it
drakonis has quit [Ping timeout: 276 seconds]
<infinisil>
(It's just a warning, not an error)
Neo-- has joined #nixos
<gchristensen>
oh, okay
<gchristensen>
+1
<infinisil>
gchristensen: If you want to get rid of the warning, "Please read `location.longitude' documentation and update your configuration accordingly" isn't very specific though hm..
<gchristensen>
yeah, it could be better
<gchristensen>
but since it is just a warning and doesn't fail the build it seems a bit less bad :)
<infinisil>
Maybe mkChangedOptionModule (which is used here) should have a customizable warning message
<evanjs>
The one I got for Compton awhile back was nice. I cant remember which setting it was but “hey you don’t need to set this option now, it’s automatic, kthxbai”
<evanjs>
Oh it did say it’s not a Boolean though
<evanjs>
s/not/now
tom__ has joined #nixos
tmaekawa has joined #nixos
_zincy has quit [Ping timeout: 244 seconds]
tmaekawa has quit [Client Quit]
Makaveli7 has quit [Quit: WeeChat 2.5]
<infinisil>
Alright got it, a command for viewing the manpage of the current '<nixpkgs>':
<infinisil>
man $(nix-build --no-out-link '<nixpkgs/nixos>' -A config.system.build.manual.manpages)/share/man/man5/configuration.nix.5
<infinisil>
etu: ^
<etu>
infinisil: Oh, not super handy but still nice when one needs that :)
<{^_^}>
Ariakenom: If a Nix file foo.nix starts with something like `{ stdenv, cmake }:`, you can build it with `nix-build -E '(import <nixpkgs> {}).callPackage ./foo.nix {}'`
<clever>
Ariakenom: callPackage then gets it from pkgs
<clever>
> pkgs.avrgcc
<{^_^}>
attribute 'avrgcc' missing, at (string):263:1
<asymmetric>
my sd-card reader does not show up in lspci, it's a realtek RTS5227 PCI Express Card Reader (rev 01). do i need to add a driver to availableKernelModules?
<Ariakenom>
clever: thanks. pkgsCross was in nixpkgs
<etu>
asymmetric: if you add -k to lspci it should print which drivers which devices are using
_zincy has quit [Ping timeout: 245 seconds]
<asymmetric>
etu: the device is not detected though..
<asymmetric>
it's a microsdhc card, if that matters
<nakkle>
Hey there! I'm trying to install NixOS on a new notebook, but the installer only shows a single underscore in the upper left corner. Anyone know how to troubleshoot this?
<clever>
asymmetric: can you pastebin the full lspci output?
<clever>
asymmetric: and the full dmesg output
<{^_^}>
[nixpkgs] @xbreak opened pull request #68007 → clipp: init at 1.2.3 → https://git.io/fjphX
spacekookie has quit [Quit: **agressive swooshing**]
<{^_^}>
[nix] @edolstra pushed to master « Support allowSubstitutes attribute in structured attribute derivations »: https://git.io/fjpjz
vmandela has quit [Quit: Leaving]
debclair has joined #nixos
dansho has quit [Remote host closed the connection]
dansho has joined #nixos
<debclair>
Hello! Sorry to interrupt, I wanted to know how haskell packages are curated for nixpkgs (specially how versions are decided/handled).
<srhb>
debclair: There are some posts you will find on mailing lists if you search for haskellngPackages (which is basically the foundation of haskellPackages nowadays) -- but the short of it is that a set of hackage packages is generated from a stackage LTS release, and committed to the nixpkgs repo.
<srhb>
debclair: These are auto generated. On top of this, various overrides and tweaks are applied. Broken packages are disabled, etc.
<sm[m]>
why is the hydra.nixos.org website so slow ?
<srhb>
debclair: Other things to look at is cabal2nix and specifically hackage2nix in that repo
<srhb>
sm[m]: Not all of it is.
<srhb>
sm[m]: The parts that are are usually slow queries to postgres.
<sm[m]>
I’m searching, and inspecting jobs & builds
<sm[m]>
It’s sitting there for minutes
<debclair>
srhb thanks a lot. I co-worker told me it was generated from a stockage LTS today but I found it hard to believe. Now I know for sure (and ow my coworker an apology). Thank you!
<sm[m]>
Not complaining, but it’s almost unusable, I wish it were a bit faster
<{^_^}>
nix#3025 (by zeecde, 4 weeks ago, open): HTTP error 200 (curl error: Timeout was reached) - unable to update and install
debclair has left #nixos [#nixos]
<_d0t>
I don't really see a solution there.
