<pbb>
hi there, I am using `wrapProgram $out/bin/my_script --prefix PATH : ${pkgs.stdenv.lib.makeBinPath [ pkgs.libfaketime ]}` in installPhase, but when I start a pure nix-shell on the result, and run the script, it can not find the `faketime` binary in the path
<pbb>
ah, nevermind, it tries to find it using `which`, but that's not installed ^^
orivej has quit [Ping timeout: 268 seconds]
<infinisil>
Rubber duck debugging ftw
the-kenny has quit [Ping timeout: 272 seconds]
<infinisil>
,zol = Linux Kernel 5 made a minor API change which people are saying is breaking ZFS. It'll be fine. ZFS isn't going anywhere. https://github.com/zfsonlinux/zfs/issues/8259
<{^_^}>
zol defined
<infinisil>
gchristensen: Fixed now, ^^
<gchristensen>
thanks, infinisil =)
<infinisil>
And with just a bit more development, I can also provide a public access point for all these commands
romildo has quit [Quit: Leaving]
orivej has joined #nixos
<{^_^}>
[nixpkgs] @orivej-nixos pushed commit from @orivej to master « bazel: fix patches after #53988 »: https://git.io/fhWWx
<ottidmes>
Ah, cool, I saw the ZFS thing mentioned a few times, but I did not know it was part of version 5 of the kernel
kvda has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
shpx has joined #nixos
<systemfault>
My knowledge on the subject is outdated but... I thought ZFS wasn't GPL so it could not be on linux..
<gchristensen>
that has never been true
protocol has joined #nixos
<gchristensen>
a single binary product containing ZFS and the Linux Kernel cannot be distributed
<gchristensen>
an end user has the freedom to use them together
<systemfault>
Oh
the-kenny has joined #nixos
<systemfault>
I wonder if there's any benefit for me to run it on my laptop.. hehe, will definitely read more on the subject
kvda has joined #nixos
<gchristensen>
it is a nice filesystem
<ottidmes>
gchristensen: do you perhaps know this, is ZFS memory consumption (outside of dedup) comparable to that of say btrfs?
doyougnu has quit [Ping timeout: 250 seconds]
acarrico has quit [Ping timeout: 268 seconds]
<gchristensen>
ZFS uses as much RAM as you'll let it for caching, I don't know what a minimum amount of RAM would be, but I don't expect it is substantially different than other FSs
<jasongrossman>
ZFS uses LESS RAM than I'd like it to. It makes its own decisions about what to cache, and I'd like it to be more aggressive.
<jasongrossman>
Also, and more importantly for ottidmes, it releases memory from its cache very readily.
<ottidmes>
lol, ok, I've got nothing to worry about then, and RAM not used is RAM wasted I guess
<jasongrossman>
ottidmes: Exactly.
<gchristensen>
you two must not use slack
<ottidmes>
I indeed do not :P
<gchristensen>
it will gladly use and waste ram ;)
<yl[m]>
I'm using nixops to deploy on VirtualBox, it's using 16.09 which is fine because I deploy with `--force-reboot` to get it to work. But how can I replace the channel in `sudo -i nix-channel --list` without executing that manually on the server as well as updating the channel so users can install their stuff without having to add the channels themselves
<ottidmes>
gchristensen: since upgrading my desktop to use 24G ram I have never again had it fill to 100%
<liamdawson_>
If I have my ~/.config/nixpkgs/config.nix, but want it to refer to a package override in ~/.config/nixpkgs/foo.nix, how do I import it in such a fashion that it can access the override?
<liamdawson_>
I tried using imports = [ ] in the config.nix file, and that didn't seem to do anything, and I can't seem to work out the syntax for the import statement
kvda has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
liamdawson_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<yl[m]>
dmj`: a.b = { }; is the same thing as a = { b = {} }; is it not?
<yl[m]>
> a.b = {};
<{^_^}>
error: syntax error, unexpected '=', expecting ')', at (string):215:5
<dmj`>
yl[m]: it is, but you're using strings
eadwu has quit [Ping timeout: 252 seconds]
* yl[m]
trying...
halfbit has joined #nixos
lassulus_ has joined #nixos
halfbit has quit [Ping timeout: 252 seconds]
lassulus has quit [Ping timeout: 244 seconds]
lassulus_ is now known as lassulus
liamdaws_ is now known as liamdawson_
aw has quit [Remote host closed the connection]
spacefrogg has quit [Remote host closed the connection]
aw has joined #nixos
halfbit has joined #nixos
spacefrogg has joined #nixos
layus has quit [Remote host closed the connection]
<infinisil>
> deepEval { a.b = {}; }
<{^_^}>
{ a = { b = { }; }; }
cyounkins has quit [Remote host closed the connection]
layus has joined #nixos
cyounkins has joined #nixos
halfbit has quit [Ping timeout: 252 seconds]
kvda has joined #nixos
cyounkins has quit [Ping timeout: 258 seconds]
trevthedevlive has joined #nixos
<trevthedevlive>
aye, hello all
<pbogdan>
why is the nixpkgs:staging jobset disabled on Hydra?
<trevthedevlive>
i just earlier screwed up, im on nixos 19.03 and thought the kernel updated or something, as it threw an error saying there was no space in my boot partition
<trevthedevlive>
so being the uh.... "genius" i am i deleted everything in boot and rebuilt the system, expecting it to download the kernel
<trevthedevlive>
well.......... it didn't and sadly now my system won't boot. am on nixos 18.03 live, is there anything i can do to fix this mess?
hooo has quit [Quit: Connection closed for inactivity]
<trevthedevlive>
i have over a hundred gigabytes of stuff id have to backup to restart from scratch, so id rather not do that
<mog>
does anyone use services.undervolt ?
<mog>
i wanted to try it but everything ive tried causes my computer to crash
<trevthedevlive>
and --install-bootloader isnt an option
<trevthedevlive>
just --install-grub
<ottidmes>
trevthedevlive: maybe it was renamed?
<ottidmes>
my manpage mentions it still
<trevthedevlive>
checked all flags for rebuild, install kernel isn't anywhere
<trevthedevlive>
theres fallback, fast, cores, keep-going, --keep-failed, --max-jobs, and some other stuff
<ottidmes>
maybe just do another nixos-install? (your normal stuff should be unaffected by it)
<hyper_ch>
Celmor: you probably need an older live cd then as the encryption format on zfs has changed. There was a transition period where old encrypted datasets could be imported as read-only
<trevthedevlive>
would that work though
<trevthedevlive>
that sounds super risky imo
<ottidmes>
done it many times in the past when I was learning NixOS, so yeah it works
<trevthedevlive>
i hope so
<ghostyy>
how do i append to a list in nix? i have a module that sets boot.supportedFilesystems but i just want it to append to whatever is already there, not to redefine it if its already defined
<Celmor>
hyper_ch, I imported it readonly anyway, I'm trying to export it in an unencrypted format
<ottidmes>
maybe backup some of the nix folder if you are super paranoid
<trevthedevlive>
is what im about to do
<ottidmes>
but nixos-install and then nixos-rebuild boot, and reboot, always worked for me and never had any lasting negative side effects for me, but better to be safe yeah
freeman42y has quit [Read error: Connection reset by peer]
freeman42y has joined #nixos
<ottidmes>
ghostyy: [ 1 ] ++ [ 2 ] == [ 1 2 ], in Nix, but you are probably mostly interested in your second question, just adding boot.supportedFilesystems = [ "btrfs" ]; for example, will merge it with what was already there automatically
<ghostyy>
ahh! i see!
