<deni>
elvishjerricco: I mean your blogpost does say that I can use --option plugin-files but I have no idea how to find where the files are located at
<elvishjerricco>
deni: You just need to install nix-plugins locally and use --option plugin-files
<deni>
It seems that a lot of tutorials are written with the premise that the start machine (ie. the one I'm running stuff on, like nixops) is also nixos.... I find when that's not the case that I hit a lot of edge cases
<deni>
Ok so I can use nix-env to install nixpkgs.nix-plugins. This will tell where in the store it's located.
<deni>
so my only option is to set --option plugin-files each time? I guess I can wrap the call to nixops with a script/Makefile
<deni>
seems I'm doing something wrong. Was it wrong to install nix plugins with nix-env -iA nixpkgs.nix-plugins ?
<energizer>
it would be cool to have a command on the local machine that runs nixops on another machine to rebuild the local machine. that way it wouldn't be necessary to write secrets to disk
<deni>
energizer: I'm not really rebuilding the local machine in my case...I'm just using nixops from my local machine to deploy to another machine. I am using a remote builder in the process though...I'm not yet sure if that will cause problems with this specific scenario with secrets. But I don't think so.
lordcirth_ has joined #nixos
<deni>
Somewhat relatedly...I asked on twitter a couple of days ago if anyone is playing around with Hashicorp vault and Nix/Nixops. Sadly I didn't get any response. I'm quite fond of Vault and have used it to great success before. But I'm at a lost how to integrate it with nixops other than with wrappers and what not
<energizer>
somebody was talking about it recently
<lordcirth_>
I'm trying to help with ZHF 20.03, but I am having trouble figuring out how to run the same tests that Hydra does. I can't find the docs on nixpkgs tests. I'm probably missing something obvious?
fendor has quit [Read error: Connection reset by peer]
<deni>
evils: energizer thanks! I see gchristensen is playing around with it. I have some experience with vault auto unsealing ... albeit in the cloud setting (AWS and GCP mostly). I'd need to think if any of that is applicable to my current non-cloud setup
<{^_^}>
[nixpkgs] @worldofpeace pushed 2 commits to release-20.03: https://git.io/JvHpi
* colemickens
googles yaml2nix
jb55 has quit [Remote host closed the connection]
<deni>
evils: energizer wow that discussion has been going on for quite some time now! :D It's very interesting to read. I wonder if peopel use https://github.com/channable/vaultenv ?
* colemickens
is surprised
<deni>
I quite enjoyed it for my non-vault-aware applications
<samueldr>
since you've read the readme, bhipple, you understand that if you run on an incompletely built eval, the next time you run it will not use the updated state unless you rm those files, right?
<evanjs>
this happens when I try to build a barebones musnix system with either nixops or morph
<bhipple>
I have to rm everything?
<samueldr>
if things change for that eval in hydra
<bhipple>
It takes so long to run from a clean slate :'(
<samueldr>
it's (too) heavy-handed in its caching
<samueldr>
due to that exact reason
<bhipple>
OK, so if I just always run it on the (N-1)th job eval, I shouldn't ever have to clean out my cache?
Havvy has quit [Ping timeout: 256 seconds]
iyzsong has joined #nixos
<bhipple>
You mean I have to remove all the build_115548183
<bhipple>
files, or just the eval_* files?
ixxie has quit [Ping timeout: 256 seconds]
chagra has quit [Ping timeout: 240 seconds]
<{^_^}>
[nixpkgs] @bhipple merged pull request #83526 → pythonPackages.scikit-build: fix test bug on python3.8 → https://git.io/JvH5B
<emily>
I'm not sure why it would be failing; I have networking.wireguard.enable etc. on
<emily>
my best guess is maybe wireguard got incorporated ahead of schedule and so the optional condition there is wrong?
<emily>
actually that makes no sense since modprobe wouldn't be failing, so I have no idea why it's not correctly finding/installing the module ^^;
mexisme has quit [Ping timeout: 256 seconds]
<{^_^}>
[nixpkgs] @bhipple opened pull request #83555 → python3Packages.apache-airflow: no-op cleanups to drv file → https://git.io/JvHhG
reallymemorable has quit [Quit: reallymemorable]
<{^_^}>
[nixpkgs] @bhipple opened pull request #83556 → <!-- Nixpkgs has a lot of new incoming Pull Requests, but not enough people to review this constant stream. Even if you … → https://git.io/JvHh8
<{^_^}>
[nixpkgs] @bhipple closed pull request #83556 → <!-- Nixpkgs has a lot of new incoming Pull Requests, but not enough people to review this constant stream. Even if you … → https://git.io/JvHh8
<deni>
well now having weeded through 2 years of IRC history it's time to go to bed. XD
<bhipple>
deni: you should write up a summary of the most interesting conversations :)
<deni>
bhipple: was mostly researching old conversations about nix+nixops+vault .... the question comes up *a lot*
<evanjs>
Oh it's dynamically linked, I just don't see it referencing SDL2
<evanjs>
adisbladis: x_x plz no
h0m1 has quit [Ping timeout: 240 seconds]
<adisbladis>
evanjs: I'm wildly speculating :)
mexisme has joined #nixos
<evanjs>
sure but still lol. scary thoughts
<tnks>
Anyone have a good NixoS Packer project I can steal from to make a VMWare image?
<Jonathan>
evanjs: I am lost.
<tnks>
Even if it makes a VirtualBox image, I think it could be helpful.
<evanjs>
Jonathan: dlopen is a nasty thing on nixos, though we don't know if that's what is being used in this case
h0m1 has joined #nixos
<Jonathan>
Oh, interesting.
<adisbladis>
How do I build an executable from that repo?
<Jonathan>
evanjs: Let me see if static linking works.
<evanjs>
add dub and dmd to your shell
<evanjs>
and just run dub
<evanjs>
adisbladis: ^
<Jonathan>
`dub test` is where it is failing (which runs the unittests).
reallymemorable has quit [Quit: reallymemorable]
<evanjs>
ohhh thank you. there's the failure lol
reallymemorable has joined #nixos
<Jonathan>
evanjs: When I switch to static linking, it has a weirder bug.
<Jonathan>
`/nix/store/ajrrkivdfvp8dp4vdg5hp1h5hblmanc9-binutils-2.31.1/bin/ld: .dub/build/sdl-d-test-library-unittest-linux.posix-x86_64-dmd_2085-936A2417E4DA8A563B14BEB7ACC10A5B/sdl-d-test-library.o: in function `_D3SDL9Exception11sdlEnforce0FiZv':/home/jonathan/Repos/sdl-d/src/SDL/Exception.d:9: undefined reference to `SDL_GetError'`
gustavderdrache has left #nixos [#nixos]
<evanjs>
adisbladis: argh. sounds closer to what you were saying ...
<evanjs>
"This does not mean they require dynamic linking. There is no link-time dependency at all. Instead, the C shared library is loaded manually at runtime with a call to a load function on a loader instance, e.g. DerelictSDL2.load."
<evanjs>
figured as much but forgot what the syntax was lol
<adisbladis>
Or you build the binary and patchelf it, but that kind of sucks for development
<evanjs>
yeah that sounds ugly
johnnyfive has quit [Ping timeout: 264 seconds]
<evanjs>
hasn't crashed yet ...
<Jonathan>
Looks like it worked.
<adisbladis>
=)
<Jonathan>
Do you mind (trying) to explain to me why.
<adisbladis>
Jonathan: LD_LIBRARY_PATH is the search path environment variable
<evanjs>
Yeah, nix needs things to be linked at build time, either so we can patchelf the dependencies to point to their respective equivalents in the nix store, or so it links the proper things to begin with
<adisbladis>
We force the loader to look in the right place
<nh2>
hpfr[m]: you got in the past `hook reply is 'decline'` when trying to use remote builders. (How) did you fix it?
<evanjs>
when libraries don't link things and use e.g. dlopen and "expect the library to by in the system path"... that doesn't really work on vanilla nix
<Jonathan>
Right, which is basically a runtime linking sort of thing?
<Jonathan>
I am lost as to how a normal linking process can find the libs but not this system. Does a normal linker/loader look use different methods of finding it?
<Jonathan>
Or does that blog post explain that.
sigmundv_ has quit [Ping timeout: 264 seconds]
<adisbladis>
Jonathan: The library is in fact not linked at all, it's being loaded through the dlopen() syscall
<adisbladis>
On "regular" distros the loader spelunks a few well known directories for the library you're trying to load
<adisbladis>
But on Nix we don't have /lib, /usr/lib and so on
<evanjs>
adisbladis++ because I suck at explaining this stuff :P
<{^_^}>
adisbladis's karma got increased to 42
<evanjs>
"global linux compatability" bleh
<Jonathan>
yeah, nix is brilliant but sometimes take quirking to be able to working with things from the stateful world.
