<samueldr>
I don't have google-chrome in my systemPackages, but it would look like `environment.systemPackages = with pkgs; [ google-chrome /* and other packages */ ];`
<samueldr>
do note that you cannot have two times the same option set in a single configuration file
<samueldr>
so you cannot do `environment.systemPackages = [ /* ... */ ];` and later another `environment.systemPackages = [ /* ... */ ];`
<tertle||eltret>
i like that one samuel
<tertle||eltret>
i will try it
<tertle||eltret>
i just wasnt sure what to do with the text that was already in the file
<samueldr>
oh, that specific file isn't "complete", I have split my configuration file into many files
<samueldr>
in this particular case, it only sets up common defaults
Acou_Bass has quit [Ping timeout: 276 seconds]
cryptomonoid has quit [Ping timeout: 276 seconds]
<tertle||eltret>
is there a simple one with some comments in it to help?
<samueldr>
tertle||eltret: you recently installed your nixos system if I understand correctly?
<tertle||eltret>
yes, im using the vbox appliance for now
<samueldr>
ah, so you did not install using "nixos-generate-config"
<samueldr>
I don't know how the default configuration.nix file in that case looks
<tertle||eltret>
no, I havent done much at all
<samueldr>
give me a sec I'm checking
<tertle||eltret>
it was already built but ive deleted the old config file now
<tertle||eltret>
it was just a template mainly
<{^_^}>
[nixpkgs] @volth opened pull request #43869 → pkgs/*: remove unreferenced function arguments → https://git.io/fNWn6
Fare has joined #nixos
wgas has left #nixos ["Leaving"]
test has joined #nixos
test is now known as Guest1754
iqubic has quit [Remote host closed the connection]
Guest1754 has quit [Client Quit]
Fare has quit [Ping timeout: 260 seconds]
<samueldr>
oh yeah
<samueldr>
just booted to be 100% sure
<samueldr>
that default configuration.nix file is... really minimal :)
Acou_Bass has joined #nixos
<tertle||eltret>
im starting over with fresh applicance
<PyroLagus>
i wonder if we will ever have nixos-unstable without week-long delays /o\
<tenten8401>
PyroLagus: I've been waiting for a package too :P
<samueldr>
tertle||eltret: if you want to see how the default "configuration.nix" generated for an install is, you can do this command "mkdir -p /tmp/nixos; nixos-generate-config --root /tmp/nixos/"
<samueldr>
and inside /tmp/nixos/etc/nixos you will have a much more furnished sample
<PyroLagus>
i heard nixos got better at that, but apparently it still isn't completely fixed
<samueldr>
tertle||eltret: do note that if you rebuild switch into it, kde will go away, as it wouldn't have anything configuring it
<tenten8401>
PyroLagus: it's mostly just the fact that a lot of the most commonly used packages are tested before it's updated (as I understand it)
<samueldr>
PyroLagus: part of it is sometimes huge rebuilds, but other parts is things breaking :/
<samueldr>
tenten8401: yeah, *everything* in the channel has to be rebuilt
<samueldr>
so if a dependency at or near the root of the dependency tree changes... it takes a bit of time
<samueldr>
and then if a test fails because of either that change or another change, it takes a bit more time until it's fixed
<tertle||eltret>
okay, thank you samueldr
<tertle||eltret>
or should i say Dr. Samuel
<samueldr>
(though, how many other distros are integration and unit tested that much?)
<tenten8401>
I know openSuse tries but I see some stuff sneak past pretty often
<samueldr>
tertle||eltret: :) it's not a title, it stands for my last names
<tertle||eltret>
oh i see
<samueldr>
what would be fun, but also *harder to implement* would be to revert commits failing
<samueldr>
*left as an exercise to the readers*
markus1199 has joined #nixos
<tertle||eltret>
samueldr: so are you telling me that my config file isnt actually the config file my machine
<samueldr>
(what is actually supposed to happen is better testing of changes, but uh, looks like nixpkgs is huge!)
<samueldr>
tertle||eltret: do any programming? what languages?
<tertle||eltret>
i mostly do networking stuff in routers and switches
<samueldr>
oh
<tenten8401>
yeah, nixpkgs is absolutely massive
<samueldr>
(I would have tried to explain using parallels into another language)
<samueldr>
tertle||eltret: the configuration.nix in the .ova is "importing" the contents of another file from nixpkgs
<tertle||eltret>
is there a way to generate a file of how my system is currently configured?
<tertle||eltret>
samueldr: i just realsized that
<samueldr>
so what this means is that the *actual contents*, what does the configuration and everything is found in another file
<samueldr>
so that's why there's so few things
<PyroLagus>
now i wonder though. do we already have a mechanism to push through critical security updates while a huge rebuild is going on?
silver has quit [Read error: Connection reset by peer]
<samueldr>
PyroLagus: the idea is to not push "untested" changes
markus1189 has quit [Ping timeout: 276 seconds]
<samueldr>
to "test" (wrong term here) there is the staging branch
<samueldr>
so stuff that is expected to break goes there first
<samueldr>
but uh, sometimes stuff isn't expected to break!
<samueldr>
and AFAIK, there is no way to cheat the system
<samueldr>
(except doing things like reverting the change, pushing, rebuilding)
<PyroLagus>
well, unless the security updates affects glibc, there wouldn't be too much to test. you could turn on a parallel test with just the patch
<samueldr>
THOUGH, in critical infras, you could realistically follow unstable, apply the patches yourself, and build yourself
<tenten8401>
I think the stable 18.03 channel got updated more recently than the unstable channel :(
<samueldr>
PyroLagus: what often happens is that *something else* changed around the same time :(
<samueldr>
tenten8401: that's expected
<samueldr>
well, not really expected, but not unusual
<PyroLagus>
samueldr: but you don't *have* to update that something else at the same time, right?
<samueldr>
what's expected is that -stable shouldn't even fail
<PyroLagus>
you could just push that aside
<tenten8401>
I guess that's true
<samueldr>
PyroLagus: first come, first served
<PyroLagus>
heh
<samueldr>
if PR for $someting merged before PR for $CVE, it's been stacked in there
<PyroLagus>
sorry firefox zero day. supertuxkart's gotta compile first
<samueldr>
PyroLagus: that's an interesting point though
<samueldr>
maybe an RFC could be written up to have a faster-tracked security story
<samueldr>
it's not a technical deficiency here, it's a process AFAIK
<PyroLagus>
yeah. there should be some notion of priority for updates imo
<PyroLagus>
not sure how to do that though
<samueldr>
PyroLagus: the commits cannot be re-ordered by priority, it would make things harder
<tertle||eltret>
samueldr: i guess i will start messing around installing it properly then
<samueldr>
tertle||eltret: yeah, get the minimal or graphical iso, install it into virtualbox
<samueldr>
that will be closer to a "real" installation
<samueldr>
and may be more edifying
<samueldr>
educamating
<samueldr>
words
<samueldr>
but, PyroLagus, it may be more realistic to figure out a process in which there is a middle-tier branch for "general stuff"
<PyroLagus>
maybe
<samueldr>
where critical stuff, and watever builds the minimal set (for those channels) always go to master/unstable, everything else into $middle-tier and breaking changes into $staging
<PyroLagus>
but i'm not talking about commits though. if you're already following master, you can just manually set critical stuff to build first
<samueldr>
"just"?
<PyroLagus>
well, "just"
<PyroLagus>
install the new package, then do an upgrade, i imagine
<samueldr>
I'm not a hydra expert, but AFAIUI, it isn't really built in a way where something specific can be queued up
alexteves has quit [Remote host closed the connection]
<samueldr>
(parts I don't know for sure will be enclosed in square brackets)
<PyroLagus>
ah, well if you actually have your own hydra set up, that's more complicated
<samueldr>
I'm thinking about the nixos hydra
<PyroLagus>
right, and for that you don't have to reorder the commits, but how hydra decides what to build when
<samueldr>
what happens is that [for each commit] pushed, hydra will start the process of building the whole world. For smaller changes, it can be as small as building the isos (commit hash has changed) and a couple packages
drakonis has joined #nixos
<samueldr>
so, even if you cancel the build for the supertuxcart commit, it still need to be built
<samueldr>
though, [unless it's a *tested* package, even if it fails channel can advance]
<samueldr>
it will still need to go through building everything that needs to be built until the channel itself advances
<PyroLagus>
yeah
<samueldr>
(-small has a smaller set of packages that needs to be built before it advances)
<PyroLagus>
i think overlays were introduced to fix address that problem, but i'm not sure what's missing
<PyroLagus>
nbp is the one who worked/works on that
<PyroLagus>
doesn't seem like anything's been happening since march, but i dunno
<PyroLagus>
i guess just having some github repo with critical security update overlays would be enough at the moment
<samueldr>
yeah, that's the kind of "process" thing I was thinking about earlier, it's not really a technical issue I think
<PyroLagus>
a bit hacky, but at least you'd be able to do a git pull, nixos-rebuild, to build the updates on your machine and install them
<PyroLagus>
yeah
<sophiag>
i just changed my stateVersion for the first time and upgraded on the slim chance it may fix a year+ old bug related to gpu switching and grub-install is reporting the error "/boot doesn't look like an EFI partition." fwiw, it had no problems with the i386 version. i feel like i've come across this before, but can't remember the fix
<samueldr>
sophiag: by default, the nixos setup for grub assumes /boot is where the ESP is mounted
<sophiag>
well, also because with stateVersion 16.09 Nix wouldn't upgrade beyond 1.11.6
<samueldr>
updating stateVersion is always a "bit risky" (you need to at least take a look at some things)
<samueldr>
,stateVersion
<{^_^}>
Setting stateVersion to the latest release doesn't update anything and can only break your setup at best. If you want to regardless, Ctrl-F for "stateVersion" in https://nixos.org/nixos/manual/release-notes.html to see things that need to be manually upgraded with the new value
<samueldr>
(but seeing 16.09 I'm assuming it won't be an issue for you :))
<sophiag>
well, i already did
<sophiag>
i mean, it's the box i'm using now...
<sophiag>
anyway, wrt to the ESP i thought boot.loader.grub.device set it?
<samueldr>
hmm, I can't tell before I started really using nixos, but AFAIK for EFI it's expected to be nodev?
<samueldr>
ah yeah
<sophiag>
didn't realize we have a wiki again :p
Supersonic has quit [Disconnected by services]
Supersonic112 has joined #nixos
Supersonic112 is now known as Supersonic
<acowley>
What's going on with hydra and nixos-unstable? It looked like there was a good evaluation yesterday, but the channel didn't advance, and today it's hung on a handful of queued builds.
<samueldr>
acowley: I was *just now* wondering
<samueldr>
(I still have a hard time being sure about what I'm looking at with hydra)
<samueldr>
sophiag: yes :) a couple users have been working on it since last year
<acowley>
samueldr: It shouldn't be complicated, but it's been pretty rocky lately with things like OOM failures.
<samueldr>
acowley: it's more like "I'm looking at "job nixos:unstable-small:tested", and there are tick marks... what other things are needed to finish before a channel would normally advance?"
<acowley>
samueldr: click on the build id at the top, then click on the evaluation link, then check the "Queued jobs" tab
<acowley>
To advance, you need all tests to pass (the check marks), and all builds to finish one way or another.
<samueldr>
tertle||eltret: you know how your computer boot, in "the past" (up to mid to late 2000) it was all "bios" based; follows an oversimplified and wrong explanation
<acowley>
samueldr: I think the people behind the curtain are all working very hard to improve things, so monitoring the channel march has dropped in priority for the moment.
<samueldr>
acowley: I do know they're working hard on stuff :) it's still a "oh no" because they'll need to take some of their precious time to figure things out :(
<hodapp>
hrm, why do my CMake builds always end up with -j1 on the actual make commandline?
<samueldr>
the CPU loads the bios, which does some initial stuff [waves hands] and then loads the first sector from the boot media into memory, "jumps" to it to start the boot proces
Guest99618 has quit [Ping timeout: 256 seconds]
<samueldr>
that is what may be called the "mbr", it then *does things* to then load your operating system...
<samueldr>
(overly simplified and probably wrong at parts!)
<tertle||eltret>
so are you sayting that EFI is superior?
<samueldr>
for UEFI based systems, the "bios" is more complex, and can do more things
<samueldr>
so instead of just loading the first (few?) sectors into memory and running it
<samueldr>
it can do more complex stuff
<acowley>
hodapp: Is enableParallelBuilding set to false on the derivation?
<samueldr>
it will read files into a partition
<tertle||eltret>
how do i know what i have? i just bought this sweet new PC?
<hodapp>
acowley: isn't that true by default?
<hodapp>
regardless, it's set true
<acowley>
hodapp: Yes, it is
<samueldr>
tertle||eltret: EFI isn't *strictly* superior, though it can do more complex things that the older scheme couldn't (due to technological advances since the first IBM PCs)
<samueldr>
and if it' recent, like past 2012, it's pretty sure it's an UEFI based system
<hodapp>
I even see "cmake: enabled parallel building"
commander has quit [Ping timeout: 240 seconds]
<tertle||eltret>
yeah i cant seem to get the rebuild switch to build
<acowley>
hodapp: Do you have NIX_BUILD_CORES set to something interesting in the environment?
<tertle||eltret>
im on EFI then
<samueldr>
tertle||eltret: if you're using virtualbox, it's almost 100% sure bios-based emulation/virtualisation
<tertle||eltret>
samueldr: i switched to VMWare
<samueldr>
oh!
<samueldr>
then AFAIK it uses UEFI by default :)
<samueldr>
tertle||eltret: what error do you see? (I'm not afraid of screenshots)
<tertle||eltret>
well, i configured it for BIOS, so i need to start over anyways
<samueldr>
:)
<tertle||eltret>
i didnt know i had EFI
<tertle||eltret>
lol
<samueldr>
you could do both installs, look at how it differs
<acowley>
hodapp: That can also be controlled in your configuration via `nix.buildCores = xxx;`
<PyroLagus>
talking about virtualization. does kvm on nixos-unstable already have virtiogpu?
<sophiag>
while everyone's discussing efi...my generated hardware config has `fileSystems."/" = {..}` but i can't just set `boot.loader.efi.efiSysMountPosition = "/"`. how do you generally handle this on nixos?
<samueldr>
(not sure if that is virtiogpu, but an article seems to say so?)
Fare has quit [Ping timeout: 240 seconds]
<PyroLagus>
that's not the same, but still cool
<hodapp>
acowley: huh, apparently I just never set buildCores
<samueldr>
PyroLagus: having trouble to find "upstream" info about virtiogpu
<PyroLagus>
okay. it is still pretty new, so i'm not surprised
<tertle||eltret>
samueldr: thats true, although EFI seems like only a few more commands
<samueldr>
tertle||eltret: pretty much
<PyroLagus>
didn't know about igvt though. looks like that would be better for intel integrated gpus
<tertle||eltret>
do you guys get hung up when doing the installs much?
<tertle||eltret>
i used gparted to partition
d1rewolf has joined #nixos
<samueldr>
PyroLagus: from a quick search, same author (kraxel) worked on both?
