<avn>
I personally used stack/cabal new-build when develop, then cabal2nix to build/deploy final versions
<gchristensen>
ah then you need to set deployment.ec2.subnetId dmj`
pie___ is now known as pie__
Lisanna has joined #nixos
<dmj`>
gchristensen: nice, it’s doing something
<dmj`>
gchristensen: had to manually create the vpc, subnet and ig
<dmj`>
gchristensen: don’t know why I can’t make it my default
mekeor has quit [Ping timeout: 252 seconds]
Dedalo has joined #nixos
iyzsong has joined #nixos
youssefhabri has quit [Read error: Connection reset by peer]
goibhniu has quit [Ping timeout: 260 seconds]
<dmj`>
gchristensen: seems to be working now, had deleted the default vpc in us-east-1 (using us-east-2) and didn’t open 22 on inbound in security groups. Thanks!
<gchristensen>
gret!
<gchristensen>
aws sure is flexible, with all the good and bad trappings
<dmj`>
gchristensen: I’m deploying from my nixos machine, will it build here first before deploying to aws? I noticed digital ocean actually built on the server I was deploying to
<clever>
dmj`: was DO done from a mac?
<{^_^}>
[nixpkgs] @wykurz opened pull request #48765 → cpuset: init at 1.6pre1 → https://git.io/fxKmH
<dmj`>
clever: from NixOS!
<clever>
not sure then
<dmj`>
clever: judging from htop, deploying to ec2 definitely builds on my local nixos first
jmeredith has quit [Quit: Connection closed for inactivity]
init_6 has quit []
sigmundv has quit [Ping timeout: 252 seconds]
sigmundv has joined #nixos
iyzsong has joined #nixos
mayhewluke has quit [Ping timeout: 240 seconds]
mayhewluke has joined #nixos
sablad has joined #nixos
<adamantium>
to upgrade to the new 18.09 do i simply change system.stateVersion = "18.03"; and then rebuild?
<samueldr>
adamantium: no
<samueldr>
,upgrade adamantium
<{^_^}>
adamantium: To upgrade nixos, read the release notes (this is important! Check ,releasenotes for details), then run (as root) nix-channel --add https://nixos.org/channels/nixos-xx.yy nixos && nixos-rebuild boot --upgrade , then reboot.
<samueldr>
stateVersion shouldn't be touched, unless someone knowledgeable explains why you would
<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
<adamantium>
thank you!
<adamantium>
samueldr: should i also remove the previous channel?
<samueldr>
it shouldn't be there after adding it correctly
<adamantium>
When should we, if ever, change system.stateVersion ?
<adamantium>
k
<samueldr>
the channel's name will be `nixos` in the channels list
<samueldr>
adamantium: possibly never
<samueldr>
most things it affects are locations of stateful files
<samueldr>
(e.g. databases)
<samueldr>
maybe down the line in a couple lines something will happen, but I would bet it's highly unlikely
<samueldr>
in a couple years* haha
<adamantium>
cool.
<adamantium>
soo, nixos-rebuild boot --upgrade, i usually just use switch. The difference is that the changes don't take affect until after you reboot, that's all right?
<samueldr>
exactly
<samueldr>
while unlikely to cause a huge mess, using `switch` with an upgrade between releases *could* cause weirdness
<adamantium>
makes sense.
<samueldr>
(none reported thus far)
mekeor has quit [Quit: digital revolution = cryptoanarchism + cybercommunism]
<adamantium>
What's the consensus regarding apparmor+nixos
sigmundv has quit [Ping timeout: 272 seconds]
Mr_Keyser_Soze has joined #nixos
erratic has quit [Ping timeout: 260 seconds]
mekeor has joined #nixos
<{^_^}>
[nixpkgs] @benley opened pull request #48767 → plex: include Plex Commercial Skipper binary → https://git.io/fxKsE
<sablad>
speaking of weirdness using nixos-rebuild switch --upgrade, I can't successfully boot 18.09.
<sablad>
Thankfully I can roll back! But it complains about not being able to load the root filesystem
<sablad>
I didn't see anything in the release notes that indicated I would have to change my setup, and I'm not sure where to start debugging
oldandwise has joined #nixos
<sablad>
When I checked 'systemctl status "systemd-fsck@dev-disk-by\\x2duuid-5BAD\\x2dDF26.service"', I see a lot of this error:
<sablad>
Failed to resolve unit specifiers on File System Check on %f, ignoring: Invalid argument
<sablad>
I've got an encrypted disk that starts by reading /dev/disk/by-uuid/5BAD-DF26, so I think it's getting hung up there
<sablad>
Does anybody know about a change in 18.09 that would mess up my boot process?