* ghostyy
wonder if = is really the right symbol to use for that sort of thing...
* trevthedevlive
backs up config
<ghostyy>
thank you though!
<ottidmes>
ghostyy: but if it did not had an option type that automerged, another trick you could have used was to define an apply: options.boot.supportedFilesystems = mkOption { apply = value: value ++ [ "my" "stuff" ]; }; but that is a bit more advanced, if you are not familiar with adding options to your modules: https://nixos.wiki/wiki/NixOS_Modules
MP2E has quit [Remote host closed the connection]
<ghostyy>
ah, i see!
<ghostyy>
i just made my first one the other day
<ottidmes>
also, if say it was a list defined as a default attribute on mkOption or via mkDefault, you could have just defined your value on the same level, so for mkDefault, just used: something = mkDefault [ "my" "stuff" ]; and then they would be on the same priority level so would merge rather than the one with priority overwritting the other
kvda has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<ghostyy>
ahhh
* trevthedevlive
still backing up, wew lads this is gonna be a while...
<ottidmes>
trevthedevlive: I would not backup the whole of /nix, that I am sure of will remain working, just maybe the /nix/var stuff
<trevthedevlive>
im just backing up config.nix and a work directory
<trevthedevlive>
work dir is 400~MiB
<trevthedevlive>
rest can go haha
<ottidmes>
ow ok
<ghostyy>
do you guys recommend the use of nix-store --optimize?
<ghostyy>
i cant help but think it reminds me of that old windows "optimizer" program that actually just deleted half your system files
<ottidmes>
ghostyy: I have auto-optimise-store = true in my nix.conf
<ghostyy>
whoa i didnt know you could enable it there
<ghostyy>
do you ever have any problems with it?
<ottidmes>
ghostyy: I believe all it does is hardlink stuff to safe space, and that can save a huge amount, I have seen it reporting saving 8G on a 22G store (so it would be 30G normally)
<ghostyy>
thats pretty good!
<ottidmes>
no problems
<ottidmes>
(that I am aware)
<ottidmes>
and have been using it for quite some time
halfbit has joined #nixos
<Yaniel>
yeah `nix-store --optimize` is very different from all those "OS optimizer" apps
<Yaniel>
it doesn't actually change or remove anything, just shuffles things around
<ottidmes>
and since you tend to have a lot of duplicated files if you are like trying to package or update something, the savings can be huge
kvda has joined #nixos
<Yaniel>
so yeah, definitely use it
<Yaniel>
unless you have excessive amounts of free space and no time
<dmj`>
yl[m]: did it work?
<yl[m]>
dmj`: no it did not
<dmj`>
yl[m]: what was the error?
<yl[m]>
dmj`: I can't remember, let me try to put the code back with your suggetion
<yl[m]>
so basically the SGs that has a dependency should be a function that take resources and the groupName should be set to the name in the resource
<yl[m]>
that way nix knows the dependency tree
<trevthedevlive>
ottimdes is download ~10GiB of stuff
<trevthedevlive>
downloading*
aswanson_ has quit [Ping timeout: 252 seconds]
benley has quit [Ping timeout: 252 seconds]
aswanson_ has joined #nixos
<ottidmes>
trevthedevlive: that is probably because the LiveCD is using a different version of nixpkgs
<yl[m]>
dmj`: do you know if nixops requires SSH access for the node deployment? What if I have a bastion, how does that work?
<dmj`>
yl[m]: yes, when master builds and tests run successfully, nixops deploy is called
o1lo01ol1o has quit [Remote host closed the connection]
<yl[m]>
that's pretty cool
<dmj`>
yl[m]: we have a single deployment with multiple resources, including ec2 instances
<dmj`>
yl[m]: we check the deployment state in encrypted using git-crypt
<yl[m]>
dmj`: would you recommend nixops at work?
<dmj`>
yl[m]: yes
<dmj`>
yl[m]: I wouldn't recommend anything else tbh
<dmj`>
yl[m]: unless you like large black box solutions
<yl[m]>
dmj`: the only unsure part is lack of "plan" as with terraform. I can't figure out what services will be restarted before actually running the deployment
<yl[m]>
is there a way to do that?
<dmj`>
yl[m]: nixops will restart one the services that have changed, they are derivations just like anything else
<dmj`>
s/one/only*
<yl[m]>
yea, but how can you tell which ones will be restarted ahead of time?
<yl[m]>
so I can build a plan if a critical service (say redis) will be restarted
<dmj`>
nixops deploy --dry-run
<dmj`>
or nixops deploy --build-only
<yl[m]>
oh nice
<yl[m]>
I'll try that out, thx!
<elvishjerricco>
yl[m], dmj`: --dry-run shows what derivations will be built, not what services will change
<elvishjerricco>
--dry-activate shows the service changes
<dmj`>
actually --dry-activate
<yl[m]>
elvishjerricco: nice, thanks
<elvishjerricco>
--plan-only will show what hardware will change, IIRC
<yl[m]>
elvishjerricco: dmj`: is there a way that instance refer to another's public IP on AWS? I have a systemd service that needs the IP (public) of another host
<yl[m]>
maybe I'll just route53
o1lo01ol1o has quit [Ping timeout: 258 seconds]
trevthedevlive has quit [Ping timeout: 246 seconds]
<infinisil>
yl[m]: I remember there being a nixcon talk about a company that implemented some smart thing to restart services
<infinisil>
Nixcon 2018 that is
_kwstas has joined #nixos
_kwstas has quit [Remote host closed the connection]
<etu>
drager: That is one way for sure, but another way would probably be to put it in your home directory somewhere. I guess that fontconfig looks there is well. But then it's not controlled by nix and that's boring ;)
<mckeanalicej>
oh minecraft doesnt run well on openjre i wonder how much oracle is paying them for that
<teto>
I am trying to use builtins.fetchGit to fetch from a private repo but it stays stuck at "fetching Git repository 'gitolite@my_domain.com:mptcp.git'" ? with wireshark, I don't see the traffic I expect (huge transfer). anyway to improve verbosity ?
<etu>
mckeanalicej: It's Microsoft that owns Minecraft nowadays...
<tilpner>
teto: -vvvvv
<tilpner>
(On the nix-* command)
mckeanalicej has quit [Quit: Page closed]
freeman42z has joined #nixos
<lassulus>
is there a "stable" link to an up2date 18.09 squahsfs?
<lassulus>
or the iso
<teto>
tilpner: I had already -vv. -vvvvv doesn't change a thing but connecting on the server, I see a git upload-pack in the background processes so it might work. Just it is slow because it's a huge (kernel) repo ?!
<tilpner>
teto: Yes, it will be slow. How much longer than a traditional git clone for the same repo did you wait?