<adisbladis>
Jonathan: With Nix binaries are patched to refer to explicit store paths in their RPATH (the "binary local" search path)
<adisbladis>
If you enter a nix-shell you'll find environment variables like NIX_LDFLAGS & NIX_CFLAGS_COMPILE which our C compilers are wrapped to be aware of
<adisbladis>
That's the same thing happening inside the nix sandbox
<Jonathan>
Oh, so Nix thingies (e.g. make) are patched to "look" in the appropriate Nix places.
<Jonathan>
But Dub, nor the derelict loader are specially patched.
<adisbladis>
Exactly =)
<evanjs[m]>
Even worse than “not patched” they’re not even linked...
<Jonathan>
Ah, see.
<evanjs[m]>
At least for some libs
<Jonathan>
Okay, I am starting to wrap my mind around parts of Nix.
<adisbladis>
So, we can either set LD_LIBRARY_PATH to get the dlopen() syscall look in the right place
<adisbladis>
Or we can use a tool called patchelf to add the correct RPATH
<Jonathan>
patchelf works for linking but not loading?
<Jonathan>
In that shell, how does `${pkgs.SDL2}` (a package) become a path?!
<evanjs>
"Too often people use it as a crutch for not doing the right thing (i.e. relying on the compiled in path). Often programs (even commercial ones) are compiled without any run-time loader paths at all, forcing you to have LD_LIBRARY_PATH set or else the program won't run."
<{^_^}>
[nixpkgs] @bhipple opened pull request #83573 → pythonPackages.effect: fix build by marking py3 only → https://git.io/JvQvs
<{^_^}>
[nixpkgs] @Ericson2314 pushed 2 commits to release-20.03: https://git.io/JvQfl
<hpfr[m]>
nh2: did you search the logs for this? I don’t remember and I don’t even know if there was something I did that did fix it. Remote builders are poorly documented and give poor feedback in NixOS. The way they fail seems nondeterministic. You can look at my dotfiles on GitHub if you think that will help
andi- has quit [Remote host closed the connection]
<hpfr[m]>
nh2: guess I'll watch that issue, ping me if something changes, I guess. but I just checked out your github and I can guarantee you'll figure it out before me... :D
wavirc22 has quit [Read error: Connection reset by peer]
johnny5 has joined #nixos
johnnyfive has quit [Ping timeout: 256 seconds]
wavirc22 has joined #nixos
kfound has quit [Ping timeout: 272 seconds]
kfound has joined #nixos
butterthebuddha has joined #nixos
ddellacosta has joined #nixos
iqubic has joined #nixos
waleee-cl has quit [Quit: Connection closed for inactivity]
<aleph->
Stupid question, does the docker containers module screw with networking at all? I see there's this docker-proxy process doing the routing from what I can see
<aleph->
Seems that despite setting ports as `0.0.0.0:53:53` or `LAN_IP:53:53` I can't actually telnet into that port from outside localhost. Already verified that my firewall is open for that port
iqubic` has joined #nixos
iqubic` has quit [Remote host closed the connection]
<pjt_014>
I can't imagine it would--that'd kind of go against the point of containers. Also If you've set it up in your config file you _shouldn't_ need to open any ports manually.
<pjt_014>
Another question--any reason you're using telnet specifically?
<jlv>
Here is the lightdm.conf https://pastebin.com/jJSuTNWg Give me a minute on the ls output. Things are a little slow working on a plain terminal with tiny text :)
<jlv>
Both `sessions-directory`s are empty.
zaeph1 is now known as zaeph
tilpner has quit [Quit: tilpner]
<{^_^}>
[nixpkgs] @Mic92 merged pull request #83475 → nodePackages: use node2nix from git → https://git.io/JvQTj
<jlv>
worldofpeace: i3 is enabled. Although it is enabled with home-manager, if that could be an issue. What's strange to me is that the same configuration has been working on another machine for a while.
<jlv>
worldofpeace: oh, you said desktop manager. I haven't specifically enabled a desktop manager in my configuration.
<worldofpeace>
jlv: I think you need to use xsession.enable = true; from home-manager
<jlv>
worldofpeace: that option is set. (P.S. thought you were refering to login manager is my last message. It's late here).
<{^_^}>
[nixpkgs] @Mic92 merged pull request #83472 → Propragate darwin go → https://git.io/JvQkI
<worldofpeace>
jlv: I believe there is a bug in home-manager that it relies on xterm the desktop manager being enabled to launch the generated .xsession
<worldofpeace>
could you enable xterm and select that as the session?
domogled has quit [Quit: domogled]
<worldofpeace>
the xsession wrapper (I believe) should then try to start the .xsession from home-manager
<worldofpeace>
I'm not sure if there is a bug in home-manager, or if it was fixed. but it would be worth an investigation
<jlv>
worldofpeace: That fixed it! For some reason, my other machine that uses the same configuration does have an xterm session, even though the option isn't enabled.
<worldofpeace>
jlv: they must have different stateVersion's
<worldofpeace>
This is actually in the release notes, but I must admit it's obscure and mostly a home-manager user issue
<jlv>
worldofpeace: Yup. That machien started on 19.03. I didn't realize lightdm sessions would be affected by that. I thoguht it was mostly sql database stuff.
<worldofpeace>
jlv: xterm used to be a default desktop manager... for some unknown reason. and somehow things relied on it, so via the release note and the stateVersion (to not break existing users setup's) we'd allow some sort of migration
<worldofpeace>
but lightdm-gtk-greeter honestly sucks and doesn't have the code to cover the base case of "no sessions exist"
<worldofpeace>
it's also partially a shame that I know all of this 🤣
<jlv>
worldofpeace: I'm glad you do XD My girlfriend will also be happy, because she recently ran into the same issue.
<worldofpeace>
jlv: if you can open a bug on home-manager it would be very helpful. this issue/bug should have been gone 6 months ago
<jlv>
worldofpeace: Before I do that. I realized that I haven't updated home-manager is a while... (I use a pinned version for the module).
<worldofpeace>
jlv: You mean the nixos module in home-manager right? I believe that has been merged into their master for a while now. It is also possible, if you are using a release version of home-manager, that it could be fixed on master
<jlv>
worldofpeace: Oh, that's nice. Is there anything special I have to do to get the home-manager options? I'm currently importing the /nixos directory from a pinned git repository version.
<worldofpeace>
jlv: I believe all you have to do is import the nixos directory of home-manager, yes.
<jlv>
worldofpeace: hmm. I'm not sure how to do that without grabbing the repository first.
<jlv>
Oh, you said it's merge into **their** master. I thought you meant it's been merged into NixOS itself.
<jlv>
I prefer not using channels like that, because it makes reproducing a configuration harder. I grab a specific revision of their respository with `fetchGit`, in the system configuration, but I have to remember to manually update the revision
<jlv>
I'm on 20.03 right now. I don't see a 20.03 branch yet.
<worldofpeace>
jlv: I guess that would actually be the master branch
<worldofpeace>
20.03 isn't released yet so home-manager can't cut a release-20.03 branch
<jlv>
Well, looks like the latest master still has the "failed to start session" bug. I'll report it.
<Athas>
Is there a way to define and use an override just by modifying my configuration.nix? Creating files seems overkill. I know how to define an overlay, but how do I actually apply it to my system?
<{^_^}>
[nixpkgs] @matthiasbeyer opened pull request #83579 → wolf-spectrum: init at 1.0.0 → https://git.io/JvQLa
<Athas>
tilpner: the version of LLVM used to compile Mesa.
pjt_014 has quit [Ping timeout: 240 seconds]
ZaraChimera has quit [Ping timeout: 258 seconds]
<tilpner>
Oh, that would do it
<DigitalKiwi>
tilpner: i'm going to not check and assume that it *is* april but it's april 1st and you saying it is not april is a joke. well played
<tilpner>
Oh no, how could you have seen through my deceit!
ottidmes has joined #nixos
linarcx has quit [Quit: WeeChat 2.7.1]
<colemickens>
,tofu
<{^_^}>
To get a sha256 hash of a new source, you can use the Trust On First Use model: use probably-wrong hash (for example: 0000000000000000000000000000000000000000000000000000) then replace it with the correct hash Nix expected. See: tofu-vim
<Avaq>
I've hit a problem, which I initially thought was https://github.com/NixOS/nixpkgs/issues/73090, but now I realize it's probably just me not understanding things. Over the years I've gotten quite familliar with Nix and its ways, but one of several topics that have eluded me is package wrappers. I don't understand what a package wrapper is: What is it good for? How does it affect a package?