<_d0t>
Other than updating, which is exactly what I'm trying to do.
growpotkin has joined #nixos
<srhb>
_d0t: If it really is that issue, you can do the upgrade with a newer Nix, but I'm not sure it is. If it is, you can also try on a faster connection. Regardless, the issue contains a number of links you could follow to gain more clarity.
<_d0t>
srhb: is there a way to update just Nix without updating the rest of the system?
tom__ has quit [Ping timeout: 245 seconds]
<clever>
_d0t: put the new nix into nix.package in configuration.nix and nixos-rebuild switch
<_d0t>
I think I have the latest nix version, 2.2.2
<srhb>
Yeah, that fix is not released in a maintenance branch yet afaik. Again, not sure that's it anyway.
<_d0t>
there's only one way to check tho
<srhb>
Sure. You can still upgrade it to some pre-release.
<alexarice[m]>
I'm getting the check phase of a package complaining about a missing library (the library that has been built) when in a pure nix-shell the tests run fine, anyone have any ideas?
nakkle has quit [Ping timeout: 252 seconds]
<srhb>
But of course, you should read carefully the changelogs if you do. I haven't checked if there's backwards incompatible changes, and there might be.
freeman42[NixOS] has quit [Ping timeout: 245 seconds]
<srhb>
alexarice[m]: Is it a mkDerivation-analog that has a separate checkInputs?
<_d0t>
related question. Is there a way to force nixos-rebuild restart on error like this one?
feep_ has quit [Ping timeout: 264 seconds]
<alexarice[m]>
srhb: it's stdenv.mkDerivation, I've just put `checkPhase = "ctest"`
<alexarice[m]>
srhb: Do you think the error is a missing checkInput?
<dminuoso>
I use nix-shell to set up a build environment for my project. Inside that build environment I use a nix supplied cabal to build my project. Now I want to invoke cabal from a git pre-commit hook.. how?
<{^_^}>
[nixpkgs] @worldofpeace pushed 3 commits to staging: https://git.io/fjheF
schjetne has joined #nixos
shibboleth has joined #nixos
<worldofpeace>
maybe set FONTCONFIG_FILE generated with makeFontsConf?
<fresheyeball>
worldofpeace: what is makeFontsConf?
<michaelpj>
would someone who doesn't have abnormal CPU usage on NixOS mind running `sudo dbus-monitor --system` for a bit and seeing if it's normal to get vast quantities of messages for NetworkManager?
<worldofpeace>
michaelpj: it's defined in all-packages.nix
<michaelpj>
it sends an array of access points in the signal, and it starts out with just one, then two, then three, until it's sending all of them at once
<michaelpj>
er well, linear in the number of access points I guess, but I wonder if it's scanning them all each time
knupfer has quit [Ping timeout: 264 seconds]
<michaelpj>
worldofpeace: nothing fancy, just kde with networkmanager
<worldofpeace>
I see michaelpj, I'm pretty sure NM disallows scans immediately following previous ones
<fresheyeball>
worldofpeace: I can find the file on github
<fresheyeball>
but I don't know how to determine the path to that function from the top level
<evanjs>
I have a set of functions that just maps over specified channels and finds a newest version of a package, but I do wonder if there's a better way
vyorkin has quit [Ping timeout: 258 seconds]
<evanjs>
Has anybody had any luck in getting powerlevel-9k and related bars working on a virtual terminal in NixOS?
<srhb>
sm[m]: You can curl -LH "accept: application/json" to get some json from whence you can get the path of the first output
<srhb>
sm[m]: Just s/linux/darwin
<srhb>
sm[m]: And then just nix-store --realise the store path from the json blob
Ariakenom has joined #nixos
<srhb>
sm[m]: (And nix-env -i that)
mexisme has quit [Ping timeout: 246 seconds]
karetsu has joined #nixos
<sm[m]>
ah, nice. And I think that's equivalent to copying the "to" hash that I see under Changes on the build page. Earlier that didn't work but I think probably because I picked the wrong architecture for this machine
karetsu has quit [Client Quit]
<srhb>
sm[m]: Yes, that's correct, the "to" is the nixpkgs version, but realizing the path directly saves you from doing any eval.
karetsu has joined #nixos
karetsu has quit [Client Quit]
<sm[m]>
I think the final piece is to find a hash that corresponds to a build that worked on all architectures (linux and darwin). So the command will work for users on either
<srhb>
sm[m]: Yeah, we don't have that information directly, you'll have to visit latest for each arch.
<srhb>
sm[m]: (They're basically two disparate packages to Nix)
karetsu has joined #nixos
<srhb>
sm[m]: Out of curiosity, why not pin a known-working one? :)
<srhb>
Of course, you have to find a rev that works first...