<PyroLagus>
huh, cool
<d1rewolf>
can someone share a working configuration.nix for i3-gaps? I have i3 working, but not gaps for some reason
<samueldr>
but I'm not even sure about what I just said
<samueldr>
tertle||eltret: other than always forgetting the grub specific "bios boot" partition on GPT for bios boot (not EFI) no, but I don't install often, don't have to!
Fare has joined #nixos
strobelight has joined #nixos
<acowley>
hodapp: defaulting enableParallelBuilding to true and buildJobs to 1 is an interesting hedge.
init_6 has quit [Read error: Connection reset by peer]
<sophiag>
oh...it seems like i may have to reinstall the bootloader :( i'm currently booted in Legacy
<tertle||eltret>
i should actualy go ahead and install it on my old 2600K now that I see how it works
<tertle||eltret>
would be cool to ssh in and do it though
<tertle||eltret>
i have it on a USB but idk if I wanna get started this late
d1rewolf has quit [Remote host closed the connection]
erratic has joined #nixos
<nh2>
can I query the flags that were given to a package?
<nh2>
like I want to check what was given as `{ , curl , ...}`
<nh2>
or do I have to `override` and put the required info into `passthru`?
hlolli_ has joined #nixos
spear2 has joined #nixos
lassulus_ has joined #nixos
<nh2>
hmm not sure how I can get at these arguments at all, it seems neither `override` nor `overrideAttrs` give them to me
hlolli has quit [Ping timeout: 240 seconds]
lassulus has quit [Ping timeout: 256 seconds]
lassulus_ is now known as lassulus
<sophiag>
i would prefer to exhaust all alternatives before reinstalling the bootloader. does anyone else here have ESP on their root partition?
<samueldr>
hm, unless your root is FAT32 (unlikely) or your bios can read $fs (ext4), it probably won't work?
<samueldr>
grub will still need to be installed into the ESP
<samueldr>
the ESP, though, can be on another physical drive, if it helps
<sophiag>
eh...not a long term fix. i typically don't mount my optical one
<sophiag>
i was wondering if i could actually specify the same device for two fileSystems? e.g. fileSystems = { "/" = { device = "/dev/sda;" }; "/boot" = { device = "/dev/sda";}};
<sophiag>
the installation docs weren't *that* bad a couple years ago that i can imagine screwing this up when i formatted the drives
<acowley>
hodapp: defaulting enableParallelBuilding to true and buildJobs to 1 is an interesting hedge.
<acowley>
Uh, sorry, emacs took over there
npmccallum has quit [Ping timeout: 276 seconds]
srl295 has joined #nixos
<{^_^}>
[nixpkgs] @nh2 opened pull request #43870 → Optional static libraries for krb5 and openssl → https://git.io/fNWWJ
<hodapp>
uh... cool. "make: *** [Makefile:130: all] Error 2" with no actual error visible
orivej has quit [Ping timeout: 264 seconds]
cryptomonoid has joined #nixos
<sophiag>
welp, i think i learned my lesson re: trying to boot off ext4 :(
aw has quit [Remote host closed the connection]
spacefrogg has quit [Remote host closed the connection]
aw has joined #nixos
spacefrogg has joined #nixos
<samueldr>
sophiag: your ESP *can* be at the end of your disk (with most UEFI implementations)
sophiag has quit [Ping timeout: 240 seconds]
<samueldr>
so, if you resize/move partitions, you can skip the tedious resize+move and instead only resize any partition
<samueldr>
if you intend to keep the kernels+initrd on your main partition, the ESP can be really small
<samueldr>
with 6 generations, I am only using ~99M according to df
iyzsong has joined #nixos
cryptomonoid has quit [Ping timeout: 240 seconds]
<{^_^}>
[nixpkgs] @hcmensch opened pull request #43871 → vim-plugin-names/default.nix: data added for vim-ledger → https://git.io/fNWWi
jb55 has quit [Quit: WeeChat 2.1]
sophiag has joined #nixos
<sophiag>
nice...after throwing my hands up, disabling efi support, and rebuilding i was thrown into a tty and can't start any window manager :p
<sophiag>
at least the screen and keyboard seem to be working fine now?
npmccallum has joined #nixos
npmccallum has quit [Remote host closed the connection]
<samueldr>
any logs for display-manager? (journalctl -b0 --unit display-manager)
wigust has joined #nixos
<samueldr>
(unless it's lightdm, which still have it stored under /var/log :( )
<sophiag>
if i'm to be totally honest, i'm trying to remember how to get out of irssi since i didn't detach it...
<samueldr>
sophiag: use another tty? alt+F2
<nh2>
sophiag: I booot off ext4 for 10 years I think, is that not supposed to work easily?
<sophiag>
i have both gdm and lightdm installed, but have been using the former until i have this sort of stuff worked out
wigust- has quit [Ping timeout: 244 seconds]
<sophiag>
samueldr: thanks :p
<sophiag>
nh2: what do you have efiSysMountPoint set to?
otlfac has quit []
hphat^ has joined #nixos
<nh2>
sophiag: I was talking about my Ubuntu system when I mentioned the 10 years, but my NixOS 18.03 server also boots from ext4. I haven't set `efiSysMountPoint` so it should be the default (/boot)
<nh2>
and my /boot is just a directory on /
<nh2>
on the / partition
<samueldr>
nh2: I from what was said, I think it's an upgrade from legacy boot to UEFI boot at issue here
<nh2>
BIOS vs UEFI and GPT vs MBR is a total pain
<sophiag>
except there's no reason it should have ever been installed as legacy boot :/
<nh2>
also took me more than a day to figure out for the first time for the server
<samueldr>
the things past $self do :)
<sophiag>
anyway, journalctl shows it did launch XServer and then gdm keeps crashing after average 250ms
<sophiag>
samueldr: lol
<samueldr>
try enabling one of lightdm, sddm or slim instead of gdm
<nh2>
so on my server, I use BIOS+GPT and a BIOS boot partition that grub uses
<samueldr>
(it may help point out whether it's gdm or X at fault)
<sophiag>
it is possible during about 24 hours of swearing and sleep deprivation i just screwed up my configuration.nix
<samueldr>
nh2: yes, that biosboot partition I always forget about :D
<samueldr>
(same situation bios+gpt on my home server)
<nh2>
sophiag: I haven't read all discussion, which combination of (UEFI|BIOS)+(GPT|MBR) do you use?
<nh2>
that is the most important thing to figure out and keep in mind when researching, otherwise it's easy to get confused with other solutions presented
<samueldr>
if you're unsure of your partition scheme, you could use this to print it out: `nix-shell -p gptfdisk --run "sudo gdisk -l /dev/sda"`
<dhess>
nh2: are you enabling bbr on all of your servers?
<sophiag>
my only confusion was whether "Disklabel = dos" == mbr as i assumed. yes?
<nh2>
dhess: yes most
<dhess>
nh2: any cases where you've seen it not perform well?
<samueldr>
sophiag: yeah, disklabel dos 99% confident is "mbr"
<sophiag>
samueldr: right. i checked that before screwing with any of this
<nh2>
dhess: no
<samueldr>
oh, looks like it's possible to convert MBR to GPT, but do know that in that case, you'll be left with an unbootable system until an ESP is set up correctly
<nh2>
dhess: but I consider that "competitive advantage" :D
<dhess>
for now anyway, until everyone turns it on :P
drakonis has quit [Read error: Connection reset by peer]
<sophiag>
samueldr: could that explain this problem i've ascribed to gpu switching for ~1yr now? just that i was trying to boot an mbr volume with bios?
<samueldr>
no idea
<samueldr>
though it's possible the firmware does different stuff for legacy boot vs. uefi boot
<sophiag>
or it's possible bumping my stateVersion fixed that and now i'm left with this problem
<nh2>
sophiag: so what did you change exactly that broke things? Sorry I'm still catching up
<sophiag>
nh2: i disabled grub efi support...
<nh2>
sophiag: and now you can't boot?
<sophiag>
but i've had a problem having to roll back to a year old build every time i reboot that seems similar so am wondering if maybe this will fix it. i didn't roll back anything now
<sophiag>
nh2: no, i just can't start a desktop manager. gdm keeps crashing repeatedly
<sophiag>
this is actually maybe a little hopeful?
drakonis has joined #nixos
<nh2>
sophiag: that's odd. Usually if you manage to boot and get a shell you have already gone past all BIOS/UEFI/GPT/MBR related problems
<nh2>
sophiag: if you start gdm manually from a shell, do you get output in the console?
<nh2>
sophiag: otherwise, try SSH in with another computer and start gdm with `strace`
<sophiag>
i'm told gdm isn't installed...
<nh2>
if it isn't installed, how can it keep crashing?
<samueldr>
looks like there are two issues, so it's best not to conflate them, 1. boot is weird (but working) 2. display-manager OR X broke; sophiag does this look like a good summary?
<samueldr>
and sophiag, this is all after updating/upgrading, right?
<sophiag>
yes. but it seems like i'm dealing with #2 now
<sophiag>
and yes
<sophiag>
i was able to boot into a tty on 18.09
<nh2>
I have to go to bed now, good luck!
<sophiag>
some display and keyboard issues for about a minute
<sophiag>
nh2: thanks!
agjacome has joined #nixos
<sophiag>
i know i can reboot, rollback a year (i did delete builds in between at least), and get display managers to launch. but if i can actually launch them from here i may actually be able to fix this
<samueldr>
yeah, fixing this shouldn't be impossible
<sophiag>
i assumed it was (at least because i couldn't isolate whatever was causing it) because i'm using this gaming laptop that hardwires an nvidia chip to the display
<sophiag>
i.e. you can't disable either integrated or nvidia...
<samueldr>
those always add a bit of "fun" when dealing with those kind of issues :/
<samueldr>
(I don't have any, so I don't have experience with configuration for them)
<sophiag>
it certainly weighs on my decision when purchasing a new one (this is a hand me down)
<samueldr>
using startx `nix-shell -p xlibs.xinit --run "startx"` does X start? if it doesn't, it's not the DM, but the X server that's at fault (probably)
<samueldr>
sophiag: you may need to use ctrl+alt+f1/f2 to get back to a tty
<samueldr>
under X, it's ctrl+alt and not only alt which is used to switch
<sophiag>
startx also isn't installed (that at least _maybe_ makes sense considering i never call it manually?)
agjacome has quit [Ping timeout: 240 seconds]
<samueldr>
using the command with nix-shell should work?
<sophiag>
it should. i was just curious
<samueldr>
yeah, it's normal to not have it installed globally
<samueldr>
the nixos module will configure the services so X, gdm, everything is available to them
<samueldr>
but not cluttering the user's environment
<sophiag>
ah ok. that makes sense. obviously i'm not listing those as packages
<sophiag>
i think i'm just so used to things disappearing every time i need to reboot!
sophiag has quit [Remote host closed the connection]
endformationage has quit [Quit: WeeChat 1.9.1]
<tertle||eltret>
its installing samueldr
<tertle||eltret>
on my real machine
<samueldr>
tertle||eltret: oh no, your sleep! ;)
sophiag has joined #nixos
<sophiag>
well, that was odd
<tertle||eltret>
it sok, im off tommorow
<tertle||eltret>
im drinknig kava and baskingin the moonlight
<samueldr>
sophiag: what kind of odd?
<tertle||eltret>
i have ot say, this is the first install oflinux ive done that actuially makes sense
<tertle||eltret>
everything is in one file
<tertle||eltret>
at least for now :P
<samueldr>
that's not an install, that's a configuration!
<samueldr>
;) and that's probably why it "makes more sense"
<samueldr>
(though there's some state left up to the admin, the partitioning mainly)
<tertle||eltret>
its more than an install
<sophiag>
samueldr: xstart launched a screen with a frozen cursor and three sort-of-windows with terminals in them
iqubic has joined #nixos
<samueldr>
sophiag: frozen cursor: weird, three sort-of-window, that's probably right
<samueldr>
(iirc the defaults for startx are to start twm with three xterm)
<sophiag>
i'm in tty. but yeah i have a friend who goes through growlers
<sophiag>
samueldr: you were suggesting i try launching gdm or lightdm from nix-shell?
<samueldr>
I was suggesting changing it in configuration
<slickshoes>
tertle||eltret is that a bag of wine?
<samueldr>
n h 2 (don't want to wake them up) was suggesting in shell
<samueldr>
gdm will probably ask to be ran as root though
<sophiag>
yes. tried that
<sophiag>
and i tried switching to lightdm in my config a while ago and still had this problem
slickshoes has quit [Quit: Peace and Protection 4.22.8]
init_6 has quit []
Parabola has joined #nixos
drakonis has quit [Remote host closed the connection]
iyzsong has quit [Read error: Connection reset by peer]
<tertle||eltret>
how to enable wireless over terminal?
<samueldr>
wpa_supplicant, as with most distros, will work
<samueldr>
I personally prefer using network-manager and nmtui
<samueldr>
only for convenience's sake
spear2 has quit [Ping timeout: 268 seconds]
<sophiag>
okay, i think i have more information...gdm is always running in the background and restarts after i sigkill it. when it starts fresh, wayland starts and crashes repeatedly as well as gnome autostart. that appears to be causing the keyboard and screen to glitch. after a while those processes die and gdm is still running in the background crashing four times a second
<tertle||eltret>
what options to iuse?
<sophiag>
i think my problem is wayland
spear2 has joined #nixos
sir_guy_carleton has quit [Quit: WeeChat 2.0]
<samueldr>
sophiag: uh, didn't know nixos' gdm used wayland, but it makes sense if it does
<samueldr>
try slim or sddm if you had issues with lightdm
<samueldr>
(though slim is possibly going away as it's not maintained?)
<sophiag>
i thought gdm has used wayland for quite a while now?
* samueldr
is out of touch with gnome-land since a good while
<sophiag>
my current hypothesis (and the first new one i've had on this issue for a while) is that wayland is the issue and something switching to it caused these problems
<sophiag>
i'm not sure i feel okay going into #wayland and asking, "hey, do you guys know which packages don't use your project?"
<sophiag>
i think this all occured much more recently than i assumed. googling on my phone i'm getting a ton of ubuntu users whining about it this year
spear2 has quit [Remote host closed the connection]
<sophiag>
i can try turning it off in gdm, but from what i'm reading i don't expect that to work
Guanin has quit [Ping timeout: 240 seconds]
init_6 has joined #nixos
<sophiag>
err...actually i can't write to that
cryptomonoid has joined #nixos
<tertle||eltret>
im moving right along here
Parabola has left #nixos [#nixos]
<samueldr>
tertle||eltret: so everything's going smoothly?
<samueldr>
oh, missed the earlier question I think
Guanin has joined #nixos
init_6 has quit [Remote host closed the connection]
<tertle||eltret>
yeah, except searching for packages
<tertle||eltret>
isnt it -q?
<tertle||eltret>
also, some graphical artifacts
<samueldr>
you can now use `nix search` iirc
<samueldr>
-q only will list your installed packages
<samueldr>
I think it was -q with -a (-qa) to list all packages
<samueldr>
--available, -a
<samueldr>
The query operates on the derivations that are available in the active Nix expression.
