<charukiewicz>
Is there any easy way for me to set environment variables in a nix-shell when I'm just using stdenv.mkDerivation to create it? e.g. is there a parameter I can pass in to mkDerivation?
<clever>
charukiewicz: every parameter to mkDerivation becomes an env var of the same name
<charukiewicz>
clever: Thanks, very simple. That worked perfectly.
meteo has joined #nixos
Qwerky has quit [Remote host closed the connection]
Rusty1 has left #nixos ["WeeChat 2.9"]
supersandro2000 has quit [Disconnected by services]
<tpw_rules>
ok, then you will have to create an overlay first
<clever>
you may need to use normal override stuff, to append to that attr
nhs has joined #nixos
<tpw_rules>
but then yeah you can say something like thePackage = super.thePackage.overrideAttrs (old: { cmakeFlags = old.cmakeFlags ++ [ "-DCOOL_OPTION=ON" ]; });
<clever>
yep
<remexre>
okay
<remexre>
I could do nixpkgs.thePackage.overrideAttrs ... too, right?
<remexre>
(if this is a quick hack, not going to be upstreamed, etc etc)
<clever>
if nixpkgs contains a copy of nixpkgs, yeah
<jesystani>
howdy, im trying to get some games working under wine, and cant for the life of me figure out how to get the 3d parts working correctly. ive made sure to enable 32bit dri support, and i installed dx9 with winetricks. is there some best practices i should be doing?
<clever>
jesystani: try running "nix-shell -p pkgsi686Linux.glxinfo" and then use the 32bit glxgears within that, does it v-sync?
<jesystani>
seems to yes
Quick_Wango has joined #nixos
<clever>
32bit dri support seems to be fine then
<jesystani>
my problem is that the game im running under wine doesnt load textures correctly, and all the graphics seem messed up, lots of straight black screens with textureless polygons in the centre of th escreen rather than where tehyre meant to be
<jesystani>
i presumed i didnt have directx9 installed properly but i cant seem to figure out the correct way to get it set up
waleee-cl has quit [Quit: Connection closed for inactivity]
<rsynnest>
hi all, building my first nix package and after a recent git pull im getting this error: anonymous function at nixpkgs/pkgs/shells/bash/4.4.nix:1:1 called without required argument 'util-linux'
<rsynnest>
was there a recent change in packaging requirements when using bash?
stree has quit [Ping timeout: 264 seconds]
<siraben>
rsynnest: what command did you use to build your package?
<rsynnest>
nix-build -A pkgname
<siraben>
try `nix-build . -A pkgname`
nhs has joined #nixos
<siraben>
utillinux was renamed to util-linux recently
<attila_lendvai>
Reventlov, i'm also seeing this. thinkpad x250. it is not too bad on my samsung sata SSD, but it is very bad on my M.2 internal SSD, to the point that i moved my nixos partition from it. it's fast when it works, but then hangs sometimes for seconds.
<attila_lendvai>
Reventlov, last time i googled around (months ago), i couldn't find anything, that's why i installed the samsung SSD as a bearable workaround.
ahmedelgabri has quit [Ping timeout: 264 seconds]
thc202 has joined #nixos
<{^_^}>
[nixpkgs] @SuperSandro2000 pushed to lscolors « lscolors: ini at 0.7.1 »: https://git.io/Jt737
<{^_^}>
[nixpkgs] @SuperSandro2000 pushed to lscolors « lscolors: ini at 0.7.1 »: https://git.io/Jt73A
<{^_^}>
[nixpkgs] @SuperSandro2000 pushed to lscolors « lscolors: ini at 0.7.1 »: https://git.io/Jt73x
<chisui>
Hey, I'm having problems with DNS and openVPN. When I'm connected to WiFi the dhcp service writes the routers DNS configuration to `/etc/resolve.conf` but when I'm starting the openVPN service that doesn't change, even if restart dhcp. Am I missing something or am I searching in the wrong place?
<attila_lendvai>
Reventlov, my condolences... :) but AFAIU M.2 is practically just a sata disk on a different physical connection. i even looked for firmware updates... my hope is that a kernel update will magically solve it eventually.
<aveltras>
When using docker with nixos, if any container is running when i shutdown my machine, i have to wait for 1m30 for the shutdown to finish because it's waiting on docker service to stop. Is there a way around it ?
