<clever>
one of my old routers claimed v6 support, yet the ISP didnt, and the router refused to stop claiming v6 support
<samueldr>
ar1a: it caused me some headaches :) e.g. CDNs thought I was in another country and gave me slow CDNs
<samueldr>
ar1a: though it also gave me the benefit of looking like I was geographically in the U.S. instead, on IPv6 enabled services
<samueldr>
(but the 360p youtube was annoying)
<clever>
samueldr: ive had reverse problems, my HE tunnel resulted in netflix thinking i'm in both canada and america at the same time
<samueldr>
wavewave: yours is "time"
<clever>
samueldr: so netflix just turns itself off
<Dagger>
the DNS client, rather than the network stack... but it sounds like the problem is more with fastly than broken ISPs
<samueldr>
clever: I think you had the *same* problem; not a reverse one :) except yours affected you in a more negative way
<samueldr>
wavewave: that 503 might be resolved now, or in a couple minutes
<clever>
samueldr: it originally didnt detect that, and i was able to watch the american lineup
<clever>
samueldr: but then it detected it, and i had to disable v6 in the tablet, to even use netflix at all
<wavewave>
samueldr: oh. somebody is working on it right now?
<samueldr>
clever: I legit thought pandora was available in canada until I tried to use it outside my home network :)
<samueldr>
wavewave: no, not really, those 503s are transient :(
camsbury has quit [Ping timeout: 245 seconds]
<clever>
samueldr: i only discovered it, because i was complaining about how startrek is just mia, then i could see it on the linux pc, but not on windows ....
<samueldr>
wavewave: at least, the few I had were
<wavewave>
hmm. so just run the same command again until it works?
camsbury has joined #nixos
<clever>
wavewave: that may do it
<samueldr>
I'll try getting hold of someone who can get hold of someone who can check into it
<samueldr>
but that's on a different time scale
camsbury has quit [Remote host closed the connection]
camsbury has joined #nixos
camsbury has quit [Ping timeout: 244 seconds]
slack1256 has quit [Remote host closed the connection]
luigy has left #nixos [#nixos]
Neo-- has quit [Disconnected by services]
Neo--- has joined #nixos
eadwu has joined #nixos
pie__ has joined #nixos
pie___ has quit [Remote host closed the connection]
<etbe>
I built and booted a kernel on 18.09 with the config options SECURITY_SELINUX and SECURITY_SELINUX_BOOTPARAM. Then I realised that I made a mistake and didn't want the bootparam enabled. So I tried to build a kernel with SECURITY_SELINUX_BOOTPARAM n but the build failed with error: unused option: SECURITY_SELINUX_BOOTPARAM_VALUE
<etbe>
Where is it getting the default values from? Looks like it might be /proc/config.gz. How do I get it to disable them? When I have SECURITY_SELINUX_BOOTPARAM n there is no SECURITY_SELINUX_BOOTPARAM_VALUE config option.
slack1256 has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @teto closed pull request #41472 → nixops: make overrides build instead of failing with 'nixops-_version_-py2-none-any.whl is not a valid wheel filename' → https://git.io/vh48l
butta has joined #nixos
rcshm has quit [Remote host closed the connection]
o1lo01ol1o has quit [Remote host closed the connection]
<kyren>
I'm recovering a system that I screwed up, and I'm trying to build a closure manually with nix-build, and (going off of the nixos-install man page) I'm using nix-build [-I <stuff>]... '<nixos>' -A system --no-out-link, and I'm getting: "expression does not evaluate to a derivation (or a set or list of those)"
<kyren>
I can't run nixos-install because my configuration is not in /mnt/etc/nixos/configuration.nix.. I could move it there if I can't figure anything else out but I would like to know how to do this manually
<clever>
kyren: you want '<nixos/nixos>'
<clever>
kyren: you can also symlink configuration.nix to the real cfg, using a relative link
vk3wtf has joined #nixos
work_ has quit [Quit: Connection closed for inactivity]
<kyren>
cool, thanks clever, the man page for nixos-install is potentially wrong though
pie___ has joined #nixos
<clever>
kyren: <nixos> refers to a channel called nixos, so it will return the entire nixpkgs set
<kyren>
clever: right, makes sense, I actually just figured that out since I needed <nixpkgs/nixos>
<clever>
nixpkgs is also setup as an alias to the nixos channel
<clever>
so <nixos> and <nixpkgs> are usually identical
<kyren>
right, it usually is, but in my case it wasn't, I could have also simply added a nixos alias of course
pie__ has quit [Ping timeout: 245 seconds]
lsyoyom has joined #nixos
<kyren>
wait, I get to do it correctly this time
<kyren>
clever++
<{^_^}>
clever's karma got increased to 104
o1lo01ol1o has joined #nixos
o1lo01ol1o has quit [Ping timeout: 255 seconds]
<{^_^}>
[nixpkgs] @hhm0 opened pull request #56397 → cuetools: symlink cuetag to cuetag.sh → https://git.io/fhbjq
<rodarmor>
srhb: I just noticed that `nixos-install` takes a `--channel` argument. This then gets passed to `nix-env <STUFF> --set $CHANNEL`, but I think it's expecting something like `/nix/store/8xbgvck01ajp25h3jgh73pr7mfi53pgi-nixos-18.09.2203.9bd45dddf81`, and not `nixos-unstable`
<srhb>
rodarmor: I didn't know about that argument and I've never used it. I know my above suggestion works :)
<rodarmor>
Okay, sweet, I will definitely go with your suggestion then
<rodarmor>
I've used nixos enough that most superficial stuff is pretty easy to understand and use, but when I get into the nitty gritty I really have no idea how anything works :P
<srhb>
rodarmor: Note that you will have to change nix-channel once you're in the final system
<srhb>
rodarmor: Heh, yeah :)
<srhb>
rodarmor: Maybe you've seen <nixpkgs> in a nix expression
<srhb>
rodarmor: That is a lookup of the nixpkgs entry in the NIX_PATH env var.
<srhb>
So by modifying that env var, you can control what <nixpkgs> is in nix exprs.
wfranzini has quit [Remote host closed the connection]
vk3wtf has quit [Ping timeout: 250 seconds]
iqubic` has joined #nixos
<rodarmor>
I'm like, vaguely aware that that's a thing, but still very unclear of the details. <nixpgs> resolves to a channel? in this case `channel:nixos-unstable` So I guess nixos knows how to resolve that into `https://nixos.org/channels/nixos-unstable`?
endformationage has quit [Quit: WeeChat 2.3]
iqubic has quit [Ping timeout: 268 seconds]
<rodarmor>
My relationship with NixOS is like that of a primitive human worshiping their vengeful god. Sometimes I get the incantations right and the harvests are plentiful, but sometimes I must mumble or something, and I get rekt by lightning strikes
cyphase has quit [Ping timeout: 268 seconds]
iqubic` has left #nixos [#nixos]
cyphase has joined #nixos
drakonis has quit [Quit: WeeChat 2.3]
jackdk has quit [Ping timeout: 272 seconds]
<slack1256>
Ok that goes to the topic roulette
iqubic has joined #nixos
slack1256 has quit [Remote host closed the connection]
hchchchchchchchc has quit []
abaiste^ has joined #nixos
orivej has joined #nixos
ibrahims has joined #nixos
<fusion809>
Hi folks, I've installed every package with telepathy in the name and I cannot find anything Telepathy-related in the app menus. Guessing I need a front-end for it?
palo1 has joined #nixos
<iqubic>
What DE / WM are you using?
<fusion809>
KDE
<fusion809>
I also have Rofi installed as I know KDE's app menus are not updated until you log out and log back in.
<fusion809>
In Rofi nothing Telepathy-related appears
<iqubic>
That's slightly worrying.
<fusion809>
I installed Telepathy packages as user and I cannot find a suitable binary in ~/.nix-profile/bin, there
vk3wtf has joined #nixos
<fusion809>
's telepathy-gabble-xmpp-console but it doesn't seem appropriate
<fusion809>
For launching Telepathy in a GUI, that is.
ibrahims has quit [Ping timeout: 255 seconds]
<iqubic>
How certain are you that the installtion was successful?
palo has quit [Ping timeout: 255 seconds]
palo1 is now known as palo
<fusion809>
Highly, no error messages occurred.
infinii has quit [Ping timeout: 245 seconds]
kvda has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<{^_^}>
[nixpkgs] @domenkozar pushed to master « nixos release: there's a wildcard protection now for release-* on github »: https://git.io/fhNve
dminuoso has left #nixos ["WeeChat 2.2"]
<sphalerite>
fusion809: empathy is one frontend for telepathy
rauno has joined #nixos
<sphalerite>
hm, I wonder how much effort it'll be to migrate this server to a new physical machine
<fusion809>
Thanks, I tried Empathy before installing Telepathy and it wasn't working, now it is.
<sphalerite>
(it's running on nixos and zfs, so that should make it relatively easy)
<fusion809>
Wasn't working as there was no back-end presumably.
<fusion809>
Nice combo, clearly data security is important to ye.
<adamantium>
sphalerite: so I upgraded last night to zfs unstable 0.8rc3 and then created an encrypted dataset after running "zpool upgrade", but then on the following reboot grub was unable to reach stage2, i'm guessing that the zpool upgrade enabled features not supported by grub (zfs on root) ?
<sphalerite>
adamantium: yes, grub doesn't support encrypted datasets I believe.
<adamantium>
ah
<sphalerite>
err, pools with encrypted datasets
<adamantium>
should have partitioned the disk with boot on ext4 or similar
<sphalerite>
yeah, or use linux as a boot loader :p
<sphalerite>
that would be a really nice feature for nixos
<adamantium>
i use linux as a bootloader for Emacs
_kwstas has joined #nixos
<adamantium>
works good most of the time, until yeeeesterday, (sorry again, channel)
fusion809 has quit [Quit: Leaving]
<sphalerite>
flokli: how's your systemd-boot parser thing going?
_kwstas has quit [Remote host closed the connection]
<sphalerite>
adamantium: ahahahaha
<etu>
adamantium: :D
<etu>
adamantium: Do you run EXWM? :)
<adamantium>
etu, yeah
<adamantium>
i mess with gnome and kde too though.
<etu>
adamantium: cool, I run exwm as well :)
_fusion809 has joined #nixos
orivej has quit [Remote host closed the connection]
<adamantium>
sphalerite: i should rewrite themelios, and add options for hybrid /boot on ext4 with encrypted dataset option
orivej has joined #nixos
<adamantium>
i guess that hibernation would be able to be added too
<sphalerite>
I'm always a bit confused by the relationship between the two
<sphalerite>
but I really like the idea of using linux as a bootloader
<adamantium>
what am i looking at
<adamantium>
just when nixos was weird enough already, you want this
<sphalerite>
u-root is a userspace for using linux as a bootloader
<sphalerite>
haha
<adamantium>
i've only heard of booting uefi stubs
<adamantium>
"linux as a bootloader" sounds confusing
ixxie has joined #nixos
<etu>
It was more confusing when there was linuxbios as well, I think that renamed to coreboot or did it go away? :p
<etu>
Yeah, it was renamed
<sphalerite>
adamantium: the idea is that your firmware loads a linux kernel directly, which then loads the actual OS kernel
<sphalerite>
etu: yep
<sphalerite>
adamantium: using kexec
vidbina has quit [Ping timeout: 255 seconds]
<adamantium>
huh. so it requires coreboot
infinii has joined #nixos
<adamantium>
sphalerite: Yeah. I don't get this! I think i'm getting old or something
infinee has joined #nixos
<iqubic>
How does one enbale horizontal scrolling on Nixos?
infinii has quit [Read error: Connection reset by peer]
<adamantium>
iqubic: need details. do you refer to the touchpad? You could mess with libinput, or, do you use a desktop environment, in which case you might set the option in a control panel
<iqubic>
I'm refering to the touchpad on my laptop.
<iqubic>
I'm using EXWM, so I don't think I have a control panel option.
<sphalerite>
adamantium: no, it doesn't require coreboot. Any old UEFI implementation can load linux.
<teto>
,locate bin/occ
<{^_^}>
Found in packages: ocserv
<adamantium>
iqubic: probably touchpad=$(xinput --list | awk '/TouchPad/ {print $7}') ; xinput --set-prop "${touchpad#id=}" 298 1 # should do what you want, you could autostart the command from exwm or .xsessionrc, or you could use a .nix file similar to mine probably
<iqubic>
except I get the output of `unable to find device`
<adamantium>
Does your touchpad work at all
<iqubic>
yes it does.
<adamantium>
are you using synaptics instead of libinput for some reason
<iqubic>
I might be. I'm not sure.
<adamantium>
for example, iqubic here is my libinput settings that works for my laptop https://ptpb.pw/5sVH/nix
<adamantium>
and horizontal scroll is enabled for mine with those settings by default
<adamantium>
so try that maybe
wfranzini has joined #nixos
<iqubic>
Oh. I am using synaptics.
<adamantium>
yeah stop that!@
<adamantium>
:)
<iqubic>
Why?