<samueldr>
and then often coupled with -P (-qaP) to also print their attribute paths
<tertle||eltret>
oh i see
<tertle||eltret>
trying to setuip x2go
<tertle||eltret>
so i can use only opne keyboard
<sophiag>
wow, i'm finding new issues about this across every display manager i can think of
strobelight has quit [Read error: Connection reset by peer]
cryptomonoid has quit [Ping timeout: 268 seconds]
freeman42x]NixOS has quit [Ping timeout: 240 seconds]
init_6 has joined #nixos
<sophiag>
from what i can gather it required multiple patches. at least gdm+nvidia seems to have been merged into 17.09, though
<PyroLagus>
so.. the chromium build failed because it timed out after 10 hours, huh
<PyroLagus>
i understand why there's build timeouts, but chromium is huge. and now we have to wait at least yet another day for that build to succeed. rip
<{^_^}>
[nixpkgs] @volth opened pull request #43872 → fix build with disallowed aliases → https://git.io/fNW4s
jmeredith has quit [Quit: Connection closed for inactivity]
<sophiag>
so, i think my entire issue should have been patched at 17.09 and i just never got it because i had my stateVersion pinned to 16.09. now after upgrading i'm in tty hell and somehow missing the symlink to my channels...
Tiez has joined #nixos
<sophiag>
it seems i'm only missing the root symlink, though. i'm able to rebuild switch as a user
<sophiag>
oh...so that's a new bug
<sophiag>
yup, i haven't upgraded in so long i didn't even know about per-user channels
igo95862 has quit [Quit: igo95862]
igo95862 has joined #nixos
igo95862 has quit [Client Quit]
<tertle||eltret>
is there any meta packagesf or Nix?
<tertle||eltret>
liek KDE base or something?
agjacome has joined #nixos
<sophiag>
you shouldn't have to declare desktop manager packages
<srhb>
sophiag: Uh oh
<srhb>
sophiag: You should never be changing your state version unless you're told to in the release notes
<srhb>
sophiag: Normally it does not prevent upgrades of any kinds
<sophiag>
srhb: i've been told that before
<srhb>
OK :)
<sophiag>
well, not certain that warrants a smiley...
<srhb>
Well, it's better that you knew already and went into it eyes open :)
<sophiag>
true. the community is doing its job at least :)
<sophiag>
i might be wrong about why i didn't get that patch. it was merged after i gave up on the issue
<srhb>
I only just caught the end of this conversation...
<srhb>
gdm and nvidia? hm
<sophiag>
well, not just that
<sophiag>
it seems to have affected multiple graphics drivers and display managers last year
<srhb>
I see.
<sophiag>
i'm hoping once i get the channels fixed gdm (or whatever, i'm not too attached) will not crash
<srhb>
fwiw I've had very few problems with lightdm for years.
<sophiag>
oh, and of course it ran out of space
<srhb>
:(
<sophiag>
yeah, people seem to like lightdm
<sophiag>
but they also reported a similar problem with it last year
<sophiag>
and slim. and etc...
agjacome has quit [Ping timeout: 260 seconds]
<srhb>
Sounds very systemic.
<sophiag>
shrb: are you up on the change to per-user channels?
<sophiag>
that's what i'm trying to sort out now
<srhb>
sophiag: I don't recall any changes, but I guess I know how it works now?
<sophiag>
ha
<srhb>
16.09 might be one version before I started using NixOS :)
<srhb>
What's the problem?
<sophiag>
the multi-user symlink disappeared
<srhb>
Never heard of such a thing indeed.
<sophiag>
i figured i'd try rebuilding as a user for diagnostic purposes and it spent an hour copying paths...
<sophiag>
now out of memory and i figure i shouldn't gc since it hasn't built
<srhb>
iirc the default (on NixOS) is to have a system profile, a root (=default) profile and a profile for every user
<srhb>
sophiag: That sounds like a sensible precaution
<sophiag>
by profiles do you mean channel list?
<srhb>
sophiag: So I guess the goal is to acquire a gc root for a system profile?
<srhb>
sophiag: I mean /nix/var/nix/profiles
igo95862 has joined #nixos
<srhb>
THings like per-user/sophiag which should reflect every thing you've nix-env'ed
<sophiag>
oh, no. that's always been the case while i've used nixos
<sophiag>
i mean users can have different channels
<srhb>
Ah, OK
<srhb>
And you lost your channels? That's far less bad
<sophiag>
not quite. i think it just didn't change them when i upgraded nixos
<srhb>
I wouldn't worry too much at gc'ing those away as long as your *profiles* are intact, so you don't lose important bootstrapping programs
<srhb>
OK, what's sudo nix-channel --list and nix-channel --list ?
<sophiag>
since you're presumably not stuck in a tty: i'm referring to issue #40165
<sophiag>
my hypothesis is that this is why i was stuck with 17.09 packages on 18.09
<srhb>
It shoudl be a symlink into /nix/var/nix/profiles/per-user/sarah/channels
juhe has joined #nixos
igo95862 has quit [Client Quit]
<sophiag>
but do you still always rebuild from root?
<srhb>
I would think that lacking it for your USER would make it default to root's channels
<srhb>
nixos-rebuild? Yes
<sophiag>
ah ok
<srhb>
(Or, well, I don't use channels at all....)
<sophiag>
same, obviously
<sophiag>
and to be clear, the other way around: i had that warning in root
<srhb>
But yes, the default is to use root's channel
<srhb>
OK, that makes it sound like ~root/.nix-defexpr/channels does not exist?
<srhb>
fwiw, here's how it looks on my system. ~root/.nix-defexpr/channels is a symlink into /nix/var/nix/profiles/per-user/root/channels and for my user s/root/sarah
<sophiag>
yes. but the warning went away after i tried rebuilding from user (even if i didn't use `sudo -i`)
<srhb>
What's your $NIX_PATH?
<sophiag>
i'm gcing now. hoping afterwards i won't have the warning and can just update channels from root and rebuild
crmlt has joined #nixos
<sophiag>
i have to remember how to copy and paste in vi :p
<srhb>
yy line copy, p paste
<sophiag>
thanks :)
<srhb>
I tried reproducing it by setting up channels for both root and my user, no luck...
<srhb>
I feel like it might be your $NIX_PATH somehow
<srhb>
According to the user you linked, and since we see a literal $HOME, it's like nixos-rebuild does not expand it, and somehow that path is erroneously on your $NIX_PATH
<srhb>
Er, the ISSUE you linked
<srhb>
More coffee.
<sophiag>
i'm still here. the path looks normal...i should have just launched emacs instead of irssi
<sophiag>
ok, that warning _is_ gone...
<srhb>
I think the error is harmless if annoying, because later on in the NIX_PATH variable there should be a literal /nix/var/nix/profiles/per-user/root/channels
<srhb>
sophiag: boggle
<srhb>
Well, that's nice, though I can't explain why :D
<sophiag>
yeah, it may have not even been the issue
<srhb>
I doubt it was.
<srhb>
By the way, the out of memory thing
<sophiag>
i did have 17.09 in addition to unstable
<sophiag>
so deleted that
<sophiag>
oh, i gced and it finished
<srhb>
That's usually a symptom of Nix 2.x having non-constant fetches
<srhb>
That is, downloading bigger files takes up space proportional to the size of the download
<srhb>
(Or something to that extent)
<srhb>
Next release hopefully comes with a fix.
<sophiag>
ah. that's good to know. what prompted me to change stateVersion was seeing that i was still on 1.11.6
<sophiag>
which is pre-16.09 i think
Fare has quit [Ping timeout: 260 seconds]
<srhb>
It's probably mostly a good thing, but that bug is really annoying.
<srhb>
I think that sounds like 17.09-territory as well
<srhb>
2.x / 1.23 was only with 18.03
hotfuzz_ has quit [Read error: Connection reset by peer]
<srhb>
Er, 2.x / 1.12
<sophiag>
i suppose i should try and rebuild again
<srhb>
No guts no glory
<sophiag>
oh, i forgot...that warning is on update
<sophiag>
so, yes, it's still there...
orivej has joined #nixos
hotfuzz_ has joined #nixos
<srhb>
If you're not *too* annoyed with it, I think you can ignore it until there's a default fix.
<sophiag>
if i had a low annoyance threshhold i don't think i'd be in this channel :)
<srhb>
Did both your channels symlinks exist by the way?
<srhb>
Because that's like the only way I can reproduce it, by nuking one of them :P
<sophiag>
do you mean both as in root and user? i deleted 17.09 so there's only unstable
<sophiag>
rebuilding just yeilds that warning five times :/
<srhb>
I mean the literal channels symlink, ~/.nix-defexpr/channels and ~root/.nix-defexpr/channels
<srhb>
if I rm that, I can see at least one of those warnings
<sophiag>
ah, so the root symlink is gone
<sophiag>
but the user one is there
<srhb>
You could try recreating it by hand, it should be harmless
<srhb>
(there should still be a literal /nix/var/nix/profiles/per-user/root/channels in NIX_PATH, so that will get used anyway if earlier entries are not found)
<sophiag>
but i thought root is symlinked to my user now
<sophiag>
at least with the hack i pulled
<srhb>
Um, while you can do that that's not what I intended to make you do :D
<sophiag>
well, it thought it was a soft link but the changes are persisting
<srhb>
Again, my setup is as follows. /home/sarah/.nix-defexpr/channels is a symlink to /nix/var/nix/profiles/per-user/sarah/channels, while /root/.nix-defexpr/channels is a symlink to /nix/var/nix/profiles/per-user/root/channels
<srhb>
That should be the default setup.
<srhb>
However, what I'm advocating is that my user simply doesn't *have* any channels (nix-channel --list is empty) and root only has one, named nixos.
<sophiag>
i agree. that's where i'm at now. my nix-path is still screwed up tho :/
<sophiag>
i'm going to type it in here for visual comparison
<srhb>
Now the ONLY place to go looking for nixpkgs is in roots channels
<sophiag>
well, that'll do it
<sophiag>
is that assuming i didn't manually create a symlink in /root?
<srhb>
Regardless of whether or not we did. We're pointing directly at the target in /nix
FRidh has joined #nixos
vandenoever has quit [Quit: ䷴៚]
vandenoever has joined #nixos
vandenoever has quit [Changing host]
vandenoever has joined #nixos
<sophiag>
oh...so now i'm getting the warning for the reason stated in that issue. i don't have any root channels. i should delete that symlink
<srhb>
The symlink makes no difference
<srhb>
is sudo nix-channel --list empty?
<sophiag>
no
<srhb>
what
<sophiag>
but that folder is
<srhb>
:P
<srhb>
Oh, then the symlink must be pointing the wrong place
<sophiag>
yeah, not that i understand this better it definitely is...
<srhb>
fwiw sudo nix-channel --update might recreate it for you on sufficiently new versions
<srhb>
And make it point to the right place
graphene has quit [Remote host closed the connection]
hxrts_ has quit [Ping timeout: 256 seconds]
graphene has joined #nixos
vandenoever has quit [Ping timeout: 264 seconds]
<juhe>
Hehe, "nix-store --delete --ignore-liveness" is quite dangerous thing... Expected it to delete only one path, though it deletes stuff recursively. Is that intended?
<srhb>
juhe: With or without a path argument?
<Dezgeg>
yes, it will delete any dependents of the path as well, otherwise your store would be inconsistent
<srhb>
Without, I think it is essentially equivalent to rm -rf /nix/store :P
<juhe>
Maybe it makes sense to put a note in nix-store about that.
<juhe>
:-D
<juhe>
I mean nix-store man page.
<srhb>
--ignore-liveness should just have a big "this is not what you want ever."
<srhb>
Hmm, it does sya "however, the path still won't be deleted if there are other paths in the store that refer to it"
<juhe>
yes... I was looking for a way how to invalidate derivation, so that it is rebuild next time I invoke the nixos-rebuild switch... Well, in the end I achieved what I wanted, and even more than I wanted :-D
<srhb>
So I don't understand how you got into that situation
cryptomonoid has joined #nixos
<juhe>
Well, it just deleted 14 other store paths besides the python-env. I did not expected that though. Never mind, lesson learned the hard way, which is OK-ish.
<sophiag>
at what point did copying paths from cache.nixos.org become so slow?
<srhb>
juhe: On the plus side you get to use nixos-enter woo
<srhb>
sophiag: Sometimes amazon is having issues in a region
<sophiag>
i didn't consider that
<srhb>
sophiag: I get good speeds in northern europe currently.
<srhb>
I wonder what the largest path we have cached is...
<srhb>
Suggestions?
<sophiag>
hmm
<tertle||eltret>
samueldr:
<juhe>
srhb: thanks for info about nixos-enter, though will it work without copying missing paths first? If it is "just" a chroot thing, I think it won't help.
graphene has quit [Remote host closed the connection]
<srhb>
juhe: Well, you're already root from the usb stick
<srhb>
juhe: So if the rest of the machinery is still there, a chroot should be just what you need.
<juhe>
right...
graphene has joined #nixos
<srhb>
That was my thought at least :)
<juhe>
ok, :-), I'll hopefully get back soon, going to get those missing paths back...
<srhb>
juhe: Good luck
<juhe>
srhb: thx
derped has joined #nixos
juhe has left #nixos [#nixos]
cryptomonoid has quit [Ping timeout: 240 seconds]
<sphalerite>
sophiag: Graham recently enabled something that improved speeds significantly for most of Asia and Oceania
<sphalerite>
and apparently Amazon's geo-DNS thingy can be a bit wonky
<sphalerite>
Maybe you're getting the wrong servers
<juhe>
Hi, I'm back. In the end it was not painfull, had over 100 generations, so booted into older one, run nixos-rebuild switch, and all is well again.
<srhb>
Nice :)
jgt has joined #nixos
<sphalerite>
srhb: ghc stuff is pretty big
<srhb>
sphalerite: A single ghc is just 85 MiB
palo has quit [Ping timeout: 240 seconds]
<srhb>
Or is it though...
<sphalerite>
huh so what was this 1GB stuff that I was thinking of
<sphalerite>
pretty sure it's involved with ghc *somehow*
<sphalerite>
might be the ghc bootstrap packages?
<sphalerite>
In any case it's big enough that I had great difficulty getting it imported on my chromebook
<srhb>
sphalerite: I'm confused as well
palo has joined #nixos
palo has quit [Changing host]
palo has joined #nixos
<srhb>
sphalerite: these paths will be fetched (85.98 MiB download, 1117.17 MiB unpacked): /nix/store/xfhxsilhlrs38m4sjx0pnlkw54jqnj5n-ghc-8.2.2
<srhb>
And then nix daemon is OOM :D
<sphalerite>
aaaaah so it's crazy compressible?
wgas has joined #nixos
<sphalerite>
oh yeah, the graphical iso could be pretty big too, and probably isn't as compressible
<sphalerite>
the biggest store path on my server is vdrift's data, but that's not in the binary cache because of its size
<{^_^}>
[nixpkgs] @vbgl pushed commit from @marsam to staging « boost167: Add lockfree next_prior patch »: https://git.io/fNW0t
<sophiag>
are you still guessing what the largest packages are? i thought of haskell stuff, but then realized that in most cases i'd be conflating build times with source
fiddlerwoaroof has joined #nixos
<srhb>
sophiag: I am :P I wanted to curl ONE big thing to see transfer rates.