<Avaq>
But since there's several layers here, I'm not sure which layer I'm supposed to be using, and I might be completely off track. So maybe what I've come here to ask is whether I'm going the right way, and also what's "package wrapping"?
<Avaq>
The reason I've never had to use derivation overrides directly, I think, is because whenever I have wanted to customize something, the options were available under nixos.programs: https://nixos.org/nixos/options.html#programs.
aveltras has joined #nixos
<tilpner>
Avaq: I'll explain in a second, but are you aware of passmenu, which types in the passwords for you, in any application?
<tilpner>
I find that to be slightly less scary than giving FF access to my passwords
turion has quit [Ping timeout: 272 seconds]
<Avaq>
I use my own dmenu script with pass
<tilpner>
And you still want passff for the auto-matching?
<Avaq>
I just wanted to give this a try. I probably won't keep it.
<tilpner>
Alright
felixfoertsch23 is now known as felixfoertsch
<Avaq>
But then when I can't install it because I don't understand things, that itself becomes the problem. :P
<tilpner>
Wrappers packages usually consist of scripts, which set environment variables, and can also start the program with extra arguments
knupfer has joined #nixos
<tilpner>
This could be used to e.g. tell Firefox where to look for things, so that it doesn't search global locations
<tilpner>
It does not affect the original (-unwrapped) package. Only the wrapper will be in PATH
<tilpner>
Does that answer your questions regarding wrappers?
bukkitgerman860 has joined #nixos
chloekek has joined #nixos
<Avaq>
So if I would install firefox-unwrapped directly, it would not be available in PATH, but I could use the binaries in another derivation to do whatever.
<Avaq>
And the wrapped version was wrapped specifically for end users?
est has quit [Ping timeout: 240 seconds]
est31 has joined #nixos
<tilpner>
If you would install firefox-unwrapped directly, it would be added to your PATH
<Avaq>
Right, I just realized that reading over my sentence a few times. :P
<Avaq>
Okay. It's clear.
<Avaq>
It's not really intended for direct use, then?
<tilpner>
But But the extra environment variables would not be set, here GIO_EXTRA_MODULES, GDK__PIXBUF_MODULE_FILE, and XDG_DATA_DIRS
turion has joined #nixos
<adisbladis>
See how the browserpass module does things
<tilpner>
No, wait, I was wrong there
<tilpner>
These environment variables are set in a wrapper even in firefox-unwrapped
<tilpner>
The ones set by the wrapper in firefox (wrapped version) are different ones, including SNAP_NAME (for some reason) and MOZ_ALLOW_DOWNGRADE
fendor has joined #nixos
o1lo01ol1o has quit [Remote host closed the connection]
o1lo01ol1o has joined #nixos
oever has quit [Read error: Connection reset by peer]
<tilpner>
You can use firefox-unwrapped directly if you want, but it does not have a .desktop file, and might not have plugins enabled
<Avaq>
Okay, so a wrapped is not much more than a regular derivation, with the difference that one of its dependencies would've added an executable with the same name.
<Avaq>
wrapper*
oever has joined #nixos
<tilpner>
That's just convention, it's not necessary that the same name is chosen
<Avaq>
And by "would've added" I mean, in case the dependency were installed directly.
m0rphism has joined #nixos
<tilpner>
Now for overrides: There's a mechanism that allows you to undo certain function calls, and then redo them with altered arguments
<tilpner>
This mechanism is provided by makeExtensible, and is used via the added override attribute
Makaveli7 has joined #nixos
<tilpner>
Wrong, it's makeOverridable
<Avaq>
Ah, so perhaps that doesn't apply here. A package would have to explicitly make itself extensible/overridable before the mechanism can be used.
o1lo01ol1o has quit [Ping timeout: 264 seconds]
apiccion has quit [Ping timeout: 264 seconds]
<tilpner>
callPackages uses makeOverridable, and since many packages are registered in all-packages.nix using callPackage, many packages end up being overridable even without an explicit makeOverrideable
<Avaq>
Ah
apiccion has joined #nixos
<tilpner>
> let add = lib.makeOverridable({ a, b }: { result = a + b; }); g = add { a = 20; b = 21; }; in (g.override { b = 22; }).result
<{^_^}>
42
<tilpner>
Avaq: Here's an example of how override allows us to redo an addition with an altered argument
<tilpner>
Note that it can only be used on sets, so directly returning an integer would not allow the use makeOverridable
<Avaq>
I see. So that would really allow me to alter any named function arguments / set attributes. So if something is overridable, I can use that to modify it to my hearts content.
<Avaq>
Correct me if I'm wrong, but that sounds like a last resort, in case the package is not configurable in any other way.
<tilpner>
Avaq: There's a catch. Many expression don't expose all required values like this, so other override mechanisms like overrideAttrs may be necessary. Sometimes, it may even be required to edit the source
<tilpner>
No, IMO .override is the prettiest mechanism we have for changing cooperative packages
<tilpner>
Modules can of course hide this from the user, but those are NixOS-specific-ish
<Avaq>
Yeah, okay. So variables local to the scope of the derivation cannot be overridden, is what you're saying?
<tilpner>
Yes, if the author uses a let ...; in ..., you can't override those usually
<tilpner>
(You sometimes can, but not really)
Andreaaa has joined #nixos
mexisme has quit [Ping timeout: 240 seconds]
<tilpner>
So firefox.override { extraNativeMessagingHosts = ...; } creates a new wrapper with that value. The rest of the arguments are not modified
<tilpner>
You can then either put that directly into environment.systemPackages or uses.users.avaq.packages, or you can add it to an overlay to assign it a name
<Avaq>
To me, this looks like I can either set this by doing: `firefox.override ({ cfg = {...} })`, or I can do something that makes the root `config.${browserName}` available.
<Avaq>
And I thought the `firefox.override` is the second best option, because the library author made an effort to default this value to `config.${browserName}`.
<tilpner>
Ahh, I see
<tilpner>
You could say that, if all you want to change are cfg attributes
<tilpner>
But not for e.g. extraPlugins
koleesch has joined #nixos
koleesch has left #nixos [#nixos]
<Avaq>
Right. Clear. So how does one set `config.${browserName}`? From earlier experience, the root config refers directly back to the object exported from `configuration.nix`?
<tilpner>
There are multiple ways, as usual
<tilpner>
1. You can set nixpkgs.config.firefox = ...; in your configuration.nix
<tilpner>
2. You can use or set a pkgs to which you passed { config.firefox = ...; } during import
hmpffff has joined #nixos
<tilpner>
3. You can add something to ~/.config/nixpkgs/config.nix (haven't used this in years, so path and attribute may be wrong)
<{^_^}>
[nixpkgs] @colemickens opened pull request #83582 → pythonPackages.roku: init at v4.1 → https://git.io/JvQqC
<tilpner>
4. You can re-import the wrapper (the outer function at line 1), and pass a different config to it
<Avaq>
Option 3 seems undesirable to be. I prefer to split my files my way. But I think it's clear: The config is passed by whatever mechanism is calling the function. So if I import the function myself and call it, then I'm looking at option 4. If I use nixos, I can set it on nixpkgs.config, and if I import my own nixpkgs I can pass it as an argument.
<bennofs>
Can nix why-depends be used to find dependency paths among derivations, without having to realise them first?