<sm[m]>
ah, and I see both latest build pages have the same hash. So that's making sense \o/
<sm[m]>
and of course I can test on both platforms
<sm[m]>
this is for install instructions, where I want to give a command that installs the latest release of the package
<evanjs>
So I'm trying to use my PR for openrazer on one of my computers... and I keep getting error: attribute 'openrazer' missing, at /home/evanjs/src/nixpkgs/nixos/modules/hardware/openrazer.nix:106:32
karetsu has quit [Client Quit]
<evanjs>
Interestingly, this only happens on _one_ computer
<{^_^}>
[nixpkgs] @worldofpeace merged pull request #68016 → treewide: don't use single quotes with placeholder → https://git.io/fjhf1
<asymmetric>
clever: it worked on a friend's thinkpad fwiw, who's not on nixos
<{^_^}>
[nixpkgs] @worldofpeace pushed 2 commits to master: https://git.io/fjhfp
karetsu has joined #nixos
<evanjs>
Not sure why it works everywhere else >.>
<asymmetric>
thanks for trying :) clever++ etu++
<{^_^}>
clever's karma got increased to 189, etu's karma got increased to 10
karetsu has quit [Client Quit]
karetsu has joined #nixos
niso has quit [Remote host closed the connection]
<sm[m]>
srhb++
<{^_^}>
srhb's karma got increased to 72
Okinan has joined #nixos
karetsu has quit [Client Quit]
byteflam1 has joined #nixos
<srhb>
evanjs: Fiddling with kernelPackages on that one?
<evanjs>
srhb: Yeah that's where it looked like it was tripping up. Not sure why on this particular computer.
<srhb>
evanjs: I meant, are you modifying kernelPackages in some funky way on that one in particular? :)
<evanjs>
No, not at all! That's why I'm sort of lost haha
<evanjs>
They all have linux_latest though
<srhb>
Ah, okay. :) Seemed like an obvious culprit..
<evanjs>
Tough, this one is not synced in terms of version yet
<evanjs>
Perhaps I should sync that first... though I'm not sure it would matter. 5.2.9 vs 5.2.11 on the others
xkapastel has joined #nixos
<sm[m]>
srhb: does nix-store --realise help me find the github archive hash ? I haven't found out how to get that other than from the build page on hydra
<srhb>
sm[m]: No, there's no way back in general from a store path to a rev.
<srhb>
Well, that's not quite true, but there's no simple way.
babic has joined #nixos
<evanjs>
srhb: The strange bit is that most settings are shared, and I'm having issues finding any major differences in those that aren't. Nothing that affects udev/etc anyway
<srhb>
evanjs: Maybe just nixpkgs changes around kernel arch? :)
<evanjs>
srhb: I mean... 2 patch versions, though??
_zincy has quit [Ping timeout: 246 seconds]
<srhb>
I don't mean the kernel itself, I mean the nix infra around those attrsets.
<evanjs>
srhb: Right right. As far as I know, they all just use mostly the same things (linuxPackagesFor) and I can't imagine any major changes, either upstream or nix infra between those two
<srhb>
evanjs: Yeah, I don't have any psychic ideas here ^^
Myhlamaeus4 has joined #nixos
evanjs| has joined #nixos
<babic>
Is it just me or has the there recently (months/weeks not days) been a change with how hashes are handled, maybe it's just a fetchFrom* thing? Basically lets say you have a package that uses fetchFromGitHub, and you change just the rev. Right now it will build using the old rev with the old hash, I'm pretty sure that used to fail.
<babic>
I'm on nixos-unstable
<srhb>
babic: It has always worked thus for all fixed-output derivations
<srhb>
babic: That is the point of them.
jgeerds has joined #nixos
<srhb>
If the output is present (per the hash) the fetch is a no-op.
<srhb>
Otherwise we would have to always confirm every single source in the world for every single dependency tree. :)
<srhb>
babic: The trick about fixed-output derivations (fetchers, for most intents and purposes) is that the inputs _don't_ matter, only the output. Which is what the hash describes.
<srhb>
mog: Have you opened an issue for it? :)
<mog>
no. thought id take a shot here. i guess dont feel right cluttering the issues with it
<srhb>
mog: You're absolutely welcome to! :)
<mog>
i assume it will be easy, just not familiar with the maven process
<evanjs>
srhb: .....
<evanjs>
srhb: guess what
<srhb>
mog: If you're interested I recently tried mavenix by icetan and was pleasantly surprised.