<adamantium>
it's fine if it works, but linux desktop ecosystem has moved toward libinput
<adamantium>
i do not know the technical reasons, why, though.
goibhniu has joined #nixos
<Ralith>
I think the technical reasons can be summed up as "it's bad"
<MayeulC>
AFAIK, lininput was needed for Wayland. A lot of effort was put into it, so that it would support most devices. Then someone realized it could work on X as well, and it soon became the default. It tends to be much better for touchscreens, non-synaptic touchpads, and weird/quirky inputs. Moreover, everyone will end up using it once Wayland is the norm.
Serus has quit [Ping timeout: 258 seconds]
<MayeulC>
Some people prefer Synaptics, notably for acceleration, though. But I haven't heard too may complaints lately.
<iqubic>
So what should I do here?
<Ralith>
I hear libinput goes to quite a lot more effort to do the right thing by default
<adamantium>
remove syntapics related config, too of course
<iqubic>
where do I put that?
<adamantium>
just steal the services.xserver.libinput = { ... } section
Serus has joined #nixos
Kakky has joined #nixos
<iqubic>
And put that into my configuration.nix?
<adamantium>
place it in the appropriate section of your configuration.nix file, and comment out any synaptics related configuration, if you have any. nixos-rebuild switch, systemctl restart display-manager and see if that fixes it
<iqubic>
cool
<etu>
adamantium: webpaste? :D
<adamantium>
etu: you got me pinned
adamt has joined #nixos
<monotux>
adamantium, so what happened yesterday with your irc client? :)
adamt is now known as Guest14940
<adamantium>
monotux: not a clue, ERC (emacs) somehow copied the entire #nixos channel buffer and sent it as a message
<adamantium>
Probably half the users now have me blocked and my nixos future is doomed
Makaveli7 has joined #nixos
<monotux>
haha
<monotux>
start having emacs display your active region
<monotux>
probably had an active region running and then you accidentally pressed yank
<adamantium>
the problem was i couldn't stop sending the messages, my bouncer received them, and i use a community bouncer so i couldn't shut it down, i also was unable to send it commands / the bouncer stopped responding to me, and i couldn't part the channel because of stickychan module
Guest14940 has joined #nixos
Guest14940 has quit [Changing host]
<{^_^}>
[nixpkgs] @marius851000 opened pull request #56401 → hackertyper: init at 1.1 → https://git.io/fhNvi
<etu>
adamantium: Hm? Just a guess since not many people seem to use ptpb.pw with language support, and since that's the top provider of webpaste... ;)
<monotux>
I'd probably have the same issue with my (web based) irc client
<{^_^}>
[nixpkgs] @xeji pushed commit from @bachp to master « gitlab-runner: 11.7.0 -> 11.8.0 (#56388) »: https://git.io/fhNvX
<adamantium>
Does anyone know how to make chromium use the gnome keyring?
<{^_^}>
[nixpkgs] @marius851000 opened pull request #56402 → ctrtool: init at 0.15 → https://git.io/fhNvy
<adamantium>
I have gnome keyring and pam working perfectly with ssh agent and gnupg agent, but even running chromium --password-store=gnome from the prompt does not work
<Taneb>
With lib.types, is there a way to express a function type?
Akii has joined #nixos
ubert has joined #nixos
fendor has joined #nixos
fendor has quit [Remote host closed the connection]
fendor_ has quit [Ping timeout: 244 seconds]
fendor has joined #nixos
dminuoso has joined #nixos
<dminuoso>
Hiya. I have installed citrix-receiver via nix-env, however I cant uninstall it: λ ~/ nix-env -u citrix-receiver; error: selector 'citrix-receiver' matches no derivations
<dminuoso>
`nix-env -q` lists that exact package.
<symphorien>
use nix-env -e
<dminuoso>
symphorien: Oh man, I feel silly now.
<dminuoso>
Thanks, I couldn't see it
ibrahims has quit [Ping timeout: 272 seconds]
lortabac has joined #nixos
qbiot has joined #nixos
<adamt>
Regarding nix-env, why is it that when installing packages one has to qualify them like nix-env -iA nixpkgs.vscode, while nix-env -e nixpkgs.vscode doesn't work? (I know nix-env -e vscode works)
<symphorien>
because nixpkgs could have been updated in the meantime
tobiasBora has joined #nixos
<tobiasBora>
Hello,
<symphorien>
the attribute name depends on the structure of nixpkgs
<symphorien>
the derivation name is intrinsic to the derivation
<symphorien>
once a derivation is installed, and nixpkgs update, nix-env only knows the derivation name of the installed packages
<tobiasBora>
I am running KDE, and I would like to enable knetworkmanager. So I added the line networking.networkmanager.enable = true;
<adamt>
So can't one imagine installing different packages, that results in the same name in the list of packages installed with nix-env?
<__monty__>
symphorien: Ah, TIL. I always figured nix-env should be able to get the derivation name from the attribute you passed.
<symphorien>
when the name is the same (modulo version) nix-env replaces the derivation instead of installing side by side
<tobiasBora>
and I added myself to the "networkmanager" group. After a reboot, I have access to network manager on the trailbar, unfortunately I cannot click on the button "configure network connections". When I click... nothing happend
<adamt>
Ah. Thanks, symphorien.
arjen-jonathan has joined #nixos
ixxie has quit [Ping timeout: 245 seconds]
<Taneb>
Is there something like "writeFileThenRunThisSedScript"?
<dminuoso>
symphorien: This is interesting, the citrix-receiver remains in my store even though I have removed it from all users. nix-store --gc is not cleansing it.
<Taneb>
I could do something with stdenv I guess?
<Taneb>
And runCommand
<symphorien>
dminuoso: it is still a dependency of your previous profile
<symphorien>
you can still rollback with nix-env --rollback
<dminuoso>
symphorien: I deleted my old generations
<symphorien>
then nix-store --gc --print-roots /nix/store/dsfsdfsdf will tell you the offender
<dminuoso>
{memory:147}
<symphorien>
a process has opened a file in this store-path
<dminuoso>
Oh, that is interesting.
<symphorien>
most likely citrix is running
clever_ has joined #nixos
sphalerite has quit [Ping timeout: 252 seconds]
<sb0>
I have a function that returns a derivation:
<fendor>
I am having regurarly problems with stack on NixOS that the PATH is modified in a way that a lot of executables can not be found. Does stack run in a pure nix-shell or something?
<symphorien>
sb0: I think callPackage adds the override attrbutes
<symphorien>
call package is for, well, packages, not attrsets :þ
<tobiasBora>
Nobody knows why I can't configure NetworkManager on KDE?
<sb0>
aah
<sb0>
symphorien: so callPackage can only be used when the file returns a single derivation?
<symphorien>
I don't know
<symphorien>
I think what you are doing happens in nixpkgs
<tobiasBora>
cannot build derivation '/nix/store/8vrrhqfwdlvr665j3k1ndvhm758gwzvm-nixos-system-myhostname-19.03pre170287.19eedaf867d.drv': 1 dependencies couldn't be built
<tobiasBora>
note that I just did a channel update to make sure everything is up-to-date
<{^_^}>
[nixops] @AmineChikhaoui pushed to master « auto-retry DescribeIamInstanceProfileAssociations api calls »: https://git.io/fhNUP
NickSeagull has joined #nixos
qbiot has quit [Remote host closed the connection]
qbiot has joined #nixos
ibrahims has joined #nixos
schjetne has joined #nixos
qbiot has quit [Remote host closed the connection]
qbiot has joined #nixos
jD91mZM2 has quit [Quit: WeeChat 2.3]
jD91mZM2 has joined #nixos
qbiot has quit [Remote host closed the connection]
qbiot has joined #nixos
nSeagull has joined #nixos
NickSeagull has quit [Ping timeout: 246 seconds]
Neo--- has joined #nixos
rprije has quit [Ping timeout: 244 seconds]
Neo---- has quit [Ping timeout: 259 seconds]
rauno has quit [Remote host closed the connection]
alter-schjetne has joined #nixos
<{^_^}>
[nixpkgs] @Mic92 opened pull request #56408 → treewide: use runtimeShell instead of stdenv.shell whenever possible → https://git.io/fhNTe
Neo--- has quit [Ping timeout: 244 seconds]
schjetne has quit [Read error: No route to host]
qbiot has quit [Remote host closed the connection]
qbiot has joined #nixos
init_6 has quit []
<Taneb>
If I'm defining a module which creates a systemd service, how can I make the service run as not-root? (preferably as a user specifically for this service)
timor has joined #nixos
<etu>
Taneb: You should be able to specify 'User = "username";'
<adamt>
Taneb: I don't have a one-liner handy for you, but you could define a host like you would with nixops or morph, and then just build the host with either tool
<Taneb>
symphorien: that's not the issue
<Taneb>
symphorien: the NixOS machine is specifically for services and I don't want to do regular development there
<johanot>
Taneb: I manage to work half a year with Nix on my Ubuntu laptop, before switching entirely to NixOS mysql
<johanot>
wow mysql = myself ..
<Taneb>
johanot: again, not the issue. This is at work, where we need to be using !NixOS for certain propietary software, but use Nix quite heavily
alex_giusi_tiri has joined #nixos
<Taneb>
We have a server running NixOS, and I am working on the configuration for the server, and want to be able to see if it works locally without copying everything to the server
<johanot>
You could build qemu vms and test with those them :)
<Taneb>
That feels wildly overkill to see if I've got any syntax errors
Ariakenom_ has joined #nixos
<Taneb>
(well, a bit more than syntax errors, but I'm hitting syntax errors often enough the way I'm doing things)
Kakky_ has joined #nixos
alter-schjetne has joined #nixos
isHavvy has joined #nixos
<johanot>
Taneb: So if you just wanna build the system, you could do something like: nix-build '<nixpkgs/nixos>' --show-trace -I nixos-config="$conffile" -A system
ub has joined #nixos
agander_ has joined #nixos
Mateon3 has joined #nixos
ggp0647` has joined #nixos
amosbird_ has joined #nixos
Piece_Maker has joined #nixos
rembo10_ has joined #nixos
regulus- has joined #nixos
<gchristensen>
sphalerite: need something from me?
<Taneb>
johanot: aha, that's (almost) exactly what I want
<gchristensen>
converting their company one person at a time
<edef>
good
<betawaffle>
how many have you got so far gchristensen?
<betawaffle>
(who?)
<edef>
i've been slimming my NixOS module for packet down
<gchristensen>
mmlb for example is here now
<edef>
the default image has an .. interesting config
<yorick>
edef: I was thinking of using fetchUrl on the metadata server :P
<yorick>
edef: what's interesting about it?
<edef>
yorick: doesn't work for me, i don't do evaluation on production servers
<betawaffle>
edef: everything at packet is slightly... interesting
<gchristensen>
betawaffle: I think also philip cristiano and tony maybe
<gchristensen>
edef: the default nixos image?
ramses_ has joined #nixos
<edef>
betawaffle: now you've piqued my interest q=
<edef>
gchristensen: yeah
<gchristensen>
edef: let's fix it, what is weird?
<edef>
gchristensen: idk, the whole thing where it hardcodes a tonne of server-specific info doesn't really work for me
NickSeagull has quit [Ping timeout: 246 seconds]
<edef>
gchristensen: there were some other nits, and the filenames are kind of odd
<betawaffle>
oh yeah this reminds me of a question i had for nixos folks... if i wanted to buy/build a machine that could just eat through nixos builds, what specs should i be optimizing?
<gchristensen>
betawaffle: my personal preference is higher single core performance over many cores
<edef>
i'm mostly sad that bonding on networkd is kind of shitty rn
<yorick>
it will bottleneck on single core performance if you're building similar stuff
<gchristensen>
as not all nixpkgs packages support parallel building
<yorick>
edef: I suppose you could rename the interfaces and bond it to if0,if1?
<betawaffle>
edef: let me share some pain points on networkd in nixos
<edef>
otoh we do parallelise well by building many packages at once
qbiot has joined #nixos
<edef>
betawaffle: sure, go ahead
<yorick>
edef: also I would really like a block storage module
<edef>
yorick: rename the interfaces why?
<gchristensen>
edef: I rebuild stdenv many times, which doesn't parallelize nicely, but otherwise yeah very true
<noonien>
hello folks
<yorick>
edef: so you don't have to deal with "predictable" interface names
<gchristensen>
hi noonien
<edef>
yorick: they're good and desirable
<betawaffle>
for .network files, the [IPv6AcceptRA] section doesn't seem to be supported, so i've had to use extraConfig. for .netdev files, the [Bridge] section doesn't seem to be supported
<edef>
yorick: if not for those i would be binding to PCI endpoint pats
<edef>
yorick: *paths
<yorick>
ew
<yorick>
edef: then how do you get it to work across different servers?
<edef>
yorick: i build images for very precise hardware setups
<yorick>
then why are you complaining that the nix is overspecified?
<edef>
yorick: that take care of everything, including partitioning
<betawaffle>
also, what's the deal with systemd.network.networks.<name>.DHCP, why not .dhcp?