Mr_Keyser_Soze has quit [Ping timeout: 264 seconds]
Mr_Keyser_Soze99 has joined #nixos
mekeor has quit [Quit: digital revolution = cryptoanarchism + cybercommunism]
<oldandwise>
i upgraded to 18.09 last week, i'm not seeing error msgs when i do `systemctl status`
<sablad>
Do you encrypt your filesystem? I suspect my problems stem from that
<drakonis>
ty Mic92 for merging the debootstrap pull
jmeredith has joined #nixos
slack1256 has joined #nixos
emmett1 has quit [Ping timeout: 252 seconds]
fusion809 has joined #nixos
<fusion809>
Hi folks, is there a way to get file managers to utilize apps installed using Nix when opening files? I installed VLC using Nix, because my distro's VLC package is not working properly, and I'd like to be able to open my media files using it from my file manager (PCManFM).
<slack1256>
fusion809: I don't know how to handle it on another distro, but on nixos that just works.
<fusion809>
I've tried symlinking my VLC desktop config file to ~/.local/share/applications, and just copying it to there and while VLC does show in the "Open with" menu, when I select it nothing happens (i.e. VLC does not start).
<fusion809>
Yeah on NixOS it works, I've tried it before, but still it'd be delightful if someone has an idea about how to get this to work on other distros
drakonis has quit [Quit: WeeChat 2.2]
drakonis has joined #nixos
endformationage has quit [Quit: WeeChat 1.9.1]
<fusion809>
Well I've found a way. I wrote a script I placed in /usr/local/bin that uses the right environment variables (XDG_DATA_DIRS and PATH pointing to Nix dirs) and executes vlc-wrapper with arguments, then referenced that in my own desktop config file and now I can launch VLC in PCManFM without a problem.
iyzsong has quit [Ping timeout: 252 seconds]
iyzsong has joined #nixos
<drakonis>
very inelegant
<drakonis>
could definitely be improved
<fusion809>
Sure could. If you've got a better solution I'd like to hear it. I've set my system environment variables (e.g. XDG_DATA_DIRS and PATH) to point to the appropriate $HOME/.nix-profile/* dirs.
<fusion809>
yet that doesn't fix this
iyzsong has quit [Ping timeout: 264 seconds]
tertl3 has quit [Quit: Connection closed for inactivity]
Czen has quit [Remote host closed the connection]
Czen has joined #nixos
Lisanna has quit [Remote host closed the connection]
revtintin has joined #nixos
<dmj`>
would exporting the state of a nixops deploy be equivalent to backing up the sqlite db that nixops uses
<CMCDragonkai>
How come `haskellPackages.ghcWithPackages (pkgs: with pkgs; [ bifunctors ])` works with `nix-shell -p` but not when it's in `shell.nix`?
<CMCDragonkai>
Arrrghhh why does it have to be different
<dmj`>
that shellHook...
<dmj`>
wonder if it works w/o the shellHook
<jonge>
hey there, i have a private git repo at my company that contains nix expressions that again checkout from private git repos. i set up fetchgitPrivate so that all nix build users and hydra can use it. it does however fail in the end because `nix-prefetch-git` crashes inbetween: `/nix/store/....hydra-2018-08-07/bin/.nix-prefetch-git-wrapped: line 160: cd: too many arguments`
<jonge>
and i am not sure how to debug this and if this might still be some error on my side or a hydra bug
<{^_^}>
[nixpkgs] @FRidh pushed 2 commits to python-unstable: https://git.io/fxKCA
alex`` has joined #nixos
<dmj`>
jonge: thanks, this is perfect
pointfourone has left #nixos ["Leaving"]
<dmj`>
jonge: been playing with nixops all day :)
<dmj`>
jonge: have you by chance setup the github PR plugin? I have my hydra fetching a private repo (using a deploy key), but would really like it if it could build PRs and provide a status indicator and link on github
<dmj`>
jonge: similar to modern CIs like travis
<jonge>
dmj`: my side project is a nixos server that builds all the projects in our company (with different compilers etc. etc.), and performs integration tests and builds docker images of everything etc. etc... in order to finally convince all colleagues that we should really really officially use nix.
<jonge>
dmj`: no i have not set up the github PR plugin. we use gitlab, so i will have to write my own gitlab PR plugin...
<dmj`>
jonge: yea, it’s hard to convince people they should understand the code that runs their company's deployment. They’d rather have it be magic I guess.
<jonge>
but what you are trying to achieve is basically what i would liek to have longterm, too.
<dmj`>
jonge: kubernetes is such a black box
<dmj`>
jonge: oh no, hydra patch failed...
<dmj`>
jonge: maybe it’s too new
<jonge>
in our company it's pretty transparent, as we use gitlab's docker-based CI. that works well, but compared to what nix is capable of it's really inferior.
<jonge>
dmj`: let me give you my patch, i think i had to fix the line numbers
<jonge>
this works on the hydra version from the nixos-unstable channel as of today
<dmj`>
jonge: seems to have worked
<dmj`>
got past "patching sources"
<dmj`>
jonge: big thanks :)
<jonge>
dmj`: no problem
<dmj`>
jonge: never thought I’d live to see the day hydra would fetch private repos behind basic auth w/ letsencrypt nginx setup
<jonge>
dmj`: hehe yeah. it's just 2 hours ago that i successfully got hydra and all nixbld users to fetch private git repos via ssh with the right key
<dmj`>
jonge: nice! using deploy keys?
justbeingglad has joined #nixos
<dmj`>
jonge: finally having nix use netrc right is great too
elgoosy has joined #nixos
<jonge>
yeah i tested it with my personal ssh key file but longterm i want a hydra user in our gitlab with its own ssh key
<jonge>
so `fetchgitPrivate` works with our gitlab repos now.