<srhb>
evanjs: 'sup :P
<evanjs>
srhb: uhhhhh.... that worked....
<srhb>
Well, at least we know it's reproducible, just not why.
<srhb>
(Which is usually the point where I go **BIGSHRUG** and move on :P)
<evanjs>
srhb: Yeah.... sorta confused but I'm sure it'll get easier once it's in tree.... and yeah basically
evanjs| has quit [Client Quit]
<evanjs>
Trying to wait a week before I ping the last maintainer that requested changes for review :P
<srhb>
evanjs: Didn't it just get merged?
evanjs| has joined #nixos
<evanjs>
What the
<evanjs>
infinisil++
<{^_^}>
infinisil's karma got increased to 136
<evanjs>
Did not notice that
<infinisil>
Hehe
rauno has joined #nixos
<{^_^}>
[nixpkgs] @srhb merged pull request #68001 → python3Packages.cherrypy: fix tests after upgrade to pytest5 → https://git.io/fjpxo
<srhb>
mog: Yeah, that's been my experience with most packages too.
<srhb>
Quite impressed because everything else I've tried to do with maven and nix has just been utter hell.
<mog>
i figured it would be simple. i think ill open issue with the default.nix and mavenix.lock
<evanjs>
srhb: same... been looking at mavenix but I think there are still some blockers for us at work right now. i.e. submodules in private repos, etc...
<srhb>
evanjs: I have seen it fail on some other complex setups, which i don't know enough maven to understand, but it seems a huge improvement at least.
bakakuna has joined #nixos
<evanjs>
srhb: wait derp I was building on the wrong computer >_> fingers crossed
<evanjs>
re openrazer weirdness
vyorkin has joined #nixos
<srhb>
#nixtroubles
<evanjs>
Worth it but bah
vyorkin has quit [Client Quit]
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
byteflam1 has quit [Ping timeout: 246 seconds]
<{^_^}>
[nixpkgs] @mrVanDalo opened pull request #68019 → terranix: init at 2.1.1 → https://git.io/fjhJo
byteflam1 has joined #nixos
<{^_^}>
[nixpkgs] @grahamc opened pull request #68020 → Vacation PR: multi-paged, styled, searchable docs with wrapper tooling and epubs for NixOS and Nixpkgs → https://git.io/fjhJi
<srhb>
"Vacation"
<gchristensen>
:|
<srhb>
"No internet"
<{^_^}>
[nixpkgs] @grahamc pushed 60 commits to samuel-docs/searchable-2: https://git.io/fjhJP
<{^_^}>
[nixpkgs] @grahamc pushed 0 commits to samuel-docs/searchable-2: https://git.io/fjhJX
<gchristensen>
srhb: I haven't left yet! I've been postponing my leaving to get this PR in :P
<srhb>
Okay. :P
schjetne has joined #nixos
<srhb>
I'm just saying, the assimilation might have progressed further than we assumed...
polman has quit [Ping timeout: 245 seconds]
wildtrees has joined #nixos
polman has joined #nixos
hmpffff has joined #nixos
dasj19 has joined #nixos
ky1ko has joined #nixos
byteflam1 has quit [Ping timeout: 245 seconds]
schjetne has quit [Ping timeout: 246 seconds]
<gchristensen>
okay bye, have a great 2 weeks. don't do anything unreproducible.
fusion809 has quit [Remote host closed the connection]
<fvh>
hi folks, can someone explain how to troubleshoot error of a form "error: value is a string while a set was expected"? I'm trying to deploy hydra with nixops and get this error, on showing trace there are more cascading values errors like "while evaluating the option `system.nixos.release' " at `.nix-defexpr/channels/nixpkgs/lib/modules.nix:336:9:`
<fvh>
polman has quit [Ping timeout: 245 seconds]
bakakuna has quit [Ping timeout: 258 seconds]
polman has joined #nixos
<srhb>
fvh: My usual method is to stare really really hard at the code.
<infinisil>
^^ same
<infinisil>
> "string".attribute
<{^_^}>
value is a string while a set was expected, at (string):263:1
<fvh>
@srhb this error coming from global modules. Is that good practice to edit them?
<srhb>
fvh: If there's a bug, sure, but does it not seem plausible that it could be in your config?
<infinisil>
This error message doesn't tell you where the error actually is
<infinisil>
module.nix is just where the error gets detected
<evanjs>
Yeah, usually I either look for any items in my config that appear in the output, or use --verbose and see go up the list of the last few files that were evaluated
<tilpner>
If your config is version controlled, triple-check the diff
<tilpner>
Or binary-search your imports
<fvh>
The thing is that, if I run everything without 1 file, related to jobset on hydra, then everything works but why jobset might need nixos version at all is not. So tilpner you suggest that bad imports?
