<{^_^}>
sshow: If you're updating a file in nixpkgs that starts with something like `{ stdenv, cmake }:`, use `nix-build -A` in the nixpkgs root with the corresponding package attribute to build it. If it's not in nixpkgs, try `nix-build -E 'with import <nixpkgs> {}; callPackage ./file.nix { }'. The mapping from package attributes to package files is in pkgs/top-level/all-packages.nix.
fendor has quit [Remote host closed the connection]
<lordcirth_>
aleph-, I think I figured it out. Restarting the service after boot seems to work. I'm betting some kind of race. I added after = ["nss-lookup.target"], I think that will fix it
<attila_lendvai>
how can i make a copy of an entire set? i know about inherit, but i don't want to list all the elements. couldn't find anything in nix pills nor the nix manual
<infinisil>
attila_lendvai: a = b?
<infinisil>
Where a is the new set, b the old one :)
<infinisil>
(But really, both are the same)
ddellaco_ has quit [Remote host closed the connection]
<attila_lendvai>
infinisil, but i want to override some attributes, too
Dotz0cat_ has joined #nixos
<infinisil>
a = b // { someAttribute = someValue; }
<pasc_>
hi! is there a way to deterministically "hash" a nix-build including the options and NIX_PATH ?
<pasc_>
I'd like to know if a particular module has already been evaluated with the same NIX_PATH
<aleph->
Hmm is there a linux_5_9 kernel image deriv somewhere?
<pasc_>
since I'm trying to cache the evaluation itself, I can't rely on the build's output already being in the Nix store
<pasc_>
is there a better way than to hash the contents of every directory in NIX_PATH ?
<pasc_>
(I suppose flakes would help here)
<pasc_>
(and of course, I'm using pure evaluation for this)
Dotz0cat has joined #nixos
nwspk has quit [Quit: nwspk]
<pasc_>
it's just to the nix interpreter can be a bit slow to evaluate big expressions, and I'd like to cache its results, so I need to be able to hash all of its outputs
attila_lendvai has quit [Ping timeout: 272 seconds]
<{^_^}>
g-w1: If you're updating a file in nixpkgs that starts with something like `{ stdenv, cmake }:`, use `nix-build -A` in the nixpkgs root with the corresponding package attribute to build it. If it's not in nixpkgs, try `nix-build -E 'with import <nixpkgs> {}; callPackage ./file.nix { }'. The mapping from package attributes to package files is in pkgs/top-level/all-packages.nix.
quinn has joined #nixos
<pasc_>
ah, all-packages.nix, not top-level.nix :P
<colemickens>
this is new: suspicious ownership or permission on '/nix/store/si796kp5rxk30fprbr7fdmakj72xfqi4-append-initrd-secrets.drv.chroot/nix/store/3616s7a47279787f4xfahpdqqfmdwicb-append-initrd-secrets' for output 'out'; rejecting this build output
<mindavi>
I'm looking at mobile-nixos, but having some issues building anything that requires asciidoc - asciidoc seems to have issues being cross-compiled. There are some paths being substituted using sed, but that uses the cross-compiled tools instead of the native tools. Is there any way to mention to nix that the tools to use should be the 'host' tools when directly substituting in a file?
<inf>
hey all! I *suspect* i've just stumbled upon a bug in /run/pulse permission race condition. After enabling systemwide pulse /run/pulse *sometimes* gets its permissions changed to 0700, and then systemctl restart pulseaudio fixes that to 0755...
fuiltilt has quit [Quit: WeeChat 2.9]
<inf>
yeah. any nixos-rebuild switch (even without any changes) changes /run/pulse permissions to 0700, unless pulseaudio gets restarted as well... so this seems like something going on in nixos activation script maybe?
<inf>
this seems like a regression in pulseaudio module as well... I have a system still running 20.03 with systemwide pulseaudio and user added to `audio` group can use that without that hack. :/ (I'm stumbling onto above issues on today's nixos-unstable)
kapil_ has quit [Ping timeout: 272 seconds]
avaq has quit [Remote host closed the connection]
avaq has joined #nixos
Darkmatter66 has quit [Ping timeout: 256 seconds]
Guest52124 has quit [Client Quit]
<inf>
though that user is exlicitly declared as "isNormalUser". not sure how groups handling in systemd services works...
<m0rphism1>
Is it a common problem to experience lags while updating with `nixos-rebuild switch`? CPU and RAM usage is pretty low, but every ~30 seconds or so my whole system freezes for 5 seconds. I guess some part of the update process is making a lot of operating system calls. Does someone experience similar issues or knows a way around it?