<bennofs>
I tried: nix why-depends /nix/store/cqsqijfs0pqm2zzni24w4gkypp8ykbd5-nixos-system-pi-cube-20.03.git.99a3d7a86fc.drv /nix/store/34x3dw3b8z2chiyr8bq9l4mm78qd5qfr-qca-qt5-2.2.1.drv but that wanted to download & build a bunch of things
<tilpner>
bennofs: No, you can't scan the outputs without having the outputs
<bennofs>
ah, but I'd like to why-depends on derivation-level only
<tilpner>
bennofs: If you want to look through the build time closure, you can call nix-store --query --tree on the derivation
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<bennofs>
ok :) that's what I did, but searching in the tree is a little bit less convenient than why-depends, which displays the single path I want to see :)
<tilpner>
I think it's good enough, it opens in PAGER, and then you can search for the dependency
<bennofs>
yeah I did do that, but I had to scroll up for quite some time until I found the "parent" of that
<tilpner>
Probably probably wouldn't complain about a PR that adds build-closure scanning to why-depends, you can probably reuse a lot of code
<tilpner>
*People probably
* tilpner
can't words
<bennofs>
ok, I may look into that. Just wanted to know if maybe I missed something
<tilpner>
I don't know a better way, which doesn't mean there isn't any
<{^_^}>
[nixpkgs] @mkg20001 opened pull request #83585 → Rework boot: Persistence, throw out isolinux, easier adding boot menu items → https://git.io/JvQm3
felixfoertsch has quit [Ping timeout: 240 seconds]
Neo-- has joined #nixos
mounty has quit [Ping timeout: 256 seconds]
opthomasprime has joined #nixos
<aveltras>
what's the obvious way to do continuous deployment in a nix ecosystem ? is https://github.com/james-preston/hail the popular way ? not sure it's still maintained
<ottidmes>
asbachb: ah, but you are overlaying your jdk in your stable nixpkgs with an unstable version and then using an unstable jetbeans (so your overlay has no effect on it), so you will have to overlay you unstable branch or probably just use unstable.netbeans.override { jdk = unstable.adoptopenjdk-openj9-bin-11; }
<asbachb>
ottidmes: I guess I want to overlay the unstable branch.
<equivrel>
Hi, I'm playing around with trying to nix package riot-desktop with the new matrix-seshat rust plugin which will allow local history search (and e2e-encrypted history search)
<betaboon>
a while ago some of my function keys stopped working and i get errors in the log along the lines of "atkbd serio0: Unknown key pressed ..." any suggestions ?
<equivrel>
upstream they are building this rust library with yarn, but do I understand correctly that for nix I will need to make a derivation for the rust library first?
zupo_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
zupo has joined #nixos
dansho has quit [Quit: Leaving]
chagra has joined #nixos
<pie_[bnc]>
clever: i asked about this before and forgot (and i still dont have a way to search my quassel logs), wasnt there somethign for fetching PRs or something like that?
<ottidmes>
asbachb: almost, thats just an attribute set, you want it to be a true overlay, i.e. a function in the form of: (self: super: { jdk = self.adoptopenjdk-openj9-bin-11; })
<MichaelRaskin>
Not sure if there is an officially packaged script around that, but wrapping this should not be hard. (upstream is how I call upstream Nixpkgs repo, no, origin is not a good name)
<pie_[bnc]>
MichaelRaskin: i vaguely remember there being a function but maybe im misremembering
<pie_[bnc]>
MichaelRaskin: i agree that origin is a bad name :P
<MichaelRaskin>
Might be, no idea
<tokudan>
,locate adb
<{^_^}>
Found in packages: axis2, oh-my-zsh, metasploit, bash-completion, python27Packages.pwntools, python27Packages.adb-homeassistant, python37Packages.adb-homeassistant
<{^_^}>
[nixpkgs] @teto pushed commit from @pacien to master « gscan2pdf: 2.6.3 -> 2.6.5 »: https://git.io/JvQOI
felixfoertsch has quit [Client Quit]
xkapastel has joined #nixos
felixfoertsch has joined #nixos
<scream>
Hello! I am having issues updating a system with matrix-synapse installed. It tries to build python3.7-pysaml2-4.8.0 as a dependency which fails. Is there a way to disable tests during build? Tests fail because of a deprecation warning.
<scream>
or somehow rollback to an earlier version of pysaml2?
<ottidmes>
asbachb: how do you build that updated netbeans? changed it in a local nixpkgs checkout? copied over the file and used callPackage? nixos-rebuild? nix-build? nix-shell?
<asbachb>
ottidmes: `nix-build -I nixpkgs=/home/asbachb/dev/src/nix/nixpkgs -p netbeans` I updated the files on my checkout nixpkgs.
<equivrel>
scream: do you know how to set doCheck = false?
<asbachb>
Mh. I'd lie if I confirm what the expression is about :D
<asbachb>
*that I know
<ottidmes>
asbachb: -p basically means use those packages as buildInputs, which is exactly what you want for nix-shell, but nix-build wants to build right then and there, and since you did not build anything using netbeans, it errors with failing to build shell.drv (apparently the default name)
ambroisie has joined #nixos
mexisme has quit [Ping timeout: 272 seconds]
<asbachb>
ottidmes: Thanks again :)
<scream>
equivrel: still runs the tests. I pasted your snipped into the configuration.nix file
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<symphorien>
andi-: is it ok if I copy my test crate in nixpkgs and use crate2nix on it ? It only depends on cc so the Cargo.nix file should be brief.
<andi->
symphorien: I don't see the need to have the crate2nix generated code in nixpkgs at all to write that test. I think you can just create those few files with the test hellpers in the buildRustCrateTests attrset
<symphorien>
my test still depends on the cc crate
iyzsong has quit [Ping timeout: 240 seconds]
<andi->
symphorien: But can't you replace that with a `pkgs.runCommand "foo" { … } "cc ${file} -o something.o" ?
<andi->
or is the CC crate part of the problem?
<equivrel>
scream: that's odd, I can't think of why it didn't work... I would open an issue
<andi->
symphorien: I also added the brotli crate generated via carnix a few years ago. So if nothing else works that is fine, I guess.
<symphorien>
also I'm sure you meant well with these utilities but they have zero comment and I'm feeling a lack of motivation when it comes to reverse engineering nix code
<scream>
equivrel: thanks for you help. I will try to muck around with more options to override and then create an issue if it fails
<pie_[bnc]>
unrelated; pudb seems pretty nice
<andi->
symphorien: Yeah, I know :/ I should probably document them better. I can try to have a shot at that and your example in like 6h or so
<symphorien>
the cc crate is not a problem, but is a call to buildRustCrate to build cc better written by hand or with buildRustCrate ?
<symphorien>
*or with crate2nix
<andi->
The problem with crate2nix in this case is the size of files we add to nixpkgs. I tried to avoid that.
<symphorien>
I see
<andi->
symphorien: feel free to make a crate2nix based attempt in the PR
<andi->
I'll review and then see if I can make it smalle
<symphorien>
let me see if I can remove the need for cc in my example
<ottidmes>
tilpner: not long ago someone had the same problem, I tested it for them, but I could not reproduce it, in the end it somehow started working again, so unfortunately no good leads as to why it then worked and not before
<ottidmes>
tilpner: I am seeing the same for the two nix-instantiate calls
alexherbo2 has joined #nixos
<tilpner>
ottidmes: Thank you, that's reassuring and disappointing at the same time
<ottidmes>
tilpner: tried with nixos-19.09, but with the same results
knupfer has quit [Ping timeout: 240 seconds]
<ivegotasthma>
hello, I'm getting an error that's deep within an application about missing symbols. Is there a fix I can do by installing packages or should I update the package? ntlm_auth: symbol lookup error: /nix/store/qjxas7v6m2hlvnmpaiy10jz6swdzwa4l-krb5-1.18/lib/libcom_err.so.3: undefined symbol: k5_os_mutex_destroy, version krb5support_0_MIT
<ottidmes>
tilpner: you got the wrong name, its part of: nvidia-x11-440.59-5.4.28
<{^_^}>
[nixpkgs] @bhipple opened pull request #83594 → aliases.nix: add aliases for removed torch packages → https://git.io/JvQsQ
<peelz_>
ottidmes: thanks, but I'm even more confused now lol
Makaveli7 has quit [Ping timeout: 265 seconds]
jakobrs has joined #nixos
<jakobrs>
Is it possible to overlayfs a store on top of /nix/store?
<peelz_>
ottidmes++
<{^_^}>
ottidmes's karma got increased to 28
<tilpner>
jakobrs: In read-only mode, yes
<tilpner>
I wouldn't try to modify it
<jakobrs>
How?
<ottidmes>
tilpner: seems like name is not defined at the right time when allowUnfreePredicate is called, cause I just debugged, and the unnamed one, actually is your nvidia-settings, but with just a pname attribute
<jakobrs>
I don't intend to modify it It's just that I need to do something where a chroot/similar won't do.
<ottidmes>
peelz_: honestly, it is a confusing topic, I just follow the rule, everything super, except when referencing to packages, than use self (or final)
<ottidmes>
jakobrs: won't the meta database of the nix store not be confused if you were to overlay the /nix/store?
<peelz_>
ottidmes: yeah that's what I got from the thread you linked... just follow the rule :P
<jakobrs>
Probably
<jakobrs>
I'm not going to build anything though; just execute a binary from another store
<jakobrs>
without copying it over
<ottidmes>
peelz_: lol, pretty much that, a case of choose your battles :P
<tilpner>
ottidmes: You're right! I looked through that file, but missed that it was passing attrs instead of derivationArg!