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
<tilpner>
fvh: I suggested commenting out imports to narrow down where your mistake is, but it appears you already know where it is
<fvh>
yes, I narrow it down by splitting in smaller files
<{^_^}>
[nixos-hardware] @globin pushed 3 commits to master: https://git.io/fjhUg
hmpffff has quit [Client Quit]
zupo has joined #nixos
freeman42[NixOS] has joined #nixos
<kraem>
clever: i didn't succeed creating the overlay which you tried helping me yesterday :( this is what i've got (am i calling rec in the right place?) https://bpaste.net/show/hFlY (i've also tried using oldAttrs) - any pointers?
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
zupo_ has joined #nixos
zupo has quit [Ping timeout: 258 seconds]
<srhb>
kraem: overrideAttrs takes a function, so ...overrideAttrs (oa: { ...
<srhb>
kraem: moreover you also need to set the rest of the fetchFromGitHub attrs
<srhb>
kraem: (You aren't somehow deep overriding into that, you're just overriding src, so you need to reconstruct it from scratch)
<srhb>
kraem: You probably also want to override name, but that's mostly cosmetics, version won't somehow propagate into the existing name string, which already has the old version substituted in.
<srhb>
kraem: Again, this is mostly understanding that things don't propagate the way you think they are :)
<srhb>
kraem: Your rec also does nothing, there's no attributes in your overlay that refer to each other as far as I can see.
<srhb>
kraem: If you wanted one of those to depend on the other, you're going to need to do that with override on top of what you're already doing.
<srhb>
I think that's it.
<kraem>
srhb: thanks! let me read that one or two times and see if i get it :)
<kraem>
srhb: so i am not able to reference the new version attribute i've set, using self, while setting rev in src= ? do you propose i hard code it in for the overlay?
andreas303 has quit [Remote host closed the connection]
<srhb>
kraem: That's what rec is for
<kraem>
as soon as i am using ..overrideAttrs (oa: {.. i can't reference fetchFromGithub
<srhb>
kraem: rec allows you to refer to another attr defined in the same set.
<srhb>
kraem: That doesn't sound right..
<srhb>
kraem: Did you mean fetchFromGitHub?
zupo_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<srhb>
kraem: (capital H)
andreas303 has joined #nixos
byteflam1 has quit [Ping timeout: 264 seconds]
feep has joined #nixos
<kraem>
yeah thanks! :) now i'm back at not being able to reference version within src. shouldn't i be able to reference self.version since that's the new attr version i've set a couple of rows above?
<srhb>
dminuoso: Yeah. It just looks up nixpkgs in NIX_PATH
<evanjs>
This works nicely for me
<evanjs>
usage: e.g. pkgs.unstable-small.linuxPackages_latest
<srhb>
dminuoso: This is why many people make a shell.nix for the nixops invocations that sets NIX_PATH to the desired commit of the deployment in its entirety.
<srhb>
k0ral: I really wouldn't, that sounds awful :P
<k0ral>
notably, how would you declare the creation of such a filesystem so that it is created only if it does not exist yet (idempotent)
<k0ral>
srhb: then how would you solve the need for a size-limited folder ?
<srhb>
quotas sound good.
<srhb>
Applying them (potentially repeatedly) also doesn't sound too scary.
domogled has quit [Ping timeout: 245 seconds]
<simpson>
k0ral: The repeated mounting can be made declarative, and one-time setup need not be declarative beyond scriptability. But really, just use more users/groups and quotas, I think.
<k0ral>
srhb: I expected that answer, but quotas are user/group-based, not folder-based (afaik), and I really need a folder-based solution
<tokudan>
k0ral, wow, are you trying to emulate zfs with ext3 there?
<srhb>
k0ral: Depends on your fs.
<srhb>
k0ral: You can certainly get per-directory (or per volume at least) quotas.
<k0ral>
srhb: don't want to use a zfs on my raspberry as it has low memory :)
<srhb>
And that's the right tool for this job, imo :)
tpanum- has joined #nixos
<srhb>
k0ral: Sure, that's fair. LVM?
tpanum has quit [Ping timeout: 246 seconds]
Guillaum has quit [Ping timeout: 268 seconds]
<tokudan>
k0ral, reduce the zfs arc and you'll probably be fine, though i admit i've never use zfs in a low memory environment
acarrico has joined #nixos
<srhb>
k0ral: I think XFS can also do something on a directory level
wfranzini has joined #nixos
<buckley310>
are per-user quotas not fit for what you're doing?