<tilpner>
lassulus: I don't think I understand your question :/
<q6AA4FD>
is it a normal practice to define a bunch of programs you want to install in bulk with buildEnv, or is there a better way to do so?
freeman42y has quit [Ping timeout: 252 seconds]
<teto>
it's just annoying not to have the feeback. With git clone, you see that it is packing, here I see nothing. I will wait Hoping that it works
<teto>
q6AA4FD: you can check out home-manager , use nix-env or define them from configuration.nix wit h users.users.XXX.packages = [ ]
<tilpner>
q6AA4FD: That depends on how you install them. If you're just going to plug it into a NixOS option, you can also concatenate lists of packages
<etu>
lassulus: There's iso's there based on latest 18.09
<q6AA4FD>
i'll give home-manager a shot, had consider it before but for other reasons. i don't use nixOS, so i'm always unsure of which features will work with my setup
<lassulus>
etu: yes, basicaally, but the isoname is so random :D
freeman42z has quit [Ping timeout: 252 seconds]
<tilpner>
Oh, that's what you mean by stable
<wedens>
q6AA4FD: there is also symlinkJoin that may work for grouping binaries
<q6AA4FD>
wedens: can you explain how that works? i'm seeing examples but they aren't really helping
goibhniu has joined #nixos
worldofpeace has quit [Ping timeout: 246 seconds]
civodul has joined #nixos
knupfer has joined #nixos
<wedens>
q6AA4FD: it "merges" paths in a single output
<wedens>
if you have pkg1 and pkg2 and you do "symlinkJoin { paths = [ pkg1 pkg2 ]; }`, result will be as if pkg1 and pkg2 were a single path
<wedens>
it'll crate symlinks for both pkg1 and pkg2 content in the output directory
<wedens>
typically useful for bundling binary with plugins and simillar things
o1lo01ol1o has joined #nixos
sheth has joined #nixos
<q6AA4FD>
wedens: oh, i think buildEnv does this too actually
<q6AA4FD>
cause i have all my binaries under a buildEnv linked together to the same store
knupfer has quit [Remote host closed the connection]
vcunat has joined #nixos
vcunat has quit [Client Quit]
o1lo01ol1o has quit [Ping timeout: 268 seconds]
thc202 has joined #nixos
the-kenny has quit [Ping timeout: 268 seconds]
ng0 has joined #nixos
bgamari has joined #nixos
<betaboon>
hello #nixos, how do you configure a 2nd screen that is only occasionaly connected ?
<etu>
betaboon: I run autorandr, just like: "programs.autorandr.enable = true;" in my config. And then as a user I set my screens up in all different places and store them as autorandr configs. Others may do it in a different way.
ThatDocsLady has joined #nixos
<jasongrossman>
I do the same as etu.
<Ralith>
autorandr seems neat, though I've never gotten it to work
<jasongrossman>
Excetpt that I didn't know about programs.autorandr.enable.
<betaboon>
etu: thanks. will take a look
<jasongrossman>
Should be services.autorandr.enable, I think.
<etu>
Recently I also found "autorandr common" and "autorandr clone-largest"
<jasongrossman>
But I haven't even been using that.
<jasongrossman>
Just installing autorandr.
<etu>
The first one clones all outputs to a single view with the lowest common resulotion for all outputs and the latter one clones them with the largest resolution and scales every output to be the same size
<etu>
Those two are great for plugging in a beamer or something kinda on the fly
<jasongrossman>
Yes.
logzet has joined #nixos
laas has joined #nixos
Celmor has left #nixos [#nixos]
<laas>
Is anyone here using linux 5.0rc1?
<laas>
I'm a new NixOS user, and when I use 5.0rc1 instead of 4.19, the perl scripts packaged with NixOS don't run anymore
<laas>
It errs saying that the OS could not execute them, leading me to think that the binfmt format changed in 5.0rc1 (breaking userspace lol)
<etu>
laas: Any good reason to run that new kernel?
<laas>
e.g. command-not-found
<laas>
Yeah, it contains a few fixes I need
<etu>
I hardly think that linux would break userspace, that's very rare
<laas>
It's either that or 4.18, but I don't know how to run 4.18
<laas>
It must be a bug I suppose, probably not intentional
<laas>
I can't check the exact error message right now, but I'm quite certain it's because it can't recognize the binfmt
<laas>
(I'm not at the machine)
<srhb>
laas: 4.20 isn't good for you either?
<srhb>
That'd be simpler to get.
r5d has quit [Quit: WeeChat 2.3]
<laas>
Unfortunately not
<srhb>
Meh..
<etu>
5.0-rc2 is out and is merged into master, that kernel may help you then
<teto>
tilpner: arg the git fetch was working but after 50mn died with "remote: fatal: Out of memory, calloc failed". I will give more memory to the VM
<{^_^}>
[nixpkgs] @Mic92 opened pull request #54044 → oraclejdk: 8.191 -> 8.201; no longer depend on requireFile → https://git.io/fhWQf
sigmundv_ has joined #nixos
_d0t has joined #nixos
<_d0t>
hi! How do I set capabilities on a specific program in configuration.nix?
<ben>
I suspect it's with the security.wrappers option but I haven't done it myself
<ben>
ok you already know more than i do, disregard me
<_d0t>
I actually don't know much. I copied and adapted this from mtr
<ben>
I'm just reading the description for the security.wrappers option :)
<_d0t>
I tried looking up docs on security.wrappers but there seems to be very little
<ben>
You're running the wrapper and not the original binary, right?
jasongrossman has quit [Ping timeout: 240 seconds]
<_d0t>
Hmmm...
<makefu>
_d0t: what are you trying to achieve? use the tool in a service or just locally as a user?
<_d0t>
makefu: the latter
teto has quit [Ping timeout: 240 seconds]
<makefu>
okay then the wrappers part should work(tm)
<_d0t>
I want to be able to run this tool without root privileges.
<ben>
I think I only ever interacted with this facility when trying to patch pmount to whitelist exfat, but I think that means I only had to be careful to patch the derivation for pmount seen by the system being built rather than a user-specific derivation
<ben>
and I didn't fiddle with that option directly
<ben>
I'm not even on a nixos system right now to help experiment though :(
<_d0t>
I just wanna see if it works before I submit a PR for the package itself.
<ben>
I guess the wrappers live in /var/setuid-wrappers?
sigmundv_ has quit [Ping timeout: 245 seconds]
<makefu>
_d0t: not sure if it really helps, but all other toos use cap_net_raw+ep instead of =ep
jasongrossman has joined #nixos
<makefu>
after setting the wrapper and rebuilding, you can check the capabilities with getcap /run/wrappers/bin/ubridge
<_d0t>
makefu: thanks. I'll check it out.
<_d0t>
Do I have to name the wrapper differently from the program itself?
<siraben>
How can I view the currently installed unfree packages?
<makefu>
_d0t: there is the subattribute "program" it seems
estrom has joined #nixos
<_d0t>
>The default name is the same as the attribute name itself.
<makefu>
yep exactly
<_d0t>
Hmm...
<_d0t>
Will the original program still be in PATH somewhere?
<makefu>
only if you add it to environment.systemPackages
<estrom>
How do I retrieve the latest nixpkgs?
<_d0t>
makefu: so if I don't, but I have a wrapper, it still will be installed, right?
<makefu>
yes exactly, this is the magic of nix
<_d0t>
Ok, thank you.
hphs^ has quit []
<makefu>
it will resolve all the attributes required to set the wrapper (and build your system essentially)
<ben>
"installed" might be a strong word :)
eadwu has joined #nixos
<_d0t>
oh yeah
<_d0t>
seems to be working
nst^ has joined #nixos
<makefu>
sweet
ThatDocsLady has quit [Ping timeout: 244 seconds]
init_6 has joined #nixos
sigmundv_ has joined #nixos
<Philonous>
I'm trying to see packages available in the current channel, so I'm running »nix-env -qa», however, it responds with "error: cannot auto-call a function that has an argument without a default value ('pkgs')". Have I misconfigured something or am I doing something wrong?
nikivi has quit [Quit: ZNC is awesome]
<__monty__>
Philonous: Seems so, did you add an override or somethinG?