<m0rphism1>
by "whole system freezes" I mean that xserver doesn't even let me move my mouse and audio playback stops.
ahmed_elgabri has quit [Ping timeout: 240 seconds]
sangoma has quit [Ping timeout: 240 seconds]
<m0rphism1>
Also I'm running NixOS on a fairly strong desktop machine, so there should be quite a bit of resources available.
<m0rphism1>
Yaniel: Thanks, good to know.
<Yaniel>
if it happens regularly, try pinpointing which part of the rebuild process it is
<adisbladis>
m0rphism1: IME that usually has to do with IO more than CPU/RAM
<Yaniel>
the part where it restarts services etc can definitely cause hiccups
<m0rphism1>
adisbladis: hmm, the system is installed on an SSD drive at least. Or do you mean IO in the sense of interaction with the operating system?
meh` has quit [Ping timeout: 246 seconds]
<adisbladis>
Well, both.
<Philonous>
I want to change filesystem options that are defined in hardware-configuration.nix among other things I'd like to use a logical path for the device rather than the UUID, but when I redfine them in configuration.nix I get "The option `fileSystems./.device' has conflicting definitions". I guess I couldjust modify the hardware-configuration.nix, but the file tells me not to do that, what's the best way to fix this?
<adisbladis>
On my old laptop I had slowdowns on the regular until I switched from the crappy bundled nvme drive to something better
<m0rphism1>
Yaniel: It happens during the package builds. I think cudatoolkit causes the worst freezes, but also other stuff like gtk+3. Now that I mention it, I'm not sure if that stuff should even be built from source.
<Philonous>
m0rphism1, I used to get similar problems on my old machine. Lots of disk IO would make it hang for seconds on end. Even just copying around large files could trigger it.
<adisbladis>
m0rphism1: If you're following nixos-unstable gtk3 should never be built from source
<inf>
Philonous: you could do ...device = lib.mkForce "newvalue";
<adisbladis>
cudatoolkit is non-free and not built on hydra
mkaito has joined #nixos
mkaito has quit [Changing host]
mkaito has joined #nixos
mkaito has quit [Client Quit]
mkaito has joined #nixos
<m0rphism1>
Is there a way to limit IO usage? I would be fine with longer build times as long as it doesn't interrupt me using the system.
<inf>
Philonous: in Your configuration.nix, that is. this is how you more-or-less solve conflicting attribute values.
<Philonous>
inf, Ah thanks, I'll try that
<m0rphism1>
adisbladis: I'm on stable nixos-20.09, that should also be cached right?
<afreak>
im having some issues with /boot device, nixos-rebuild complained it was readonly, so i umounted it, fsckd it and now i cant mount it again, more info here https://kopy.io/HkcuX#7Bx6QKD28Lm98B
ahmed_elgabri has quit [Ping timeout: 240 seconds]
n-osborne has quit [Remote host closed the connection]
o1lo01ol1o has quit [Remote host closed the connection]
n-osborne has joined #nixos
johnw has quit [Remote host closed the connection]
Qwerky has joined #nixos
kfm has quit [Quit: ciao kakao]
lorwin has joined #nixos
lorwin has quit [Client Quit]
domogled has quit [Ping timeout: 272 seconds]
kmein has joined #nixos
raghavsood has quit [Quit: Connection closed]
stree has quit [Ping timeout: 240 seconds]
vidbina_ has joined #nixos
bitmapper has quit [Quit: Connection closed for inactivity]
<twink0r>
Hi! Is there a good way to start writing own service files? I want to use vmalert and the binary if already part of the victoriametrics package
<zn60>
For my own user environment I often end up writing derivations that basically have a builder script that just does `cp -a "$src" "$out" ; patchShebangs "$out"`. This seems to fail with permissions problems during `patchShebangs`. So I know I could do two `cp -a` phases through a working directory to do `patchShebangs` but is there a more
<zn60>
standard/streamlined way to do this simple base case?
<clever>
zn60: $src is read-only, so the copy in $out will also be read-only
<clever>
zn60: you need to run a `chmod -R +w` over it to fix that
jluttine has joined #nixos
o1lo01ol1o has quit [Ping timeout: 240 seconds]
<zn60>
Is `chmod -R u+w "$out"` a standard idiom?
<zn60>
Thanks!