<edef>
yorick: no, for *hardware setups*
<gchristensen>
yorick: she's saying the packet.net generated config is overspecified, whic his true
<edef>
yorick: give me a hundred of those servers and i will deploy precisely the same image to them
<edef>
yorick: the packet.net one contains the hostname
<gchristensen>
edef: please please plaese open a ticket describing how you use it
<noonien>
i wonder if nix would make a good pakage manager given fuchsias features
<gchristensen>
edef: so I can make it better for you
<edef>
betawaffle: the mappings are a bit of a mess, i personally just use the fooConfig ones because it means i can write from systemd.{link,net{work,dev}}(5) instead of having two manuals open
<edef>
betawaffle: i'd like to make them more obvious eventually, and the RA section is definitely a thing i can go fix in a bit
<edef>
betawaffle: same goes for bridge
<betawaffle>
edef: yeah, good point... also, you don't get the excessive newlines in the generated files that way :P
Makaveli7 has joined #nixos
<betawaffle>
does lib have something to convert duplicate newlines into one?
<simpson>
noonien: Maybe. Just like with other capability platforms like Genode, Nix sounds like a good idea on top, but Nix wasn't designed to be portable to a non-POSIX kernel, and that's likely going to complicate things.
<edef>
betawaffle: idk what the deal with those is, or why we generate them
<yorick>
edef: does packet just give you dhcp, then?
<betawaffle>
yorick: only in ipxe mode
<noonien>
simpson: i see, that makes sense
<edef>
yorick: on netboot setups, yes
<yorick>
edef: so how do you get the IP?
<edef>
yorick: but my intention is to just write a systemd-networkd drop-in on first boot
<betawaffle>
i wrote that, btw
<gchristensen>
betawaffle: that chainsed me so hard
<yorick>
betawaffle: how do you get the IP without having one? :P
<edef>
betawaffle: the iPXE support is really good by the way, thank you so much for that <3
<gchristensen>
yes :)
<betawaffle>
well, the installer gets DHCP, it has to for PXE
<betawaffle>
it can then call packet's metadata api
<edef>
i remember badgering zach over it endlessly, and you a few times
<yorick>
and then the DNS is extracted from that :(
<gchristensen>
edef: thank you for doing all the hard work on the badgering :P
<edef>
haha
<gchristensen>
edef: oh, and thank you for introducing me to Packet, 'cus whoa
<betawaffle>
edef: i didn't add the part where the user can supply ipxe, but i did all the backend stuff to actually use ipxe internally for packet's installers
knupfer has quit [Ping timeout: 246 seconds]
<betawaffle>
but that was my idea, they just didn't do it while i was there
<gchristensen>
yorick: DNS info is fetched from metadata
<edef>
letting the user twiddle the URL is kinda the small bit impl-wise
<yorick>
gchristensen: no, it's taken from /etc/resolv.conf during installation
<betawaffle>
edef: i'm currently trying to use nixos on a router, using only networkd, and nftables
<edef>
yorick: right, but i don't want to have that on prod servers
<betawaffle>
but now that they're building a proper R&D department... that's going to be on the wish list
<gchristensen>
betawaffle: I forget which, but that hardware which would enumerate ipxe on every. interface. and. wait. for. time. out.
<betawaffle>
yep!
<edef>
yorick: and i already have a rigorously maintained PGP setup
<edef>
yorick: so with only one weekend wasted on gnarly libgcrypt internals, i've now saved myself this minor pain in the ass forever
<yorick>
edef: I'll read up on ssh ca stuff, thanks
<edef>
i'm kind of sad it doesn't let me sign CAs recursively
NickSeagull has joined #nixos
<edef>
but i'll probably write the SSH CA for work to recognise my personal CA
agander_ has joined #nixos
<tobiasBora>
gchristensen: hum, this syntax for tests looks awesome: $machine->waitForText(qr/Alice Foobar/);
<gchristensen>
isn't that great? :D
<tobiasBora>
what is nix using, and how does it work?
<gchristensen>
its Perl
<monotux>
betawaffle, what is networkd supposed to do in that setup? I've run a router using nftables but no networkd
xkapastel has joined #nixos
<tobiasBora>
gchristensen: and it manages to run several VM in parallel?? amazing. What is the name of this library? (if I want to use it on other projects)
<betawaffle>
monotux: eliminate the scripted-ness? i want everything to be declarative. but i may be ahead of the curve a bit
<gchristensen>
tobiasBora: yeah, it is just NixOS's test framework
<gchristensen>
tobiasBora: not really a separate library
gustavopr has joined #nixos
<monotux>
betawaffle, ah, ok. I don't recall having to use anything but declarative stuff in my configuration.nix plus a nftables rules file
<edef>
the current networking setup is backed by a bunch of shell scripts
qbiot has quit [Remote host closed the connection]
<monotux>
it was only a test setup (used for a week or so) so I'm not even sure I still have the config anywhere
<betawaffle>
monotux: well, nobody else has either. i'm just a crazy person, probably
countingsort has joined #nixos
qbiot has joined #nixos
<monotux>
haha
<tobiasBora>
gchristensen: oh I see. The underlying system is qemu?
<betawaffle>
i honestly hate shell/perl scripts
<betawaffle>
wish i could eliminate them
<betawaffle>
but networkd still has a few rough edges
<betawaffle>
we'll see if i can get it to work completely
<edef>
i've been aiming to write more perl rather than less lately
<edef>
i've instructed $partner to prevent me from using `if` or `match` in awk and make me use perl instead
<betawaffle>
i'd like nixos a lot better if all the glue was Go, rather than perl/shell
<edef>
agree
<monotux>
you need help
<betawaffle>
monotux: probably
<edef>
although i'm kind of down for like, execline
<betawaffle>
what's that?
<countingsort>
hey, while were on the topic, is there a way to switch /bin/sh away from bash on nixos? im assuming its just an option, i couldnt find it in the option list or in the wiki though
<edef>
skarnet has this big suite of things that are basically continuation passing style unix utilities
<edef>
countingsort: there is an option!
<edef>
countingsort: it's undocumented
<edef>
countingsort: like explicitly undocumented
<countingsort>
whys that?
<countingsort>
does it break everythinf?
<edef>
countingsort: it's environment.binsh
<edef>
countingsort: there's also environment.usrbinenv which i added years ago
<monotux>
I'm guessing that it'll break pretty much everything
<countingsort>
thanks
<edef>
idk, i've been wanting to switch it to ash on my system
<edef>
or maybe dash
earldouglas has joined #nixos
<countingsort>
ill come back and tell you not to if wverything stops working :p
<monotux>
you'd be surprised how much bash is used everywhere
<edef>
betawaffle: so like, execline is basically a #! that just .. reads the file body and makes an argv from it
<edef>
monotux: you're only promised a POSIX shell on /bin/hs
<betawaffle>
interesting
<edef>
s/hs$/sh/
Makaveli7 has quit [Quit: Leaving]
<edef>
betawaffle: and just execs into that
<monotux>
edef, fair enough
<edef>
betawaffle: and every utility in the suite takes a continuation
<gchristensen>
oh dear tobiasBora don't go there, I'm goig to delete it
<hyperfekt>
countingsort: just curious, why would you want to switch away? is changing your own shell and the shebangs in your scripts not sufficient? or is there something about bash that you dont like for executing others' scripts?
<countingsort>
monotux: bash as /bin/sh does a bad job at being posix only, and some scripts rely on it though
<edef>
i'd prefer that we eventually move to a POSIX shell
<gchristensen>
edef: what, your /bin/sh isn't fish? :)
<countingsort>
hyperfekt: im writing a lot of posix sh scripts and i feel better about not accidentally using a bashism when i cant. also speed
<edef>
i added the usrbinenv one with the aim of eventually pushing for a minimal POSIX env(1)
<edef>
given that we cannot expect GNU env on macOS either
<edef>
or illumos, if we ever resurrect that
<tobiasBora>
gchristensen: indeed it works ^^
<hyperfekt>
countingsort: doesn't one usually specify the interpreter in the shebang anyway? excuse the basic questions, i haven't been using unix for long
<edef>
(idk if anyone but me and vizanto ever even ran nix-on-illumos)
nD5Xjz has quit [Ping timeout: 255 seconds]
<edef>
hyperfekt: yes, but some people do #!/bin/sh and expect bash
<edef>
i *think* that shouldn't override the existing entry
<gchristensen>
edef: debian's is dash ... we could switch maybe.
<pie__>
i think the build script for the nix utility assumed /bin/sh is actually a bash
<betawaffle>
what do i need to read to better understand how config merging works?
<gchristensen>
countingsort: use shellcheck and it'll make sure you're posix
<hyperfekt>
countingsort: oh, i see! so you want portable scripts by using sh but if you yourself have bash you wont notice if it isnt actually portable, is that correct?
<edef>
gchristensen: there's also the thing where it'd be neat to have those be static
<pie__>
it was using bashisms in the autotools script
<edef>
gchristensen: because /bin/sh gets brought into the sandbox last i checked
<pie__>
autotools explicitly deliberately only supports sh
<edef>
gchristensen: and right now it drags glibc in
<gchristensen>
ah
<countingsort>
gchristensen: i am. but im also lazy sometimes. also dash is still faster. (im not exactly sure this is a 100% rational choice, and not just preference because it feels cleaner or something)
<symphorien>
edef: there was an experiment to replace it by statically linked busybox, but it broke too many things and was reverted
<countingsort>
hyperfekt: correct
<edef>
symphorien: oh, huh
<edef>
symphorien: i missed that one
<edef>
i'd rather we break a bunch of stuff and then figure out how to fix it, tbqh
<countingsort>
^
<edef>
but expecting this is kind of why i've been saving any attempt at it for when i have more spoons
<countingsort>
~time to confuse people sayong that~
nSeagull has joined #nixos
<symphorien>
TIL
alter-schjetne has joined #nixos
NickSeagull has quit [Ping timeout: 255 seconds]
schjetne has quit [Ping timeout: 250 seconds]
grumble has quit [Quit: in lieu of the innuendo in the end know my intent though]
<edef>
symphorien: as far as i can tell the current binsh setting was introduced in b3f40405 and never changed since?
<edef>
symphorien: that was 2013
<symphorien>
the /bin/sh in the sandbox is not the same as the /bin/sh in the system
<hyperfekt>
is there anything to read about the state of implementation for the Nix UI draft? The linked issue is closed but a lot of cool things in there don't seem to have come about. https://git.io/fhNIY
<symphorien>
that was a nix level change iirc
alter-schjetne has quit [Ping timeout: 244 seconds]
grumble has joined #nixos
nD5Xjz has quit [Ping timeout: 246 seconds]
<edef>
chrootPaths/sandboxPaths was my next place to look, yeah
nD5Xjz_ has joined #nixos
<edef>
ah, i see, nix 2.0 moved the responsibility into nix
<gchristensen>
hyperfekt: many of those notes are done in Nix 2
<edef>
symphorien: nope, appears to be busybox just fine here
<hyperfekt>
gchristensen: I see. I'll go through the Nix manual and see what's actually there so I can replace my Nix-1-isms (which is funny because i started using Nix after Nix(OS) 2 was released). I guess what's not implemented has either been discarded or has an issue open.
<gchristensen>
hyperfekt: and Nix 2's UI is still "beta"
<gchristensen>
best to stick to Nix 1 for automation I think
<kai_w_>
how are .dtb and .dtbo files managed on nixos for a raspberry pi? I need to enable a 1-wire peripheral, and can't work out what to do.
<kai_w_>
"dtoverlay=w1-gpio,gpiopin=4" does nothing, presumably because /boot/overlays doesn't exist...
<gchristensen>
#nixos-aarch64 might be a lower-volume, more-focused place to ask that, kai_w_ :)
<symphorien>
I have a hard time finding how it was solved
<yorick>
symphorien: grahamc_ofborg
<symphorien>
?
<yorick>
symphorien: there's automation and checks on pull requests now
<gchristensen>
we always do a zero Hydro failures ticket before each release, though
ayerhart has joined #nixos
<{^_^}>
[nixpkgs] @xeji merged pull request #56409 → Document the addresses Alertmanager will listen on → https://git.io/fhNTL
<{^_^}>
[nixpkgs] @xeji pushed commit from @tomfitzhenry to master « Document the addresses Alertmanager will listen on (#56409) »: https://git.io/fhNId
<{^_^}>
#51858 (by YuMingLiao, 11 weeks ago, closed): error: value is null while a set was expected, at pkgs/top-level/default.nix:63:5 (when nixops deploy)
<therealwaphire[m>
yeah, I'm using grub and my ESP would be at /boot/efi
<therealwaphire[m>
I checked my other systems and the actual EFI files don't even reach 1 M
<ldlework>
perhaps I just need to move to a more recent nixpkgs?
nD5Xjz has joined #nixos
<therealwaphire[m>
currently, the efi files of systemd-boot on arch are taking 196K + 16 K
<ldlework>
or install nixops via override
<therealwaphire[m>
<freenode_Mic "therealwaphire: if you use grub,"> so I should be fine with a small ESP, right?