<nefix>
Hey there! I'm trying to use again NixOS (this is my second attempt in 6 months). I need to make a script or something that depending on the network I'm connected, the /etc/hosts file change. How could I archieve that?
ubert has quit [Ping timeout: 244 seconds]
<Philonous>
__monty__, Not that I'm aware of
<__monty__>
Philonous: Hmm, if you didn't add an override or an overlay I'm out of ideas.
<__monty__>
nefix: Sounds like an XY problem.
<wirew0rm>
nefix: that seems not completely nixos related, nixos does the static configuration... there are currently mutltiple ways of doing network configuration in nixos but for your usecase i would use networkManager which can run hooks for connections.
<Philonous>
Oh, I found the problem. I added the nixos-hardware channel. If I remove it, nix-env -qa works again.
<Philonous>
OK, they have it in the issue tracker.
<nefix>
wirew0rm: well, since I configured the /etc/hosts file with the nixos configuration file and I know it can be scripted (in some ways) I thought I would be able to do something like that
<mpickering>
Is there a way to specify the file to run from a tarball when pointing "nix-shell" at a URL?
<sphalerite>
mpickering: not that I know of
<symphorien>
with IFD, i imagine
<sphalerite>
you just need to have a default.nix in it
<sphalerite>
oh yeah
<mpickering>
so I need to add a default.nix with just import ./file.nix?
<sphalerite>
yes, or symlink it
<wirew0rm>
nefix: yeah i just saw that hosts is in the store... but i think you do not want to rebuild your system each time you change network
<Philonous>
Where does the per-user nixpkgs config go? ~/.config/nixpkgs/config.nix or ~/.nixpkgs/config.nix ? I've seen contradictory information on the web and neither seems to pick up my »{allowUnfree = true;}«
<duairc>
Philonous: Both work, but ~/.config/nixpkgs is the new one
<duairc>
I don't know why allowUnfree wouldn't work though
silver has joined #nixos
<{^_^}>
[nixpkgs] @bjornfor pushed to master « gitolite: change patchPhase to postPatch »: https://git.io/fhWFC
ottidmes has joined #nixos
<Philonous>
It didn't work because the file name had a space at the end. I don't even want to know how that happened -_-
<{^_^}>
[nixpkgs] @bjornfor pushed to release-18.09 « gitolite: change patchPhase to postPatch »: https://git.io/fhWFi
<Guanin>
Hi, is there a way to generate a random variable inside the nix language? My usecase for this would be the following: I want to setup prosody (Jabber daemon) with biboumi (IRC transport). They need to share a secret value. Reusing the secret itself is easy to do (just access the attribute). But it would be nice if I do not even have to generate a secret (and maybe make it change when rebuilding the system)
<gchristensen>
no
dramforever has quit [Client Quit]
laas has quit [Quit: WeeChat 2.3]
init_6 has quit []
<tilpner>
teto1: If you can, fetch shallowly or fetch archives of git repos
<ottidmes>
Guanin: is it OK if the value is world readable? otherwise you could just start with some default value and run some service or activation scripts that generates a new value on every rebuild which is used the next time
<Guanin>
World readable is okay for my usecase (only I have access to anything that is not within a specific chroot), but I basically just wanted to know if there is something like a "getRandomString" builtin
dramforever has joined #nixos
<gchristensen>
no
<duairc>
You'd think readFile /dev/random or something would work, but it just returns an empty string
<{^_^}>
[nixpkgs] @Mic92 pushed to master « conan: fix distro package version »: https://git.io/fhWNr
<dramforever>
You can probably use a build script that passes --argstr secret $fresh_rand_string
<ottidmes>
Guanin: builtins.currentTime might be sufficient for your use case
<mikeplus64>
how do i, using callCabal2nix, depend on an external package/lib? i *thought* `callCabal2nix "foo" ./foo { tar = nixpkgs.libtar; }` would work, because callPackage seems to work that way, but that results in an error that the "tar" attr is unexpected.
<ottidmes>
Guanin: you then do not even have to store it, it produces the same timestamp for every evaluation, but different per evaluation
<{^_^}>
[nixpkgs] @bjornfor pushed 2 commits to release-18.09: https://git.io/fhWxt
ThatDocsLady has joined #nixos
o1lo01ol1o has quit [Ping timeout: 258 seconds]
ng0 has quit [Quit: Alexa, when is the end of world?]
tA| has joined #nixos
tA- has quit [Ping timeout: 272 seconds]
sinner has joined #nixos
sinner has quit [K-Lined]
<dramforever>
Is anyone on nixos-unstable on a personal machine?
<dramforever>
I'm thinking of living on the bleeding edge on my laptop, but I'm wondering about the rate of upgrades.
<srhb>
dramforever: I usually am.
Guanin has quit [Ping timeout: 272 seconds]
<srhb>
dramforever: Upgrades are usually less frequent but more severe.
<dramforever>
Like, would I get like large upgrades a lot more frequently?
<srhb>
The upgrades are less frequent but often _bigger_
<dramforever>
Hmm
<srhb>
Less frequent because of failures on Hydra, bigger because mass rebuilds much more often.
<ben>
Hang on they're bigger where relative to where?
<srhb>
A stable channel.
<dramforever>
hundred-meg upgrades is a problem because at my location, the download speeds of cache.nixos.org are not very fast
<dramforever>
averages to about 100 KiB/s
<srhb>
Then you're probably better off with stable.
<dramforever>
Whereas a very close local mirror of Arch Linux (for example) can get to 10MiB/s
<srhb>
dramforever: Where are you located?
leotaku has quit [Ping timeout: 258 seconds]
<dramforever>
srhb: okay, thanks
<dramforever>
I'm in Beijing, China
<srhb>
dramforever: Hm. Didn't know our CDN situation was that bad there. Might be worth mentioning on discourse or something.
<srhb>
Or just asking fastly, maybe..
ikhoi has joined #nixos
<dramforever>
fastly has been slowly the whole time, for everyone
<dramforever>
We all know it here
<dramforever>
everyone -> everyone here
cyounkins has joined #nixos
<ikhoi>
hello there, everything was working very well on my nixos install until today: for example "nix-env -i firefox" says "error: cannot auto-call a function that has an argument without a default value ('pkgs')". Any idea what's wrong?
<dramforever>
anyway, I'm currently fine with stable and I don't really need anything from the bleeding edge, so I guess I'll keep my current situation untouched
<__monty__>
ikhoi: Do you have the nixos-hardware channel?
<ikhoi>
yep
<__monty__>
Came up earlier, something regressed there probably.
<ikhoi>
I am importing <nixos-hardware/dell/xps/13-9370> from there
knupfer has joined #nixos
jD91mZM2 has quit [Quit: WeeChat 2.3]
moet has joined #nixos
ikhoi has quit [Quit: Page closed]
lfish has joined #nixos
fendor has joined #nixos
Guanin has joined #nixos
<{^_^}>
[nixpkgs] @Ma27 opened pull request #54051 → nixos/prometheus: make `source_labels` optional → https://git.io/fhWjY
fusion809 has quit [Remote host closed the connection]
ddellacosta has quit [Remote host closed the connection]
<Mic92>
I think any attribute set without the recurseIntoAttrs function would do
<mpickering>
I'm not particularly sure why NUR is trying to evaluate the src anyway?
<mpickering>
I suppose the problem is that I'm using builtins.fetchurl
<mpickering>
so the evaluation forces that to be fetched
ubert has joined #nixos
r5d has quit [Quit: WeeChat 2.3]
fendor_ has joined #nixos
fendor has quit [Read error: Connection reset by peer]
<lfish>
When I run ssh-agent, it works for programs run from that terminal. Is there a way to run it just once? Is that okay security-wise?