<supersandro2000>
aveltras: a not so nice way would be to set the timeout to 1
<aveltras>
i think it was the same thing when i used nixos-containers but haven't done so recently
nhs has joined #nixos
<aveltras>
supersandro2000: which option is this ,
<mvwm>
I flashed the latest build on the SD card and started the Pi
<mvwm>
On this guide they mention the nixos-generate-config command at first. So is this the first thing Im gonna have to to? Also, what is "and dont forget to set a user" supposed to mean?
<tejing>
mvwm: the main installation guide is in the nixos manual. that wiki page is mainly describing the parts that are different from a normal installation
<tejing>
I've done an install on a raspberry pi 4, but I wouldn't reccommend it as a first install. there are some shenanigans with the boot process and kernel which require a bit more understanding of how nixos works to do correctly, and last I checked, the accelerated graphics hardware support didn't work right, so you were stuck with framebuffer graphics.
<siraben>
s1341_: "now that there are some people around can anyone help me with the packaging question I have?", you should just go ahead and ask your question
<qyliss>
s1341_: you can use a seperate python output, or just keep them in $out. Either way, put them in the standard location for Python
<siraben>
ah sorry the {^_^} chat made it hard to see
sangoma has joined #nixos
<s1341_>
qyliss: they want to go in site-packages. which is in nix-store... what's the 'standard' python location. or Rather how do I use a separate python output?
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
orivej has joined #nixos
nhs has joined #nixos
<qyliss>
s1341_: putting them in site-packages is correct
<qyliss>
I recommend not using a seperate output unless you find a good reason you need to
<s1341_>
qyliss: but how do I do that under nix:
<s1341_>
?
mvwm has quit [Ping timeout: 246 seconds]
<s1341_>
site-packages is in nix-store and is read-only.
ddellacosta has joined #nixos
<qyliss>
s1341_: they should be to $out/${python.sitePackages}
<immae>
I cannot find how I can add a parameter with space in it in curlOpts in fetchurl, is there any indication how I could achieve that? (in the "builder.sh" it does `curl+=($curlOpts $NIX_CURL_FLAGS)` but I don’t know how the nix string is translated into a bash string in the first place)
<aforemny>
s1341_: Are your nixpkgs and home-manager sources in sync? Ie. are you using nixpkgs unstable and home-manager master, or both using the 20.09 releases, etc.?
avn has quit [Read error: Connection reset by peer]
justanotheruser has quit [Ping timeout: 272 seconds]
<aforemny>
s1341_: You might be able to work around that by setting `manual.manpages.enable = false;` in your home configuration. What commits are you on nixpkgs and home-manager? I might be able to check.
nhs has quit [Ping timeout: 260 seconds]
orivej has quit [Ping timeout: 240 seconds]
aveltras has quit [Quit: Connection closed for inactivity]
<aforemny>
For my NixOS configurations, I have forked nixpkgs and home-manager. For each repository, I have two remotes: origin = my fork and upstream = the original repository. I usually rebase my forks against upstream, so that my changes are on the top of the commit log. However, then I have to force push to my origin and I always feel very uneasy about that. What if I accidentally force push to the upstream repositories? How do you
<aforemny>
approach that? Can I prevent Git from pushing to upstream?
<aforemny>
s1341_: What was the issue/ solution?
<adisbladis>
aforemny: You could separate fetch/push
demize has quit [Quit: Do Androids Dream of Electric Sheep?]
demize has joined #nixos
stree has quit [Ping timeout: 260 seconds]
nhs has joined #nixos
dx_ has joined #nixos
tgys[m] has joined #nixos
<dx_>
hi, i need correct configuration so when turning on laptpo I login to terminal bash console. And then I do 'systemctl start display-manager.service' to start my GUI. And when I exit from GUI i go back to bash console.
Qwerky has joined #nixos
<dx_>
my prob is, when I exit in my fluxbox WM, I am thrown to a GUI login manager.
<etu>
Unaduda: If you refer to #114116, I'm a bit sceptical about it, especially the last comment by heroku-miraheze since that would mean that the contributor haven't actually built it on NixOS?