<adamt>
therealwaphire[m: You have a 1TB disk, right? Why not just use the default 512MB for the EFI partition, so you can change your mind later without mucking around with your partition layout? :P
agander has joined #nixos
<therealwaphire[m>
ah, no, right now I'm talking about my laptop with 512 GB NVMe, haha
<therealwaphire[m>
<freenode_ada "therealwaphire: You have a 1TB d"> thing is I want to give as much space as I can to ZFS
<therealwaphire[m>
I'm used to having a 512 MB ESP at /boot but want to try something else for a change :P
<samueldr>
therealwaphire[m: if you don't deviate from the default, every generation could add files to the ESP
<samueldr>
10MB for the initrd, 4MB for the bzImage (kernel) for me here (I'm including more modules than the default)
<therealwaphire[m>
right, I want to know if Nix adds files to the ESP or just creates entries to the bootloader
<samueldr>
it adds files to the ESP
<therealwaphire[m>
<freenode_sam "10MB for the initrd, 4MB for the"> yeah but uh, the kernel stuff will be at /boot which is gonna be ZFS
<samueldr>
(unless you configure it not to, but it will default to)
immae has quit [Quit: WeeChat 2.3]
<samueldr>
ah, then it wouldn't
<therealwaphire[m>
if GRUB is built with ZFS, would it not be capable to load generations by default
<therealwaphire[m>
yeah
<therealwaphire[m>
how did you get such a small bzImage btw
<samueldr>
you'll have to configure the EFI system mount point
<samueldr>
(I have a netboot image built to be bootable in the ESP, which acts like a recovery image; it's basically the nixos installer available in the boot menu)
_kwstas has quit [Client Quit]
<ldlework>
nm
<ldlework>
I think I understand
<therealwaphire[m>
<freenode_sam "(I have a netboot image built to"> ooo this is new
<samueldr>
it's not *that* convenient, but it's been useful twice, to date, always to fix things my dumb self did
agander has quit [Ping timeout: 255 seconds]
<therealwaphire[m>
haha
<therealwaphire[m>
does keeping multiple generations not help there?
<therealwaphire[m>
I mean, think about this
<samueldr>
against removing storage and misconfiguring it?
<samueldr>
:)
<therealwaphire[m>
lmao
<therealwaphire[m>
sorry
<therealwaphire[m>
I was gonna say, keeping a nix generation for rescue :p
_kwstas has joined #nixos
<samueldr>
I removed a drive, thus previous generations wouldn't really work as they were (mount point required for boot) and the new generation had a misconfiguration (used /dev/disk/by-path/uuid-here) which caused the same kind of issue :D
_kwstas has quit [Remote host closed the connection]
agander_ has quit [Ping timeout: 258 seconds]
_kwstas has joined #nixos
_kwstas has quit [Client Quit]
<adamt>
therealwaphire[m: You probably want a completely different system for rescue. The rescue system can also be wastly smaller than your normal system that way
<therealwaphire[m>
<freenode_sam "I removed a drive, thus previous"> damn, unfortunate
<therealwaphire[m>
<freenode_ada "therealwaphire: You probably wan"> true, guess I'll live dangerously then
<edef>
you can have multiple profiles
<samueldr>
anyways, it's not really important, the installation media is only one dd away
<therealwaphire[m>
yup
<therealwaphire[m>
I've thought a lot of times about keeping a rescue system that way but I always end up using it for something else
camsbury has quit [Remote host closed the connection]
domogled has joined #nixos
winem_ has quit [Remote host closed the connection]
camsbury has joined #nixos
iqubic has left #nixos ["ERC (IRC client for Emacs 26.1)"]
<{^_^}>
[nix] @edolstra merged pull request #2703 → Support parameters in S3 URLs → https://git.io/fhbQI
<pie__>
eh, why not make it less? ;P <samueldr> anyways, it's not really important, the installation media is only one dd away
<pie__>
hmmmm nixos netboot server image
hyper_ch2 has quit [Quit: Page closed]
<edef>
pie__: they work pretty well ^_^
<edef>
pie__: i install my servers for work with a ~300M EFI binary, that i boot over EFI HTTP boot
<pie__>
edef, hi im noob
<edef>
pie__: built by nix, contains a kernel/initrd that contain a working system and an install.sh
<edef>
pie__: which just partitions the lot and does nixos-install --closure
* pie__
looks up this efi http boot
<edef>
it's like PXE but less hassle
<pie__>
is that a thing so you dont have to be on the same subnet
<edef>
i still need DHCP in the deployment env unfortunately
<edef>
but i can set the URL through the LOM on most servers
<pie__>
im probably thinking about this all wrong but im guessing you dont have to have everything on a vpn at that point this way
<pie__>
though that sounds kind of sketchy
<edef>
i'm not running HTTP boot over the internet, fwiw
<Taneb>
I have a script which has a nix-shell shebang. I want the same script, but in the nix store either wrapped or patched to be able to find anything.
knupfer has joined #nixos
<Taneb>
Is there an easy way to achieve this?
<pie__>
edef, i didnt think you were
<edef>
although i think the firmware will probably do TLS
<edef>
but whether i trust that TLS stack to not be pwnt as hell..
<pie__>
edef, which is why "idk" what youre doing
<adamt>
edef: Lucky you. We have to include the redistributable firmware, so our image is way bigger than that :P
<edef>
the only hardware i need working is like, the SAS HBA and the network card
<siers>
What's the command to find file in any of the packages?
<siers>
an executable in this case
<pie__>
siers, nix-locate maybe?
<edef>
command-not-found knows where to find stuff
<srhb>
,locate bin readlink
<{^_^}>
Found in packages: busybox, coreutils, klibcShrunk, netbsd.stat, coreutils-full
<adamt>
edef: Our people responsible for hardware buys whatever HP/IBM/Dell server is the cheapest give some specs, so we have so much weird hardware :P
<pie__>
you can also pm the bot^
domogled has quit [Quit: domogled]
camsbury has quit [Ping timeout: 244 seconds]
<siers>
,locate bin cal
<{^_^}>
Found in packages: busybox, libuuid.bin, utillinux.bin
domogled has joined #nixos
<srhb>
siers: Avoid installing busybox globally though
<siers>
utillinux is probably the one
<srhb>
I'm not sure it's even still possible, but it's Bad. :)
domogled has quit [Client Quit]
<pie__>
edef, so how do you connect to the image server
<siers>
srhb, I've done that on an arch installation once. :)
<srhb>
siers: Sounds.. Fun.
<pie__>
i bet that made you real busy
<siers>
Yeah. :D But I have utillinux already! No cal thoguh.
ibrahims has joined #nixos
<siers>
haha
* pie__
shows himself out
<siers>
I've the /usr/bin/cal, but that's different.
<siers>
nix-index – that's what it's called
<srhb>
readlink -e $(which cal) on my system gives: /nix/store/mnnx1yny4b43djzpcbi5z7236bcsfr1b-util-linux-2.33.1-bin/bin/cal
<srhb>
So, a global util-linux should indeed do the trick.
<edef>
pie__: i use the netboot for our test rack, so it's just another server in the rack
<siers>
srhb, attribute name's utillinux
<siers>
:P
<siers>
it's bulding nix-index nooooo
<srhb>
Oh, right..
<srhb>
siers:It's cached for both 18.09 and nixos-unstable, so must be local overrides.
<srhb>
Argh, why is my weechat suddenly not autocompleting with a space after the <nick>:
<siers>
maybe an old channel? could it be?
<pie__>
edef, oh ok makes sense, so its not remote install
<srhb>
siers: Sure..
<siers>
btw, I've always wondered – for how long do the caches live?
<therealwaphire[m>
question, do you need `acpid` when you have `logind`?
<srhb>
siers: Forever.
<srhb>
siers: (So far...)
endformationage has joined #nixos
<siers>
srhb, How much is it now?
<srhb>
siers: Many, many years.
<srhb>
siers: It's a ginormous S3 bucket :)
<pie__>
lol
<siers>
Sure, but how many Gs or Ts?
<pie__>
srhb, now im curious
viric has quit [Read error: Connection reset by peer]
<srhb>
siers: I don't actually know that.
viric has joined #nixos
<pie__>
and how much of it is GHC lmao
<siers>
:D:D
<siers>
I'm not around s3 buckets much any more, but didn't it show its size in the AWS panel?
<therealwaphire[m>
<freenode_pie "and how much of it is GHC lmao"> lol
<srhb>
pie__: Shhh, don't make our cache go away...
<pie__>
D:
<srhb>
siers: Yes, I don't have access to that though. :)
<siers>
It would sure be interesting to find out :P
fendor_ has quit [Ping timeout: 245 seconds]
<srhb>
In the last 12 days, it looks like we grew it by 1.6 TiB.
<siers>
srhb, % ls $(nix-build '<nixpkgs>' -A utillinux --no-out-link)/bin => col fdisk fsck getopt hexdump mount script umount whereis write
<siers>
no cal!
<srhb>
siers: how about utillinux.bin? :)
<siers>
still no! but what is that
<srhb>
siers: It's there.
<siers>
ok, sorry, wait
<srhb>
It's a multi output derivation, which has a bin output.
init_6 has quit []
<siers>
What's the code to get to that derivation?
<siers>
> (import <nixpkgs {}).utillinux
<{^_^}>
undefined variable 'nixpkgs' at (string):218:10
<srhb>
nix-build '<nixpkgs>' -A utillinux.bin
<siers>
okay, not going to spam the channel
<siers>
is that correct? that doesn't work for me
<siers>
is it a linux thing? i'm on osx
<siers>
yeah, looks like it
<srhb>
siers: Oh, that might be.
<srhb>
siers: In that case, busybox version might not be that crazy after all.
<srhb>
I wonder if we should keep it in -- was it bothersome to build?
stepcut has joined #nixos
<srhb>
Oh, great, it's npm now...
<ddellacosta>
not strictly a NixOS question but: why should I use udisks vs. writing some udev rules for mounting drives? It's not clear to me why I need both systems, and udev seems more general
<alex_giusi_tiri>
lejonet: I think I may have gotten ceph's WITH_TESTS=ON working; WITH_SYSTEM_BOOST=ON and WITH_SYSTEMD=ON seems to have worked too; used boost 1.69.0, g++ 8.3.0
<jonreeve[m]>
Anyone here use NixOS on a laptop? I'm getting bad battery drain in suspend mode.
schjetne has quit [Ping timeout: 246 seconds]
<emily>
jonreeve[m]: which laptop?
<alex_giusi_tiri>
jonreeve[m]: was it the same with another distro/os?
<jonreeve[m]>
It's the new Lenovo Yoga C930.
<jonreeve[m]>
I haven't tested it with other distos, really.
<srhb>
jonreeve[m]: Are you sure it's actually suspended? It seems surprising to me that there would be any difference at all.
<srhb>
iff it's suspended, that is.
<jonreeve[m]>
All the signs of suspense (?) are there.
<jonreeve[m]>
Black screen, flashing power button, unresponsiveness
<srhb>
Okay, weird, I don't know then.
<Taneb>
I still need to figure out the issues with my laptop, which, tbf, I couldn't get working properly with Ubuntu either
<jonreeve[m]>
I read somewhere that there is more than one suspend setting for systemd
<jonreeve[m]>
That is, suspend2RAM and one other
<jonreeve[m]>
And one, for my laptop, is really battery-draining for some reason
<jonreeve[m]>
I've been trying to find the appropriate option for it, but no luck yet.
schjetne has joined #nixos
<jonreeve[m]>
So far, I've tried `powerManagement.enable` and using `powertop`.
<srhb>
I only know of suspend to ram and hibernation (suspend to disk, which is a real shutdown)
<krav1>
srhb: I think maybe it just uses npm to build and we can snatch if from a deb or something, trying to fix it is a nightmare it's full of compoents downloading other components. I originally thought the dashboard server was node, but I think it might actually still be python.
krav1 is now known as rkav
<alex_giusi_tiri>
jonreeve[m]: were you able to try with other OSes? how did it behave with them?
<goibhniu>
jonreeve[m]: I believe that's a more common problem alright
o1lo01ol1o has joined #nixos
<therealwaphire[m>
<freenode_srh "I only know of suspend to ram an"> How come there is no nix attribute for logind's Suspend then hibernate?
<jonreeve[m]>
There's apparently sus2idle and sus2ram:
<jonreeve[m]>
alex_giusi_tiri: I have an ubuntu USB, but I haven't really tried using it for more than an hour at a time, so I don't know how battery drain works on other distros
<goibhniu>
have you tried the latest kernel already?
<samueldr>
but since both a recent, maybe a good starting point
<rkav>
srhb: yes, I think the dashboard is a rather recent improvement, but I concur. I could give it another shot tomorrow, if it's just the frontend js then it might be fine to steal from a binary package.
<jonreeve[m]>
> [jon.reeve](https://matrix.to/#/@jon.reeve:matrix.org): you could switch to linuxPackages_latest
<jonreeve[m]>
How would I do that? Is it a channel thing?
<{^_^}>
error: syntax error, unexpected $end, expecting ')', at (string):219:1
<srhb>
rkav: You've done a lot, I can take a look too. :) By the way, sorry for picking it up so late. For the record, I didn't get notified when you removed the WIP tag, so I completely missed it. (I don't think notifications ever happen from that)
schjetne has quit [Ping timeout: 255 seconds]
<rkav>
srhb: I'm just happy to contribute something, thanks for your feedback!
<alex_giusi_tiri>
would WIP mean "work in progress"? :)
<NinjaTrappeur>
Hey, I find myself using a combination of readFile and writeText quite often when I want to save a local file into the store when writing derivations. I was wondering if there was a function (either builtin, either in nipkgs) doing that in one call?
<srhb>
alex_giusi_tiri: Yep.