<lfish>
by the way, I'm running on nixos 18.09 with xmonad
<spacefrogg>
lfish: Today, you normally use gpg-agent as a replacement for ssh-agent. There are systemd user-units available that just need to be activated.
<srhb>
lfish: programs.ssh.startAgent
<Mic92>
mpickering: yes. the normal fetchurl would not have this problem.
<mpickering>
I can't use the normal fetchurl as I want the impurity?
<Mic92>
the normal fetchurl is a fixed-input derivation.
<Mic92>
it is only executed when you actually build something.
r5d has joined #nixos
nakkle has quit [Ping timeout: 252 seconds]
nakkle has joined #nixos
<mpickering>
But don't you have to specify the sha256?
endformationage has joined #nixos
<lfish>
thanks, i'll set ssh-agent and look into gpg-agent when I have more time
<gchristensen>
should we mark rustc as big-parallel?
knupfer has quit [Client Quit]
phreedom has quit [Remote host closed the connection]
<Mic92>
gchristensen: we could. It does not scale beyond 7 cores or so with our settings however.
<gchristensen>
oh really? interesting
<Mic92>
the concurrency was limited due to memory consumption
<gchristensen>
it isn't finishing a compile on aarch64 these days, and big-parallel would get it more cores
phreedom has joined #nixos
<Mic92>
Interesting. I thoughed it would be faster now that we no longer compile their llvm.
<Mic92>
Is this on master or stable?
<Mic92>
betaboon: I think it is possible to skip build scripts.
<Mic92>
and do those manually.
<betaboon>
Mic92: yeah. but those are required for that to work
reinhardt has quit [Quit: Leaving]
<Mic92>
betaboon: are the assets shipped in a released version or so?
<Mic92>
we do this for grafana
<Mic92>
compiling javascript yourself is acedemic in the most case anyway.
<Mic92>
oh, it is nodejs only
<Mic92>
forgot what I said
<betaboon>
Mic92: nope assets are not in the release-package (provided on github)
<Mic92>
npm2nix and skip the build scripts maybe.
o1lo01ol1o has joined #nixos
kisonecat has quit [Ping timeout: 250 seconds]
<Mic92>
or are the dependencies not available during the build?
<betaboon>
Mic92: just trying to work with yarn2nix as they state in their readme to use yarn because of reasons
<betaboon>
Mic92: they are packaging a webfrontend for their stuff within the repo as a separate package (called in the make-assets script) and that (ofc) doesnt find the dependencies
nikivi has quit [Read error: Connection reset by peer]
Nikita is now known as Guest38017
Guest38017 is now known as nikivi
kisonecat has joined #nixos
bara has joined #nixos
<unacceptable>
If a derivation has a lib.optionalString, how do I supply the value for that in nixos?
<srhb>
unacceptable: It's not clear from your description. Is it an argument to a top level name? eg. something you'd normally adjust with overrideAttrs? Or maybe a nixpkgs.config value?
<clever>
that chunk of code runs: nix-build '<nixpkgs/nixos/release.nix>' -A options -o whatever
<clever>
which generates the json file used by the website
<nefix>
Hey again. Could someone help me? I'm trying to build a nix-shell script for Golang using https://github.com/MatrixAI/Golang-Demo this. The issue comes because it can't find lib. Could anyone give me a hand? Thanks!
<clever>
my installer GUI was then reusing that, to make a full tree view of the options, with type-sensitive helpers
<clever>
Taneb: cant remember what the exact problem i ran into was though...
fendor has joined #nixos
<Taneb>
clever: that sounds slightly more ambitious than my idea
<{^_^}>
[nixpkgs] @dotlambda pushed commit from @r-ryantm to master « python37Packages.w3lib: 1.19.0 -> 1.20.0 (#54063) »: https://git.io/fhlma
<fuzzy-id>
in a `shell.nix' file, can i set the initial `nixpkgs ? import <nixpkgs> {}' so that `nixpkgs' points a certain release?
<wedens>
what's the expiration policy on official binary cache?
<{^_^}>
[nixpkgs] @peti pushed 3 commits to haskell-updates: https://git.io/fhlmw
<fuzzy-id>
background is: i develop in haskell on nixos 18.09, build server is nixos 18.03 lots of libraries are a bit dated on the build server in the default nixpkgs set
<{^_^}>
[nixpkgs] @FRidh pushed commit from @r-ryantm to master « urh: 2.5.3 -> 2.5.4 »: https://git.io/fhlYj
reilithion has joined #nixos
Judson has joined #nixos
<adisbladis>
Point taken though, needs better README
<nefix>
Oh it's in nixpkgs. Though I want to learn how to install it. So I'm using rycee's Home Manager. Is there any way to automatize this build?
Judson is now known as Guest68469
<wedens>
nefix: so, just add `vgo2nix` to `environment.systemPackages` (for system-wide on nixos) or `nix-env -iA nixpkgs.vgo2nix` (for imperative per-user)
<nefix>
wedens: yeah, but I was saying if it weren't in the nixpkgs repo
<reilithion>
Hey, what's the deal with packages that are broken by the libupnp 1.6 -> 1.8 version bump? Is there any reason not to have a libupnp16 in nixpkgs to fix those while their upstreams ponder porting to libupnp 1.8?
<wedens>
nefix: yeah, you can use home-manager to have per-user configuration. also `users.users.<name?>.packages` in nixos config if you just want to manage packages per-user
<nefix>
adisbladis: You could also include https://github.com/MatrixAI/Golang-Demo as an example in the README (or create your own example folder), it would help a lot to novice people
<{^_^}>
[nixpkgs] @dotlambda pushed commit from @r-ryantm to master « python37Packages.uncompyle6: 3.2.4 -> 3.2.5 (#54064) »: https://git.io/fhlOj
<adisbladis>
nefix: Oh great :) I hadn't seen that one before
<nefix>
I'm lucky I found it, I wouldn't make it work without it :D (though I made some changes [the pkgs.nix file isn't needed])
<nefix>
adisbladis: ^
<simpson>
reilithion: Okay. What have you tried so far? What are you stuck on?
<reilithion>
simpson, Basically, it won't compile. There are about two dozen errors in at least one file. It /looks/ like RetroShare /may/ have tried to support compiling against libupnp 1.8, but I'm not sure. There's a line in a config file that detects whether they are on a Debian system with a fakey 1.8 API that's actually 1.6 under the hood. It detects this poorly, and in particular would turn up a false positive on Nix. I tried adding a sed patch that
<reilithion>
neuters this detection, but it didn't fix the compilation errors for some reason.
<{^_^}>
[nixpkgs] @dotlambda pushed commit from @r-ryantm to master « python37Packages.vine: 1.1.4 -> 1.2.0 (#54065) »: https://git.io/fhl31
<{^_^}>
[nixpkgs] @Ekleog pushed to master « opensmtpd: add comment about progress of upstreaming our patch »: https://git.io/fhl3Q
<simpson>
reilithion: Mm. Yeah, at this point I'd be trying to write a patch that fixes the build too.
ThatDocsLady has quit [Ping timeout: 258 seconds]
kisonecat has quit [Ping timeout: 250 seconds]
ToxicFrog has quit [Ping timeout: 250 seconds]
fendor has joined #nixos
<reilithion>
I don't know how to fix the build. I don't know what libupnp does; I don't know what it's even doing in RetroShare; I don't know how the API changed from 1.6 to 1.8.