<sophiag>
it doesn't seem to be bandwith-limited. i think this was just a gigantic channel update for me
<srhb>
Right right.
<sophiag>
like now that i'm getting build errors, it's pretty fast
<sophiag>
presumably shouldn't it always break one one, which leaves an error code towards the end?
<srhb>
rebuild?
<sophiag>
the problem is that one in this case is not _my_ dep
<sophiag>
yes, rebuild
<srhb>
I think the original error can be higher up. In reverse order: x broke because of failed dependency y broke because of ....
agjacome has quit [Ping timeout: 268 seconds]
<srhb>
The last error will be the direct dependency, the first error will be the direct build error.
Anton-Latukha has quit [Ping timeout: 264 seconds]
<juhe>
Hm, how can I debug the problem of a package providing python library not being included into python environment even though it is specified in python27.withPackages ?
<sophiag>
srhb: thanks. different packages, but all from the same devs
<srhb>
juhe: My favourite approach is this
<srhb>
juhe: Paste the nix expression to #nixos and go Y U NO WORK
<FRidh>
juhe: are these Python bindings?
<juhe>
to be specific: (python27.withPackages (ps: with ps; [ kicad-unstable wxPython ]))
<FRidh>
and I suppose the issue is with kicad?
<juhe>
the kicad-unstable doesn't get into the python-2.7.15-env/lib/python2.7/site-packages/
<FRidh>
right
<juhe>
FRidh: right
<FRidh>
that's because it's not marked as providing Python modules
<FRidh>
python.buildEnv has been changed to only consider packages that are explicitly marked as having modules
<FRidh>
see python.pkgs.toPythonModule
<FRidh>
basically, all Python bindings should be offered via python.pkgs
<juhe>
FRidh: ok, thanks, I'll try it
thc202 has joined #nixos
<sophiag>
well, i better sure hope i don't need to switch channels again since that took two hours. and ironically 18.03 seems to have coincided with one company breaking most of their libraries :p
<srhb>
sophiag: If you stay relatively up to date it's pretty fast!
sophiag has quit [Remote host closed the connection]
<infinisil>
Imagine if nix were to have incremental updates somehow
<LnL>
yeah, we could drop it with stateVersion 18.09 or later
<sophiag>
is it possible to define hardware options in my configuration.nix? i'm getting an error on rebuild, but theoretically shouldn't edit hardware-configuration.nix
<srhb>
sophiag: Sure.
<infinisil>
sophiag: Any module is treated the same, configuration.nix and hardware-configuration.nix are no different
<srhb>
I swear it used to work though, so I suspect another driver had access to the sensor/fan earlier.
<etu>
I tried for a while to run a GTX 970 TURBO without driver to have PCI Passtrough with it. That damn thing was spinning away to the max TURBO speed of the fan all the time
<srhb>
haha
<etu>
(TURBO means "turbo sound" not performance, I learned that the hard way)
<srhb>
etu: Note, that it's not max speed here
<srhb>
It's "whatever speed was set when you card switched off"
<srhb>
If I switch it off with fan off, fan stays off
<sphalerite>
etu: placebos are real! :p
<etu>
srhb: That's slightly better but still annoying :D
eskytthe has left #nixos [#nixos]
<srhb>
Yep.
<sophiag>
i was also wondering whether i could just put "intel" in xserver and it would turn nvidia off (i don't have that option in bios)
<etu>
sphalerite: hah :)
<srhb>
Just give me the damn controls, nvidia :)
crmlt has quit [Quit: Lost terminal]
<sphalerite>
I bet many people would be impressed if they had a TURBO BUTTON for their graphics card that doesn't change anything besides spinning the fans up to full 😂
<srhb>
sophiag: Hmm, if you don't have the BIOS option I'm not sure, but it's easy to check
<sophiag>
i had to check the pci ports for optimus and the nvidia 960m is listed as "3D something" whereas intel is VGA...
agander has joined #nixos
<etu>
sphalerite: But the classic turbo-button was a button to clock down the CPU...
<srhb>
sophiag: With my card off, I get this from lspci -vvv 01:00.0 VGA compatible controller: NVIDIA Corporation GF108M [NVS 5400M] (rev ff) (prog-if ff)
<srhb>
!!! Unknown header type 7f
<srhb>
sophiag: With the card on, I get actual output instead of Unknown header type
<sphalerite>
etu: I don't mean a turbo-button, I mean a TURBO BUTTON
<sphalerite>
®™
<sophiag>
shrb: you mean if you turn it off you don't see it in lspci?
<srhb>
I see its header, but not the full output. let me paste for comparison...
<sphalerite>
huh, mine shows up in lspci even when off
<Izorkin>
infinisil: default PubkeyAcceptedKeyTypes +ssh-dss and HostKeyAlgorithms +ssh-dss
<{^_^}>
[nixpkgs] @FRidh closed pull request #43828 → stage: use fix' to allow deep overriding of `pkgs` from within Nixpkgs → https://git.io/fNczU
<infinisil>
Yeah but when the list is empt it's just a line "PubkeyAcfeptedKeyTypes" without any value on the right, which might throw an error when used
<LnL>
yeah, but you get a "PubkeyAcceptedKeyTypes " line in the config
<LnL>
checking if that's ok
johanot has joined #nixos
<srhb>
I think it is, and means "no pub key types are accepted"
<Izorkin>
LnL: if pubkeyAcceptedKeyTypes = []; to result - "PubkeyAcceptedKeyTypes "
sophiag has joined #nixos
<sophiag>
sometimes i feel like using nixos is a form of self-harm
agjacome has joined #nixos
<srhb>
sophiag: You seem to be going through more pain than most of us, indeed.
<infinisil>
It gets better over time :)
<srhb>
Like a good cheese. Once the rot sets in.
agjacome has quit [Client Quit]
<infinisil>
LnL: So is it a problem for ssh or not?
coot_ is now known as coot
<sophiag>
if i just put intel instead of nvidia services.xserver i get a tty. no lightdm. the display doesn't glitch and the keyboard works immediately, which seems great. but then it starts displaying cpu stall messages if i try to do...anything. like i really broke my entire system. i'm on a ~15 month old build of 16.09 that works now
<sophiag>
it's an asus "republic of games" (i didn't buy it)
<srhb>
I wasn't about to judge you :P
<sophiag>
ha
<sophiag>
it's mostly just really heavy and gets no battery life
<sophiag>
i should get a new one by the end of summer
<sophiag>
but it makes me question those razors, which are still maybe better than the new macbook pros. i can't do apple now regardless
<srhb>
Your config looks pretty minimal, but I'd still feel tempted to try once more with the absolute minimal configuration, just with an intel driver for the xserver...
<srhb>
At some point :)
<sophiag>
that's what this is
<sophiag>
i commented most things out
<srhb>
Oh well.
<srhb>
No good ideas then.
<srhb>
I want to get one of those core i9 hexacores, but... none of them are nice.
<zgrep>
I feel like nixpkgs should have a recommended way of doing things, or some sort of style guide, because there are so many ways to do, well, everything.
<infinisil>
zgrep: I agree, got a specific example?
jensens has joined #nixos
<manveru>
i'm still hoping for a nix linter and formatter though
<sphalerite>
manveru: from what specifically this time? :p
<sphalerite>
oh right so many ways to do everything?
<manveru>
yeah :)
<manveru>
i still love the perl manpage
<infinisil>
I'd like to have a library function for declaring a nixos service module, which creates a user, sets dataDi, declares options, conig file skeleton, etc.
<manveru>
"Perl actually stands for Pathologically Eclectic Rubbish Lister, but don't tell anyone I said that."
<manveru>
or >> The Perl motto is "There's more than one way to do it." Divining how many more is left as an exercise to the reader.
<manveru>
yeah
<manveru>
that'd actually make it much easier to support different init systems...
<zgrep>
infinisil: Currently I'm trying to package up https://www.sourcetrail.com/ , and the only way I've managed to get it working is a chroot/fhs env. So to figure out how to do what I need, I looked at its documentation, and through nixpkgs to see what others have done. I ended up mimicking what conda/common.nix does, because that seems to work. :/
<srk>
there's some prior work to support different init systems
<zgrep>
manveru: Obviously the linter and formatter should be written in perl. It's only natural.
<{^_^}>
#26067 (by copumpkin, open): Make a service abstraction layer
<{^_^}>
#26075 (by copumpkin, open): [WIP] One way to do other types of services
<zgrep>
infinisil: I believe not, it is a commercial endeavour.
<infinisil>
> sourcetrail, a source code explorer – doesn't provide source code
<{^_^}>
error: syntax error, unexpected ',', expecting ')', at (string):162:12
<zgrep>
Haha.
<infinisil>
Kind of ironic :P
<infinisil>
srk: neat!
<zgrep>
Yeah, it is ironic. :P I thought I'd try it out on a messy C++ codebase I need to do some stuff in.
<zgrep>
And I accidentally ended up learning a bit about how to package things using nix.
mekeor has joined #nixos
<manveru>
is it just one executable?
sam__ has joined #nixos
<infinisil>
My first instinct when I discover a new cool program is to try package it for nix
jensens has quit [Ping timeout: 244 seconds]
<zgrep>
manveru: No, it pulls in Qt libraries and whatnot. I tried initially setting the rpath, but it wasn't happy with nix's version of Qt, or something like that. Their precompiled .so files that they shipped seem to not work if I add them to the rpath (maybe they expect things to follow the FHS?). So my next step was to wallow around a bit, and I ended up manually making a chroot and running it in that. Until I found out that nix can do
<zgrep>
chroots too, which makes sense in retrospect.
<manveru>
my favorite package is still steam :)
<manveru>
even though it's hacky as hell
<manveru>
it's amazing how well it works in practice
<srk>
hehe, was looking at how it's done few days ago :)
<zgrep>
Eeeh. Not yet, I don't think so. It'd be nice if I could figure out the necessary magic to have it download the right binaries for 32 bit, and possible even if I could make it do the right thing for macOS, seeing as how they have binaries for those. But that's some other time, probably.
sam__ has quit [Ping timeout: 276 seconds]
<sphalerite>
zgrep: it's a start and it's more useful to other people in NUR than it is not in NUR, so I'm +1 for adding it :)
<manveru>
well, that's what i do for my quick and dirty packages, so someone else can at least try them easily, until i get around to clean them up
<zgrep>
Uh. Well. Uh. Sure, I guess.
<manveru>
otherwise i'd suggest nixpkgs :D
<sphalerite>
^
<manveru>
plus it's a fun exercise in distributed packaging
<infinisil>
NUR is just a link to you repo really, gets updated automatically
<infinisil>
Then you can just tell people to build nur.repos.infinisil.foo to get a package :)
<tilpner>
Except you don't have a repo yet! :/
<infinisil>
Yeah, i was just preparing one to add yesterday :)
<tilpner>
And got stuck finishing the module PR?
<infinisil>
Yes :D
<tilpner>
Just testing it now
<infinisil>
Neat, should be 100% compatible, only adds more possibilites
<tilpner>
There's now nur.repos.tilpner.modules.auto-tinc, just have to figure out how to import it
<tilpner>
(And without fetchTarball-ing NUR in every module)
<infinisil>
Unfortunately that's not possible nicely
<d1rewolf__>
guys, I sym link a file into my config directory (ln -s /home/d1rewolf/nixos-config/common/applications.nix /etc/nixos/applications.nix) and then import it from /etc/nixos/configuration.nix like imports = [./harware-configuration.nix ./applications.nix];
<srk>
LnL: the quote need to be before Pubkey - ${optionalString (cfg.pubkeyAcceptedKeyTypes != []) PubkeyAcceptedKeyTypes "${concatStringsSep "," cfg.pubkeyAcceptedKeyTypes}"}
<LnL>
yes, 9d72403
<d1rewolf__>
however, in this case, /etc/nixos/configuration.nix is also a symlinked file (ln -s /home/d1rewolf/nixos-config/machines/onemachine/configuration.nix /etc/nixos/configuration.nix). the problem is, at nixos-rebuild time, it tries to import applications.nix from /home/d1rewolf/nixos-config/machines/onemachine *instead* of /etc/nixos
<srk>
LnL: ah, ok :
<d1rewolf__>
so it tries to import from whatever directory configuration.nix lives in (really) and not its symlink..
<srk>
:)
<d1rewolf__>
how does one get around this or what's the better way of doing this?
<d1rewolf__>
should I not use symlinks and simply copy my configs over to /etc/nixos each time?
<LnL>
srk: I tested with build-vm but didn't add the quote changes before pushing -_-
<{^_^}>
nix#2109 (by Infinisil, open): imports don't resolve symlinks absolutely, leading to unexpected behaviour
<infinisil>
?
<d1rewolf__>
infinisil, that looks about right. so it's a bug :-/
ixxie has quit [Quit: leaving]
ixxie has joined #nixos
ixxie has quit [Client Quit]
<d1rewolf__>
does anyone have a good simple example of keeping your configs modularized? I have about 5 machines I'll be running nix on occasionally, and the differences between machines are small compared to the commonalities in my environment I wish to have across them. I had started down the path of using symlinks, but apparently that's buggy
ixxie has joined #nixos
<infinisil>
d1rewolf__: well not necessarily bug, more like potentially unconvenient semantics
<d1rewolf__>
infinisil, well, if I symlink into a directory, I'd expect nix to treat the symlink as if it were a real file...or a replacement for whatever file it's linked for. that doesn't seem to happen
<d1rewolf__>
infinisil, how do you handle machine-specific configs?
<infinisil>
d1rewolf__: i do, small how to: declare enable options for everything, use those instead of imports (import everthing by default to get the options available)
<d1rewolf__>
infinisil, instead of importing applications.nix from ~/nixos-config/common/appications.nix, it tries from ~/nixos-config/machines/holodeck/applications.nix. does that make sense?
<infinisil>
Hmm no i think that makes sense
<infinisil>
Wait what are your imports?
orivej has joined #nixos
<d1rewolf__>
infinisil, import ./applications.nis
<d1rewolf__>
er...
<infinisil>
Yeah i think that's expected behaviour
<{^_^}>
[nixpkgs] @FRidh opened pull request #43886 → vim_configurable: get rid of composableDerivation → https://git.io/fNW2Q
<sphalerite>
d1rewolf__: one thing I'd suggest for cross-machine modularisation is not to do too much too soon.
<sphalerite>
I made that mistake and it just results in a mess. Much more so than just keeping completely separate configs.
<d1rewolf__>
sphalerite, yeah, that's what I was trying. my symlink approach seemed workable, until it wasn't :(
<d1rewolf__>
maybe my answer for now is eliminating the relative import and importing with full path
<infinisil>
You can also have a tree of imports, clever does that
<d1rewolf__>
clever, when you're around, please share.
iyzsong-x has joined #nixos
<d1rewolf__>
so importing the full path works, but I'd like to avoid symlinks to something in my home directory. is there anything wrong with storing it somewhere like /etc/nixos-setup?