<tilpner>
<3 ottidmes
<{^_^}>
ottidmes's karma got increased to 29
<{^_^}>
[nixpkgs] @bhipple opened pull request #83596 → [20.03] treewide: remove torch and related packages; add throw aliases → https://git.io/JvQGY
Peter_Storm has quit [Quit: Connection closed for inactivity]
knupfer has joined #nixos
xelxebar has quit [Ping timeout: 240 seconds]
<pkolloch[m]>
#NixOS Does anyone have experience with remote building on Mac OS? I'd love to use my Linux machine for development but run tests on Mac OS. I am running into trouble with impure paths: https://gist.github.com/kolloch/361ad07886c4a5f2df4633297d7293a2
zarel has joined #nixos
<pkolloch[m]>
It looks like it tries to resolve the impure paths on my linux machine instead of the remote builder.
<{^_^}>
[nixpkgs] @bhipple pushed 2 commits to release-20.03: https://git.io/JvQGr
lunik1 has quit [Quit: :x]
<ottidmes>
pkolloch[m]: not sure if you can make that work like you want it too, one workaround on top of my head, if its mostly /usr/lib paths, why not use a sshfs to your mac from the Linux machine?
<{^_^}>
[nixpkgs] @bhipple pushed 2 commits to release-20.03: https://git.io/JvQGP
<pkolloch[m]>
<ottidmes "pkolloch: not sure if you can ma"> Thank you for the suggestion. Maybe I'll do that but it'll make everything slower for me, unfortunately. (The Mac is in the same Wifi but even ssh is a bit laggy, don't know why)
<ottidmes>
peelz_: why not just: with builtins; rather than a let inherit (builtins) map;? I personally would not overlay packages with unstable packages if they are meant to go in systemPackages, I just directly use the unstable packages in systemPackages, its more clear that way, but more importantly you won't get accidental rebuilds of other packages if they might depend on them
teto has quit [Ping timeout: 246 seconds]
timmw has joined #nixos
<est31>
danderson: as a laptop buddy, I have a question to you. If you connect your t495 to the power supply, are you getting tiny electric shocks when touching it? Because that's my experience with this device and I don't know whether it's common or not
<peelz_>
ottidmes: oh yeah, I'll change it to "with builtins;". I can't overlay those packages if I refer to them directly in my systemPackages.
<bhipple>
jakobrs: yes I have used overlayfs (read-only) with a Nix store successfully. The NixOS installer CD also uses OverlayFS to compose stores as of https://github.com/NixOS/nixpkgs/pull/35188
<{^_^}>
#35188 (by sorki, 2 years ago, merged): use overlayfs by default for netboot and iso
<est31>
danderson: maybe not right away because I just got it and have current need for it and also I want to study it further so that I can write a full report about when exactly the issue occurs
dermetfan has quit [Ping timeout: 272 seconds]
<est31>
but idk in a month or so I'll send it back
<{^_^}>
[nixpkgs] @mmilata opened pull request #83598 → test-driver.py: don't leave vde1.ctl around → https://git.io/JvQZe
<ottidmes>
peelz_: but did you need to? like I have environment.systemPackages = with pkgs; [ nixos-unstable.vscode ]; I don't need to overlay vscode with nixos-unstable.vscode, I can just use it like that, now with vscode it does not really matter, but packages that might also be dependencies to others, overlaying them might cause unnecessary rebuilds
<Cadey>
how much disk space does the nixos hydra use?
<Cadey>
like the current production one
<peelz_>
ottidmes: right, I'll keep that in mind. I don't think many programs depend on my unstable packages, so it doesn't make much of a difference in this case. Also most of these packages are available in master but not backported to 19.09
<peelz_>
so I guess it makes sense for packages that aren't available in my current channel
xelxebar has joined #nixos
<aveltras>
is there some kind of integration between hydra and nixops ? for example, triggering a nixops deployment whenever a build finishes on hydra
<ottidmes>
peelz_: just saying why I consider it best practice to do it like I do, I have been in situation like yours, but after updating later on, the situation changed and a few of them caused rebuilds that I then had to figure out why they were being rebuild only to find out eventually it was caused by such overlays
<peelz_>
ottidmes: hmm... I just don't like the idea of those packages not being overlayable
<{^_^}>
#83540 (by stigok, 15 hours ago, open): unbound: don't implicitly enable local resolver in resolvconf
<peelz_>
ottidmes: is there a way to have overlays for packages from a different channel?
<ottidmes>
peelz_: they still would be, but on your unstable channel
<peelz_>
ottidmes: oh lmao
<peelz_>
so pkgs-unstable.overlays ?
<{^_^}>
[nixpkgs] @orivej-nixos pushed commit from @orivej to master « midicsv: init at 1.1 »: https://git.io/JvQZU
teto has joined #nixos
<ottidmes>
peelz_: so yeah :P when you import your unstable branch you pass it a set of arguments, one of which is overlays, which default to the impure value for it (<nipxkgs-overlays> or one of them directories), but you can also explicitly pass a list of overlays
<pie_[bnc]>
anyone have any idea how I can inspect the value of an argument to something called via callPackage?
ixxie has quit [Ping timeout: 256 seconds]
<peelz_>
ottidmes: I'm using home-manager, so I'm not sure how that would work...
<pie_[bnc]>
i.e. in `callPackage ({a,b,c}: mkDerivation stuff) {}` how do I inspect a b c?
<ottidmes>
pie_[bnc]: call override on it and trace it
butterthebuddha has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<pie_[bnc]>
oh hm i guess i can figure out what i want by looking at the drv
<peelz_>
ottidmes: what is the parameter set passed to configuration.nix? what does "config" contain?
<ottidmes>
pie_[bnc]: ah you want the final set, yeah, then checking how its used is the only way I can think of
butterthebuddha has joined #nixos
<ottidmes>
peelz_: configuration.nix is just unlike any other NixOS module
<ivegotasthma>
nix search causes my laptop to hang, is this a known issue?
<ivegotasthma>
it only occurs on my t440p with an SSD, I'm following unstable
<peelz_>
ottidmes: oh gotcha
<peelz_>
ottidmes: so should I pass `config.nixpkgs.config` to pkgs-unstable?
<ottidmes>
peelz_: since they are general options you likely want to do that, I mean, if you want to allow unfree in your stable branch, you likely also want it for your unstable, so I indeed copy it over
<ottidmes>
ivegotasthma: heard of it before, think it is really memory hungry or running into some other limitations, I could not find it myself, but within the last few days someone showed a way to update it directly based on some prebuilt index, maybe look for that in the logs?
<peelz_>
ottidmes: okay thanks. Sometimes it's a bit hard to follow what all of these references resolve to lol
<ratsclub>
I'm having a little trouble with lxc here. I have virtualisation.lxd.enable = true; on my configuration but I can't really start any container
<ottidmes>
jluttine: with builtins.exec might work, although the sandbox might still prevent you from getting the info, not sure
<ottidmes>
infinisil: neither would I, but is there an alternative, other than preparing the output of the call beforehand?
<infinisil>
jluttine: ottidmes: There is `lib.commitIdFromGitRepo` at least
<jluttine>
infinisil: yes, that's why i thought there might be support for even `git describe --tags` type of output
<ottidmes>
jluttine: its just using what available on the file system, that is why it can do it for the commit id, so you should be able to do the same
<infinisil>
Alternatively you could use IFD to evaluate a git command within a derivation
<infinisil>
And import the result into nix
<jluttine>
infinisil: what is IFD?
<infinisil>
,IFD
<{^_^}>
import-from-derivation (IFD) is when you evaluate nix from a derivation result, for example `import (pkgs.writeText "n" "1 + 1")` will evaluate to 2. This is sometimes problematic because it requires evaluating some, building some, and then evaluating the build result. It has been described as "such a nice footgun."