<simpson>
I don't know any of those things either. Again, you *could* revive the older version of libupnp, if you like. This sounds like a good time for multiple library versions.
<yl[m]>
infinisil: I don't fully understand `deployment.hasFastConnection`. If I set it to true the nix-closure will be copied from my computer to the aws instance or would the instance get them from the cache?
<infinisil>
If true, it will copy it from the deployment machine directly instead of going through the cache
<nefix>
Ok, so now I'm trying to migrate my Neovim configuration. There's a plugin I use that is the fzf. It requires an external dependency (the fzf program itself). How am I supposed to add it? There are two vim packages related with fzf: fzfWrapper and fzf-vim. Which should I use?
<ingenieroariel>
Hi everyone, I have created a few nix packages for golang applications (go-spatial/tegola, ory/hydra, ory/keto, ory/oathkeeper) and would like to contribute them upstream - is the window for sumitting packages to 19.03 open ?
<Synthetica>
ingenieroariel: You can always submit to master, this should automatically be included in 19.03 at this point
<ingenieroariel>
perfect, for now I went through the nur-packages route to test them - are modules mandatory (packages are supposed to be services)
<ottidmes>
ingenieroariel: not mandatory, but very much appreciated when they are meant to be run as services
<Synthetica>
ingenieroariel: No, most packages don't have modules. When your package implements something like a server, it might be nice to add a module
PLPD-Bot has quit [Remote host closed the connection]
fusion809 has quit [Remote host closed the connection]
<ingenieroariel>
ok, then I'll try my luck with the bare packages first and then learn about modules - thanks a lot
fusion809 has joined #nixos
PLPD-Bot has joined #nixos
fusion809 has quit [Remote host closed the connection]
jasongrossman has quit [Read error: Connection reset by peer]
Judson has joined #nixos
Judson is now known as Guest38142
<kirikaza>
Hello! I wrote a derivation with udev rules and now I wonder how to test it. AFAIU `sudo nix-build -i $pkg -f $repo` isn't enough, I need to add $pkg to config.services.udev.packages. Is it correct?
<symphorien>
or udev.packages = [ ((import /path/to/nixpkgs/checkout {}).packageattr) ];
<kirikaza>
Oh, looks too easy :) Will try
<symphorien>
the first one uses stable's deps and the second one uses your checkout's
juliendehos has quit [Quit: Leaving]
<rnhmjoj>
this mechanism must break sometimes because i'm sitting right in front of the pc but apparently the system think i'm not and it doesn't give me permission to do stuff like audio or shutting down
ThatDocsLady has quit [Ping timeout: 250 seconds]
o1lo01ol1o has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @kirikaza opened pull request #54090 → moolticute: init at 0.30.8 → https://git.io/fhlat
hedning has joined #nixos
mizu_no_oto has quit [Quit: Computer has gone to sleep.]
dermetfan has joined #nixos
esmerelda has joined #nixos
MP2E has joined #nixos
<esmerelda>
I read issue #8 (about how to manage secrets on Nix) and I'm wondering, why isn't it possible to keep derivations housing secrets in a separate capability-checked store?
<esmerelda>
like have /nix/secretstore, where all secrets and derivations referencing secrets go
<Xyliton>
trying to open the mentioned URLs in my browser prompts a download so I assume it to be a problem with curl?
<LnL>
esmerelda: I've played around with that before, and I think that idea makes much more sense then transparent encryption or other things like that
<esmerelda>
O_o what a strange issue @Xylitnlon. can you reproduce the issue with curl manually, with -v turned on?
<esmerelda>
also, try running the diagnostic script for nixpkgs
<ottidmes>
lol, he seems to _really_ approve of the changes
<esmerelda>
LnL: did you fork the nix code to make it work? that's interesting
<gchristensen>
it is just a bug in github, if I reload all but one go away, but haha
ixxie has joined #nixos
<LnL>
esmerelda: no, I just made a second store in a path that's only accessible to one user
<esmerelda>
oh! @gchristensen I read your blog post about buildLayeredImage and the algorithm for deciding which layers to merge and it's terrifically cool! it came up on HN a while ago
<Xyliton>
esmerelda: diagnostic tool? After googling the only thing I could find was the `-v` flag
esmerelda has quit [Read error: Connection reset by peer]
jabranham has quit [Quit: ERC (IRC client for Emacs 27.0.50)]
esmerelda has joined #nixos
esmerelda has joined #nixos
<esmerelda>
LnL: neat, thanks! I'll give it a try
<gchristensen>
oh dear
<gchristensen>
:)
<LnL>
only used it for something very simple
<Ralith>
rycee: `home-manager news` is now telling me on 18.09 that I need to add a services.dbus.packages config entry, but home-manager itself still complains that no services.debus option exists
<Ralith>
exact same thing that was happening on master a while ago
<LnL>
needed a simple dynamic file with a secret and didn't want to use an awkard templating language
<Ralith>
still happens on master, too
<Xyliton>
Looks like the curl problem I'm getting is not limited to xsession-wrapper
<ingenieroariel>
Ok, my first pull request is done and I am now packaging others. Unfortunately, one of the packages I want in nixpkgs is called ORY Hydra (hydra!!) https://ory.sh - it's from an oauth / openid solution.
<{^_^}>
[nix] @veprbl opened pull request #2635 → manual: "sandbox" option default is "true" → https://git.io/fhlrR
<ingenieroariel>
what's are my options there? should I create a pkgs/servers/ory folder and call the packages ory_hydra ory_authkeeper and ory_keto ? should I rename the binary to avoid collusions with the other hydra?
<rycee>
Ralith: services.dbus is in the system configuration. Home Manager does indeed not have that option.
<samueldr>
ingenieroariel: with dashes, but prefixing them with ory sounds fine to me
<Ralith>
is this another one of the things where you have to reboot before it works?
<ingenieroariel>
thanks samueldr
estrom has quit [Remote host closed the connection]
<samueldr>
ingenieroariel: as for the binaries, not sure if necessary
<rycee>
Ralith: But this option was always needed if you were using, e.g., the gnome terminal module.
<ingenieroariel>
samueldr: I rather have them with their original names
<samueldr>
ingenieroariel: it would require that *both* hydra and ory-hydra would be in use in the same scope
<Ralith>
it's just recently started throwing up errors on switch afaict
<samueldr>
ingenieroariel: which I guess is not going to happen
<ingenieroariel>
right
<rycee>
Ralith: I'm not sure actually. I think it might at least be necessary to relog.
<ingenieroariel>
ok, those are good pointers, I'll create a PR with the three tools
<rycee>
Ralith: I'm guessing because you are configuring gtk through Home Manager? Recently that module was changed to also set the options in dconf (which is necessary for wayland).
blahlexh has joined #nixos
<{^_^}>
[nixpkgs] @rnhmjoj opened pull request #54094 → treewide: use ${stdenv.shell} instead of /bin/sh where possible → https://git.io/fhlrH
<Ralith>
yeah
<{^_^}>
[nixpkgs] @aanderse opened pull request #54095 → Enable memory hotplug support → https://git.io/fhlok
<Ralith>
oh joy, switching to master broke my ability to switch :/
<Ralith>
aand dconf is happy again after a relog, wonderful
<Ralith>
thanks for the help!
estrom has joined #nixos
<rycee>
Yeah, it basically just runs `home-manager switch` so it is sometimes useful this sort of circumstanses.