<etu>
Unaduda: But overall there's not much wrong with the code itself, but the commit history is not following guidelines.
<notgne2[m]>
uh, why do you even want it merged? you don't need anything upstream to use it, and it sounds like you can't use it either way anyway 🤔
<etu>
Unaduda: "maintainer-list: Add maintainer joaquinito2051" and "trafficserver: init at 9.0.0" would be the only two commits I'd expect in that commit log.
<shla>
hi, where can I find more information about this error: error: Package ‘ena-2.2.11-5.10.17’ in /nix/store/23qyhqsbs79xzpgfavp4q3hyi1dj6bcj-nixos-20.09.3246.d4189f68fdb/nixos/pkgs/os-specific/linux/ena/default.nix:35 is marked as broken, refusing to evaluate.
amk has quit [Ping timeout: 256 seconds]
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
waleee-cl has joined #nixos
gustavderdrache has joined #nixos
dev_mohe has joined #nixos
<supersandro2000>
shla: you or some package you want to use is depending on a package that got marked broken because it fails to build most likely
hiro98 has quit [Ping timeout: 272 seconds]
<shla>
where can I check build status of particular package?
<shla>
after it's merged, how long it will take to build the new package and propagate it
<shla>
?
Qwerky has quit [Remote host closed the connection]
<rmcgibbo[m]>
I think a few hours at least, but also it's only merged into unstable, so we'll need to do a backport PR to 20.09 if you want to use it in a prior release.
jimmiehansson has quit [Remote host closed the connection]
dev_mohe has joined #nixos
anderslu1dstedt has joined #nixos
chisui has quit [Quit: Connection closed]
mvwm has quit [Ping timeout: 240 seconds]
anderslundstedt has quit [Ping timeout: 240 seconds]
sangoma has quit [Ping timeout: 256 seconds]
zupo has joined #nixos
<ekleog>
Hmm… I suppose there's no way for nix to directly fetch data that lies behind GCS authentication?
zupo_ has joined #nixos
gustavderdrache has quit [Quit: Leaving.]
<raboof>
I'd like to recreate https://hydra.nixos.org/build/137470087#tabs-details - but when I check out nixpkgs c7d0dbe094c and run `nix-instantiate NIXOS_LABEL_VERSION=21.05pre271938.c7d0dbe094c nix-instantiate nixos/release-combined.nix -A nixos.iso_minimal.x86_64-linux` I get a different store path on different machines. any idea what would cause that?
davidv7_ has quit [Remote host closed the connection]
<dev_mohe>
raboof: What does `echo $NIX_PATH` print?
<dev_mohe>
disregard, shouldn't matter as you used a local path
davidv7 has joined #nixos
<codygman__>
Can I rerun the nixos hardware scan? I think with my new hardware that hardware-configuration.nix might change. Overall I'm trying to get my nixos machine with no internet to use my new usb wireless dongle.
<rembo10_>
~.~
davidv7 has quit [Client Quit]
<rembo10_>
~.
davidv7 has joined #nixos
ahmedelgabri has joined #nixos
zupo has joined #nixos
ahmedelgabri has quit [Ping timeout: 272 seconds]
<clever>
codygman__: just run nixos-generate-config again
stree has joined #nixos
<codygman__>
Alright. I think I might be able to just run `nix-env -iA nixos.linuxPackages.rtl8812au` to solve my full problem but thats TBD. Good to know I can just do that.
cole-h has joined #nixos
yoctocell has joined #nixos
<rmcgibbo[m]>
Is it possible to have a flake depend on a non-flake (like a github repo that just contains a default.nix)? Maybe that defeats the point of flakes?
cwslimy[m] is now known as cwsticky[m]
<infinisil>
rmcgibbo[m]: Yeah, just set `flake = false` for the input
<codygman__>
Does anyone here have experience or knowledge about downloading a nix expression to one machine and installing on a machine without internet by just copying the nix store via usb drive or something?
vidbina_ has quit [Ping timeout: 265 seconds]
<codygman__>
After doing `nix-env -iA nixos.linuxPackages.rtl8812au` the command seems to be hanging at something I've never seen before: `waiting for the big garbage collector lock...`
<codygman__>
Is this a concern and can I learn about it anywhere?