<srhb>
ninjatrappeur: Just refer to the file.
ibrahims has quit [Ping timeout: 250 seconds]
<srhb>
ninjatrappeur: As in, foo = /path/to/your/file, or, in a string, "${/path/to/some/file}"
<alex_giusi_tiri>
srhb: Thanks. :-)
<NinjaTrappeur>
mmh, fair enough. Thanks srhb
<srhb>
ninjatrappeur: If your reluctance is because you want to rename it or something before adding it to the store, there's builtins.path
<srhb>
Or, what I perceive to be reluctance :-)
<mdash>
jonreeve[m]: I have a new lenovo X1C6, it had this problem
<NinjaTrappeur>
It's not reluctance, I wasn't aware that pointing to a file was implecitely copying it in the store
<srhb>
NinjaTrappeur: Ah :)
<NinjaTrappeur>
srhb, I just read the writeText implementation and realize how wrong I was :)
<goibhniu>
mdash: did you manage to fix it?
<srhb>
NinjaTrappeur: It's quite useful! :)
<mdash>
jonreeve[m]: Look at dmesg output and see if 'ACPI: (supports S0 S3' is in there
<mdash>
jonreeve[m]: The fix was for Lenovo to put out a BIOS revision that added an option for enabling ACPI S3
knupfer has joined #nixos
stepcut has quit [Remote host closed the connection]
NickSeagull has quit [Ping timeout: 246 seconds]
<mdash>
jonreeve[m]: and then for me to flip the 'sleep mode' setting under 'Power' in BIOS from 'windows 10' to 'linux'
<tokudan[m]>
i've got a strange phenomen on 18.09 since a couple of days with tab completion in bash with ssh. assuming srv01 and srv02 are known and i enter "ssh s<tab>", I expect bash to complete everything up to "ssh srv0". this is working fine. hitting tab twice more should give me a list of possible completions (listing srv01 and srv02 on a new line). this doesn't work anymore. instead bash inserts two literal tabs at the current
<tokudan[m]>
position.
NickSeagull has joined #nixos
<srhb>
tokudan[m]: Try in a pure shell with bashInteractive?
<srhb>
therealwaphire[m: Unsets most of the environment
<therealwaphire[m>
I see
<tokudan[m]>
what surprises me is that the completion from s<tab> to srv0 works, but as soon as there are multiple options, the completion fails
<srhb>
tokudan[m]: Yeah, it _sounds_ like some completion script is barfing.
<tobiasBora>
Hello,
<therealwaphire[m>
guys, so like there is a system configuration at `/etc/nixos/configuration.nix`, can I also have a user configuration that adds / modifies things like zsh and tmux?
<tokudan[m]>
it does work for commands though, for example c<tab> lists lists call the command starting with c: cal, cancel, cat, ...
<srhb>
,home-manager therealwaphire[m
<{^_^}>
therealwaphire[m: home-manager allows you to manage your per-user configuration similarly to NixOS's configuration.nix: https://nixos.wiki/wiki/Home_Manager
ibrahims has joined #nixos
stepcut has joined #nixos
<mdash>
and in fact you can put them in the same file
<tokudan[m]>
i have little knowledge about bash completion, is there some place I could start?
<tobiasBora>
I'd like to configure a VPN, however for now networkmanager does not allow me to configure it (no idea why), so before the issue is fixed I was thinking to use services.openvpn.servers to add my vpn to nix. Any idea if it's going to work, or conflict with network manager?
<srhb>
tobiasBora: It'll work. You'll juts be managing the systemd jobs manually.
<therealwaphire[m>
damn, these guys have it all figured out lol
<tobiasBora>
srhb: what do you mean? I need to start them or they will start automatically?
<therealwaphire[m>
what do you do for options that are not present as nix attributes?
<srhb>
tobiasBora: You can do either.
<therealwaphire[m>
for example, a package or something in emacs
<srhb>
tobiasBora: The point is that managing the systemd jobs will be what enables/disables your vpn.
<therealwaphire[m>
once nix touches a configuration file, is it advisable to then manually modify it?
<srhb>
therealwaphire[m: There are nix options for that, so poor example :-)
<ldlework>
how long would you guess that takes to make it into the unstable channel?
<mdash>
therealwaphire[m: you wouldn't do that, yeah
<srhb>
therealwaphire[m: They will be imported to the store, which should be read-only
<therealwaphire[m>
wdym
<tobiasBora>
srhb: oh, I missed the full options like .autostart. Will try that, thank you!
<srhb>
ldlework: Between a few days if things are going very smoothly to a week and a half, if things are going poorly
<ldlework>
srhb: from now?
<srhb>
ldlework: Things are going smoothly now, so usually a day or two.
<therealwaphire[m>
no what i mean is, see I have a local tmux conf with home manager right? so then I want to add something which isn't present as a nix attribute
<therealwaphire[m>
what can I do in that case?
<srhb>
ldlework: By the way, *which* unstable channel?
<srhb>
therealwaphire[m: Usually you can do things like home.file.".tmux.conf".source = /path/to/your/source/file
<srhb>
therealwaphire[m: At the point in time where you do home-manager switch, the file will then be copied into the nix store (read only) and your ~/.tmux.conf will be a symlink to that store path
<srhb>
therealwaphire[m: Read the boxes at the top. Then: nixpkgs has a different test set that must succeed before the channel is advanced
<srhb>
therealwaphire[m: In other words, both channels represent the "master" git branch, but at different points in time. The nixos-$foo channels generally have a lot of system related tests and only advance once those pass.
<srhb>
You can see which tests for each channel in the rightmost column of that page.
<therealwaphire[m>
I see
<therealwaphire[m>
so, the attribute that you have enabled
<therealwaphire[m>
allowUnfree
<therealwaphire[m>
is that limited to the `nixpkgs` channel?
<srhb>
No, it's not.
<srhb>
There's a conflation of terms here, unfortunately. :)
<srhb>
Each of the channels are all "nixpkgs" regardless of their name.
<therealwaphire[m>
ohhh
<therealwaphire[m>
right
<therealwaphire[m>
that got me confused
<srhb>
Understandable.
<mdash>
jonreeve[m]: If the kernel can see S3 then it should work
<mdash>
jonreeve[m]: at least, that was the issue I faced, the BIOS just didn't advertise that sleep state
<therealwaphire[m>
since `nixos-unstable` would have a lot of system related packages, I assume `nixpkgs-unstable` would be a better choice if I only want mostly updated apps
<srhb>
therealwaphire[m: nixpkgs-unstable doesn't care whether systemd, for instance, is horribly broken.
<therealwaphire[m>
I am
<srhb>
therealwaphire[m: So it might hose your system with no fanfare.
<therealwaphire[m>
ooo
<infinisil>
ldlework: See the description of the config option
<therealwaphire[m>
so the safe option is to include `nixos-unstable` as a secondary channel, right?
<srhb>
therealwaphire[m: nixos-unstable, on the other hand, does care, so you're less likely to break everything.
<srhb>
therealwaphire[m: Um.. Depends what you are trying to achieve really.
<therealwaphire[m>
oh, awesome!
<therealwaphire[m>
err, let's say I want git 2.21
<ldlework>
infinisil: services.znc.config? I don't see that option...
<therealwaphire[m>
which is not in nixos-18.09 yet
<{^_^}>
[nixpkgs] @BenSchZA closed pull request #56413 → nixos/jupyter: Fix documentation example for `jupyter.kernels` → https://git.io/fhNmR
<srhb>
therealwaphire[m: It's not in nixos-unstable either.
<betawaffle>
does anyone know how to ensure nftables doesn't try to start until a *specific* network interface is created? (like a vlan interface, for example)
<therealwaphire[m>
as far as I understand, my option is to add nixos-unstable to my list of channels and get the updated packages from there
<infinisil>
ldlework: Ah yeah, that list on the web page is on the stable version
<srhb>
therealwaphire[m: This is basically what you're asking.
<therealwaphire[m>
yes!
<srhb>
therealwaphire[m: I can't in a short timespan explain all the ways this can be achieved though, since the way we can override things gives us a whole programming language at our fingertips :)
<srhb>
therealwaphire[m: So examples are as varied, almost, as the people in this channel.
<edef>
betawaffle: no nftables knowledge, sorry
<therealwaphire[m>
totally understand
<betawaffle>
but networkd/systemd knowledge?
<betawaffle>
i'd be the same with iptables
<therealwaphire[m>
you've been really helpful as is
<betawaffle>
it'd*
knupfer has quit [Ping timeout: 244 seconds]
<edef>
betawaffle: i assume it's gated on systemd-networkd-wait-online.service
<edef>
betawaffle: you can confgure which interfaces have to be up for that
<betawaffle>
right, but i've got vlan interfaces that are created by networkd
fendor_ has quit [Ping timeout: 246 seconds]
<betawaffle>
and my firewall rules depend on those interfaces
<edef>
betawaffle: interfaces default to RequiredForOnline=yes
<edef>
betawaffle: generally you want firewall rules set before the interfaces come up, though
<lejonet>
srhb: You mean the module error it has during startup?
<betawaffle>
right, but i need the interfaces to be *created*, then the firewall, then come up
<srhb>
lejonet: Yeah.
<lejonet>
alex_giusi_tiri: cool :D
<srhb>
lejonet: I figure it's harmless, just noisy.
<lejonet>
srhb: yeah, that is my take on it too. My guess is that it enumerates all the modules, disabled or not, during startup
<srhb>
I might get time to try and build it properly tomorrow..
<srhb>
Because otherwise we definitely should be conscientious and add it to release notes.
<edef>
betawaffle: hmmm
<srhb>
That it's missing, I mean.
<{^_^}>
[nixpkgs] @BenSchZA opened pull request #56415 → nixos/jupyter: Fix documentation example for `jupyter.kernels` → https://git.io/fhNYI
ikitat has quit [Ping timeout: 246 seconds]
<lejonet>
srhb: definitively, it should be noted that its missing
ayerhart has quit [Read error: Connection reset by peer]
<edef>
betawaffle: apparently you can do WantedBy=sys-subsystem-net-devices-br0.device
<lejonet>
srhb: I'm gonna go nuts and push ceph-13 to my VM host, what could possibly go wrong? ;)
ayerhart has joined #nixos
<edef>
betawaffle: (in your unit that configures nftables)
<srhb>
lejonet: Nothing! Clearly! :D Let us know how it goes
<edef>
betawaffle: not sure how to stop networkd from configuring the interface, though
<lejonet>
srhb: will do
<betawaffle>
edef: where'd you find that? that's like exactly what i was looking for
<sphalerite>
gchristensen (or anyone else): can I run the installer with the daemon mode thing fully-automatically?
<infinisil>
ldlework: Are you using my module rework?
<sphalerite>
(nix installer)
<ldlework>
Yeah, prettttyy sure
<gchristensen>
pipe yes to it and have passwordless sudo
<ldlework>
I used to have a fork of it in my repo but I'm trying to update it to use the one in unstable
<sphalerite>
perfect, thanks!
<infinisil>
ldlework: Yeah so the legacy module required you to set this in `confOptions.passBlock`
<gchristensen>
you may not even need to pipe yes to id, sphalerite
the-kenny has quit [Ping timeout: 272 seconds]
the-kenny has joined #nixos
<ldlework>
infinisil: are you saying you think i'm actually using the legacy one?
<infinisil>
ldlework: There is the option `useLegacyConfig` you can use to disable all legacy settings, then you should set the pass block in the `config` option (as indicated by the example)
<ldlework>
infinisil: that worked
<ldlework>
thanks
nSeagull has joined #nixos
domogled has quit [Quit: domogled]
NickSeagull has quit [Ping timeout: 255 seconds]
<Taneb>
How is the nixpkgs-channels repo updated? What picks up that a build's succeeded on Hydra?
<edef>
for some reason opening links by ctrl-clicking works like, 50% of the time
<edef>
i should steal that wipefs -a for my netboot server bootstrap
<therealwaphire[m>
@clever: amazing stuff
<clever>
edef: nixos will also activate all lvm's, so you must `vgchange -a n` or it will fail to delete those partitions
<therealwaphire[m>
Going through it rn
<edef>
clever: i don't use LVM
<clever>
edef: if any pre-existing lvm exist on the disk your trying to nuke
<edef>
in fact i've been contemplating overriding pkgs.lvm2 to null
<clever>
such an override would prevent nixos from activating them, which also works, lol
<edef>
trying to whittle down the size of my images is hard, at this point it seems the bulk of it is just .. all of the things adding up
<edef>
rather than anything individually being particularly big
ub has quit [Quit: Leaving]
lejonet has quit [Ping timeout: 245 seconds]
<clever>
edef: not-os is another option, just start over from basically nothing
<edef>
yeah, i've been considering it, though specifically i want to continue working on s6-based things
<therealwaphire[m>
So you've combined shell and nix into one
<therealwaphire[m>
Which is what I wanted to do from the start
<edef>
i was kind of sad you disappeared in London before i figured out you were around
<therealwaphire[m>
How do you go about using it?
sigmundv has quit [Ping timeout: 250 seconds]
<clever>
therealwaphire[m: justdoit.nix is a nixos module, you add it to the imports section of an "installer" configuration.nix, and also set options like kexec.justdoit.nvme = true;
<clever>
therealwaphire[m: and when the installer is built, it will generate a justdoit bash script, that lands in $PATH at runtime
<edef>
nikesOS
<therealwaphire[m>
umm
<therealwaphire[m>
suppose I want to do it from a live boot
<therealwaphire[m>
I don't use discard while mounting though
<therealwaphire[m>
`fstrim.timer` is sufficient
<therealwaphire[m>
as ArchWiki said but with ZFS, I read somewhere that I don't even have to do that
<lejonet>
Yeah, I usually ignore discard regardless, but that is mainly due to the fact that on most NVMe devices I have, I run disk encryption
pie_ has quit [Remote host closed the connection]
rfold has joined #nixos
<therealwaphire[m>
oh that's cool
pie_ has joined #nixos
<therealwaphire[m>
how much of a performance hit there is with encryption?
ldlework has quit [Quit: co'o ro do]
<lejonet>
Highly depends, but its not enough to be noticable in normal usage scenario even on SATA SSDs
gagbo has quit [Ping timeout: 250 seconds]
<{^_^}>
[nixpkgs] @Ericson2314 opened pull request #56421 → pinentry: Use lib.string.{enable,with}Feature to avoid duplication → https://git.io/fhN3z
ldlework has joined #nixos
<lejonet>
therealwaphire[m: http://dpaste.com/0ACCXD7 (this is a Samsung 960 Pro, through 2 layers of disk encryption (don't ask why 2 layers, old habit that I just haven't changed :P ))
pie_ has quit [Ping timeout: 255 seconds]
<srhb>
lejonet: Why 2 layers tho.