<tokudan>
another idea: run one web server per user and just proxy to the right instance
<k0ral>
simpson: it might look like a one-time thing, setting up the file-filesystem, but I have multiple raspberry pi on which I will need to do it, so it would be nice if I can have nixos do it for me
<srhb>
k0ral: Yeah, that's the reason I think quotas sound nice. Applying them again and again with some job sounds safe.
<{^_^}>
[nixpkgs] @alyssais pushed commit from @manveru to master « ruby.withPackages: init »: https://git.io/fjhkL
<k0ral>
srhb: can you point at the specific nixos options that are related to quota setup ?
<simpson>
k0ral: However you decide to do it, you may want to make a golden SD-card image, and then copy that image using a relatively high-bandwidth SD-card reader. RPis are not known for their bandwidth.
tpanum- has quit [Ping timeout: 245 seconds]
npmccallum has quit [Quit: npmccallum]
alp has quit [Ping timeout: 246 seconds]
dasj19 has quit [Quit: dasj19]
<qyliss>
so happy we finally have ruby.withPackages
<srhb>
k0ral: I don't know of any existing settings. I'm literally thinking "run this job at boot" or something like that :)
zupo has quit [Ping timeout: 258 seconds]
<qyliss>
manveru++ amazing work
<{^_^}>
manveru's karma got increased to 24
Zer0xp has joined #nixos
<manveru>
qyliss++
<{^_^}>
qyliss's karma got increased to 6
Guillaum has joined #nixos
<{^_^}>
[nixpkgs] @vbgl opened pull request #68025 → ocamlPackages.janeStreet: init at 0.12 (for OCaml ≥ 4.07) → https://git.io/fjhkY
matthuszagh has joined #nixos
<simpson>
manveru++ qyliss++
<{^_^}>
manveru's karma got increased to 25, qyliss's karma got increased to 7
alp has joined #nixos
xkapastel has quit [Quit: Connection closed for inactivity]
k0ral has quit [Ping timeout: 276 seconds]
eyJhb has joined #nixos
<matthuszagh>
i'm currently using startx with exwm as a window manager. this mostly works great, but occasionally i run into issues which I think are related to needing a more full-fledged desktop environment. is it possible to keep my startx exwm config and sometimes run kde plasma from startx?
k0ral has joined #nixos
<k0ral>
srhb: LVM would certainly answer my need, although it looks unnecessarily powerful, I hoped for something simpler
feep has joined #nixos
Guillaum has quit [Ping timeout: 264 seconds]
chr15 has joined #nixos
ixxie has quit [Ping timeout: 245 seconds]
Guillaum has joined #nixos
Stuck_ has joined #nixos
feep has quit [Ping timeout: 244 seconds]
andgate has joined #nixos
<tokudan>
k0ral, LVM is definitly more common than having a bunch of filesystems lying around as files and mounting them through loopback devices.
<Stuck_>
Hi, I've a haskell package (random) installed, but the package is hidden, I've tried to expose the package using `sudo ghc-pkg expose random` but it says I don't have permission... even in root
<{^_^}>
[nixpkgs] @andir pushed 10 commits to release-19.03: https://git.io/fjhku
<clever>
Stuck_: what is claiming its hidden?
<clever>
Stuck_: did you add random to your cabal file as a dependency?
<Stuck_>
the ghci
knupfer has joined #nixos
<clever>
Stuck_: how are you running ghci?
o1lo01ol1o has joined #nixos
<Stuck_>
in the terminal
<tokudan>
k0ral, if it's just about php, it might be an idea to sping up a php-fpm pool for each user, so their own files get created with their user and then you could use quotas
<Stuck_>
in no particular project
feep has quit [Ping timeout: 246 seconds]
<clever>
Stuck_: that should just work..., what does `type ghci` return?
<Stuck_>
ghci is /run/current-system/sw/bin/ghci
<clever>
Stuck_: i also prefer to not install ghc system-wide, and to only use it from within a nix-shell
byteflam2 has joined #nixos
<k0ral>
tokudan: LVM is necessarily setup before installing nixos, unless I am mistaken, so I won't get the nice declarative/automated setup
<Stuck_>
do you think that would work?
<clever>
Stuck_: it works perfectly fine on this end
<tokudan>
k0ral, not if you dedicate a partition to it and create that during bootup if it hasn't been setup yet
<Stuck_>
so I can't have a package exposed if I'm using ghc system-wide?
<symphorien>
,libraries Stuck_
<{^_^}>
Stuck_: Don't install libraries through nix-env or systemPackages, use nix-shell instead. See https://nixos.wiki/wiki/FAQ/Libraries for details.