<d1rewolf__>
instead of /home/user/nixos-setup, as my user changes across machines?
<infinisil>
You can just set nixos-config=/custom/path/configuration.nix in NIX_PATH btw
<d1rewolf__>
or, perhaps it would be better to use the value of $HOME?
<d1rewolf__>
sphalerite, so do you "cp ~nixos-config/servers/kakuna.nix /etc/nixos/configuration.nix"?
<d1rewolf__>
or something else?
<sphalerite>
I have a symlink /etc/nixos/configuration.nix -> /nixos-config/servers/kakuna.nix
<d1rewolf__>
sphalerite, k. so my approach isn't unusual...it's just that i'm using other symlinks besides configuration.nix, and nix only seems to support symlinking configuration.nix
<d1rewolf__>
well, easily with relative imports
<sphalerite>
why do you need multiple symlinks?
<infinisil>
A symlinked file behaves like a file in the resulting place
<sphalerite>
yep
<d1rewolf__>
sphalerite, " guys, I sym link a file into my config directory (ln -s /home/d1rewolf/nixos-config/common/applications.nix /etc/nixos/applications.nix) and then import it from /etc/nixos/configuration.nix like imports = [./harware-configuration.nix ./applications.nix];"
<d1rewolf__>
it was an attempt to let applications.nix vary across machines as well
<d1rewolf__>
so that I could symlink in the appropriate one
<sphalerite>
any reason why you wouldn't just have a single symlink into the repo?
<sphalerite>
at that point I think it would make more sense to just have no separate applications.nix and put those in the individual machine files.
wangoe has joined #nixos
<d1rewolf__>
sphalerite, you mean use the entire repo as /etc/nixos?
<d1rewolf__>
sphalerite, yeah, seems like the easiest way at the moment
<sphalerite>
that, or the same way I use the nixos-config repo
<d1rewolf__>
while I learn to walk ;)
<sphalerite>
if the applications aren't the same across machines it doesn't make sense to have shared files for them :p
<d1rewolf__>
sphalerite, yeah, perhaps I'm abstracting too much
<sphalerite>
or another way of looking at it is: "applications" isn't a good abstraction.
<d1rewolf__>
right
<d1rewolf__>
;)
<sphalerite>
You want to look at the roles that the machines are fulfilling, not so much the things that get set as part of that role
ambro718 has joined #nixos
<d1rewolf__>
sphalerite, got it. thanks
* d1rewolf__
smacks head
<d1rewolf__>
I guess I could just play along with the way it's importing and do "import = [ ../../common/applications.nix]" instead
<d1rewolf__>
same functionality, and avoids the symlink. that'd let me keep it simple for now. thanks guys ;)
<d1rewolf__>
I am so freaking HAPPY with nixos so far. it's a far bit of work up front, but oh my GOD the pay off. It's what I've wished for for so long ;)
<{^_^}>
iqubic: Fancy way to uninstall packages, needs fzf installed: nix-env -q | fzf | xargs -I{} nix-env -e {}
<d1rewolf>
I really have a little tear in my eye right now
<d1rewolf>
that may be the kratom I ingested an hour or go, but I think it's more likely nixos ;)
<iqubic>
infinisil: I kinda deleted the .nix files I used to install the packages that I now want to uninstall. What should I do?
<infinisil>
Have you tried fancy-uninstall
peterzky has joined #nixos
<iqubic>
no.
<infinisil>
then do that
<infinisil>
d1rewolf: Haha
<infinisil>
d1rewolf: There is a huge problem with NixOS though
<d1rewolf>
infinisil, oh no
<d1rewolf>
infinisil, wait....let me ride this wave of bliss for a few more seconds...
<d1rewolf>
ok, I'm ready: tell me
<infinisil>
d1rewolf: Which is that you won't be able to use any other distro without longing for NixOS' features
<srk>
:))
<d1rewolf>
infinisil, I'll take that risk! \o/ ;)
<d1rewolf>
has anyone started using nix as a general purpose scripting language?
<d1rewolf>
wondering ways to force myself to learn it more deeply
<infinisil>
No, and for good reason, it's not a general purpose language
<d1rewolf>
infinisil, k. I'll take that as sage wisdom ;)
<srk>
d1rewolf: I've had the same feeling recently, like a new breeze since everbroken fedora/EL world
<infinisil>
It's a language specifically designed to build derivations and it can't do anything besides that (by default at least)
<iqubic>
Well, I really wish that I could have some swap on zfs.
<iqubic>
nix daemon runs out of memory when the only change to configuration.nix I have made is to remove nodePackage.eslint and add fzf.
<srk>
heh, how much ram?
<d1rewolf>
srk, I've been running linux as my default desktop since 1998. over the past few years, I've had three linux laptops and one iMac. It really killed my productivity because each machine was different. I finally got rid of the iMac, so one huge productivity sink was gone (yes, I'm far more productive on linux).
graphene has quit [Remote host closed the connection]
<srk>
:)
<d1rewolf>
but now nixos has given me consistency across machines. there is no going back
<iqubic>
srk: 8G
<infinisil>
I guess it would be possible to implement a RTS on top of Nix though lol
<iqubic>
8 gigs of ram.
peterzky_ has joined #nixos
graphene has joined #nixos
<srk>
sounds plenty, do you have a lof of stuff/packages in configuration.nix?
<iqubic>
Yes.
<iqubic>
I do actually.
<srk>
I've moved most of my user packages to ~/.config/nixpkgs as I rebuild the system frequently from master and didn't want to wait
<iqubic>
I have a bunch of packages listed in systemPackages.
<iqubic>
srk: Will that fix my current problem?
<infinisil>
iqubic: why not just add some zvol swap?
<iqubic>
zfs can have swap? I didn't think that was possible.
<{^_^}>
[nixpkgs] @leenaars opened pull request #43887 → ipgrep: init at 1.0 → https://git.io/fNWVv
<srk>
don't know really, removal of eslint and addition of fzf doesn't sound like something that memory heavy
<infinisil>
iqubic: Read the zpool manpage, section on "zpool create" with the -V flag
peterzky has quit [Ping timeout: 256 seconds]
<srk>
but it depends, sometimes if you have tons of stuff in caches nix will fail to allocate enough mem
<srk>
at first
<srk>
then the system flushes some of that and it works again..
<infinisil>
This memory fix should land in stable nix already damnit
<iqubic>
infinisil: Don't I need to have some free space in order to make a swap partition?
<srk>
you can use swap file as well
<iqubic>
with zfs?!?!
<infinisil>
iqubic: no need for partitioning, just use zfs
<srk>
yeah, swap is pretty agnostic
<srk>
swap file is just a regular file on the filesystem
<d1rewolf>
manveru, you around?
<infinisil>
Zfs zvols
<manveru>
yeah
<infinisil>
srk: i think i heard that zfs has problems with that though, not sure
jensens has joined #nixos
<iqubic>
infinisil: I'm looking at man zpool, in the section where it talks about zpool create. I'm seeing nothing related to a -v flag.
<d1rewolf>
manveru, I'm trying to decipher your polybar config. I currently use i3blocks by changing the i3 status_command parameter. I don't see you doing that (yet) for polybar. can you tell me where you hook polybar in?
<infinisil>
iqubic: Big V
<d1rewolf>
also, what fonts are you using (manveru)?
<infinisil>
iqubic: or search for "zvol"
<manveru>
d1rewolf: i start polybar in each machine differently, so for example there's one in kappa.nix
<manveru>
home-manager has a service module for it
<iqubic>
I just searched the zpool man page for the string "zvol" no results found.
<d1rewolf>
manveru, ok, cool. not using home-manager yet but interested in a good starting polybar setup
<d1rewolf>
guys, I notice that installing virtualbox doesn't install kernel modules. what's the appropriate way to do this on nix?
<d1rewolf>
sphalerite, I'm curious, for my own learning...how did you know in the case of virtualbox that you need this specific line instead of just installing it to systemPackages?
<d1rewolf>
just experience?
<iqubic>
infinisil: How is the -V flag supposed to help me?
<infinisil>
iqubic: how about googling "zfs swap"
<sphalerite>
d1rewolf: systemPackages is (roughly) just "stuff that's supposed to be in $PATH". Anything that needs further config like kernel modules needs to be modules
<infinisil>
iqubic: you need to improve your googling skills
<d1rewolf>
sphalerite, right, but as a user, when I'm looking for package "A", how do I know "A" needs more configuring than just being in the path?
<{^_^}>
[nixpkgs] @Ekleog opened pull request #43888 → matrix-synapse: bump to 0.33.0 → https://git.io/fNWVl
<iqubic>
Can I decrease the size of zpool while my machine is running?
ixxie has quit [Ping timeout: 244 seconds]
<infinisil>
d1rewolf: You have a point, it would be neat if there could be warnings "the option foo already adds bar to systemPackages, no need to do it in your config"
<sphalerite>
d1rewolf: yeah I see what you mean
<sphalerite>
d1rewolf: I guess one way to do it would be to always check if there's an option first
<sphalerite>
but yeah it's not wonderful
<infinisil>
iqubic: did you create a zvol now
<iqubic>
no.
<infinisil>
Then do that
<d1rewolf>
sphalerite, what do you mean "check if there's an option first"? ho wwould I do that?
<iqubic>
I'm not sure I have enough free space to do tha.
<iqubic>
All the guides I have seen online reference a swap command. I don't have that command available.
<infinisil>
iqubic: create a zvol
<infinisil>
Do it now
<infinisil>
Use google if you don't know how
<iqubic>
I will. Let me just look up how to do that.
<infinisil>
Come on, I won't hold hands for every single thing, you gotta be able to stand on your own
<d1rewolf>
sphalerite, I got you. thanks ;)
slaytor has joined #nixos
<iqubic>
infinisil: I'm actually looking up the relevant info now.
slaytor has left #nixos [#nixos]
<{^_^}>
[nixpkgs] @Ekleog opened pull request #43889 → matrix-synapse: bump to 0.33.0 → https://git.io/fNWVu
<iqubic>
How big do I want this swap space to be?
knupfer has quit [Ping timeout: 256 seconds]
<iqubic>
Also, I think I want "zpool add" and not "zfs create -V"
<d1rewolf>
iqubic, the general advice used to be that swap space is twice the size of physical ram, but I'm not sure that really holds up any more
<d1rewolf>
I still often use it tho
<iqubic>
so I'd want 16Gigs of swap? That sounds like a large amount.
<d1rewolf>
iqubic, if you don't even know what size to make your swap, I'm wondering if you're pretty new to linux? If so, maybe zfs is a little too complicated right now? I don't know
<d1rewolf>
iqubic, it's only a lot of if you don't end up using it
<d1rewolf>
but yeah, i hear you. create it as 8GB for now. you can always change later
derped has quit [Remote host closed the connection]
<iqubic>
infinisil: The internet tells me that I can't shrink a zfs pool once created. Does this mean that I can't create any swap right now?
<infinisil>
iqubic: Alright this is my last reply to you: Learn to google (again), you want zfs create -V, and also you only want 1GB swap
<iqubic>
infinisil: Why do you have to be so rude/
FRidh has quit [Quit: Konversation terminated!]
ixxie has joined #nixos
<infinisil>
iqubic: Because you're not learning anything and I don't have time to answer every question of yours multiple times, especially when you often just ignore the advice
<d1rewolf>
iqubic, infinisil is giving good advice. he's teaching you to fish rather than just handing you a trout
<d1rewolf>
;)
<iqubic>
So I created a new thing to my zfs set-up.
<iqubic>
This article by oracle suggests that I use "swap -a"
<iqubic>
But I can't because the "swap" command is not found.
<iqubic>
termbin.com/b8v1
<iqubic>
That's the result of "zfs list"
<infinisil>
iqubic: GOOGLE
<infinisil>
"swap command not found"
<infinisil>
It's simple
<infinisil>
Come on
<infinisil>
The solution is that the command is swapon on linux
<d1rewolf>
iqubic, you're reading a blog post which uses solaris
<iqubic>
Oh.
<d1rewolf>
as vsftab is specific to solaris
<d1rewolf>
but truly, you're out of your depth. you're trying to fly an 777 by reading a few paragraph blog post
<d1rewolf>
you *may* get it off the ground, but a crash will soon come. as infinisil stated, you really need to spend more time reading/googling about what you're doing
<iqubic>
Will running "mkswap /dev/zvol/latitude-tank/swap" hose my system?
<d1rewolf>
and I sure hope you're doing it in a vm so you don't kill your system
johanot has quit [Quit: Leaving.]
<d1rewolf>
iqubic, I think it's safe to say your system will be hosed soon unless you seek to understand what you're doing (no offense)
<iqubic>
d1rewolf: I am working in a vm.
<d1rewolf>
but if you've created a swap partition, and it's located there, then probably you'll be ok.
<iqubic>
Which I cloned before starting this proccess.
<d1rewolf>
iqubic, good ;)
<d1rewolf>
so, I'm not using zfs, but if that's a real parition, then mkswap <partition path> and then swapon <partition path> should do it i think
<srk>
yup
ixxie has quit [Ping timeout: 244 seconds]
tenten8401 has joined #nixos
<iqubic>
At this point I'm not sure if it is a real partition or not.
<iqubic>
clever: Can I tell it that my partition also supports discard to get added performance?
<clever>
i dont see an option for that
<iqubic>
Oh. I see.
<vaibhavsagar>
how do I get all the build inputs of a derivation?
<vaibhavsagar>
e.g. the contents of drv.env for a haskell derivation
<iqubic>
clever: We should really add an option for that.
<d1rewolf>
those of you using i3-gaps...do you occasionally see black boxes behind certain windows?
<sphalerite>
vaibhavsagar: all the buildInputs -> drv.buildInputs
<sphalerite>
vaibhavsagar: all the dependencies, I'm not sure
<d1rewolf>
also, quick survey: what terminal are you using?
<manveru>
ocharles_: you around?
<sphalerite>
d1rewolf: st
<sphalerite>
+tmux
<infinisil>
d1rewolf: alacritty + tmux :)
<iqubic>
emacs's eshell.
* d1rewolf
uses terminator, but looking for a better alternative
<manveru>
urxvt(d|c)
<iqubic>
clever: Can I specify a device by path?
<iqubic>
when setting up swap.
<d1rewolf>
manveru, I've seen a number of folks using urxvt. I thought that was an older terminal which was aging. what's the appeal of it out of curiosity?
<manveru>
well, it's super fast, can be scripted with perl, can make sure you don't paste dangerous stuff, works well with unicode, has server-client mode...
<TheBrayn>
What are your thought about nixos on btrfs? Are there any btrfs features which might be interesting for nixos or should I just stick to LVM with ext4?
<d1rewolf>
Dezgeg, can you explain further?
<d1rewolf>
what you mean?
<Dezgeg>
it was response to the "server-client style with an open json protocol"
knupfer has joined #nixos
<infinisil>
Dezgeg: Oh yeah that might be a bit slow..