<therealwaphire[m>
<freenode_lej "Highly depends, but its not enou"> oh I've always worried about the performance imact
<therealwaphire[m>
oh I've always worried about the performance impact
<lejonet>
srhb: xD
<lejonet>
srhb: bad srhb, bad! :P
stepcut has joined #nixos
<srhb>
:)
<lejonet>
therealwaphire[m: unless you have very bandwidth requiring loads, you'll not really notice encryption on modern hardware (modern here meaning you have AES-NI instructions on your CPU and you run AES for the crypto)
<ldlework>
my server resize worked like a charm
Neo-- has quit [Ping timeout: 268 seconds]
<lejonet>
even tho it was a looong time since I ran it on spinning rust disks (mechanical disks) I would dare say that even on mechanical harddisk you wouldn't really notice the encryption layer with normal, desktop-y usage
<ldlework>
Does anyone know how to massage the nginx options to provide mime types?
<siers>
qyliss, qyliss^work, hi, did you see the "ruby: export generic builder" changes yet?
<siers>
By the way, can I become a @NixOS/user/siers? :)
markus1189 has joined #nixos
<lejonet>
srhb: even more fun because its connecting to a 12.2.0 cluster :D
wolfmantjg123 has joined #nixos
<srhb>
lejonet: Funtimes.
<srhb>
lejonet: Well, I think we should take one last stab at getting dashboard in, then call it good enough whether we succeed or not.
<lejonet>
srhb: wasn't the main problem with the dashboard that it wants to pull in all kinds of npm bs?
<srhb>
Apparently yes.
<lejonet>
(as in stuff through npm)
<srhb>
Maybe there's a way to tweak the build to make it not do that. or maybe we can inject it through a regular build in another "package" and inject it later on.
<lejonet>
The latter is probably the best way to solve it
<srhb>
I hate their build instructions... "./install-deps.sh"
<lejonet>
Haha yeah, the build instructions for ceph historically have never been useful or descriptive :P
<wolfmantjg123>
Hey guys yesterday I was able to install nixos on my laptop, but my the dpi is all messed up so I see the words tiny, do you guys know the command to fix this problem?
<wolfmantjg123>
therealwaphire[m: thanks I’ll check that out
<therealwaphire[m>
wolfmantjg123: the above should fix most problems
<therealwaphire[m>
look at GTK and QT variables if you want to tweak it more to your liking
ikitat has joined #nixos
<wolfmantjg123>
therealwaphire[m: ok, sounds good
<{^_^}>
[nixops] @AmineChikhaoui opened pull request #1106 → introduce a persistOnDestroy option for s3 buckets → https://git.io/fhN37
<lejonet>
srhb: I'm gonna make dinner, then I'll see about diving into at least enumerating what node modules that dashboard needs, hopefully they aren't many
<{^_^}>
[nixpkgs] @pSub pushed 2 commits to install-llpp.inotify: https://git.io/fhN35
<srhb>
lejonet: I also feel like that's not what's causing the backend (python) error, which krav also suspected
<srhb>
lejonet: I can't help tonight, I'll check it tomorrow. Enjoy dinner. :)
freeman42x has joined #nixos
<lejonet>
srhb: do you have the error easily pastable somewhere? I suspect you guys are completely correct on that
<srhb>
Yeah, sec.
<lejonet>
worst case I just run the testcase and get it from there :P
<{^_^}>
#27183 (by 0xABAB, 1 year ago, open): ZFS dataset with specified non-legacy mount point does not automount
amosbird has joined #nixos
alex`` has joined #nixos
<therealwaphire[m>
the last reply's by me who's doing this with a kind of a hack :P
<srhb>
therealwaphire[m: Regardless, you're taking away hints from NixOS if you really do this. It's probably not going to be great. And legacy mounts work just fine :)
<{^_^}>
[nixpkgs] @Mic92 merged pull request #56417 → glibc: fix cross builds on Darwin → https://git.io/fhNOm
<srhb>
therealwaphire[m: (There's a ton of zfs users in here, disproportionately so)
<therealwaphire[m>
but I want my aesthetic `zfs list` to be pretty
<srhb>
You can still have that ;-)
<srhb>
Just name them according to their mount path!
<therealwaphire[m>
shows up as legacy though
<srhb>
True.
<gchristensen>
legacy is beautiful!
<srhb>
legacy is a horribly opinionated naming of them
<srhb>
It should be called "standard"
<therealwaphire[m>
that's where my problem is
<srhb>
:-)
<therealwaphire[m>
hahaha
<therealwaphire[m>
I don't know, man
<therealwaphire[m>
I wanna have that luxury real bad
<therealwaphire[m>
no real reason
<gchristensen>
you could write a patch to rename `legacy` in the output
<therealwaphire[m>
so yeah, can I make nix ignore the root mount?
<srhb>
therealwaphire[m: That's not going to stop me from trying to dissuade you.
<therealwaphire[m>
xD
<therealwaphire[m>
I know, man
<therealwaphire[m>
I buy into the nix philosophy 100%
<therealwaphire[m>
everything but this
<srhb>
therealwaphire[m: Like, the more info you give NixOS, the smarter it can be about your system. if you take this information away from it at build time, and make it a runtime dependency, NixOS knows less and can thus make less powerful decisions about your system!
<therealwaphire[m>
normally this wouldn't be a problem for me, imo
<srhb>
See? That's bloody convincing. :-)
<therealwaphire[m>
it's just this one thing though...
* lejonet
makes a ZFS wrapper for therealwaphire[m that turns legacy into "nixos STARPOWER!"
<srhb>
Okay, well, I also don't know where this particular footgun -- er, option, is located...
<therealwaphire[m>
man stop doing this to me, haha
<therealwaphire[m>
have to change my whole setup if I listen to you xD
<srhb>
:-)
dgarzon has joined #nixos
<therealwaphire[m>
I understand what you're saying, man
hyperfekt has joined #nixos
<gchristensen>
she's worth listening to =)
<lejonet>
Indeed
<therealwaphire[m>
but everytime I create new datasets, I have to do `nixos-generate-config`, no?
<therealwaphire[m>
didn't know you were female, srhb
<therealwaphire[m>
sorry there
<gchristensen>
I only do that for a few important ones: /, /home, /nix. the rest I let magically mount
<therealwaphire[m>
anyways
<srhb>
therealwaphire[m: No worries, and thanks :)
<dgarzon>
Hi, I am running into some compatibility issues between Darwin and Linux and openjdk. While the Darwin distribution has `*/bin/javac` on it’s path, the Linux one does not. I am using nix-shell with openjdk, and when I look at the derivation path the binary for javac is missing. Is there any way to get it?
<{^_^}>
[nixpkgs] @xeji pushed commit from @WilliButz to master « grafana: 5.4.3 -> 6.0.0 (#56412) »: https://git.io/fhNs7
<therealwaphire[m>
hmm, nah, I'll live on this hack until the functionality is merged into core :P
<hyperfekt>
Anyone know how to get the parent directory of a path in a variable? 'pathvar/..' does not work.
<srhb>
"${path}/.." ?
corpix[m] has joined #nixos
<srhb>
path + "/../" ?
<srhb>
It's all a bit ew.
NickSeagull has joined #nixos
<hyperfekt>
well in either of those the /.. would be a string not a path modifier it seems to me
<srhb>
hyperfekt: Yes. Though, paths are just fancy strings in some sense..
<srhb>
But you do lose an existence guarantee no matter how you do it.
<hyperfekt>
oh, of course. I don't even need to turn 'a/b' into 'a' in Nix if the tools used to build the new derivation are just going to handle 'a/b/..' for me to achieve the same
<srhb>
True! Much better :)
<tilpner>
> builtins.dirOf /foo/bar/quux
<{^_^}>
/foo/bar
<srhb>
dgarzon: This sounds wrong to me. How are you concluding that it's missing?
<hyperfekt>
tilpner: That's it, thank you! :D
Thra11 has joined #nixos
<kreisys>
Is there a way to make nix-build fail immediately if a dependency isn't available in binary cache? I want to write a script that would search for a nixpkgs revision to pin in a shell.nix file that recently started trying to build stuff.
<srhb>
samrose: Probably something trying to use $HOME
<samrose>
srhb: is there a way around this $HOME issue?
<srhb>
samrose: It much depends on the package in question. You may just need to set it to the build directory, but the fact that it's trying to muck with git sounds like there's more to the story.
<srhb>
samrose: Like, it's trying to fetch something outside of a fetcher.
<therealwaphire[m>
one question, if I do legacy mounts, parent properties are still applied right? compression, xattr, atime and such
<samrose>
srbh: yes it is a cargo file trying to use git to fetch something
<srhb>
samrose: Normally that's not allowed.
<samrose>
hmmm I guess I should bring this into my expressions somehow
<srhb>
therealwaphire[m: Yep, that's only related to the dataset hierarchy
<therealwaphire[m>
cool, cool
<samrose>
srhb: I wonder if I try to use carnix on this stuff if that might help
<abbafei[m]>
just want to notify the powers that be that last I checked, `flightgear` is broken in 18.09. There is pull #49617 which likely fixes this in unstable (testing now), but it seems this has not been backported.
<lejonet>
srhb: it would seem like the reason why cherrypy returns "unknown" for version is because it cannot import pkg_resources (it does a try-except for __version__ = pkg_resources.require('cherrypy')[0].version)
mkoenig has quit [Ping timeout: 245 seconds]
dbmikus_ has joined #nixos
drakonis has joined #nixos
jabranham has joined #nixos
domogled has joined #nixos
mkoenig has joined #nixos
<lejonet>
if I do a nix-shell -p python2Packages.withPackages(ps: [ ps.cherrypy ]) I get its version with __version__ tho
drakonis has quit [Read error: Connection reset by peer]
<ottidmes>
I did replace your fixed nodejs vs for the standard version 8, because I did not want to build nodejs just for this
<lejonet>
gaah, brainfart, how do I enter into a nix-shell "during" a build of a package?
<ottidmes>
samrose: are you on nixos-unsable? I am on 18.09
<symphorien>
lejonet: I would just patch __version__ to be a hard coded value but correct value
wolke23 has joined #nixos
lassulus_ has joined #nixos
<lejonet>
symphorien: yeah, I'm thinking about that, to get it to shut up, just sed something in, but I want to do some basic debugging just to see if its something stupidly easy to fix that is causing it
<samrose>
ottidmes: I am on 18.09
abaisteabaiste has joined #nixos
<{^_^}>
[nixpkgs] @pukkamustard opened pull request #56430 → nixos/guix: init at 0.16.0 → https://git.io/fhNGS
<samrose>
ottidmes: 18.09.153843.aabc61049c0
<samrose>
ottidmes: I am running this as nix-build rather than nix-shell
lassulus_ is now known as lassulus
<samrose>
ottidmes: nix-shell does work, but I am trying to build this as a package that is destined for my hydra server
<samrose>
so i am doing standalone nix-build for now
drakonis has quit [Read error: Connection reset by peer]
drakonis has joined #nixos
<ottidmes>
right of course, $HOME is set in nix-shell, should have known
<mog>
i want to do something dumb via nix-shell. i deploy to some machines from a nix environment. the machines are completely dynamic so the keys and ips never match so when im in nix-shell --pure environment i want ssh to have StrictHostKeyChecking=no UserKnownHostsFile=/dev/null enabled for all calls to ssh
Thra11 has quit [Ping timeout: 240 seconds]
<ottidmes>
mog: cant you just write a wrapper script that exposes $out/bin/ssh with those settings? maybe you need buildEnv to make it work, but that approach should work
<mog>
i cant just pass it at cli as some of the commands and scripts are calling ssh for me
<sphalerite>
mog: what about the ssh client config?