<clever>
Stuck_: i'm not sure why it isnt working when you install it system wide, but ^^
zeta_0 has joined #nixos
<zeta_0>
`
wfranzini has quit [Remote host closed the connection]
drakonis has joined #nixos
wfranzini has joined #nixos
byteflam1 has quit [Quit: Lost terminal]
<zeta_0>
i can't find a `protonvpn` package in nix pkgs, so how do i install and setup/configure protonvpn in nixos ?
<hyperfekt>
What actually is the reason for .drvs to be in the store? When do you actually want to create a drv but not build it at the same time?
<Stuck_>
so if I'm to remove from the config ghc and the packages, and installed them with nix-shell it should work, right?
<symphorien>
with remote builders, hyperfekt
matthuszagh has left #nixos ["ERC (IRC client for Emacs 27.0.50)"]
<clever>
hyperfekt: nix-copy-closure can copy the drv files to a remote machine, nix-store -r /nix/store/foo.drv can re-build a drv, and skip the expensive eval, nix-shell /nix/store/foo.drv can shell into a drv to reproduce failures
<clever>
hyperfekt: nix-shell is handy when the thing is burried in let blocks and imposible to access with -A
<clever>
Stuck_: probably
<{^_^}>
[nixpkgs] @jtojnar closed pull request #47009 → openrazer: init at 2.3.1 (driver, daemon, lib and NixOS module) → https://git.io/fAQHY
<clever>
zeta_0: it doesnt appear to have a nixos module yet
<hyperfekt>
symphorien: If it was just for remote builders you could send the drvs straight to them instead of either of the machines putting them in the store
<boxdot>
When specifying buildInputs = [ cmake ninja ]; automagically ninja injects itself as build tool for cmake. How is this done? I can't find the place in nixpkgs where the injection happens.
Myhlamaeus has quit [Remote host closed the connection]
<clever>
boxdot: the setup hooks, run `nix-build '<nixpkgs>' -A ninja` and then `ls -l result/nix-support`
<clever>
boxdot: then repeat the above on cmake also
<hyperfekt>
clever: So, essentially eval caching and that's it (apart from the 'not properly accessible' thing)?
<{^_^}>
[nixpkgs] @fpletz pushed to gcc-8 « highlight: build with standard llvm version on darwin »: https://git.io/fjhki
npmccallum has joined #nixos
emptyfl4sk has quit [Ping timeout: 244 seconds]
<clever>
hyperfekt: yeah
<zeta_0>
clever: so the `protonvpn` package is still being developed ?
<clever>
zeta_0: i think the package is done, it just needs a module to be built
<boxdot>
clever: ah, this is outside of nix actually.
<clever>
boxdot: sorta, its a convention setup within nixpkgs, that happens inside the nix builds
<clever>
boxdot: the ninja setup hook, will overwrite the buildPhase, and the cmake setup hook will detect that, and change the cmake flags
<zeta_0>
clever: ok, thanks for the info
<zeta_0>
clever++
<{^_^}>
clever's karma got increased to 191
vyorkin`` has joined #nixos
vyorkin` has quit [Ping timeout: 246 seconds]
<disasm>
zeta_0: we'd happily accept a PR if you want to get it in, but freeze is coming soon.
<Stuck_>
so I've just remove all ghc, and random stuff from the config, rebuilded it, `nix-shell -p 'haskellPackages.ghcWithPackages (p: [p.ghc p.random])'`, ran ghci, tried to import System.Random and still hidden, then ran `sudo ghc-pkg expose random` and still with the same permission denied :(
<clever>
Stuck_: ghc expose will never work under nix
<clever>
Stuck_: what if you remove p.ghc from the list?
drakonis1 has joined #nixos
<Stuck_>
if I remove it how can I do ghc-pkg?
<clever>
Stuck_: ghcWithPackages gives you a ghc already
<{^_^}>
[nixpkgs] @aristidb pushed commit from @averelld to master « vim: 8.1.1547 -> 8.1.1967 (#68011) »: https://git.io/fjhkF
<clever>
Stuck_: try removing x86_64-linux-8.6.4
<pbb>
What could be causing my block devices to be "busy"? They are not mounted, lsof knows nothing, but wipefs or make-bcache say "Device or resource busy".
<clever>
pbb: was lvm previously on the device?
<pbb>
no
<clever>
pbb: mdadm? luks?
<pbb>
no
<clever>
pbb: what was on it then?