<iqubic>
there should really be a way to specify options for swap devices, like we do with other fileSystem stuff that we want to add our fstab
<d1rewolf>
Dezgeg, ah, sry ;)
<sphalerite>
Dezgeg: but there's already IPC on every keystroke (X11) ;)
<Dezgeg>
x)
<d1rewolf>
infinisil, how have you found alacritty? that's the one written in rust iirc?
<manveru>
that input latency is the only thing i dislike about compton...
<infinisil>
d1rewolf: Yeah, it's decent, few problems
<infinisil>
but generally pretty good
<sphalerite>
TheBrayn: btrfs features certainly make sense. I've heard a bunch of horror stories about it though. Personally I use zfs which has similar features without the history of data loss :p
* d1rewolf
nods
<Dezgeg>
doesn't look like terminator can do scripting
<d1rewolf>
Dezgeg, manveru: I'm curious...what sort of things do you typically use the perl+urxvt capability?
* srk
for url matching :D
<Dezgeg>
I have a script that can open URLs with hotkeys without using the mouse
<d1rewolf>
Dezgeg, ah, I see
<infinisil>
Oh, tmux can implement lots of such functionality
<infinisil>
It also has a plugin for opening urls iirc
<sphalerite>
yeah I like tmux because it allows using the same features with or without X
<infinisil>
I'll eventually make a tmux plugin manager with nix
<sphalerite>
then a simple terminal like st just serves as a bridge between X and tmux
<manveru>
d1rewolf: changing font size, copy/pasting, confirming pastes, search scrollback, and matcher... nothing too fancy
eskytthe has joined #nixos
<manveru>
and i don't write any of those myself
MinceR has quit [Ping timeout: 240 seconds]
<TheBrayn>
sphalerite: which features would you consider especially relevant for nixos?
ekleog has quit [Quit: WeeChat 2.0]
<TheBrayn>
snapshots are pretty much irrelevant as far as I understand
<sphalerite>
TheBrayn: they're as relevant as on any other OS
<sphalerite>
just not for the nix store
<sphalerite>
but they're still very useful for all the mutable data that isn't nix's/nixos's responsibility to manage
<sphalerite>
e.g. your $HOME
<TheBrayn>
hmm, true
MinceR has joined #nixos
<TheBrayn>
I guess I'll use that again
<d1rewolf>
manveru, mind if I pm you an i3 question (since you're around an experienced with it)?
<sphalerite>
resilience against silent data errors is very nice too
<sphalerite>
and compression! I love my compression.
<manveru>
d1rewolf: go ahead :)
<TheBrayn>
d1rewolf: I have been using i3 for quite some time, maybe I can help
* sphalerite
also uses i3
Fare has joined #nixos
<d1rewolf>
well, ok: if I have four windows in vpslit (stacked on top of each other) and I select the bottom one, and then press $mod+s, it changes to stacked for all four windows (as if the parent container). I just want to set stacking for the bottom window. how can I do that?
<sphalerite>
off-topic is usually welcome here, within reason. If there are on-topic discussions where it would get in the way there's always #nixos-chat
<d1rewolf>
in other words, how can you cause operations to create a new container for the currently selected window and apply to that, instead of applying to the parent container?
<d1rewolf>
sphalerite, last time, I promise ;)
<sphalerite>
what I'm saying is it's fine to ask here :)
<d1rewolf>
I did ask in #i3, if anyone cares to answer there
<d1rewolf>
ah, k ;)
<TheBrayn>
you can create a new container on the bottom one, move the one above down and then apply stacking to the container
<manveru>
stacking always stacks at least 2, if i see right?
<d1rewolf>
TheBrayn, maybe the "create new container" function is what I'm missing
<d1rewolf>
TheBrayn, how does one create a new container? I typically just let i3 do that for me
<d1rewolf>
manveru, my problem is that I have one container with four windows, so if I change the mode of the bottom window, it applies to all the windows (changes to stacking for example for all windows)
<d1rewolf>
what I want to do is leave the top three in splitv, and then make the bottom window stacked as I add others to it
<TheBrayn>
mod+h
<d1rewolf>
interestingly, if you're in splith, this seems to be the right behavior
johanot has joined #nixos
<d1rewolf>
TheBrayn, $mod+h changes to splith for current container, right?
<TheBrayn>
uhm
ekleog has joined #nixos
<d1rewolf>
so in splith, if you have three windows side by side [] [] [], and select the right most window, then lets
agander_ has joined #nixos
<d1rewolf>
say $mod+s to set the right most window to stacking, then go back to middle window and move it right, it goes into the stacked container on the right
<d1rewolf>
that's what I want to do in a similar way in splitv
<iqubic>
So I added the swap stuff to my configuration.nix, and it is taking a while too rebuild the system configuration.
<TheBrayn>
I'm confused because my shortcuts are all diferent, I don't use qwerty
<iqubic>
Wait, there it goes.
<d1rewolf>
actually, it seems that it doesn't always work in splith either. hmm.
<d1rewolf>
i'm missing something here
<sphalerite>
d1rewolf: open empty workspace, open terminal, $mod+v, open terminal, $mod+h, open terminal, $mod+s
<sphalerite>
d1rewolf: I think this does what you want?
<sphalerite>
assuming default keybindings of course
orivej has quit [Ping timeout: 268 seconds]
agander has quit [Ping timeout: 276 seconds]
<iqubic>
Adding swap to my machine fixed my nix daemon out of memory erros.
FRidh has joined #nixos
<iqubic>
Even just 2 gigs made a big difference.
<sphalerite>
d1rewolf: maybe also $mod+w instead of $mod+s so you have the title bars taking up less space
cryptomonoid has joined #nixos
<iqubic>
,fancy-install
<{^_^}>
iqubic: Did you mean fancy-uninstall?
<{^_^}>
Fancy way to uninstall packages, needs fzf installed: nix-env -q | fzf | xargs -I{} nix-env -e {}
<d1rewolf>
sphalerite, close, yes. let me play with that a bit. I more interested in when you have existing windows.
<d1rewolf>
I guess what I'm looking for is the opposite of "select parent"
<d1rewolf>
sphalerite, let me try that. my bindings are slightly different
<iqubic>
What are we talking about here? What program are these bindings for?
<sphalerite>
hsplit, focus up, move down, change to tabbed
<d1rewolf>
iqubic, i3
<sphalerite>
actually that's easier to write as well x)
<iqubic>
fancy-uninstall is only giving me two options to select from when it gets to the fzf stage of the pipeline
<iqubic>
,fancy-uninstall
<{^_^}>
Fancy way to uninstall packages, needs fzf installed: nix-env -q | fzf | xargs -I{} nix-env -e {}
<d1rewolf>
sphalerite, that definite does it. thanks. however, I'm confused, as it appears sometimes when I select the bottom window, it appears to select the parent instead ($mod+h in your example ends up affecting parent container instead of just bottom window)
<iqubic>
So IDK what's up wit that.
<sphalerite>
how do you select the bottom window?
<d1rewolf>
sphalerite, that's what I'm trying to figure out now. I'm wondering if when it's not worked in the past I've been selecting with the mouse
<d1rewolf>
I am not using compton at all yet. I don't want to pollute this channel with more off-topic...I'm off to google compton+i3, but if you have any good links, feel free to pm them my way ;)
<d1rewolf>
infinisil, i see an opacity setting in alacritty.yml
i-am-the-slime has joined #nixos
<manveru>
d1rewolf: see my config ;)
<infinisil>
manveru: I have that fancy kawase blur compton version, but unfortunately it's suuuper laggy on my NVidia GTX 980Ti :( (It works just fine on my crappy laptop with intel hd graphics, lol)
<infinisil>
d1rewolf: Ah, haven't heard of that :O
<d1rewolf>
I am using an nvidia card, but not the nivida drivers because I get better battery life with the intel chipsets (I think this is a prime card). would nvidia drivers be required to use compton?
<infinisil>
d1rewolf: Btw there's #nixos-chat for offtopic things :)
<d1rewolf>
infinisil, joining now
<{^_^}>
[nixpkgs] @periklis opened pull request #43892 → skhd: bump 0.1.1 to 0.2.2 → https://git.io/fNWrG
<d1rewolf>
infinisil, I'll carry on the discussion in #nixos-chat
<zgrep>
I'm trying to make my own nix channel, because why not, and I can't seem to get nix to work with it properly. I do nix-channel --add $url && nix-channel --update, and unlike nixpkgs which is in /nix/store/...-nixpkgs.../nixpkgs/default.nix, my things ends up being /nix/store/...-nameofchannel/default.nix. Any ideas on what I might be doing wrong?
juhe has left #nixos [#nixos]
Mic92 has joined #nixos
<clever>
zgrep: the nixpkgs channel has a nixpkgs subdir
<clever>
in the tar itself
<zgrep>
Hmm. It wasn't there when I downloaded it?
<zgrep>
I deleted everything using "ls /nix/store | grep znur | xargs -I{} nix-store --delete /nix/store/{} --ignore-liveness", and... things worked. Before that I only tried running "nix-store --delete /nix/store/*-znur --ignore-liveness", and it did the wrong thing consistenly (and probably didn't delete the .drv file). As best as I can tell, that's the only thing that's changed.
<clever>
if you --ignore-liveness, you risk seriously breaking the machine
<clever>
it can delete the current nix-env profile, and basically uninstall ALL programs at once
goibhniu has quit [Ping timeout: 256 seconds]
<clever>
zgrep: and because of how nix works, deleting things like that is pointless, nix will perfectly recreate the exact same files at the same paths every time
<clever>
thats the whole point of nix
<infinisil>
zgrep: Yeah that's bad, you should a `sudo nix-store --verify --check-contents --repair` or so to fix any things you potentially broke
<clever>
its far more likely, that you messed up the tar on your first test, and ~/.cache/nix/tarballs was caching the contents of the tar for 1 hour
<clever>
and nix-channel was refusing to download the new tar until an hour had passed
philippD has joined #nixos
<zgrep>
Aah.
<zgrep>
That would be it, yes.
<sphalerite>
infinisil: no, nix-store --delete is safe
<sphalerite>
oh wait no
<sphalerite>
I didn't see the --ignore-liveness
<clever>
--ignore-liveness still wont corrupt the store, but it can delete important paths that leave the machine unusable
<clever>
rotaerk: i think nixon is special, us president
<rotaerk>
:P
<{^_^}>
[nixpkgs] @peti pushed to master « hackage-packages.nix: automatic Haskell package set update »: https://git.io/fNWKf
<d1rewolf>
hmmm... nix-env -qaP '*.android-stu.*' shows android-studio, but putting this in user.users.d1rewolf.packages states "undefined variable android-studio". what should I use there in this case?
<{^_^}>
value is an integer while a string was expected, at /var/lib/nixbot/state/nixpkgs/lib/attrsets.nix:283:5
<infinisil>
Goddamnit
<d1rewolf>
clever, If I add it to packages for my user, I get error: The option value `users.users.d1rewolf.packages.[definition 1-entry 2]' in `/etc/nixos/configuration.nix' is not of type `package'.
<clever>
so when you try to lookup .foo.bar, it will find foo in the hash table, then use its index to search the set
<infinisil>
clever: Doesn't that only happen with static attributes, not dynamic ones?
<clever>
infinisil: it also happens with dynamic ones
<tilpner>
clever - I'm not going to generate all possible keys (unless that can happen lazily). I was just wondering if there was an __index
<d1rewolf>
clever, ok, so I think I follow. However, when I search for webstorm, it returns 'nixos.jetbrains.webstorm webstorm-2018.1.5'. However, searching for nix-env -qaP '.*jetbrains.*' doesn't return this package. so how can I match the jetbrains in nixos.jetbrains.webstorm when searching?
<infinisil>
(jetbrains does get recursed into though, not sure why you can't find the packages, I can find them)
<tilpner>
infinisil - I wanted local-nur.modules.* to evaluate to an empty module, as an eval-time optimisation (because eval here is two stages). Now I just pass the full local-nur for both stages, which is a little slower, but I'll manage. It was mostly a "I wonder if this is possible" question
<infinisil>
The regex thing is something conceptually totally different
<Mic92>
d1rewolf: nix search jetbrains ?
<infinisil>
The regex search uses derivation names, which are not the same as attr paths (as output by -P)
<infinisil>
(-P outputs attr names on the left, derivation names on the right)
<d1rewolf>
Mic92, yeah, that seems to work, although "nix search" is new to me. I've been told to use nix-env -qaP by others
<Mic92>
d1rewolf: the nix command was added with nix 2.0
<d1rewolf>
infinisil, ah, so it's matching the regex against names and not attributes.
<infinisil>
Yeah
<d1rewolf>
Mic92, ah, k. thanks
<infinisil>
I never used the regex thing because it sucks imo
rev_strangehope has quit [Ping timeout: 260 seconds]
<d1rewolf>
infinisil, fzf seems well equipped
<infinisil>
I'd recommend nix search or `nix-env -qaP | fzf`
<infinisil>
But actually I always just rg the nixpkgs tree for stuff :P
<d1rewolf>
sphalerite, and now I'm back with i3. I select a window, and press $mod+s, but it stacks the parent. /me scratches head
i-am-the-slime has quit [Remote host closed the connection]
poapoa has joined #nixos
stratact has joined #nixos
jD91mZM2 has joined #nixos
<stratact>
jD91mZM2: o/
<jD91mZM2>
\o
<poapoa>
Heya, trying my luck here - there was a repo that installed nixos alongside existing distro with an option to revert, but I can't remember it's name...
<stratact>
Hi, what's the recommended way to update packages through nix? I'm using the nix package manager on gentoo.
orzo has quit [Ping timeout: 260 seconds]
<jD91mZM2>
There was something about nix-env -u updating using names instead of attributes, but I can't find it
<{^_^}>
[nixpkgs] @FRidh opened pull request #43900 → php: get rid of composableDerivation → https://git.io/fNWiy
<sphalerite>
stratact: nix-env -u will update packages in your user env to the latest versions (after updating your package info using nix-channel --update), but doesn't work well for all cases
<sphalerite>
I want to add a method to FSAccessor in nix/src/libutil/fs-accessor.hh for fully resolving symlinks. I'm not sure how to return the various possible results though — there's either A, a final store path, B, a final path outside the store, or C, a loop. Anybody who knows C++ and nix's C++ style who could suggest the best way to do this?
iqubic has quit [Remote host closed the connection]
<clever>
sphalerite: splits exists within a tab on vim, so each tab has its own split layout
<sphalerite>
clever: we were talking about i3
<clever>
ah
<ghasshee>
Does anyone know how to include libudev.h in a formal way ?
<ghasshee>
I could figure out with locate but are there any variables to include libudev ?
<clever>
ghasshee: i believe you just add libudev to the buildInputs and then #include <libudev.h>
iyzsong-x has quit [Read error: Connection reset by peer]
agander has quit [Ping timeout: 264 seconds]
<sphalerite>
ghasshee: for building stuff in nix what clever said will be enough. On non-nix systems you'll probably need to do some pkg-config stuff or something
Cale has quit [Remote host closed the connection]
aanderse-laptop has joined #nixos
<mjrosenb>
Is there a channel for haskell on nixos?