Qwerky has quit [Remote host closed the connection]
dev_mohe has quit [Remote host closed the connection]
dev_mohe has joined #nixos
<rmcgibbo[m]>
infinisil: that kindof works. thanks.
<jschievink>
is there any way to revert to a previous generation when using the Raspberry Pi bootloader (not uboot)? perhaps from another (non-AArch64) machine?
meh` has quit [Ping timeout: 240 seconds]
<infinisil>
codygman__: nix-store --export and --import is useful for transferring store paths
<infinisil>
(see man pages)
Qwerky has joined #nixos
tomberek has joined #nixos
<codygman__>
infinisil: Thanks!
<tomberek>
having problems with running nixos-container's. The `nixos-container start` files after 105 seconds due to some timer that i can't find.
<ekleog>
welp now it's “writing to file: Connection reset by peer”
* ekleog
discovers interesting new failure modes of the evaluator
<ekleog>
aaand looks like trying these again eventually makes things work, so… probably some evaluator or FFI communication bug? I couldn't see any earlyoom / oom killer messages so probably not that, but given I have no clean way to reproduce…
<{^_^}>
[nixpkgs] @SuperSandro2000 merged pull request #114121 → remove unused inputs, remove stale substituteInPlace, other cleanups → https://github.com/NixOS/nixpkgs/pull/114121
<ekleog>
on an unrelated matter: do we have any way to have the traceback to a warning? trying to figure out where I'm getting stdenv.lib messages from, and grep isn't very helpful
<cole-h>
ekleog: you could make stdenv.lib `throw`, which might give a location, I think?
<cole-h>
Usually happens because they're using argv[0] or something and the binary was `makeWrapper`'d
<cole-h>
makeWrapper moves the original binary to `.<original name>-wrapped`
i1l has joined #nixos
<tomberek>
infinisil: i'm starting a container with almost nothing in it.. Possible some service is timing out? I get "Startup finished in 31.143s". but every time at 105s:
<tomberek>
```
<tomberek>
Feb 23 14:38:51 tom systemd[1]: container@flake-test.service: Killing process 1196 (systemd-journal) with sign>
<tomberek>
Feb 23 14:38:51 tom systemd[1]: container@flake-test.service: Killing process 989 (systemd) with signal SIGKIL>
<tomberek>
Feb 23 14:38:51 tom container flake-test[980]: nixos login:
<tomberek>
Feb 23 14:38:51 tom systemd[1]: container@flake-test.service: Killing process 980 (systemd-nspawn) with signal>
<tomberek>
Feb 23 14:38:51 tom systemd[1]: container@flake-test.service: State 'stop-sigterm' timed out. Killing.
<tomberek>
Feb 23 14:37:21 tom systemd[1]: container@flake-test.service: start operation timed out. Terminating.
tomberek has quit [Killed (Sigyn (Spam is off topic on freenode.))]
<gchristensen>
at least it was not a kline
justanotheruser has quit [Ping timeout: 272 seconds]
<srk>
when you forget this is not a matrix chat window .. D
<cole-h>
x)
alexherbo2 has quit [Ping timeout: 260 seconds]
<srk>
jschievink: with u-boot you can adjust /boot/extlinux.conf by hand
<jschievink>
not using u-boot
<jschievink>
managed to do it by editing cmdline.txt by hand, seems to work well
<asbachb>
Hi. I'm struggeling setting a custom dns server for my wireless connection. I added the dns server to the nmtui connection but in /etc/resolvd.conf it has no priority.
shibboleth has joined #nixos
<jess>
gchristensen: sigyn does a kill and then a k-line
<gchristensen>
ack!
<tomberek>
regarding nixos-container from manual "This command will return as soon as the container has booted and has reached multi-user.target. On the host, the container runs within a systemd unit called container@container-name.service. Thus, if something went wrong, you can get status info using systemctl." It looks like multi-user.target is reached, but
<gchristensen>
jess: it isn't common we get someone pasting like that. would freenode staff prefer I come begging for forgiveness when it happens, kick out Sigyn unless we're seeing spam, or something else?
<tomberek>
lordcirth_: I get "Reached target Multi-User System. Startup finished in 31.154s." But the command still hangs on the host.