<ottidmes>
but that would require the git repo to be in store, right? you can't access arbitrary paths in abuild
<{^_^}>
[nixpkgs] @nh2 pushed 2 commits to haskell-updates: https://git.io/JvQZh
<infinisil>
ottidmes: Yeah
romildo has quit [Ping timeout: 256 seconds]
<infinisil>
jluttine: What do you need version for btw? Because if you don't need it during eval time, it won't be as problematic
<ottidmes>
I have a similar issue with wanting to have the date of the last git tag, which is easy when getting a tarball from github, cause the file creation dates matches that, but when using a local path, no way to get any of such info that I am aware
<infinisil>
In my system config I also have something like this, but I just use a small wrapper script that calls the `git` commands, whose results then get passed to Nix
<Avaq>
Is it possible to set the value of an attribute for all properties in an attribute set? To illustrate, like: `network.wg-quick.interfaces.*.dns = [ nodes.dns.deployment.targetHost ]`? So I don't know the names of the attributes on `interfaces`, if any, but for each that will exist, I want to shove my property into there. :P
cole-h has joined #nixos
captn3m0 has quit [Read error: Connection reset by peer]
captn3m0 has joined #nixos
<atemu12[m]>
Avaq: builtins.attrNames
remirol is now known as lorimer
zaeph1 has joined #nixos
reallymemorable has quit [Quit: reallymemorable]
<ivegotasthma>
tilpner: CPU starts spinning and never finishes
LysergicDreams has quit [Ping timeout: 264 seconds]
spacekookie has quit [Quit: **agressive swooshing**]
spacekookie has joined #nixos
<infinisil>
Avaq: It's a bit complicated, because if you try to assing `interfaces = <something to enumerate all interfaces>` you'll get infinite recursion, because interfaces now depends on itself
<infinisil>
Avaq: What you need to do instead is to modify the interfaces option itself to include a submodule that sets dns as a default value
<infinisil>
Avaq: This looks about like this: `options.networking.wg-quick.interfaces = lib.mkOption { type = attrsOf (submodule { dns = [ ... ]; }); }`
<Avaq>
Right. I figured I was probably doing something weird. Thanks. :)
<{^_^}>
[nixpkgs] @primeos pushed 2 commits to release-20.03: https://git.io/JvQcM
<balsoft>
First, I somehow need to compile an ocaml compiler that will be compiling packages for my target, since AFAIU ocaml doesn't support cross-compilation?
<aleph->
Man, getting a pihole docker container set up almost makes me wanna switch this host back to ubuntu... almost >_> I can't tell if it's the docker-proxy process interfering with stuff or not.
<Avaq>
aleph-: I just completed setting up my Raspberry PI with pihole via NixOps last week.
<ivegotasthma>
how do I get the dev output from a package?
<aleph->
Huh, how'd that go? I'm trying to run it via the docker-containers module as a systemd service. For whatever reason and with my firewall currently turned off. I can't get hosts to connect to it
<aleph->
Unless I specify `--net=host`
<tilpner>
ivegotasthma: lib.getDev or .dev
<tilpner>
lib.getDev more for when you're not sure there is a .dev output
<ivegotasthma>
tilpner: my problem is that the top-level package (lutris) has an error when starting a game. It fails because one of the libraries (krb5) cannot find symbols. I'd like to patch lutris to use the krb5 dev output.
<tilpner>
Uh oh
<srk>
Avaq: I'm interested in something like that as well, is the project public?
<Avaq>
aleph-: That's how I'm running it. I'll create a gist with the relevant config. Hold on.
<srk>
oh wait docker on rpi? :)
<Avaq>
Yes
<aleph->
Avaq: Danke
linarcx has quit [Quit: WeeChat 2.7.1]
oever has quit [Read error: Connection reset by peer]
linarcx has joined #nixos
<srk>
I've considered packaging pihole but didn't start working on that (yet)
oever has joined #nixos
oever has quit [Remote host closed the connection]
<aleph->
Yeah I wish someone would.
oever has joined #nixos
<aleph->
I'm tempted but it seems like a pain to write a service for that
oever is now known as vandenoever
<srk>
why why
<srk>
aleph-: we can co-op on that if you have time during next week
<Avaq>
I'm still working on that gist. Jut adding some comments to the parts that might not make sense.
LysergicDreams has quit [Ping timeout: 265 seconds]
xkapastel has joined #nixos
<aleph->
Ya ya, tbh I might just build my own container and just set the lighttpd port to 8080 by default so I can run it with `--net=host`
growpotkin has joined #nixos
jumper149 has joined #nixos
<aleph->
srk: Not a lot of time lately, but if you have a branch I don't mind helping
lsix has quit [Ping timeout: 240 seconds]
<srk>
aleph-: yeah, I'll take a look at the project again and let you know if I work something out
<aleph->
:thumbsup:
<tilpner>
FWIW, kresd+hblock makes a decent replacement for pihole, if you don't need the web UI
* tilpner
had pihole running in docker on nixos on a pi too, but couldn't get it to consult stubby outside docker for upstream DNS
<Avaq>
tilpner: I was working on that, by combining this setup with cloudflared. But I haven't finished it, and decided to come back to it later.
<aleph->
Ugh DNS shouldn't be over http >_>
<aleph->
tilpner: Yeah I kinda like the web UI otherwise this would be way easier
<aleph->
A native package would be great though
<tilpner>
Avaq: That direction is good too, but I meant that outgoing queries from your pihole should be encrypetd
<Avaq>
Yeah, I know.
<tilpner>
aleph-: Any encrypted method is fine, and DoT isn't HTTP
mexisme has joined #nixos
<Avaq>
That was the idea, to trick the pihole into sending dns queries via cloudflared.
<Avaq>
But I haven't really done much besides reading up on cloudflared and defining a systemd service for it.
<tilpner>
Avaq: You can use services.stubby.* for the same effect, no packaging effort
<aleph->
tilpner: True, DOT isn't http at least
mexisme has quit [Ping timeout: 265 seconds]
<aleph->
srk: Yeah you're config still isn't working. If I do `dig @192.168.1.175 github.com` can't resolve still. Only gotten it to resolve if I set `--net=host` >_>
<aleph->
So guess I'll just make my own container to run with --net=host then
LysergicDreams has quit [Ping timeout: 265 seconds]
<tilpner>
aleph-: How did you configure the firewall of 192.168.1.175?
<atemu12[m]>
NemesisD: As I said, you need to escape it, so `nix-env -e \*`
<ivegotasthma>
abbradar[m]: hey, are you the same abbradar that did work on lutrix?
<ivegotasthma>
I'm getting an error while running a game, coming from krb5 about a missing symbol k5_os_mutex_destroy I can send you a log
delli3 has joined #nixos
zupo has joined #nixos
<NemesisD>
oof i should have researched this more: `Package ‘home-manager-2019-06-25’ in /nix/store/vifxvc1qnb4g672zp3nf9bm6jr2p631x-nixpkgs-19.09pre187222.002b853782e/nixpkgs/pkgs/tools/package-management/home-manager/default.nix:34 is not supported on ‘x86_64-darwin’, refusing to evaluate.`
<pie_[bnc]>
tbenst[m]: I don't rememeber why I "needed" the PYTHONPATH to work
<aidenholmes>
i don't understand, this is an xml fil
<aidenholmes>
file*
vandenoever has joined #nixos
vandenoever has joined #nixos
vandenoever has quit [Changing host]
hmpffff has joined #nixos
<pie_[bnc]>
do suggested changes on github still get notifications if the issue is closed?
<pie_[bnc]>
s/issue/pr
<tbenst[m]>
pie_[bnc]: thanks for making major progress on DeepLabCut! I’m not sure if your comment re: wxPython build.py is still current. I should have some time later today to dive into things
<pie_[bnc]>
tbenst[m]: not sure id call that major, but seems like something I could do for once ;P
o1lo01ol1o has joined #nixos
<pie_[bnc]>
tbenst[m]: also pudb is pretty nice
<pie_[bnc]>
tbenst[m]: a trick i learned recently that really decreased my threshold for wanting to mess with crap is that instead of running individual phases you can just run genericBuild in a nix-shell
<pie_[bnc]>
so i just did that and then patched the file to add a pudb call
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
ambroisie has joined #nixos
Darkmatter66 has quit [Ping timeout: 265 seconds]
o1lo01ol1o has quit [Remote host closed the connection]
o1lo01ol1o has joined #nixos
<pie_[bnc]>
tbenst[m]: im open to collabing, ping me
<pie_[bnc]>
tbenst[m]: how much later?
<aidenholmes>
i fixed the problem
<pie_[bnc]>
though i have like no ram so idk about running builds :D
<symphorien>
with home-manager, how would I go at enabling the builtin service systemd-tmpfiles-setup.service ?
cr4y1 has joined #nixos
CRTified has quit [Ping timeout: 240 seconds]
acarrico has quit [Quit: Leaving.]
<tbenst[m]>
Piepie_[bnc]: in a meeting until 2pm ET, so in 2.5 hrs if you’re still around?