<rycee>
No worries :-)
jabranham has joined #nixos
<estrom>
Is the discord package only available in a version >0.0.5 on the unstable channel? I have ran nix-channel update and discord is still on version 0.0.5, `nix-env -qa discord`
<Ralith>
I think I was confusing "dbus" and "dconf" in the news item
guest14 has joined #nixos
guest14 has quit [Client Quit]
<estrom>
Also, can I query the unstable channel using nix-env?
<rycee>
Ralith: Well, they are pretty tightly coupled so you weren't so far off :-)
cyounkins has quit [Remote host closed the connection]
<blahlexh>
noob question, I'm loading a vim_configurable via nix-shell on nixos which already has vanilla "vim" included
reinhardt has quit [Quit: Leaving]
<blahlexh>
I noticed that in the shell, the vanilla vim's path precedes vim_configurable so the wrong vim is getting launched unless I call it explicitly or change PATH
EoF has joined #nixos
<blahlexh>
another way to get the correct vim into the path is if I run e.g. `env -i bash`, then `nix-shell`, then `vim`
<blahlexh>
I kind of assumed that nix-shell would prepend the paths required for the packages inside default.nix, but that is evidently not the case
<blahlexh>
what's the preferred way of doing this? i.e. loading a custom build of vim, or some other program, on top of a system with a different version?
ddellaco_ has quit [Remote host closed the connection]
agander has quit [Ping timeout: 258 seconds]
_d0t has joined #nixos
<_d0t>
hi! Is it a good practive to provide git versions of software instead of releases?
<_d0t>
In this case, gnuradio-osmosdr hasn't seen releases in many years, but the development is still on, and the last released version is just plain outdated.
<etu>
Depends
<_d0t>
For one, it doesn't build with the latest libbladeRF, so bladeRF devices are unusable.
<Synthetica>
_d0t: Sounds like a case where it could be strongly considered
jabranham has joined #nixos
<_d0t>
soo... what do I do? File a github issue for discussion?
<etu>
_d0t: If the project is making much progress without doing releases, then it's probably a good case to pick random git commits at a reasonable update frequency or something like that
<etu>
and set a date (commit date) as version
<etu>
Just make a PR :)
<_d0t>
I wouldn't say they make MUCH progress. They seem to just update whatever gets broken along the way.
<etu>
That's probably good enough
<Synthetica>
_d0t, etu: The general naming scheme is `pname-unstable-YYYY-MM-DD`
<_d0t>
ok.
<_d0t>
etu, Synthetica, thanks. I'll try making a PR.
<{^_^}>
[nixpkgs] @hedning pushed commit from @averelld to master « nextcloud-client: fix build with qt 5.12 (#53881) »: https://git.io/fhlXt
<{^_^}>
[nixpkgs] @benley closed pull request #51609 → emacs: make libXcursor available at runtime for non-GTK builds → https://git.io/fpMqU
lush has joined #nixos
<{^_^}>
[nixpkgs] @ngerstle-cognite opened pull request #54101 → Update postman to 6.7.1 → https://git.io/fhlXW
<lush>
hey everybody. I just bought a new network card, and so far I've always used ethernet to connect to the internet. I somehow can't seem to get my network card working.. what should I do?
<lush>
ifconfig won't show me any interfaces that start with "w..." (I read those are the wlan-interface)
<lush>
*+s)
<simpson>
lush: Normally I'd first check the bus that you connected the card to. `lspci`, `lsusb`, etc. These tools have flags to help you figure out which kernel modules are loaded for them, too.
<simpson>
Once armed with that, I'd start reading about the specific kernel driver for the card, and also inspect dmesg for details.
<Church->
simpson: Here's a stupid question. In my event loop a key-combo press spawns say a xterm instance. Would I then want to walk the xcb equiv of XQueryTree to get access to the spawned sub window so I can then say change it's geometry and position? Or is there something else I should do?
<simpson>
Church-: XCB at least offers more structure, as well as an API that can be easily moved to other languages; XCB's generated by turning some XML into C.
<Church->
Eyep
<Church->
So maybe I will use it then.
<simpson>
Oh, you're writing a WM. I don't have any good advice, except maybe "don't" and "I'm sorry even though it's not my fault"
<Church->
Even though I'm going for a super stripped down case of "I only need four windows spawned at all times, one Firefox instance and three gnome-terminal instances. All taking up exactly a quarter of the root window."
<lush>
simpson: so you mean adding "hardware.enableRedistributableFirmware = true;" to my configuration?
<simpson>
Yes.
doyougnu has joined #nixos
<simpson>
Church-: Speaking as somebody who's written GPU drivers: It's way easier to just learn a tiling WM and have a script that does your setup.
* simpson
lazy
<Church->
simpson: Oh definitely, I use i3.
<Church->
I just wanted to try writing my own super minimal use case.
ingenieroariel_ has quit [Client Quit]
<Church->
Like writing my own software. My shell, terminal, CI pipeline and time series db/log shipper were all written in my off time.
<Church->
As well as the editor I'm working on.
doyougnu1 has quit [Ping timeout: 250 seconds]
<Ralith>
Church-: have you considered targeting wayland instead?
<Church->
Ralith: I've been looking at it, but holy frack is there so much extra stuff to write.
<Ralith>
fewer ridiculous legacy design issues to work around there
_kwstas has quit [Quit: _kwstas]
<Church->
Since you have to write the renderer and everything.
<Ralith>
honestly I'd be surprised if it wasn't less work in the long run
<Church->
Maybe
_kwstas has joined #nixos
<Ralith>
simple compositing isn't that hard
<Church->
I've been debating on it.
_kwstas has quit [Remote host closed the connection]
<Ralith>
it's practically an OpenGL hello world problem
<oldandwise>
anybody using nixos for rpi3 ? is it usable?
<lush>
simpson: I tried both enableRedistributableFirmware and enableAllFirmware = true; it still shows the same error in dmesg and doesn't show any "w..."-interfaces in ifconfig (it shows a "enp3s0"-interface though)
<simpson>
lush: Hm. You might have to go on a journey to find the firmware package, then. The dmesg error had a .bin filename which will help you.
jabranham has joined #nixos
<lush>
"[ 14.865362] brcmfmac 0000:06:00.0: Direct firmware load for brcm/brcmfmac4366c-pcie.bin failed with error -2" So you mean I have to look for some nixos package or sth. that offers that brcmfmac4366c-....bin?
o1lo01ol1o has joined #nixos
silver has quit [Ping timeout: 268 seconds]
<{^_^}>
[nixpkgs] @fpletz pushed commit from @dtzWill to staging « freetype: re-enable freetype-config, don't break all the things »: https://git.io/fhlyW
kisonecat has joined #nixos
<simpson>
lush: Yeah. FWIW it looks like linux-firmware (the kernel's official firmware package) may have it, but in that case I'm not sure why things aren't working for you. This git repo appears to have several extracted bits of firmware without license information: https://github.com/Hill-98/phicommk3-firmware
sigmundv_ has joined #nixos
nokomprendo has quit [Quit: Leaving]
<lush>
simpson: any chance "boot.extraModulePackages = [ config.boot.kernelPackages.broadcom_sta ];" could help? Just some random guess from looking at a github pull request
slack1256 has quit [Remote host closed the connection]
npmccallum has quit [Ping timeout: 250 seconds]
silver has joined #nixos
<djahandarie>
I have a really big file (~1.5GB) I need in my build environment. If I do the naive thing in the nix expression, expression evaluation takes a really long time (with the message warning: dumping very large path (> 256 MiB); this may run out of memory). Is there any more efficient way to do this?