<clever>
yeah, that should work
rajivr has quit [Quit: Connection closed for inactivity]
<sterni>
clever: Thra11: haskell.lib.overrideSource is more idiomatic I guess (also allows you to set version which allows to easily confirm that the override is working)
<NickHu>
Hi, is anyone using rustc from mozilla nixpkgs with the wasm32-unknown-unknown target? As of recent, I get the following errors: rust-lld: error: unable to find library -lstdc++ rust-lld: error: unable to find library -lgcc
<NickHu>
Hi, is anyone using rustc from mozilla nixpkgs with the wasm32-unknown-unknown target? As of recent, I get the following errors:
<NickHu>
rust-lld: error: unable to find library -lstdc++ rust-lld: error: unable to find library -lgcc
<NickHu>
Hi, is anyone using rustc from mozilla nixpkgs with the wasm32-unknown-unknown target? As of recent, I get the following errors: rust-lld: error: unable to find library -lstdc++ rust-lld: error: unable to find library -lgcc
<NickHu>
It used to work fine but it suddenly stopped working
<NickHu>
(Sorry for message spam, freenode bridge being flaky)
<installnixos[m]>
Hmm maybe try giving gcc as a build input?
<NickHu>
installnixos: already doing that
<installnixos[m]>
Maybe add gcc headers to LD_LIBRARY_PATH? If that doesn't work I'm out of ideas..
FRidh has quit [Quit: Konversation terminated!]
<NickHu>
I tried with `export LD_LIBRARY_PATH=$(nix-build -E 'import <nixpkgs>' -A 'gcc.cc.lib')/lib64` but that didn't fix things
<davidak[m]>
ryantm: why is the PR now merged into staging? that leads to more time people on unstable have to wait for the latest releases https://github.com/NixOS/nixpkgs/pull/113743
<davidak[m]>
ryantm: can something be done about that? should i open an issue?
o1lo01ol1o has quit [Remote host closed the connection]
<installnixos[m]>
Idk if gcc even has that but if it does that's where the headers probably are.
<NickHu>
installnixos: it doesn't seem to have that
<{^_^}>
[nix] @domenkozar pushed to installer-artifact « Generate installer script for each PR/push »: https://git.io/Jt9UY
o1lo01ol1o has joined #nixos
<ryantm>
davidak: Yeah, I'm sure something can be done. I'm focusing on improving the nix ecosystem documentation situation right now though. Feel free to open an issue.
<installnixos[m]>
Nick Hu add zlib or zlib.out to propagated build inputs
<installnixos[m]>
Sorry I cant actually try it. I'm on my phone : ( will try later
kw has joined #nixos
<NickHu>
installnixos: same issue
superherointj has joined #nixos
<kw>
How do I determine what packages in my config.nix depend on go-1.14?
sangoma has quit [Read error: Connection reset by peer]
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<OneLegend>
a good 75% of messages are from the bot
<KarlJoad>
Is there a best way to structure my NixOS configuration files for multiple machines?
<figgyc>
hi, i'm doing nix flake update and for some reason flake.lock seems to think the latest nixos-unstable is 6b1057b452c55bb3b463f0d7055bc4ec3fd1f381 ? im not sure why it seems to think 9816b99e71c3504b0b4c1f8b2e004148460029d4 doesnt exist, is it a caching thing? wouldnt normally care, just want to get past the commit which fixes the broken bluetooth :P
<tpw_rules>
are you actually updating the input?
dev_mohe has quit [Quit: dev_mohe]
teddyc has joined #nixos
mallox has quit [Quit: WeeChat 3.0.1]
<sterni>
KarlJoad: I usually split them into multiple modules where I have absolute base settings, settings for machine types (servers, workstations, ..), machine specific settings
<sterni>
KarlJoad: also you can split program / service specific configuration into own modules as well if wanted
<tpw_rules>
i.e. passing --update-input nixos-unstable or whatever
<sterni>
but there is really no right or wrong way
<sterni>
whatever works for your use case
<figgyc>
tpw_rules: not until now, that did the trick :) thanks
<tpw_rules>
yeah, nix flake update will not update anything that's already locked unless you specifically name it with that flag
<tpw_rules>
or use some other flag called like --regenerate or something that updates everything at once
<tpw_rules>
--recreate-lock-file
<KarlJoad>
sterni: I started doing that. I'm asking because I've written something into my hardware-configuration.nix that I do not want to lose, and is machine-specific. Do I just add it to the imports list in configuration.nix?
<sterni>
KarlJoad: I usually copy the hardware-configuration.nix into the machine cnofiguration first thing I set up the machine tbh
fendor has quit [Remote host closed the connection]
<sterni>
there's no reason why it should be separate
<KarlJoad>
sterni: The only reason I'm hesitant is because I symlinked hardware-configuration.nix to the machine-specific configuration, and I need to install a new NVMe to the system for auto-mounting.
<{^_^}>
#30742 (by ljli, 3 years ago, open): buildRustPackage: found duplicate version of package
<rmcgibbo[m]>
I'm trying to get "vector" working, which is supposed to be a tool that can forward logs from journald -> amazon cloud stuff, but it looks like the hydra build is failing and I hit #30742 when I try to upgrade to the last upstream release