<tbenst[m]>
Sorry 2pm PT
<duairc_>
What's the best way to make a derivation that's basically fetchFromGitHub, but additionally applies some given patches to the result?
duairc_ is now known as duairc
o1lo01ol1o has quit [Remote host closed the connection]
LysergicDreams has quit [Ping timeout: 265 seconds]
<{^_^}>
[nixops] @grahamc pushed 0 commits to revert-1266-parallel-send-keys: https://git.io/JvQBF
LysergicDreams has joined #nixos
johnnywho has joined #nixos
gustavderdrache has joined #nixos
ZaraChimera has joined #nixos
johnnywho has quit [Client Quit]
johnnywho_ has joined #nixos
johnnywho_ has left #nixos [#nixos]
miscellaneousbus has joined #nixos
<{^_^}>
[nixpkgs] @petabyteboy pushed commit from @vincentbernat to master « nixos/nginx: use mailcap mimetypes in all cases (#83611) »: https://git.io/JvQRI
<{^_^}>
[nixpkgs] @petabyteboy merged pull request #83611 → nixos/nginx: use mailcap mimetypes in all cases → https://git.io/JvQ8e
<miscellaneousbus>
hey, trying to do a nixos-install with an LVM on LUKS setup. Following this guide https://gist.github.com/martijnvermaat/76f2e24d0239470dd71050358b4d5134. When i run nixos-install i get syntax error, unexpected $end, at /dev/nvme0n1p1 (my LUKS encrypted partition). Using nixos 20.03 beta (since it has my wifi drivers)
<miscellaneousbus>
gchristensen if i remove the boot.initrd.luks.devices, it will install fine no error. It seems like its trying to include the device in the configuration? Im not sure though as im new to nixos
<colemickens>
It's amazing how productive I can be when avoiding a task I don't want to do.
<{^_^}>
[nixpkgs] @petabyteboy pushed commit from @vincentbernat to master « nginx: increase types_hash_max_size to 4096 (#83609) »: https://git.io/JvQRa
Avaq has quit [Quit: Leaving.]
jbgi_ has joined #nixos
<{^_^}>
[nixpkgs] @vcunat pushed 48 commits to staging-19.09: https://git.io/JvQRD
<miscellaneousbus>
gchristensen ok i see the problem. When i initially had boot.initrd.luks.devices = [ {...} ];, nixos gave me an error about the syntax. i realize now it was a future deprication warning and not in use at the moment
<miscellaneousbus>
nixos installs when i reverted
<gchristensen>
ahh nice
turion has quit [Ping timeout: 272 seconds]
<peelz_>
I have a dead symlink... is that normal? /nix/var/nix/gcroots/profile -> /mnt/nix/var/nix/profiles
<{^_^}>
[nixpkgs] @andir opened pull request #83616 → buildRustCrateTests: Fix link order test on darwin → https://git.io/JvQ0v
<peelz_>
s/profile$/profiles
<pie_[bnc]>
tbenst[m]: any idea if the python build and test phase can be separated into two derivations so we dont have to recompile all of wxpython if we're only working on tests?
<tbenst[m]>
piepie_[bnc]: perhaps you can enter shell with doCheck=false, then manually run tests via python setup.py test / however the tests run?
<tbenst[m]>
Could copy the test scripts to a writeable directory outside of nix-store
<tbenst[m]>
Only machine I can give access to is the nix-data hydra, which is memory optimized so only 4 cores. Happy to give access if helpful? But honestly it’s pretty slow
<pie_[bnc]>
this might work <tbenst[m]> piepie_[bnc]: perhaps you can enter shell with doCheck=false, then manually run tests via python setup.py test / however the tests run?
lsix1 has joined #nixos
<pie_[bnc]>
tbenst[m]: well im running on 4 cores right now but its cooking my machine so that might not be bad :D
<{^_^}>
[nixpkgs] @Ma27 opened pull request #83617 → nixos/release-notes: mention that mongodb is unfree now → https://git.io/JvQ0l
otulp_ has joined #nixos
alexherbo2 has joined #nixos
otulp has quit [Ping timeout: 246 seconds]
otulp_ is now known as otulp
oida has quit [Remote host closed the connection]
<pie_[bnc]>
pytest-timeout seems kinda crappy
<{^_^}>
[nixpkgs] @FRidh pushed 622 commits to staging-next: https://git.io/JvQ0g
<pie_[bnc]>
like, can i query a list of tests and just run them individually?
nobrak`` has joined #nixos
<pie_[bnc]>
ill do some googling but i gotta run off for a sex
<pie_[bnc]>
err
<pie_[bnc]>
sec
lsix1 has quit [Read error: Connection reset by peer]
<pie_[bnc]>
(the keys are like right next to eachother)
wavirc22 has joined #nixos
lsix1 has joined #nixos
<xfix>
if you can help it, don't test performance as part of your tests
<energizer>
pie_[bnc]: pytest has tons of options for selecting tests by module or name or partial name or @mark
<pie_[bnc]>
energizer: yeah, what im going to try is see if i can list them then run each one as a separate process
johnny101 has quit [Quit: Konversation terminated!]
nobrak` has quit [Ping timeout: 258 seconds]
<mdash>
any of y'all used nixos-generate to create an openstack image? i've created a qcow2 image but upon boot it gets to "Booting the kernel" and stops with no further output
<pie_[bnc]>
xfix: i wouldnt count "test hangs" as a performance bug :P
<xfix>
yeah, i get that you may want to test whether your algorithm is O(1), and not O(N!^3), but I don't think timeouts are a way to go
<tbenst[m]>
pie_[bnc]: re `(opencv3.override { enableFfmpeg = true; })`, I think you needed `overrideAttrs`. I'm trying that now, modifying in callPackages so users can still override IIRC
slack1256 has joined #nixos
<{^_^}>
[nixpkgs] @Ma27 pushed to master « vorta: init at 0.6.24 »: https://git.io/JvQzc
<{^_^}>
[nixpkgs] @worldofpeace pushed 2 commits to master: https://git.io/JvQza
<tbenst[m]>
pie_[bnc]: ah right you are. Hmm adding ffmpeg to opencv didn’t fix for me. I already had an override for that it turns out
slack1256 has quit [Ping timeout: 265 seconds]
<pie_[bnc]>
tbenst[m]: i think there might be some python screwery involved or something, notice how using the overlay worked and the override inside the nix expression buildinputs didnt
mexisme has quit [Ping timeout: 256 seconds]
<pie_[bnc]>
but you might want to verify that
<tbenst[m]>
Overlay*. I wonder if enableCuda for opencv is screwing things up
<tbenst[m]>
pie_[bnc]: you saw testscript work successfully with your overlay? I haven't tried your nix-build command yet,
<tbenst[m]>
but I have the same overlay in my ~/.config/nixpkgs/overlays/ with no luck. still stalls on kmeans
<pie_[bnc]>
tbenst[m]: well it didnt finish because it was stuck on some training stuff
<pie_[bnc]>
but it got past the error
reallymemorable has joined #nixos
<tbenst[m]>
ah I see, I never saw that error as already had opencv overlay with enableCuda and enableFfmpeg. Script is supposed to take only a couple minutes even on CPU, so I bet we stalled at same KMeans step...
shafox has quit [Remote host closed the connection]
<pie_[bnc]>
tbenst[m]: yes it was kmeans
<{^_^}>
[nix] @edolstra pushed to flakes « nix flake info --json: Dump attr-style flakeref »: https://git.io/JvQgm
mounty has quit [Ping timeout: 258 seconds]
mounty has joined #nixos
ixxie has quit [Ping timeout: 256 seconds]
work_ has quit [Quit: Connection closed for inactivity]
Rusty1 has quit [Remote host closed the connection]
<pie_[bnc]>
tbenst[m]: you should build these from a separate branch so you dont get accidentally working stuff
<tbenst[m]>
I'm going to add intel-openmp wheel
<pie_[bnc]>
tbenst[m]: get the PR version and do -I nixpkgs=path on the commands you run
<tbenst[m]>
pie_[bnc]: sorry I didn't quite follow that?
<pie_[bnc]>
tbenst[m]: i mean you said you have other local changes
<pie_[bnc]>
re: local overlays
<tbenst[m]>
only in overlays
<pie_[bnc]>
actually im not sure how to disable that now that you mention it
<pie_[bnc]>
tbenst[m]: well, nixpkgs doesnt have those overlays so you want to test without them no?
nikitten has quit [Quit: WeeChat 2.6]
<tbenst[m]>
agreed, my current command does not ignore overlays as it's annoying to recompile all dependencies but this is a perfect example of why we should. I have notes on how to do this somewhere...