<pbb>
just a filesystem
<pbb>
I also just did wipefs -af and rebooted
<pbb>
Of the two devices I want to use sometimes the first is busy and sometimes the other one
nhey_ has joined #nixos
<{^_^}>
[nixpkgs] @globin pushed to gcc-9 « gcc: switch default to gcc9 »: https://git.io/fjhkA
<pbb>
clever: actually on the partition that's busy right now has never existed before, so the header should have been in the middle of a previous filesystem before I did wipefs -af
<{^_^}>
[nixpkgs] @globin opened pull request #68029 → [WIP] Default gcc to gcc9 → https://git.io/fjhkx
o1lo01ol1o has quit [Remote host closed the connection]
<clever>
pbb: what does `dmsetup ls` report?
<averell>
well then. might even get gcc9 done before branch-off :P
<nhey_>
How do I get man-pages for C library functions on NixOS? I have gcc installed - but for example man 3 printf or man scanf are missing
<averell>
you should install "manpages"
<clever>
,libraries nhey_
<{^_^}>
nhey_: Don't install libraries through nix-env or systemPackages, use nix-shell instead. See https://nixos.wiki/wiki/FAQ/Libraries for details.
<pbb>
clever: "dmsetup ls", says "No devices found"
zupo_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
schjetne has joined #nixos
<Yaniel>
oh joy, now some steam games are broken too
<nhey_>
nix-shell is awesome :D but I find when taking university courses where I am to do exercises and work in groups, sometimes it is preferable to have libraries in systemPackages
<evanjs>
Speaking of nix-shell, I still need to check out lorris
<nhey_>
Thanks for the link though, I understand why I would want to avoid installing systemwide gcc
<{^_^}>
[nixpkgs] @aristidb pushed commit from @Provessor to master « Add large hunspell dicts (#67424) »: https://git.io/fjhIG
Myhlamaeus has joined #nixos
<Stuck_>
sry, I've been distracted
<Stuck_>
Have you sugested something?
nhey_ has quit [Quit: Leaving]
<clever>
Stuck_: try removing x86_64-linux-8.6.4
Soo_Slow has joined #nixos
<Stuck_>
expose still doe'nt work
<clever>
Stuck_: expose will never work
<clever>
Stuck_: does import work in ghci?
<pbb>
ok, after wipefs -a on the whole drive, repartition and wipefs -a on the partition, and then reboot it works
<Stuck_>
0.0 yes it did
<Stuck_>
why?
Myhlamaeus has quit [Ping timeout: 246 seconds]
<clever>
Stuck_: you had something in your ~/.ghc that hid all packages
<pbb>
but now I created the bcache device, it says state is clean, but spams the dmesg with "bch_count_backing_io_errors() md127: Read-ahead I/O failed on backing device, ignore"
<Stuck_>
hoo, I'm gonna have to see more about that but thank you very much, I think that's all of my problems ^^
<Stuck_>
just for reference, it works even with ghc system-wide
<clever>
Stuck_: yeah, so it was likely ~/.ghc/ that has been causing all of your problems
UltraPhil has joined #nixos
Stuck_ has quit [Quit: leaving]
selfsymmetric-pa has quit [Remote host closed the connection]
Miyu-saki is now known as Miyu-chan
Zer0xp has quit [Ping timeout: 264 seconds]
chr15 has quit [Ping timeout: 245 seconds]
<{^_^}>
[nixpkgs] @rycee pushed commit from @sikmir to master « josm: 15238 -> 15322 (#68022) »: https://git.io/fjhIw
shibboleth has quit [Remote host closed the connection]
the-kenny-w has quit [Ping timeout: 244 seconds]
the-kenny-w has joined #nixos
vyorkin`` has quit [Ping timeout: 245 seconds]
<jasom>
Is there an easy way to get nix to build a certain set of programs with debug symbols? I'm getting a segfault that doesn't reproduce on other distros and was hoping to diagnose.
<emily>
there's a fancy filesystem that automatically fetches debug info for you
<clever>
jasom: if you need a certain library with debug, you can do `--arg config '{ packageOverrides = pkgs: { libfoo = pkgs.enableDebugging pkgs.libfoo; }; }'`
<clever>
and it will then rebuild whatever it needs to
<evanjs>
How do we get the same behavior as ignoreConfigError in the repl? e.g. if I'm trying to evaluate config.boot.kernelPackages.openrazer and I get `
<evanjs>
error: assertion failed at /nix/store/klq1dhdzhc60nai0cnqyjl0knn7q8bqi-nixos-unstable-small-19.09pre189595.147b7ab692b/nixos-unstable-small/pkgs/os-specific/linux/blcr/default.nix:5:1`
wfranzini has quit [Remote host closed the connection]