<gchristensen>
djahandarie: use fetchurl on it, and specify a sha256
lush has quit [Quit: Page closed]
<{^_^}>
[nixpkgs] @evhan opened pull request #54109 → chicken: fix egg installation by preserving compiled libdir → https://git.io/fhlSX
pddalm has joined #nixos
<djahandarie>
It is a slightly different file ~every evaluation.
spacekookie has quit [Quit: **more agressive swooshing**]
cyounkins has quit [Remote host closed the connection]
<djahandarie>
I suppose that isn't stopping me from sha256ing it on the fly every time, but would that work?
<gchristensen>
no
<gchristensen>
that counts on it being stable
<gchristensen>
what is the file?
cyounkins has joined #nixos
<infinisil>
djahandarie: But if it changes you need to reimport it, which is the thing that takes the long time..
<djahandarie>
infinisil, right, so I think my question is if there's a faster way to import something into the nix store (especially a way that doesn't seemingly involve loading the entire thing into memory???)
<djahandarie>
gchristensen, I think you won't be happy with my answer so I'll just leave that question unanswered :D
<gchristensen>
now I have to know! :o
<djahandarie>
Hah. It's a rust 'target' directory. This is definitely not something one should ever be doing normally, but...
<gchristensen>
...but?
<infinisil>
djahandarie: Well, I hate to suggest it, but you *could* turn off the sandbox, which would allow the build access to the file directly, without going through the store.
<djahandarie>
gchristensen, but I have been frustrated by my nix build times enough that it seems like a reasonable enough hack to do for local nix builds.
kisonecat has quit [Ping timeout: 268 seconds]
<gchristensen>
what are you trying to actually accomplish?
<djahandarie>
See, this is why I didn't want to answer!
<djahandarie>
Cliff notes of why I'm switching off carnix: 1. EPIC nix expression evaluation times (over a minute), 2. constant breakage on more complicated cargo.toml files, 3. when I fix it, it's hard to push patches back because carnix source code is managed on a nonstandard SCM/host and the author doesn't respond to emails, 4. it doesn't allow you to use cargo in a nix-shell, which is what I really really w...
<djahandarie>
...ant for things like a super fast cargo check
<djahandarie>
I'm sure there's a way to make carnix work. I've exhausted my emotional stores on persuing it any further.
<djahandarie>
On the other hand, buildrustpackage basically does everything I want with no issues, and I can iterate very quickly in a nix-shell. I just want to be able to speed up my nix builds for when I need to do them to build my docker container (not to deploy it but to test it), where I'm fine with the 'impurity' introduced by throwing the target directory in there instead of rebuilding all my dependencie...
<djahandarie>
...s (takes a *long* time)
npmccallum has joined #nixos
<djahandarie>
So I basically already have my solution to X. Y is how to make it 10% faster. That's why I didn't come in with the entire original problem statement.
<infinisil>
djahandarie: I hope the situation with carnix improves sometime, would love to have the haskell equivalent in nixpkgs
_kwstas has joined #nixos
cyounkins has quit [Remote host closed the connection]
<lush>
simpson: just one thing left: is there any kind of "standard" approach in finding the neccessary firmware? So I know I need that "brcm/brcmfmac4366c-pcie.bin" thing, but is there like any comfortable way to finding a packages/option that offers that? Thanks for everything btw :)
cyounkins has joined #nixos
<djahandarie>
Yes, me too. It's conceptually a far better more nix-like solution, it's just really not where it needs to be in terms of maturity IMO. I've been using it for almost a year and am finally giving up.
<symphorien>
djahandarie: "4. it doesn't allow you to use cargo in a nix-shell," why ?
<simpson>
lush: The steps I sketched are how I find firmware. I don't know if there's a better way.
<symphorien>
I mean, cargo as buildInput is usually enough to use cargo on most projects
<djahandarie>
symphorien, well sure, but not use cargo in any way that utilizes the carnix derivations for your deps.
__monty__ has quit [Quit: leaving]
<symphorien>
right
<symphorien>
still, since cargo is incremental, the cost to pay is not that steep, is it ?
<symphorien>
I share your feeling about buildRustPackage though
kisonecat has joined #nixos
<djahandarie>
I suppose, but if I'm already specifying all my native dependencies manually and using cargo for all my development, the marginal utility of bringing carnix into the picture and dealing with all its imcompatabilities and slowness seems very low
<lush>
simpson: alright thx... is asking for such a firmware worth posting sth. on github? because I feel that is where the ppl gather, that could possible know where I can find my firmware
cyounkins has quit [Ping timeout: 244 seconds]
<lush>
simpson: sorry if that's a stupid question.. how should you know what's worth to be asked and what not lol
<symphorien>
I see
<simpson>
lush: Possibly, but I think that we've already found the firmware and now it's a matter of getting it into the right /lib spot.
<simpson>
FWIW the ultimate problem is the hardware manufacturer, both for requiring firmware and for making it hard-to-obtain.
<lush>
that's true, won't help me making my network work though :D
<lush>
but if the firmware has got installed in the "wrong" spot shouldn't that be a problem regarded with some error in nix-code that was executed by my changed options?
<ivan>
I enabled boot.initrd.network.ssh and it doesn't seem to have started an ssh server in initrd, any ideas
jomik has joined #nixos
<lush>
so... shouldn't that be considered a bug or am I missing sth simpson? :')
<simpson>
lush: Fuck if I know.
<lush>
hahahaha alright
<lush>
thanks! :)
<lush>
for everything so far :) I'll continue tomorrow. you really helped me a lot, thank you for that simpson. cya
<simpson>
Peace.
<ivan>
oh I forgot to boot.initrd.network.enable
lush has quit [Quit: Page closed]
<symphorien>
djahandarie: so I was curious, and it seems that cargo check is like normal building but adding the flag --crate-type=metadata
_kwstas has quit [Quit: _kwstas]
<systemfault>
Anyone using zfs on root? What about the old comment saying to not use a zvol for swap?
<symphorien>
so maybe something like RUSTFLAGS=--crate-type=metadata $buildPhase would work
<systemfault>
I have a difficult time just taking an advice that was given nearly 5 years ago.
<systemfault>
Or should I simply not bother with zfs and use good old ext4
<gchristensen>
systemfault: it'd be fine to use ext4. probably still shouldn't use a zvol for swap. ZFS needs RAM to write, swap is used when RAM is low. if ZFS can't allocate RAM, system can't swap
<{^_^}>
[nixpkgs] @telotortium opened pull request #54113 → xrdp: fix clipboard for non-ASCII characters → https://git.io/fhlQj
<gchristensen>
systemfault: I use ZFS for my root, and a separate partition for swap
<{^_^}>
[nixpkgs] @dtzWill opened pull request #54122 → llvm7: patch to fix PR39427 → https://git.io/fhlFc
<ivan>
can I mkOption in my nixos configuration or does that not work? (I put an mkOption in a file early in the imports = [ ] list but it doesn't seem make the option exist.)
<clever>
bsima: a function it will run on the drv, before using it
<clever>
ivan: either `lib.mkOption` or put `with lib;` near the top
leo_ has quit [Disconnected by services]
<ivan>
clever: the lib.mkOption seems to work but not the user
<ivan>
error: The option `hardware.ethernetKernelModule' defined in `/home/at/code/system/configuration/common/configuration.nix' does not exist.