<pie_[bnc]>
tbenst[m]: easy way:
nikitten has joined #nixos
<pie_[bnc]>
-E import <nixpkgs> { overlays = [] }
<pie_[bnc]>
well, approrpiately quoted
<pie_[bnc]>
i assume you know ow -E works
<tbenst[m]>
ah, right! thx!
<pie_[bnc]>
alternatively check how impure.nix imports overlays
<pie_[bnc]>
osmeting somethin NIX_PATH IIRC
CptCaptain has joined #nixos
<{^_^}>
[nixpkgs] @worldofpeace opened pull request #83625 → nixos/ibus: add ibus portal if portals are enabled → https://git.io/JvQgD
butterthebuddha has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
hmpffff has quit [Quit: nchrrrr…]
butterthebuddha has joined #nixos
<cole-h>
You can also NIX_PATH=nixpkgs-overlays=/var/empty:$NIX_PATH
<cole-h>
cc pie_[bnc] tbenst[m]
<pie_[bnc]>
right
<energizer>
`home.packages = [ (pkgs.python3.withPackages (ps: with ps; [pendulum]))]` but `python3 -c 'import pendulum` -> ModuleNotFoundError: No module named 'pendulum' There are no packages in site-packages, but there's a sitecustomize that adds things from NIX_PYTHONPATH (which is unset).
<cole-h>
(or -I nixpkgs-overlays=/var/empty)
<tbenst[m]>
thx cole-h
<energizer>
How do I get python3 to see the packages I asked for?
<pie_[bnc]>
energizer: tell me if you figure it out, i have weird problems with python sometimes in a similar vein and i havent figured out when it fails
<pie_[bnc]>
sometimes things work with nix-build but not nix-shell I think?
<energizer>
`$ nix-shell -p 'python3.withPackages(ps: with ps; [ toolz ])'` does work
<{^_^}>
[nixpkgs] @worldofpeace pushed 22 commits to master: https://git.io/JvQ2L
dermetfan has joined #nixos
<pie_[bnc]>
energizer: do you know, if i set -n 1 for xdist, if it will actually spawn a new python for each test?
nikitten has quit [Quit: WeeChat 2.6]
<pie_[bnc]>
seems like stuff gets submitted to a worker pool but i dont know how to enforce clean environments
jumper149 has quit [Quit: WeeChat 2.7.1]
<pie_[bnc]>
the whole point is that i seem to have been told tests might have some buggy crosstalk
<pie_[bnc]>
which is what im trying to get around
* pie_[bnc]
waits for a nix-shell doCheck=false build meanwhile...
<energizer>
pie_[bnc]: i doubt it
<energizer>
pie_[bnc]: if you want to just run one test at a time you can do that by specifying the name of the test
<pie_[bnc]>
energizer: yeah i think ill try that first
<pie_[bnc]>
gotta wait for the build though...
<tbenst[m]>
pie_[bnc]: forgot that nix-data hydra box is a bit weird since it's a nix-ops deploy, don't think it's a good option for local dev, sorry!
<energizer>
pytest mypackage.mymodule::test_foo
<pie_[bnc]>
tbenst[m]: ah well :I
<pie_[bnc]>
energizer: yeah google gave something like that
<tbenst[m]>
maybe ask on nix-community channel?
<pie_[bnc]>
tbenst[m]: i guess thats an idea?
<{^_^}>
[nixpkgs] @worldofpeace merged pull request #83425 → nixos/tests/hibernate: disable for platforms other than x86_64 → https://git.io/Jv9Bu
<{^_^}>
[nixpkgs] @worldofpeace pushed 2 commits to master: https://git.io/JvQ2Z
<pie_[bnc]>
energizer: apparently --collect-only is supposed to list them
UltraPhil has left #nixos ["Leaving"]
<{^_^}>
[nixpkgs] @worldofpeace pushed commit from @xfix to release-20.03 « nixos/tests/hibernate: disable for platforms other than x86_64 »: https://git.io/JvQ2n
<ivegotasthma>
I'm looking for anyone with familiarity with the wine / lutris / samba packages, there's some library that's not configured correctly
<pie_[bnc]>
now what do i do while waiting for the build..
<energizer>
how do i find out which package made `python3` accessible in my environment?
<pie_[bnc]>
ivegotasthma: i probably cant answer but ask
<pie_[bnc]>
energizer: realpath will give you the nix path
<pie_[bnc]>
energizer: or rather run which pyton
civodul has quit [Quit: ERC (IRC client for Emacs 26.3)]
__monty__ has quit [Quit: Saving my tailfeathers!]
<pie_[bnc]>
that will give you a path, if its a nix store path you win
<{^_^}>
[nixpkgs] @worldofpeace pushed commit from @ericdallo to staging-next « gnomeExtensions.drop-down-terminal: v24 -> unstable-2020-03-25 »: https://git.io/JvQat
<pie_[bnc]>
if you want to see what built that, i think you can directly run show-derivation on it
<ivegotasthma>
when I try to start a game through lutris, wine is called. wine somehow makes a call to krb5 which cannot find a symbol for the ntlm_auth library
<ivegotasthma>
ntlm_auth: symbol lookup error: /nix/store/qjxas7v6m2hlvnmpaiy10jz6swdzwa4l-krb5-1.18/lib/libcom_err.so.3: undefined symbol: k5_os_mutex_destroy, version krb5support_0_MIT
<ivegotasthma>
0042:err:winediag:SECUR32_initNTLMSP ntlm_auth was not found or is outdated. Make sure that ntlm_auth >= 3.0.25 is in your path. Usually, you can find it in the winbind package of your distribution.
<energizer>
pie_[bnc]: ok but why is that package installed?
<pie_[bnc]>
nix show-derivation $(which python) | less
<pie_[bnc]>
right
<ivegotasthma>
I don't know how to disect the stack of packages correctly and find the offending option / package
<pie_[bnc]>
im not sure theres a way to get that out of the blue
<pie_[bnc]>
and even if there is, youll only get derivations, not nix expressions
<pie_[bnc]>
but it can help
<pie_[bnc]>
you can look at the dependency path between the derivation and your system configuration or user configuration however
<pie_[bnc]>
hold on
justanotheruser has quit [Ping timeout: 258 seconds]
<pie_[bnc]>
'/nix/store/0w7jpx112b914pf3lzpxc8mxxw6ycyfn-nixos-system-nixos-20.03pre-git' does not depend on '/nix/store/k8lhqzpaaymshchz8ky3z4653h4kln9d-coreutils-8.31'
<pie_[bnc]>
energizer: i thought there was a way to get a nice tree view out of that like --tree but i dont remember
<pie_[bnc]>
tbenst[m]: no idea >.>
alexherbo2 has quit [Ping timeout: 264 seconds]
<tbenst[m]>
think I have to run pudb in a shell
<energizer>
pie_[bnc]: `nix-store -q --tree my.drv` does give a tree, but that doesn't help, it's a huge tree and i don't know which of those things is in my config
<pie_[bnc]>
energizer: the tree is something else,
<pie_[bnc]>
you need --referrers (see man page)
<pie_[bnc]>
thats the "this is depended on by" relation
<{^_^}>
[nixpkgs] @jonringer pushed commit from @colemickens to master « plex-mpv-shim: v1.7.12 -> v1.7.14 »: https://git.io/JvQad
linarcx has quit [Quit: WeeChat 2.7.1]
linarcx has joined #nixos
<pie_[bnc]>
energizer: ok i just dont know how any of this works lol
<pie_[bnc]>
energizer: try this nix why-depends /run/current-system/sw $(which ls)
<pie_[bnc]>
(i was so close! :P)
<energizer>
$ nix why-depends /run/current-system/sw $(which python3) -> '/nix/store/2vgiq3k0pxzga5jqf3ad2v19xbzc5fkh-system-path' does not depend on '/nix/store/xb1mp0f16lv23rwhfdc1665v5r20jfmm-python3-3.7.6'
<energizer>
how did it get on my PATH then
<pie_[bnc]>
well then thats probably going to get removed on garbage collection and you havent refreshed your shell
<energizer>
yeah i have
<pie_[bnc]>
ok its entirely uncertain it will be removed from garbage collection
<pie_[bnc]>
the awk command will give you garbage if a file name has ' -> ' in it but whtever
<pie_[bnc]>
i should follow my own advice and get to know nix-store better ;_
<pie_[bnc]>
;)
knupfer has quit [Remote host closed the connection]
<pie_[bnc]>
might be better to use $1 instead of $2 so you get the more meaningful path but i think that includes the -> a well so gotta figure out how to fix that