<mjrosenb>
or, I guess in this case, nixpkgs
orivej has joined #nixos
<sphalerite>
mjrosenb: this one ;)
cryptomonoid has quit [Ping timeout: 248 seconds]
<ghasshee>
clever: sphalerite: Thank you! I confused since I could not find it with `nix-env -qa '.*udev.*'`, and how to find it ?
<mjrosenb>
so, I used cabal2nix to generate default.nix, then wrote test.nix which instantiated it, overriding some annoying stuff with heist. I ran nix-shell test.nix, got a shell, but now cabal build fails because it is missing a bunch of deps.
agander has joined #nixos
<sphalerite>
ghasshee: it's provided by systemd
<clever>
mjrosenb: use nix-shell test.net -A env
<sphalerite>
ghasshee: also, `nix search` is a lot nicer to use than nix-env -qa :)
<mjrosenb>
well, I used nix-shell test.ni -A project, since that was the name of the thing where the output of default.nix was bundled.
<clever>
mjrosenb: by default, nix ignores the versions in the cabal file
<clever>
mjrosenb: the simplest thing is to run haskell.lib.doJailbreak over it, or just remove them from the cabal file
<clever>
or adjust them to include nix's version
igo95862 has quit [Quit: igo95862]
<mjrosenb>
clever: that may have worked this time. If the package doesn't work with the new version at all, do I have any recourse other than to upgrade thw project I'm trying to build?
periklis has quit [Ping timeout: 276 seconds]
<clever>
mjrosenb: you can add a package override to your test.nix to change the version of the other package
<mjrosenb>
that is good to hear.
kerrhau has joined #nixos
kerrhau has quit [Changing host]
kerrhau has joined #nixos
freusque[m] has quit [Read error: Connection timed out]
freusque[m] has joined #nixos
zduch4c has joined #nixos
<zduch4c>
hullo
<tenten8401>
Anyone know if there's a good way to manipulate a .nix file from the command line?
<tenten8401>
for instance, changing or adding values to it
<tilpner>
ed c.c
<zduch4c>
when launching eolie, I get an error: ```ValueError: Namespace GtkSpell not available```. this is obviously because of a missing dependency, but there is no such package (python-gtkspell) in NixOS repos
<tenten8401>
ed doesn't have actual parsing as far as I'm aware
<zduch4c>
im not sure if a broken pacakge could land in stable repos, so this might be some other problem
<tilpner>
tenten8401 - You were just asking about editing the text, not semantic editing support
<tenten8401>
zduch4c: there's gtkspell2 and 3 as packages
<tilpner>
tenten8401 - Consider manipulating JSON, importing that from your .nix, acting depending on imported value
<tilpner>
JSON is easier to manipulate than Nix
<tilpner>
If you really need to, look at hnix or Nix plugins
kerrhau has left #nixos ["WeeChat 2.1"]
<zduch4c>
tenten8401: yes, I have tried installing them but it doesn't work still, there must be a python-gtkspell package or maybe its some other reason
<tenten8401>
is there a way to convert Nix to JSON or would I have to make a configuration that parses the json and does it's thing?
<tilpner>
> builtins.toJSON config
<{^_^}>
"{\"allowUnfree\":true}"
<tenten8401>
hmm... I might be able to work with that
<sphalerite>
tenten8401: functions can't be serialised into json though
<tilpner>
Which is probably okay
<sphalerite>
tenten8401: your best bet if you want something properly powerful and reliable is probably hnix
<joepie91>
[18:54] <sphalerite> tenten8401: functions can't be serialised into json though
<sphalerite>
but yeah if you just need to manipulate plain data JSON is probably easiest :)
<joepie91>
challenge accepted? :D
<sphalerite>
joepie91: ok ok they can't be serialised trivially
<sphalerite>
:p
<joepie91>
hehe
<tenten8401>
I was thinking of making a program that could manipulate a configuration.nix from a GUI of some sorts, would hnix be my best option for something like that or would exporting to json probably work alright?
<zduch4c>
the only thing I can see about this is in some french site
<zduch4c>
"Il manquait une dépendance dans le paquet."
<joepie91>
tenten8401: so, there's one other option... depending on your requirements...
<zduch4c>
from my broken french this seems to mean "missing dependency"
<joepie91>
one sec
astrofog has quit [Quit: Quite]
<tilpner>
tenten8401 - What would be the intended usage of such a program?
<joepie91>
tenten8401: this does some JS-to-Nix stuff, and I *believe* that it can also produce valid Nix source code from a structure declared in JS: https://www.npmjs.com/package/nijs -- however, afaik it has no Nix *parsing* capabilities, so it'd only work one-way
<tilpner>
(I believe there's better options for all usecases I can imagine)
<tenten8401>
tilpner: I don't quite know yet, but I was imagining it as like a YaST type thing for NixOS (YaST is from openSuse)
<sphalerite>
zduch4c: yep seems like the packaging is wrong
<sphalerite>
hang on
<tilpner>
tenten8401 - Your goal is graphically editing a NixOS configuration. The big question is, if the resulting configuration should be ready to be taken over for manual editing when the user becomes comfortable enough with Nix
<tenten8401>
joepie91: very interesting but it looks like development may be halted
<tenten8401>
tilpner: I'd hope so eventually, either that or just having a section for the user to edit their own custom nix statements
<joepie91>
I feel like the merging logic in nixpkgs would accommodate that pretty well
<tilpner>
tenten8401 - You can go a very long way with just { config = builtins.fromJSON (builtins.readFile ./config.json); }
<joepie91>
by storing generated configuration separately
<joepie91>
and just merging that low-prio with custom Nix config on top
<tenten8401>
tilpner: yeah, I feel like the json route would be the way to go for something like this
<sphalerite>
zduch4c: it seems that this bug has been fixed on master already, so you can install a working eolie with `nix-env -f channel:nixos-unstable -iA eolie`
aanderse-laptop has quit [Ping timeout: 268 seconds]
<joepie91>
ie. there does not need to be a hard cut-off between 'use GUI' and 'write Nix yourself'
<tilpner>
tenten8401 - I have thought about this too (and would be interested in further discussion/updates). One approach would be writing profiles and just managing imports from a GUI
<zduch4c>
sphalerite: can I install unstable packages declaratively? like in users.users.<>.packages?
<zduch4c>
also, if a package is broken, shouldn't this be also pushed to stable?
<sphalerite>
zduch4c: yes, although it's a bit fiddly
<sphalerite>
and yes, it should
<tenten8401>
tilpner: as in making a bunch of JSON configs and managing the imports through the GUI?
<tilpner>
tenten8401 - No, hand-writing various Nix modules, and managing the imports in a sensible manner (mutually-exlusive sets, optional ones, at-least-one-of-these sets) from a GUI
<TheBrayn>
is there no vim mode for the nixos config on the livecd?
<tilpner>
tenten8401 - nix-instantiate --eval -E 'builtins.toJSON { foo.bar = 42; }' works too
<samueldr>
(you may need to use '<nixpkgs>' otherwise bash will want to redirect stuff)
<tenten8401>
yeah, zsh was having issues if I didn't put quotes around it
<tenten8401>
repl is exactly what I was looking for, thanks
orivej has joined #nixos
<sphalerite>
TheBrayn: no, but that would be nice to have!
<sphalerite>
I might just make a PR for that :D
<TheBrayn>
I'll check if there is an issue for that or a pr
<TheBrayn>
can do
<sphalerite>
or yeah if you can make it
mojjo has quit [Ping timeout: 260 seconds]
<takeda>
is there a way for me to either get the original source directory or copy source to temporary mutable place. Basically before performing a build I need to cwd to the location and run script to obtain parameters, but it's possible that this script might download some dependencies and store them in CWD :(
<manveru>
takeda: the original source directory of what?
<manveru>
in a derivation?
<takeda>
source code, I'm writing nix expression for python packaging that is extracting information from setup.py and creates derivation for python package
<manveru>
for that i usually do something along the lines of `mkdir -p $out/tmp; cp -r $src $out/tmp; chmod -R +w $out/tmp; do_stuff; rm -rf $out/tmp`
<manveru>
not pretty, but it works
<takeda>
I have issues when setuptools_scm is used that gets package version through git describe
<manveru>
even with fetchGit?
<Orbstheorem>
Does anyone have a ctags grammar for the nix language?
<Orbstheorem>
s/have/has
<clever>
manveru: fetchGit deletes .git
<manveru>
then fetchgit?
<manveru>
one of those had an option to keep the .git
<{^_^}>
universal-ctags/ctags#261 (by blueyed, open): Support for the Nix expression language (Nix / NixOS)
<takeda>
I also want to run it in local checked out directory
<clever>
pkgs.fetchGit has the option, but it can cause problems
<manveru>
though i think it wasn't deterministic
<juhe1>
Is there a way to automagically copy already built closures from other machine over ssh specified on command line while runing nixos-rebuild command? I do not want to modify configuration.nix since there's not plan to do this very often. But sometimes it can be handy...
<manveru>
juhe1: you can specify substituters
<takeda>
"mkdir -p $out/tmp; cp -r $src $out/tmp; chmod -R +w $out/tmp; do_stuff; rm -rf $out/tmp" <- I guess this would work, but there's no nicer way?
<manveru>
takeda: any nicer way would do the same thing, so i guess not
<{^_^}>
[nixpkgs] @WilliButz opened pull request #43909 → nvtop: init at 0.2.2 → https://git.io/fNWMX
<takeda>
manveru: good point :)
<manveru>
you can't make $src writable because it's in the nix store
<manveru>
and you can only write to $out...
<manveru>
so yeah :)
<takeda>
I see
<clever>
the patchPhase runs after the stdenv has copied (and unpacked) $src to the working directory
<clever>
and that allows you to patch the unpacked copy of the source before the configurePhase runs
<manveru>
hmm
<manveru>
so prePatch?
<clever>
that also works
<manveru>
i forgot about that :D
<manveru>
clever++
<{^_^}>
clever's karma got increased to 13
<takeda>
clever: what I'm trying to do though is to generate derivation automatically by obtaining data from setup.py so that's even before derivation even exists
<clever>
takeda: something like pypi2nix?
<clever>
> pkgs.pypi2nix.meta.description
<{^_^}>
"A tool that generates nix expressions for your python packages, so you don't have to."
<tenten8401>
does anyone know why it's trying to download a NixOS 16.09 image when I deploy from NixOps to virtualbox?
<takeda>
clever: I'm writing my own, because I believe it can be done better :)
<tenten8401>
tilpner: that's interesting
<clever>
takeda: your tool will need to fetch the source and parse it as it generates the derivation
<tilpner>
tenten8401 - The source confirms it's downloading 16.09. That may be just because nobody updated it though
<tilpner>
(And there's file listing on the web server, so hard to tell what other vmdks are accessible, if any)
<tilpner>
*no file listing
<FRidh>
takeda: will you monkeypatch `setup` to retrieve the values that are passed in?
<takeda>
clever: so far I didn't need to fetch the code, just communicate with PyPI to get extra information, the nix would be fetcing and parsing the code during building
<FRidh>
or statically?
<nh2>
man nix-store shows `nix-store --add-root /home/eelco/bla/result --indirect -r ` but `-r` isn't documented. What does it do?
<takeda>
FRidh: there's setup.cfg which python seems to be moving toward which makes things so much better because it is deterministic, but setup.py has hooks to obtain information from it as well
<FRidh>
takeda: setup.cfg is indeed easier, but way too few packages are using it
<takeda>
yeah, the other method should work on any package that uses setuptools/distutils
<FRidh>
the only way really is to execute python setup.py or you'll miss far too many requirements
drakonis has joined #nixos
<takeda>
FRidh: that's what I'm doing, but there are hooks that you can execute it without doing installation and obtain Distribution object where you can get information
<FRidh>
Ah, right, nice.
<takeda>
anyway once I have something that works well I will open source it
<samueldr>
nh2: isn't it --realise?
<nh2>
samueldr: oops you are right, my grep skills are poor today
<{^_^}>
[nixpkgs] @FRidh pushed 190 commits to staging-next: https://git.io/fNWDB
<clever>
sphalerite: ok, something very fishy is going on with my desktop....
<clever>
lately, nixos has been having a lot of overheating problems, going into thermal shutdown repeatedly
<clever>
i powered up the motherboard monitoring software, and found that the cpu is drawing over 80 amps
<clever>
without a baseline, i just assumed that was normal, it matches up with the watt rating and the voltage in the same util
dejanr_ has joined #nixos
<clever>
windows is pulling 10 amps at max load on all 8 cores...
<dejanr_>
hi anyone using libvirt, and managed to passtrough gpu, when i use OVMF i only get UEFI shell, it want boot from drive or iso
<dejanr_>
i was trying for few weeks now to setup gpu passtrough for windows, and its not working, always getting nvidia error 43 in windows guest
Tiez has quit [Quit: WeeChat 2.1]
<clever>
dejanr_: ive had a lot of trouble getting gpu passthru to work, i had to use a special linux param to entirely blacklist linux from touching the pci device, which meant you dont even get a text console to login at
<clever>
dejanr_: and even then, windows can only use the card once per boot, then the host has to do a full powercycle
<dejanr_>
clever: i made that work, my second gpu has pci-vfio kernel driver loaded
<clever>
so even with a linux host, your uptime is bounded by the stability of windows
<dhess>
80 amps? That can't be. That's more than an arc welder draws.
<dhess>
not to residential mention circuit breakers max out at 20A, unless it's for a washer/dryer
<clever>
dejanr_: its a 125 watt cpu, running at 0.8 volts right now
<clever>
dhess: remember, this is 80 amps at about ~1 volt
<clever>
80 amps at 1v is very different from 80 amps at 120v
<clever>
hmmm, its only at 0.86 volts on windows...
PLPD-Bot has quit [Remote host closed the connection]
<dhess>
ohhhh I thought you were measuring this at the power supply :)
PLPD-Bot has joined #nixos
<clever>
dhess: the motherboard has some monitoring software that runs over usb, and lets me monitor the cpu amps, many voltages, fan speeds, and temps
<infinisil>
Is there a symlinkJoin that warns when there are collisions?
<clever>
infinisil: and buildEnv wont do?
<infinisil>
Ohhhh
<clever>
dhess: and you can remotely change the clock freqs
<infinisil>
Thanks
<infinisil>
clever: Oh actually, I need to link all files, not just specific dirs
<clever>
dhess: heh, during shutdown it rises to 1.3 volts on the cpu
<clever>
infinisil: buildEnv links everything by default
<infinisil>
Ohhh, `pathsToLink ? ["/"]`
<infinisil>
I see
<clever>
infinisil: if only one thing provides a given directory, it links the directory, if 2 things provide a dir, it creates one, and links all of its contents
<infinisil>
Yeah that's perfect
<takeda>
clever: runCommand's $out is just a single file, is there a way to get a directory instead?
<infinisil>
When I set `src = ./path/to/some/file.nix`, mkDerivation fails, because it doesn't know how to unpack it
<infinisil>
Can I make it work regardless? unpack it if possible but don't do anything if that doesn't work?