<cole-h>
gchristensen: https://github.com/freenode/Sigyn#requesting-sigyn-to-your-channel "If opped in your channel you can ask Sigyn to unkline an user, /msg Sigyn unkline <nick>, you have a dozen minutes to do so after the kill/kline, it only works if the user was banned due to abuse detected in your channel."
<lordcirth_>
tomberek, 30s seems a long time for a start?
<jess>
gchristensen: you can op up, then /msg sigyn unkline <nick>
<gchristensen>
oh cool
<cole-h>
gchristensen: might help?
* jess
nods
<gchristensen>
thank you both :D
<cole-h>
jess++
<{^_^}>
jess's karma got increased to 2
<cole-h>
hehe
<jess>
you can also just come find staff and ask, someone's usually around :)
bgamari_ has quit [Ping timeout: 256 seconds]
<gchristensen>
y'all are wonderful, thank you
zupo_ has joined #nixos
<{^_^}>
[nixpkgs] @SuperSandro2000 pushed to sagemaker « pythonPackages.sagemaker: add pandas for analytical feature »: https://git.io/Jt5kO
bgamari has joined #nixos
<tomberek>
lordcirth_: possible. I can root-login to it immediately, things like dhcp and some background stuff still goes for a few seconds... After the 30s is hangs for another minute until the command times out.
<tomberek>
hrm... i just tried with normal config, worked ok. The failure case seems to be when the nixos configuration is a flake.
zupo has quit [Ping timeout: 264 seconds]
<tomberek>
it starts fine, the flake env seems to work as expected (correct things installed and in path)
<lordcirth_>
I had a different issue with flake-based containers - I couldn't remove them. It said they were declarative.
<__monty__>
cole-h: Yes, but wrapProgram uses `exec -a $0` where $0 is the wrapper's name to set .program-wrapped's $0. This fails for me and I'm trying to understand why.
<tomberek>
lordcirth_: i may have fixed it with a networking.useDHCP=false;
<lordcirth_>
tomberek, that is recommended in general
slack1256 has joined #nixos
<tomberek>
how does one allow it outbound network access?
<jschievink>
how do I make a systemd user service auto-start for one specific user only?
<spinlock[m]>
jschievink: user services are user-specific from the get-go. If that user is you, you can do something like `systemctl --user enable appname.service`. If you prefer to do things via your nixos config, https://github.com/nix-community/home-manager is a convenient way to manage user services
<jschievink>
oof, no way to do this in the NixOS configuration?
<lordcirth_>
Because the point of hibernation was largely to be faster than booting. Booting is fast now.
<lordcirth_>
Most programs also save what you were doing
<ashkitten>
i have reasons to want this
<lordcirth_>
However, if you have swap > 1x your RAM, it should work
<lordcirth_>
What happens when you try?
<simpson>
ashkitten: It works same as any other distro for me; make sure that there's enough swap, and then `echo disk > /sys/power/state` to go sleep.
<simpson>
The kernel will tell you whether it thinks that this is feasible if you `cat /sys/power/state` to see the various options.
<spinlock[m]>
Anyone have any experience running **docker** containers from within **nix-shell**? I'm trying to write some "all-in-one" `shell.nix` that my colleagues can use to run our project. I'd like to start a docker container via `shellHook`, but (of course) the docker daemon isn't running.
<spinlock[m]>
The wiki suggests adding `virtualisation.docker.enable = true` to `/etc/nixos/configurationnix`, but is it possible to install/manage docker entirely via my `shell.nix`? It would be ideal if my colleagues didn't have to go through manual setup steps just for docker (they're quite lazy! lol)
<tomberek>
i'd guess you need to enable that setting, that would be the cleanest, this can be done via an overlay
<ashkitten>
how do i debug why it's not resuming from swap?
<ashkitten>
swap on xfs should work fine, right?
fendor_ has quit [Remote host closed the connection]
apache8080 has quit [Ping timeout: 246 seconds]
<spinlock[m]>
fair enough. My goal was to be able to say "put this `shell.nix` file in the project dir and run `shell-nix`". it's been quite the journey, but I've got everything going except for this docker bit 🤔
apache8080 has joined #nixos
<rmcgibbo[m]>
docker needs root privileges at some point, right? that's why it's gotta be enabled in configurationnix i presume.
fendor has joined #nixos
<spinlock[m]>
hmmm. that makes a lot of sense. It think you're right rmcgibbo. I understand that nixos has ways to let users to things that previously only root could do (install/run arbitrary apps, for example), but I support there are limits!