<mog>
so i dont see how to point ssh config to something other than .ssh/config
<mog>
i dont want my whole environment insecure
<sphalerite>
mog: also, would you consider using CheckHostIp=no and not using the settings you mentioned, for better security, if the hostname/key mapping doesn't change?
<lejonet>
symphorien: annoying part is if I get into a python shell inside of the build and import cherrypy, cherryp.__version__ has the correct one
<mog>
it all changes
drakonis_ has joined #nixos
<mog>
its random ec2 machines that get spun up and down
<lejonet>
cherrypy*
<sphalerite>
mog: aww damn. I'd guess you could set $HOME to point ssh elsewhere
<mog>
sphalerite, i did try that but its still using the ssh-agent thats running global so old version
<mog>
or thats what im thinking happened
work_ has joined #nixos
<sphalerite>
ssh-agent isn't involved in checking host keys
<sphalerite>
AFAIK
<mog>
it was cool to have $HOME set to something other than home, i dont think id ever done before
<ottidmes>
samrose: first I have to finish my dinner, but unexpectedly I did get: /nix/store/khgviadj4gxrk9702l6w7rslrnfkq7h5-holochain-rust succesfully build while using nix-build
<samrose>
whoa cool!
<samrose>
ottidmes: it turns out I was on the wrong version of holochain-rust, so I am updating these repos
<samrose>
and trying again
<gchristensen>
[grahamc@Petunia:~]$ nixos-version
<gchristensen>
19.03.170684.07e2b59812d (Koi)
<samrose>
chances are if you had found a solution ottidmes it will be relevant
drakonis has quit [Ping timeout: 258 seconds]
statusbot has quit [Remote host closed the connection]
statusbot has joined #nixos
<sphalerite>
gchristensen: um, shouldn't there be a "beta" in there? D:
<sphalerite>
did I do something wrong?
drakonis_ has quit [Read error: Connection reset by peer]
<samueldr>
I'm guessing the wording could have been "the first letter, in alphabetical order", but then at Z either we branch-off out of the 26... uh... usual letters
<samueldr>
or start back at A
<gchristensen>
Ape'
<lejonet>
Start going through nordic characters? :P
<sphalerite>
We've got some time until then, and we can discuss how to proceed in an RFC :D
<cransom>
you could letter up front. aardvark is easy. i'm at a loss for any 'bb' animals though.
<ottidmes>
samrose: all I did was: nix-prefetch fetchFromGitHub --owner samrose --repo conductor, to get the latest rev+sha256 and update the fetchFromGitHub in derivation.nix accordingly, then I updated the cargoSha256, I also replaced your fixed nodejs with pkgs.nodejs-8_x, but I doubt thats relevant
<cransom>
*double letter
<edef>
i vote we move to кириллица after we've exhausted those
Glider_IRC_ has joined #nixos
drakonis_ has quit [Read error: Connection reset by peer]
<samueldr>
cransom: bad news B.10. Release 13.10 (“Aardvark”, 2013/10/31)
<edef>
haha
drakonis_ has joined #nixos
vk3wtf has quit [Ping timeout: 264 seconds]
<cransom>
that was some real inconvenient thining back in 2013.
<gchristensen>
lol
<cransom>
*thinking.
<edef>
so with two release a year i guess we have 13 years of letters
<samrose>
ottidmes: thnx
<edef>
27.03 will be when we run into trouble
drakonis_ has quit [Read error: Connection reset by peer]
drakonis_ has joined #nixos
Glider_IRC__ has quit [Ping timeout: 245 seconds]
<johanot>
"Y2.2703K"
<symphorien>
we could have the name start with a lowercase letter
<symphorien>
this is greater than an uppercase one iirc
<{^_^}>
[nixpkgs] @veprbl merged pull request #56080 → openafs_1_8: Include support for Linux 4.20 → https://git.io/fhdMB
<sphalerite>
I wonder what people will use when we get to ~
<edef>
oh, i'm mis-mathing
<sphalerite>
because as we know from IBM, the alphabet goes abcdefghijklmnopqr~stuvwxyz
<symphorien>
?
<edef>
nevermind
<samrose>
ottidmes: interesting this did not make a difference for me
<samrose>
oh wait
<edef>
or, yes, 26.03 will be the last one we have a letter for
<samrose>
ottidmes: how did you update cargosha256?
_kwstas has joined #nixos
wedens has quit [Quit: Connection closed for inactivity]
<samrose>
I see ottidmes you put a fake sha256 in there, build and then get the result
dbmikus__ has joined #nixos
<edef>
lib.fakeSha256 is there for when you need one, fwiw
adamt has joined #nixos
adamt has joined #nixos
adamt has quit [Changing host]
dbmikus_ has quit [Ping timeout: 272 seconds]
domogled has quit [Quit: domogled]
<petersjt014>
anyone here have access to the arm server?
<gchristensen>
what's up, petersjt014?
<petersjt014>
I've been trying to compile ghc for armv7 and my vm is grinding at it pretty hard
<gchristensen>
eek, armv7
<petersjt014>
im trying to build cachix for arm
<gchristensen>
armv7, specifically?
<petersjt014>
ghc seems to be the main roadblock
<petersjt014>
I'm building it for an armv7 vm so the vm can build it for armv6
<gchristensen>
oh dear
<petersjt014>
mmmmhm
<petersjt014>
it's the fastest way i have
<gchristensen>
I suppose upgrading to armv8 is not much faster... :P
<petersjt014>
prob wouldnt be, cpu is the bottleneck
<petersjt014>
im capped to 2gb of ram and its hardly even tapping out
<gchristensen>
(right, but Hydra builds aarch64, armv8 already so you wouldn't need to build at all -- it was mostly a joke, suggesting you could just decide toreplace armv6 with armv8))
<petersjt014>
like...maybe i guess? but last time i tried aarch64 ghc was marked as unsupported, and it isnt under armv7
<petersjt014>
because reasons i guess
<samueldr>
(gchristensen, I'm possibly wrong, but I think armv8 as a target relates to 32 bit, at least, according to uname and such tools)
<petersjt014>
hmm. i *swear* last time i got a 'marked as unsupported; refusing to evaluate'
<petersjt014>
weird
johnnyfive has quit [Quit: I quit now.]
stepcut has quit [Remote host closed the connection]
hamishmack has quit [Ping timeout: 245 seconds]
<lejonet>
srhb: in an off chance that something behaves differently on python3, I'm trying that :)
<eyJhb>
Dose anybody have som good foo tricks for a program that accepts a input after you ahve run it. But does not read the input if you do "cat test | ./bin", "./bin < test", etc.? Anybody that might have some good tips?
arjen-jonathan has quit [Ping timeout: 258 seconds]
<clever>
petersjt014: i also looked at ghc on arm (aarch64 to be specific), and it fails to build if you specify more then -j1
<clever>
petersjt014: its due to a lack of NCG, and bugs in the llvm backend
<clever>
eyJhb: strace it?
<clever>
int isatty(int fd);
<clever>
eyJhb: is it using this syscall?
<petersjt014>
good to know
<eyJhb>
2 secs, docker is not happy with strace
ayerhart has quit [Remote host closed the connection]
kvda has joined #nixos
hyperfekt has quit [Ping timeout: 256 seconds]
<eyJhb>
clever: it uses read(1,
* edef
tilts head
<eyJhb>
I always forget all the piping tricks and stuff
<petersjt014>
always be pipin'
GryphOne has joined #nixos
<petersjt014>
play your cards right and you dont even need no variables :sunglasses_face:
<petersjt014>
(also lotsa xargs)
halfbit has quit [Ping timeout: 250 seconds]
eadwu has joined #nixos
<GryphOne>
Hello. I am new to this. I am stuck out of the gate with ---- oh no! - Please do not run this script with root privileges. and adduser username gives me
<GryphOne>
adduser: Please enter a username matching the regular expression configured via the NAME_REGEX[_SYSTEM] configuration variable. Use the `--force-badname' option to relax this check or reconfigure NAME_REGEX.
drakonis has joined #nixos
<GryphOne>
Ubuntu 18.04.1 on linode
<petersjt014>
this is the NixOS channel--it's a different distro (confusing name I know). you may wanna try #ubuntu
<adamt>
Maybe he's installing nix on ubuntu? (never installed nix manually, so dunno)
vk3wtf has joined #nixos
Ariakenom has quit [Quit: Leaving]
<petersjt014>
I've installed it on macos and didnt need to do anything like that
<petersjt014>
no permission errors that i can recall
<adamt>
I'll just go back into hiding then.
drakonis_ has quit [Ping timeout: 244 seconds]
kvda has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<samueldr>
curious what NAME_REGEX is configured as under ubuntu
<dmj`>
clever: I found a cool trick
<petersjt014>
buuut that was nearmly a year ago so idk
<symphorien>
GryphOne: you can edit NAME_REGEX in /etc/adduser.conf
<dmj`>
clever: journalctl cat <name>.service
<etu>
dmj`: I usually pipe that through cat to get rid of the usage of the pager :D
<symphorien>
GryphOne: the needed names are in the form nixbld1, nixbld2 and so on
<dmj`>
etu: ok I see you
guest3333 has joined #nixos
<dmj`>
etu: very nice
<callahad>
So, my nixos-rebuild fails due to a collision between btrfs-progs and dosfstools... any ideas for how I can resolve this and get my laptop updated? https://github.com/NixOS/nixpkgs/issues/56432
<{^_^}>
#56432 (by callahad, 3 minutes ago, open): Collision between btrfs-progs and dosfstools, prevents nixos-rebuild
<etu>
dmj`: if you mean `systemctl cat` ofc :)
guest3333 has quit [Client Quit]
<petersjt014>
I think there's an override feature that is relevant here
kvda has joined #nixos
<petersjt014>
,search override
<petersjt014>
hmm
<petersjt014>
how do i used bot
<petersjt014>
,help override
<symphorien>
,overlays
<{^_^}>
override: Use `,` to list all commands, `,foo = Foo!` to define foo as "Foo!", `,foo =` to undefine it, `,foo` to output "Foo!", `,foo somebody` to send "Foo!" to the nick somebody
<{^_^}>
symphorien: Did you mean overlay?
<{^_^}>
Overlays look like `self: super: { foo = ...; }`. Use the self argument to get dependencies, super for overriding things and library functions (including callPackage). More info: https://nixos.org/nixpkgs/manual/#sec-overlays-install
<tilpner>
callahad: If you're desperate for a dirty solution, override either package globally to remove the colliding file
<edef>
you can also just modify their .meta.priority
dbmikus has joined #nixos
<symphorien>
callahad: I would say nixpkgs.overlays = [ (self: super: { btrfs-progs = lib.hiPrio super.hiPrio; }) ];
<callahad>
symphorien, tilpner: Cool, I can locally override it and I should be good, but this seems like it might be a pretty common occurrence... how does the NixOS community tend to resolve these kinds of breaks-the-build issues?
<callahad>
(Like, how can I help out in a way that makes this better for everyone else, rather than just a local hack on my machine?)
<{^_^}>
[nixpkgs] @matthewbauer pushed 3 commits to staging: https://git.io/fhNnY
<callahad>
(And are these kinds of breaks common in general, or just when tracking the unstable channel?)
<symphorien>
in your case, I would say that it depends on the use case.
<gchristensen>
matthewbauer[m]: here now?
<tilpner>
callahad: First you need to check if the completions are identical. If they are not, you'd need to figure out some way to have both coexist and contribute that to nixpkgs
<symphorien>
so I don't know if there can be a general solution
<lejonet>
srhb: created a branch ceph-13-python3 to track my progress with python3 (We might want to do the bump to python3 regardless, so far it seems to be building nicely)
agander has quit [Ping timeout: 244 seconds]
<tilpner>
callahad: But it's usually packages that don't build, not buildEnvs, I haven't encountered that yet (mine just complain and build anyway)
<samueldr>
smells like both should be something like "$package-mkfs.fish" or something
fendor has joined #nixos
<edef>
i'm somewhat surprised we don't catch this in tests
<edef>
i'd hope we have btrfs tests, and dosfstools is default-installed
<{^_^}>
[nixpkgs] @nyarly opened pull request #56433 → cadre: init at 1.0.4 → https://git.io/fhNnn
<callahad>
I've run across a surprising number of broken packages in my two weeks with NixOS. :( I want to help make things better since Nix is such a compelling idea, but it takes so long to come up to speed.
<callahad>
...but this is the first time a channel update has broken me
<tilpner>
edef: It seems to be fish specific, the tests probably don't have it installed
<adamt>
callahad: Keep reporting issues when you run into something :-) As a counterpoint I haven't run into any issues like that (at least none i remember) after more than a year of NixOS on my laptop and servers, so just hang in there :-)
<dmj`>
clever: sometimes, when I do journalctl -f | ncat --ssl <ip> <port> I still get ncat Input/Output errors :(
<dmj`>
despite my firewall rules allow for HTTPS traffic
jabranham has quit [Quit: ERC (IRC client for Emacs 27.0.50)]
<callahad>
adamt: I definitely plan to. This is more "roll-up-my-sleeves" frustration than "abandon it" frustration. Just annoying to not be totally self-sufficient while I'm still learning the ecosystem.
<asymmetric>
is there a way to get the current platform, i.e. "linux" or "darwin"
<asymmetric>
like builtins.currentSystem but without the "x86_64"
<clever>
asymmetric: from nix or bash?