<clever>
infinisil: why is the src another nix file?
<infinisil>
Well it's actually a .sh file
<infinisil>
Just used the first extension that came to my mind :P
<clever>
juhe1: what is process 20938?
<clever>
infinisil: youll want to replace the default unpackPhase then, or just switch to runCommand
<infinisil>
Hmm, I'll want to get the default unpack phase
graphene has quit [Remote host closed the connection]
<infinisil>
Oh, I guess I could runHook or so
<mconstant>
zduch4c: still no luck. apparently sbux hates Linux
<clever>
infinisil: the default unpackPhase only works for tars, zips, and directories, it wont know what to do with a file
<infinisil>
Yeah, but I could || true it
<infinisil>
So it hopefully won't fail
graphene has joined #nixos
<zduch4c>
mconstant: it really should work, networkmanager should just make everythign work by default, it's probably fault on starbucks end
<clever>
infinisil: what are you trying to actually do?
<zduch4c>
hell they might be blocking any non-macintosh computers ;)_
aanderse-laptop has quit [Ping timeout: 260 seconds]
<jtojnar>
zduch4c: will also need 79092439ca6a243e609319a0cf3514b5c4551c9d
<infinisil>
clever: make a function that takes a source as input and have it extract all files of a certain extension to $out. It should also support a single file (source = ./path/to/file), to be put into $out/file
<infinisil>
I have the extract part done by using $src
<zduch4c>
jtojnar: hm?
offlinehacker[m] has quit [Ping timeout: 260 seconds]
<clever>
infinisil: then youll want a custom unpackPhase that will check $src, and decide if it should copy it directly, or run the original unpackPhase and filter it
<infinisil>
Yeah makes sense, original unpack phase where?
<juhe1>
clever: no such process runs... "ps ax | grep 20938" returns only the line with "20944 ? Ss 0:42 nix-daemon 20938"
<clever>
juhe1: sounds like a bug in nix then, it should have terminated that process automatically
<juhe1>
clever: ok, I'd guess the 20938 was the nix build command
tenten8401 has quit [Ping timeout: 240 seconds]
<jtojnar>
zduch4c: gah, I hate release branch, as if weekends need more merge conflicts
<zduch4c>
there's also an issue (at least for nvidia users) with webkitgtk not rendering anything, though this is probably more of a webkitgtk issue than a nix issue
<{^_^}>
[nixpkgs] @jtojnar pushed to release-18.03 « gtkspell3: add gir »: https://git.io/fNWSz
mconstant has quit [Ping timeout: 248 seconds]
<jtojnar>
zduch4c: yep, I think we have an issue open for that
<zduch4c>
thats probably the biggest bummer, I'd use eolie/epiphany if not for that. i have to use chromium :~(
<zduch4c>
here's what eolie has to say when opening google.com and not rendering it: Cannot create EGL window surface: EGL_BAD_MATCH
dejanr_ has quit [Quit: WeeChat 2.0]
agander has quit [Ping timeout: 260 seconds]
<zduch4c>
and yeah, the switch does 'fix' everything
ihar has quit [Ping timeout: 245 seconds]
ihar has joined #nixos
tertle||eltret has joined #nixos
<adelbertc>
sphalerite: thanks for helping me w/ the docker issue a couple days ago - we found out what was causing it. this seems to fix it https://github.com/NixOS/nixpkgs/pull/38837 - our theory is since we were on a nixpkgs revision before that change, and because it's possible the docker daemon(s) in our k8s cluster run as non-root, when it untar's it untar's with the default umask which may have had insufficient permissions
<juhe1>
manveru: thanks for pointing out the substituters thingie, took me a while to find and understand the stuff around it but it was worth of the effort
<{^_^}>
[nixpkgs] @matthewbauer pushed 2 commits to staging: https://git.io/fNWQm
<{^_^}>
[nixpkgs] @matthewbauer merged pull request #43224 → [staging] substitute(): --subst-var was silently coercing to "" if the variable does not exist. → https://git.io/fNvj6
genesis has quit [Ping timeout: 256 seconds]
<tertle||eltret>
anyone been able to unstall x2goserver on Nix?
<tertle||eltret>
install
<tertle||eltret>
are there any package types that work with nixos?
<tertle||eltret>
liek debs?
<tilpner>
There are some packages built from .debs, but it's not automatic
<hlolli>
Hi, I made a fetchClojar function made mostly with fetchMaven, I need to use it for another derivation later on. I want to create a PR for it, into which directory would it make sens to add a fetchClojar derivation to?
<fresheyeball>
for some reason any interaction with nix-shell results in "error: writing to file: Broken pipe"
<fresheyeball>
as well as installing new packages with nix-env -i
<fresheyeball>
I rebooted, am at a total loss
<selfsymmetric-mu>
Hey #nixos. My system is completely freezing every day or so. Nothing shows up in `journalctl/dmesg`. I have an AMD processor which doesn't support `mcelog` so I can't use that. Memory checks pass with flying colors. How can I diagnose further?
<sphalerite>
fresheyeball: that sounds scary
<sphalerite>
fresheyeball: could you try `strace -o /tmp/trace nix-shell -p hello` and share the log?
<fresheyeball>
sphalerite: I just want my nix back :(
<fresheyeball>
error: writing to file: Broken pipe
<sphalerite>
for running strace..?
<fresheyeball>
yes
<`_>
Hold on
<`_>
fresheyeball: /tmp/trace will be the relevant log
<sphalerite>
oh yeah that
<sphalerite>
selfsymmetric-mu: did you leave the memory test running for 2 days or so?
__monty__ has quit [Quit: leaving]
<fresheyeball>
it blorks out a mess of dubstep or something
<sphalerite>
fresheyeball: that mess of dubstep could be helpful for working out what the issue is :)
<sphalerite>
so if you could make a gist of it or something
kyren has joined #nixos
<sphalerite>
selfsymmetric-mu: has this started occurring recently?
<selfsymmetric-mu>
sphalerite: I left it running overnight, no issues. It's been happening since the first day.
<`_>
Huh, TIL linux runs on dubstep
<selfsymmetric-mu>
This is a new build.
<selfsymmetric-mu>
I'm considering whether I should replace the CPU/mobo to Intel. At least I can rule out CPU issues that way (and I'll have `mcelog`).
<fresheyeball>
I can see the bass drop in this file, somewhere around line 1497
<sphalerite>
fresheyeball: hm ok that's not helpful so far :/ could you try adding -f to the strace command (before nix-shell) and pasting all the /tmp/trace.* files?
<selfsymmetric-mu>
While I'm building that (sorry, I triggered a full rebuild-switch), what does it means when I get an error like `Specified group 'render' unknown`? I'm trying to eliminate my highest-priority errors in `journalctl` just in case.
<sphalerite>
tldr: use `nix-shell -p` to get a shell with make and a C compiler etc
<tertle||eltret>
oh ok
<fresheyeball>
solved it!
<sphalerite>
fresheyeball: what was it?
<tertle||eltret>
whats the diff between a library and a binary? why shouldnt i install it?
<nh2>
infinisil: I didn't realise you are at ETH. Did we meet?
<selfsymmetric-mu>
Is that so! Excellent. Which text console would you recommend? Also, it's strange that it would succeed in rendering the text, but not in dumping the crash info.
<selfsymmetric-mu>
Is there something other than `journalctl` that might record fatal errors?
<infinisil>
nh2: I probably won't go this soon, but maybe at a later date, I am a person who pretty much exclusively spends time on his own at home lol
<sphalerite>
well journald runs in user space
<selfsymmetric-mu>
Probably a silly question, but is there a kernel space analog?
<sphalerite>
and if stuff goes so horribly wrong it's best not to touch the disk any more to avoid catastrophic data loss or whatever
<nh2>
infinisil: if you come to the meetup you will get a free dinner and save time from making dinner yourself so it's a good time investment ^^
<sphalerite>
no persistent log I know of
<selfsymmetric-mu>
sphalerite: Ah, I understand completely now, thank you.
<selfsymmetric-mu>
So the kernel tries to be careful.
<nh2>
infinisil: many of us there spend a lot of time at home, as there are many working-from-home people (me included)
<selfsymmetric-mu>
But it leaves me in the dark, unfortunately.
<sphalerite>
but yeah if you can connect another machine up to that machine's serial port that would be the best way to get some info
<sphalerite>
even on modern machines without a proper serial port there's still usually a header on the motherboard
<sphalerite>
of course you need appropriate hardware for it :/
<selfsymmetric-mu>
sphalerite: I like that idea. I'll see if I can do that.
<infinisil>
nh2: That does sound pretty interesting, hold on, let's go to #nixos-chat
<selfsymmetric-mu>
SMa
<nh2>
ah I didn't know about that channel
<selfsymmetric-mu>
Maybe somebody's built a dongle that just reads serial-port dumps.
<tertle||eltret>
make isnt in bash shell
<sphalerite>
But for a start, with no extra hardware, you can try just switching to tty1 and waiting for it to crash again to see if anything interesting pops up there
<selfsymmetric-mu>
sphalerite: Yeah. I'll try keeping my monitors on and unlocked. Leave them for a while in tty1..
<sphalerite>
selfsymmetric-mu: there are USB-RS232 adapters out there. Not sure about how to connect that up to the header on a motherboard though, or what that header looks like for that matter
<selfsymmetric-mu>
Yeah, it's not a great situation. I appreciate you talking to me about it.
<selfsymmetric-mu>
Another thing is I've heard `rasdaemon` recommended, but I don't think there's a nix package for it.
wchresta has quit [Quit: Yaaic - Yet another Android IRC client - http://www.yaaic.org]
<sphalerite>
never heard of that, and yeah doesn't seem to be packaged
<selfsymmetric-mu>
Another possible strategy. I have _almost_ the same NixOS configuration on my work machine. I might swap the drives for a day, see how that changes things.
<selfsymmetric-mu>
That might help disambiguate configuration problems from hardware problems.
<sphalerite>
yeah, although I doubt a config difference would cause this
<selfsymmetric-mu>
sphalerite: So you do suspect a hardware issue?
<selfsymmetric-mu>
That's your gut feeling, anyway?
<sphalerite>
oh yeah, you could also check for store corruption with nix verify --all --no-trust
<sphalerite>
yes
<selfsymmetric-mu>
sphalerite: Do I run that as superuser?
<selfsymmetric-mu>
as sudo, I mean?
<sphalerite>
nah you don't need to
<selfsymmetric-mu>
Okay, cool. Running that now.
<sphalerite>
if there are any corrupted paths you will need root to repair them though, sudo nix-store --repair-path <path>
<sphalerite>
yeah… I think the way to go in this case is to run nix-store --dump-db on the old nix, then nix-store --load-db with the output from it
<sphalerite>
s/old/new/
<sophiag>
oh, nice
<sphalerite>
or ideally just use the new nix…
<sphalerite>
but, well.
<sophiag>
i have something like the cap theorem with current versions of nixos, this laptop's graphics chips, and the actual packages i need :(
<{^_^}>
[nixpkgs] @jfrankenau opened pull request #43934 → mpd: add nfs and smb support → https://git.io/fNWNW
<sophiag>
sphalerite: i think for where i'm at right now (and being cautious wrt to time on rebuilds that leave me with just xterm) i'd prefer just udpating packages from 18.03 and _not_ upgrading nixos. it seems your recommendation should do that?
<sophiag>
modulo having bumped the stateVersion...
<sphalerite>
well you shouldn't really bump the stateVersion at all
<sphalerite>
ever
<sphalerite>
in principle
<sophiag>
yes, i was aware before i did it the first time yesterday
<sophiag>
i figured it couldn't be more broken and just wasted a ton of time going down that route
<sphalerite>
ah ok
<sphalerite>
lol
<sophiag>
it's relevant now, though, because i probably want to change it _back_ in order to ensure this build works
<sphalerite>
idk at this point I don't know.
<sophiag>
did you mean to say i need to boot from one of my half-broken newer builds in order to dump the nix-store schema?
<sphalerite>
I don't think so, you should be able to use /nix/var/nix/profiles/system-xyz-link/sw/bin/nix-store
cryptomonoid has quit [Ping timeout: 260 seconds]
hiratara has joined #nixos
<sophiag>
oh, so i can just pick one of the newer builds i rolled back from and load the schema from that?
<sphalerite>
yeah that should work I think
<sphalerite>
you might need to unset NIX_REMOTE for it
<sophiag>
omg, no one ever gets my lawnmower man jokes!
<sophiag>
but no, in this case i meant as in biblical :)
<sphalerite>
:')
<sophiag>
as root i'm being told "error: you don't have sufficient rights to use this command"
<sophiag>
`nix-store --load-db`
<sphalerite>
try it with NIX_REMOTE=local
joehh has quit [Ping timeout: 244 seconds]
<sophiag>
oh right, thanks
aanderse-laptop has quit [Ping timeout: 244 seconds]
<epta>
I'm trying to apply release.nix approach by Gabriel Gonzalez for referencing local dependency, and got strange error 'liba-0.1.0.0-inplace is unusable due to missing dependencies' https://github.com/Gabriel439/haskell-nix/issues/56
<{^_^}>
Gabriel439/haskell-nix#56 (by dmalikov, open): Referencing source dependencies lead to shadowed dependency error
<sphalerite>
right it's really time I got some sleep
<sphalerite>
good luck sophiag!
<sophiag>
thanks. i'm still getting the same error :/
<tertle||eltret>
this is fun
<tertle||eltret>
im install nvidia driver now
<sophiag>
tertle||eltret: you're still at it :)
joehh has joined #nixos
<tertle||eltret>
yeah ive wasted my whole weekend on this lol
<sophiag>
welcome to SunkCostOS :D
<tertle||eltret>
lol
<Jason_Grossman>
sophiag++
<{^_^}>
sophiag's karma got increased to 1
<tertle||eltret>
tertle||eltret: ++
<{^_^}>
tertle||eltret:'s karma got increased to 1
<epta>
epta: ++
<{^_^}>
epta:'s karma got increased to 1
<epta>
nice
<epta>
Any idea what this shadowed dependency error is about from the link above?
<sophiag>
(at this point just the word "nvidia" is like a pavlovian trigger for me)
<tertle||eltret>
lol
<tertle||eltret>
u start foaming?
<sophiag>
more like nauseated
<sophiag>
i'm roughly at the bargaining stage i'd say
Fare has quit [Ping timeout: 244 seconds]
<tertle||eltret>
y?
<{^_^}>
[nixpkgs] @marsam opened pull request #43937 → ia-writer-duospace: init at 20180721 → https://git.io/fNWAR
<sophiag>
i'm mostly joking. ironically nvidia appears to not be my problem. it's way _more_ screwed up if i have services.xserver.videoDriver = "intel"
<sophiag>
im
<sophiag>
oops...pretty sure if i could just disable integrated i'd be fine
<joepie91>
there's a thing you don't hear every day
<joepie91>
:)
<sophiag>
which part? most people i know using nixos seem to have one or the other disabled
<sophiag>
bumblebee doesn't seem to work on all hardware. i suppose i could be hopeful about the optimus patch that was recently merged in 18.09