<spinlock[m]>
thanks for the advice!
<slby[m]>
If I nix search nixpkgs#searchterm with flakes enabled I feel like I am getting much less answers than before, am I doing something wrong?
<rmcgibbo[m]>
maybe there are other container-running tools other than docker that don't need root anywhere, not sure.
<shla>
nixos runs out of memory when rebuilding on t3a.nano
jkachmar has joined #nixos
<jkachmar>
is the `$NIX_PATH` behavior documented extensively anywhere? in particular, i'm trying to understand what exactly sets `/home/<username>/.nix-defexpr/channels` in my configuration
Cushingar[m] is now known as Cushingar[m]
<colemickens>
jkachmar: I think you're looking for nix-channel?
<jkachmar>
I'm pinning everything with Flakes and then setting my nix path with `nix.nixPath = lib.mkForce [ ... ]`, but I can't seem to remove the channel entry at all
<colemickens>
oh, that's strange. I force set nixPath to empty like that in my flakey configs and wind up with an empty NIX_PATH
<colemickens>
I wonder if something else in your shell config might be setting it?
<colemickens>
```
<colemickens>
❯ echo $NIX_PATH
<colemickens>
```
apache8080 has quit [Ping timeout: 256 seconds]
<jkachmar>
I elided it but it's really like `nix.nixPath = lib.mkForce [ "nixpkgs=${inputs.nixpkgs}" "unstable=${inputs.unstable}" ]`
<jkachmar>
but I'm seeing the channels entry show up regardless; maybe something in home-manager is doing this but that's the only other thing I can think of that would be touching my shell env
gustavderdrache has quit [Quit: Leaving.]
strikerlulu has quit [Ping timeout: 246 seconds]
<figgyc>
This is probably stupid and/or not the right way to do what I want, but is there a way to like build a derivation inline and get the path of it? like say I have a pkgs.writeTextFile function, that on its own just outputs <derivation /nix/store/whatever.drv>, how would i make it build and be just /nix/store/whatever/ as a path which I can set in an attrset for some package that expects a path (like ` configDir = ./config; `
<figgyc>
or whatever, but more programmatic) ?
<stevenroose>
Anyone experience with running NixOS on ZFS? I'm reading the page for that but it doesn't say what the boot.loader.grub.device value should be. nixos-generate-config somehow failed to detect the root partition as well, but I could fill that manually. But what should be the grub boot device? I have a zpool called rpool and the / is on the rpool/root volume.
<stevenroose>
boot.loader.grub.device = "rpool/root"; doesn't work
<cole-h>
Everybody I've talked to uses systemd-boot with ZFS
<stevenroose>
Should it perhaps be the plain /dev/disk/by-id/... path for the actual device?
<colemickens>
figgyc: uh, I'm pretty confused, the output of `pkgs.writeTextFile` *is* the out path.
<stevenroose>
Oh. Any reference on how to set that up? The NixOS on ZFS page doesn't talk about that and assumes nixos-generate-config does all that..
<cole-h>
figgyc: If you're doing it in `nix repl`, you can prefix it with `:b`
<tomberek>
stevenroose: i've used it. My current device is set to /dev/sda, but i've also used the /dev/disk/by-uuid....
<rmcgibbo[m]>
figgyc: do you want to store a directory containing a single file in the store rather than just storing a the file itself in the store?
<colemickens>
the grub config would depend on how you've set the disk up
zenguy- has joined #nixos
<tomberek>
:q
<zenguy->
/!\ this chat has moved to irc.crimeircd.net #pp /!\
<figgyc>
rmcgibbo, yeah probably
<cole-h>
stevenroose: `boot.loader.systemd-boot.enable` for systemd-boot
<jkachmar>
stevenroose: I've got a few systems configured with NixOS & ZFS; typically I make a separate EFI partition (either on the same drive or on a separate drive) and mount it to `/boot` in my `hardware.nix` declaration
<jkachmar>
e.g. on my dual boot desktop, I have my EFI boot partition, windows install, and NixOS swap on my NVMe SSD, and my zpool on a SATA SSD
<colemickens>
I do the same ^ and then just use systemd-boot since it's simpler / easier to configure / I like systemd (bootctl is nice, etc)
<stevenroose>
jkachmar: hmm yeah the wiki page mentions that option but it kinda makes it seem optional so I didn't do it. I'm gonna try tomberek's suggestion and use grub with the /dev/disk/by-id for the partition
<colemickens>
uh
<stevenroose>
hm mthat doesn't seem to work
<colemickens>
if you're EFI booting I wouldn't expect it to
<stevenroose>
so it an efi boot partition a requirement for systemd-boot?