<asymmetric>
from nix
<adamt>
callahad: I know the feeling. I know a few in here, myself included, who were very lucky to have a coworker help us getting started. :-)
<clever>
> stdenv.isLinux
<{^_^}>
true
<clever>
> stdenv.isDarwin
<{^_^}>
false
<asymmetric>
but no builtins.currentPlatform or something?
<adamantium>
Does anyone know how to make chromium store the password in gnome-keyring?
<adamantium>
chromium --password-store=gnome doesn't work
hyperfekt has joined #nixos
<tilpner>
asymmetric: Not strictly, no
<adamantium>
ssh and gnupg are being stored and the keyring is being autounlocked by lightdm/pam , so i'm pretty sure it is already configured properly.
<tilpner>
asymmetric: I think you should not use currentSystem if possible
<tilpner>
(I just built a derivation where currentSystem was x86_64 and buildPlatform was aarch64)
<clever>
asymmetric: currentSystem, can break if you are trying to configure a cross-build from darwin->windows, while doing the eval on linux (hydra)
<asymmetric>
(this is for an overlay that fetches historical versions of the solc package from many commits of nixpkgs, and needs to be know which versions work for macOS and which don't)
<asymmetric>
so it's not for nixpkgs or anything, will never hit hydra
kvda has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<asymmetric>
currently i do builtins.getAttr super.system (import versions.nix)
<asymmetric>
but i'd like not to have keys like "x86_64-linux", and instead just have "linux"
petersjt014 has quit [Read error: Connection reset by peer]
<hyperfekt>
clever: Do I understand correctly that not-os is basically a derivation built with nix, and the NixOS is the same with nixos-rebuild taking the result and realizing it on the system? I presume it's probably easier to set disabledModules to the content of module-list.nix and instead import all the modules from my patched nixpkgs.
<clever>
hyperfekt: it may be simpler to call <nixpkgs/nixos/lib/eval-config.nix> and pass it a moduleList param, then to disabledModules every single thing
<hyperfekt>
clever: I thought so too, but I can't just put that call into my configuration.nix, can I?
<hyperfekt>
clever: So unless I want to operate on a different nixos channel (which trying to avoid is the reason of this whole exercise) with different arguments to eval-config.nix I guess using disabledModules seems like my only choice.
<hyperfekt>
At least if I want to keep the amount of dirty hacking (like patching nix-channel or nixos-rebuild to patch nixos) to a minimum
<hyperfekt>
wth is up with the unstable channels today!!? i remember them being almost two weeks old when i started using NixOS. love it!
<samrose>
anyone run into the need to set `RUST_SODIUM_LIB_DIR` in a package? looking for pointers on that
drakonis has quit [Quit: WeeChat 2.3]
<tilpner>
hyperfekt: What are you trying to accomplish?
johanot has quit [Quit: WeeChat 2.2]
random_yanek has quit [Ping timeout: 245 seconds]
<hyperfekt>
tilpner: Applying diffs to nixos so I can have changes active on my system without having to pull updates to the modules I changed all the time.
<samueldr>
hyperfekt: if you don't like it I can add breakage to keep the unstable channels from updating ;)
kreisys has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
kreisys has joined #nixos
<infinisil>
I'm trying to run docker with `virtualisation.docker.enable`, but I get `Error starting daemon: Error initializing network controller: list bridge addresses failed: no available network` :/
<tilpner>
hyperfekt: And how much effort would you be okay with? I have a setup that allows me to specify the nixpkgs version for my system with Nix
<hyperfekt>
tilpner: But your solution specifies the nixpkgs to use by providing a derivation? That would be pretty neat, I'd be okay with some impurity to get the current channel given that it's just on my system.
<hyperfekt>
edef: Yeah, that's definitely the easy solution. I don't know how far I trust my git abilities to keep several branches with changes merged into one. Also I really like the idea of the low friction of adding a branches name to my config being the only thing necessary to incorporate its changes.
o1lo01ol1o has quit [Remote host closed the connection]
knupfer has quit [Ping timeout: 264 seconds]
<tilpner>
hyperfekt: The idea is to redirect <nixpkgs/nixos> to a Nix expression that evaluates your system yourself, but twice: once with your current nixpkgs version to get desired nixpkgs version, and then with that determined version for the final system
<tilpner>
But if the git checkout sounds like an acceptable solution to you, you should probably take that
jackdk has joined #nixos
<hyperfekt>
tilpner: No, I really like that idea. I'm having fun hacking on this, after all. :) If I decide it's not a good fit I can still do the checkout.
aanderse_ has joined #nixos
aanderse has quit [Ping timeout: 250 seconds]
<tilpner>
hyperfekt: Okay, here goes. You set the NIX_PATH element "nixpkgs" to point to a directory you created manually (inside the git repo for your config)
<tilpner>
hyperfekt: In this directory, you have a file default.nix, and a directory "nixos" with another default.nix
<adamt>
There's a minor update to the version of kafka currently in the 19.03 branch. What is the correct way to PR an update? Make a PR against master as usual, get it merged, and then...?
<tilpner>
So now when nixos-rebuild calls <nixpkgs/nixos>, you get to decide what the outcome will be
<adamt>
gchristensen: So cherry-pick and PR against release-19.03?
<gchristensen>
yea
<adamt>
gchristensen++
<{^_^}>
gchristensen's karma got increased to 76
<tilpner>
hyperfekt: Your custom <nixpkgs/nixos> now calls lib.evalModules with check = false; and a module that adds an option to set your system nixpkgs version
<tilpner>
hyperfekt: From this first evaluation, you only take the value of that new option. The value should be a derivation that builds to your patched nixpkgs versions
<tilpner>
hyperfekt: Then, you pick the baseModules from that patched version and lib.evalModules again
stepcut has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @dotlambda pushed to revert-52464-fish_generate-completions « Revert "nixos/fish: generate autocompletions from man pages" »: https://git.io/fhNcM
<{^_^}>
[nixpkgs] @dotlambda opened pull request #56439 → Revert "nixos/fish: generate autocompletions from man pages" → https://git.io/fhNcD
<{^_^}>
[nixpkgs] @veprbl opened pull request #56440 → qt5.qtwebengine: fix on darwin (again) → https://git.io/fhNcH
stepcut has joined #nixos
ixxie has quit [Ping timeout: 240 seconds]
drakonis has joined #nixos
drakonis has quit [Client Quit]
drakonis has joined #nixos
stepcut_ has joined #nixos
rfold_ has quit [Quit: leaving]
GryphOne has quit [Quit: Page closed]
stepcut has quit [Ping timeout: 246 seconds]
kreisys has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<{^_^}>
[nixpkgs] @Ericson2314 merged pull request #56408 → treewide: use runtimeShell instead of stdenv.shell whenever possible → https://git.io/fhNTe
<{^_^}>
[nixpkgs] @Ericson2314 opened pull request #56441 → treewide: use runtimeShell instead of stdenv.shell whenever possible for 19.03 → https://git.io/fhNCI
<{^_^}>
[nixpkgs] @Ericson2314 closed pull request #56441 → treewide: use runtimeShell instead of stdenv.shell whenever possible for 19.03 → https://git.io/fhNCI
matix has joined #nixos
<{^_^}>
[nixpkgs] @dotlambda merged pull request #56439 → Revert "nixos/fish: generate autocompletions from man pages" → https://git.io/fhNcD
<{^_^}>
[nixpkgs] @dotlambda pushed to master « Revert "nixos/fish: generate autocompletions from man pages" (#56439) »: https://git.io/fhNCL
aanderse_ has quit [Ping timeout: 240 seconds]
<{^_^}>
[nixpkgs] @Ericson2314 pushed 2 commits to release-19.03: https://git.io/fhNCq
<{^_^}>
[nixpkgs] @Ericson2314 opened pull request #56443 → treewide: use runtimeShell instead of stdenv.shell whenever possible for 19.03 → https://git.io/fhNCO
ddellacosta has quit [Ping timeout: 250 seconds]
erasmas has quit [Quit: leaving]
<dmj`>
So
<dmj`>
When I do a deploy with nixops, my systemd unit gets restarted, even if its on a timer to run once a day. Is there a way I can deploy such that the systemd unit doesn't get restarted
<dmj`>
Does nixops have to restart my service?
<adamt>
dmj`: That's probably not nixops specific, but just how NixOS handles systemd timer units
<dmj`>
if so, how can I get the service to not run the script
<elvishjerricco>
dmj`: Yea, display-manager.service on desktops does what you want I think, so you could look there to see how it does it
<dmj`>
tilpner: omg, so wait, the next time it runs, will it use the newly deployed code?
<dmj`>
tilpner: the next time the timer runs
ddellacosta has joined #nixos
<tilpner>
I don't know, you should try it
kvda has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<tilpner>
And then tell us if it did
<elvishjerricco>
Id be surprised if it didn't use the new code. The switch should do systemctl daemon-reload, which should be all that's necessary to switch to the new version
<adamt>
How would it use the new code if the systemd unit wasn't reloaded?
<adamt>
Oh. Of course. the unit is reloaded.
<elvishjerricco>
adamt: daemon-reload will reload all unit files without triggering any restarts or anything
<adamt>
Yeah I know, I'm just tired and angry at VMware, sorry.
<elvishjerricco>
aren't we all? :P
<adamt>
I can brick VMware guests by just loaded iPXE on them. Totally bricked. Can't even enter the EFI firmware. That's like an unbeatable achievement, I think. :P
<dmj`>
I redeployed, it didn't restart, yay
__monty__ has joined #nixos
<dmj`>
But I assume the systemd timer will restart it to use the new code...
eadwu has quit [Quit: WeeChat 2.4]
random_yanek has quit [Ping timeout: 246 seconds]
luigy has joined #nixos
<dmj`>
also, if I stop the service, will the systemd timer start it when the time comes?
<gchristensen>
you'd have to mask the service
shu9 has joined #nixos
<dmj`>
gchristensen: what does masking do
<dmj`>
what is masking I guess
<adamt>
dmj`: Prevents it from starting
<dmj`>
adamt: I'd like it to be such that only the timer can start it
<dmj`>
is that was masking would allow?
<dmj`>
and the timer can only start it at 3AM
<gchristensen>
I don't think youcan make it so *only* the timer can start it
<dmj`>
and if it fails with a 0 exit code, it will restart, so Restart=on-failure
<dmj`>
I want the service to be active, but waiting for the timer
<dmj`>
to start it
<gchristensen>
maybe I'm missing what you're having trouble with?
<adamt>
Doesn't a systemd timer basically just start a oneshot unit (or similar) on a schedule? I mean, nothing fancy about it, and root will always be able to also start the unit
kvda has joined #nixos
<gchristensen>
right
work_ has quit [Quit: Connection closed for inactivity]
<dmj`>
when root starts the unit, it immediately executes the script
<dmj`>
I'd like to have the unit active, and not execute the script until the timer says so
<dmj`>
Am I better off keeping the unit inactive, and then letting the timer make it active?
<gchristensen>
ah, you've overspecified the unit
<gchristensen>
don't make the unit WantedBy anything / don't make anything Want the unit
<dmj`>
woa
<gchristensen>
systemd doesn't start anything until something asks it to
<tilpner>
No wantedBy = [ "timers.target" ];?
<gchristensen>
maybe that one
<tilpner>
That's what I put on all my timers
<dmj`>
when I do, "systemctl stop <service>" it says, "Warning: Stopping <service>.service, but it can still be activated by <service>.timer
<dmj`>
I just have the Unit = "<service>.service" in the systemd.timers.<service> unit
stepcut_ has quit [Remote host closed the connection]
<dmj`>
I removed the wanted by
<dmj`>
but when I do systemctl start <service>.service, it still executes the script section
<dmj`>
I'm just gonna stop it, and keep it dead, see if the timer starts it
<dmj`>
thanks for the help guys
<tilpner>
dmj`: systemctl list-timers --all
<tilpner>
dmj`: And systemctl status <service>.timer
<{^_^}>
[nixpkgs] @Ericson2314 merged pull request #56443 → treewide: use runtimeShell instead of stdenv.shell whenever possible for 19.03 → https://git.io/fhNCO
<{^_^}>
[nixpkgs] @Ericson2314 pushed 2 commits to release-19.03: https://git.io/fhNCd
<dmj`>
tilpner: it's there so
<dmj`>
tilpner: if the unit it is watching is inactive, will it start it?
<tilpner>
The important part is that the timer is active
aramiscd has joined #nixos
<{^_^}>
[nixpkgs] @Infinisil opened pull request #56444 → hueadm: init at 1.2.0 → https://git.io/fhNCA
fenedor has quit [Read error: Connection reset by peer]
matix has quit [Quit: WeeChat 2.4]
matix has joined #nixos
njha has joined #nixos
<njha>
Is there an easy way to specify docker-compose files to run in configuration.nix?
random_yanek has joined #nixos
<njha>
I'm running debian right now, but I want to be able to deploy a copy of the same containers to any computer, so NixOS seems like the right choice.
shibboleth has joined #nixos
<mdash>
njha: don't think there's a one-line solution for it
<mdash>
njha: i use docker-compose just because the rest of my team does at work, for personal stuff i've used nixos-container