<stevenroose>
I can quickly redo the setup
<colemickens>
We really, really need to know if it's GPT/MBR and EFI or not.
<jkachmar>
^
<colemickens>
Or yeah the section of the manual cole-h linked :)
<stevenroose>
it's GPT but without efi
<colemickens>
isn't that like the worst option?
<stevenroose>
I honestly don't know the difference and odn't know what to pick for that :D
<tomberek>
(oh,,, wait... stevenroose: i retract my statement ,that was for btrfs)..... my ZFS system uses device = "rpoot/root/nixos"
<colemickens>
I'd give a strong second to jkachmar's recommendation. I think that's a common setup.
<stevenroose>
k I'll redo gpt with efi. I think my favor for bios is that I had many laptops in the past struggle with gpt. But this is a server anyway
<tomberek>
fsType = "zfs"
<colemickens>
But I do know that some folks with ZFS use MBR partition so that they don't have to partition the disk.
<stevenroose>
tomberek: for grub? hmm for me the nixos-install refuses to take grub devices that don't start with a slash
<colemickens>
But I also think that's maybe cargo-culted knowledge and I haven't seen good evidence for justifying that setup over a normal GPT+EFI setup.
<stevenroose>
tomberek: oh that's the filesystem. what's the boot.loader.grub.device?
zebrag has quit [Quit: Konversation terminated!]
<tomberek>
i do have systemd-boot enabled, but it seems grub is running too
<stevenroose>
oh
zebrag has joined #nixos
<stevenroose>
kay you know what, I'll just start over using gpt and make an efi partition and maybe nixos-generate-config will pick it up automatically
<stevenroose>
thanks for all the suggestions, I'll keep you posted :)
<colemickens>
grub or systemd-boot this time?
<jkachmar>
My general recommendation when using NixOS + ZFS on systems that support EFI boot is: 2 GB (or thereabouts) `/boot` partition, 0 GB -> 8 GB of non-ZFS-backed swap space, and the remainder of the drive allocated for the zpool
<jkachmar>
In reality, I end up with a `vfat` `/boot` partition and the remainder of the drive allocated in a contiguous block for some over-complicated LVM stuff but I'm not sure I'd recommend that for most folks
<jkachmar>
I ended up with 2 GB of `/boot` because drive space is cheap these days and NixOS generates a lot of entries if you don't collect garbage frequently; the default amount of 256 MB or whatever seemed a bit small
<jkachmar>
For NixOS specifically, once you've got the boot stuff sorted out, I'd recommend making a dedicated dataset for the `/nix` store with auto-snapshotting disabled (since Nix can take care of that sort of thing itself, and I want `nix-collect-garbage -d` to actually clear out my old store entries) and `atime=off` (since the `/nix` store shouldn't need access time information)
<stevenroose>
Learned that if you have a zpool on a partition, you cna do `fdisk` and `g` + `w` to write a new table to the disk and the zpool will stay existing :F
tokudan has joined #nixos
prusnak has quit [Ping timeout: 264 seconds]
<{^_^}>
[hydra] @grahamc opened pull request #876 → Create an ephemeral PostgreSQL database and Nix store per test → https://github.com/NixOS/hydra/pull/876
<cole-h>
Maybe in a separate PR... I'm not familiar enough to create a test like that atm.
<colemickens>
jkachmar: nooo, that's too good to only be in a gist
__monty__ has quit [Quit: leaving]
<jkachmar>
colemickens: haha okay I'll make it a priority this week to at least get this stuff polished up into a proper blog post at least; then perhaps there will be a good path towards making updates to the NixOS wiki
<jkachmar>
All of the procedural filesystem stuff was easy to document, but for some reason I ran out of steam when it came to trying to explain my `configuration.nix`
<tomberek>
cole-h: i'm learning more about it
rsynnest has quit [Quit: Connection closed]
<tomberek>
cole-h: i'm also interested in adding more tests for packages (not NixOS services). There's the passthru.tests feature
civodul has quit [Quit: ERC (IRC client for Emacs 27.1)]
leotaku has joined #nixos
<jonreeve[m]>
Does `nix search` not work with flakes enabled?
stree has joined #nixos
<jonreeve[m]>
I used to be able to do, e.g. `nix search transmission` and get the name of the Nix package for Transmission. Now everything I try just says "cannot find flake 'flake:transmission' in the flake registries"
mruest has joined #nixos
Thra11 has quit [Quit: WeeChat 3.0.1]
mruest has left #nixos [#nixos]
<jkachmar>
does `nix search nixpkgs#transmission` work as expected?
<jkachmar>
jonreeve[m]: ^^
<jonreeve[m]>
Oh I guess it does. Maybe the syntax just changed
<cole-h>
That only finds the exact attr
<cole-h>
You probably want `nix search nixpkgs transmission`
<jkachmar>
aaa yeah that's the right syntax, sorry
<cole-h>
They both work
<cole-h>
But `nix search nixpkgs transmission` doesn't show exact matches
<jonreeve[m]>
Thanks!
<cole-h>
(Which is probably good when you don't know exactly what you're looking for)
<stevenroose>
hmm, so update on my adventure :D I tried all over with GPT, EFI partition on /boot etc and the install went all smooth, I disabled GRUB and enabled systemd-boot and the rebooted and the machine doesn't seem up. it's a dedicated server so can't see a screen, but can't SSH into it (I enabled SSH etc). so it seems that something must have gone wrong with booting
<samueldr>
it's the only service that is allowed to do so
<stevenroose>
it's so amazing about nixos that you can just check the config file for the openssh service :D
<samueldr>
because it would be... quite inconvenient otherwise
<Yaniel>
samueldr: you mean by default?
lordcirth has joined #nixos
<samueldr>
not sure if any service grew an option to open firewall ports, but yes, by default
<stevenroose>
samueldr: is that a special rule? or does the nixos repo just not accept services that do that
<Yaniel>
seeing how syncthing does have an openDefaultPorts setting
<Yaniel>
(which defaults to false)
<samueldr>
stevenroose: not a special rule as in firewall rule, but yes
<samueldr>
so any other service's `enable` option will not change firewall rules
<samueldr>
just saying it since it *could* be a head scratcher :)
<stevenroose>
samueldr: yeah I meant nix derivatoin validation rule or so, but ok it seems that it's just a repo enforcement rule and services are allowed to have a opt-in flag that opens the port
<samueldr>
stevenroose: social rule, yes
<samueldr>
convention
<stevenroose>
samueldr: do that might change with flakes etc if nixpkgs doesn't stay the only place to primarily get modules
lordcirth_ has quit [Ping timeout: 240 seconds]
<samueldr>
flakes could do whatever they like, just like you can do whatever you like in your config
<stevenroose>
Yaniel: so say it did boot and there was a network issue (it's a manually ocnfigured non-dhcp network so quite likely I messed up), where would I look for an indication of that?
<samueldr>
as far as nixos modules go, there is nothing "special" about a flake, no?
<samueldr>
so I'd say that a third party flake, if they want to be a good citizen, shouldn't change firewall options, except with specific options for the firewall :)
<Yaniel>
samueldr: technically no but flakes could definitely sprout different social groups with different norms
<samueldr>
Yaniel: yeah, my point exactly
yilin has joined #nixos
<stevenroose>
Yaniel: no I mean without being able to log in the system. or do you mean somewhere in /var/log/ whatever convention systemd uses to place log files?
<yilin>
Hello; is there an easy way to inspect the channel expressions?
<Yaniel>
stevenroose: that'd be /var/log/journal
<Yaniel>
but those aren't human-readable
<stevenroose>
Yaniel: eh, really? but well if they are not empty, then the system booted and it's probalby a network issue, right? as opposed to if the boot failed they should be empty, correct?