<dmj`>
awesome!
<jonge>
dmj`: do you have any link to the netrc thing?
<elgoosy>
hi, how do i increase the boot screen timeout (where all the versions are shown + windows boot)?
<hyper_ch>
elgoosy: grub?
<{^_^}>
[nixpkgs] @c0bw3b merged pull request #47775 → nixos/munin: move from /var/run to /run → https://git.io/fxLr5
<dmj`>
jonge: I just hope github webhooks work amidst basic auth
<jonge>
dmj`: so who exactly is using this netrc file? `nix-build` in your personal shell, or hydra, or one of the nixbld users?
<elgoosy>
hyper_ch: systemd-boot
jasongrossman has quit [Ping timeout: 252 seconds]
<hyper_ch>
elgoosy: no idea
init_6 has joined #nixos
<dmj`>
jonge: I think nix itself uses it, if users specify my hydra as a cache, nix will call into libcurl and needs a NETRC file to get around the basic auth
<dmj`>
jonge: Oct 21 08:09:30 hydra hydra-evaluator[5857]: error: writing to file: No space left on device
<dmj`>
lol
<jonge>
dmj`: oops.
<dmj`>
jonge: guess those ebs volumes don’t come by default :)
<jonge>
dmj`: i just wondered about the scenario where maybe all your private git repos are hosted behind this http basic auth. and then a job description in hydra points to such a repo. then hydra needs to have the netrc file in its own config. no idea how to inject that.
<dmj`>
jonge: if hydra is poll-based it shouldn’t have to worry, but for gitwebhooks being sent to hydra, you might have to put the basic auth creds in the webhook url and pray it gets sent encrypted
hotfuzz has quit [Ping timeout: 245 seconds]
<dmj`>
jonge: oh I think I misread what you meant
<dmj`>
jonge: you want a git server behind basic auth
<jonge>
dmj`: yeah i assumed you have that. and then need hydra to build projects from there.
<dmj`>
jonge: I’m just using GitHub for now
<jonge>
dmj`: and then hydra needs to use this netrc file itself and i wondered how to setup that.
<dmj`>
jonge: it should work transparently if you set an environment variable that points to the netrc file, hydra would call into libnix and it would “just work"
<fusion809>
The homepage for Nix (https://nixos.org/nix/) says, "Nix runs on Linux, macOS and other systems." I'm just wondering what "other systems" it runs on. The *BSDs per chance?
<fusion809>
Thanks, guessing support for many of these is quite experimental.
<dmj`>
jonge: I assume gitlab has a notion of deploy keys like github, which is tantamount to giving a user read-only access (and potentially write) if they specify a public ssh key
elgoosy has quit [Ping timeout: 264 seconds]
<jonge>
dmj`: yes, you're right
<Izorkin>
How to change nixos/modules/tasks/network-interfaces.nix so if enableIPv6 = false; add to dhcpcd.extraConfig = "\nnoipv6rs \nnoipv6"; ?
<dmj`>
jonge: and to only a single repo
<jonge>
dmj`: the reason i first went the SSH way was that i found the fetchgitPrivate function along with documentation, and not much about the netrc thing
<jonge>
dmj`: also didn't know that netrc is something to look for
<dmj`>
jonge: thing it was a relatively recent advancement
<dmj`>
s/thing/think
<jonge>
dmj`: that line came in with a patch 2 years ago according to git blame
<dmj`>
emphasis on relatively :)
<jonge>
dmj`: i am slowly slowly getting a hang of it all. currently my nix knowledge is good enough to understand the nixpkgs libs and build my own stuff. i guess the next step is understanding the c++ and perl level of nix/hydra :)
jtojnar has quit [Ping timeout: 272 seconds]
<dmj`>
jonge: yikes :)
<dmj`>
jonge: let’s hope both of those get replaced before it ever comes to that :)
<jonge>
dmj`: i don't particularly like perl, but i am comfortable with c++
<dmj`>
jonge: I’d gladly take C over C++
<jonge>
so you would like to do all the resource handling manually? the error handling, too? no type safety anywhere? with no performance advantage over c++?
__monty__ has joined #nixos
<dmj`>
yes, c++ is super complicated
<jonge>
that is true and you need some specific level of c++ mastery while C is a language that you could learn in one day. but then, it's much easier to master complexity while at the same time having the compiler reducing the assembly to something as simple as C build output. minus the programmer mistakes.
<dmj`>
All the best stuff is in C, postgresql, sqlite, linux.
<jonge>
they could have been in C++ as well. maybe they would, if these projects were not older than c++
<dmj`>
jonge: personally I would use haskell for everything application related, C only if necessary for performance bits where the law of diminishing returns sets in, and then nix for deployments.
<jonge>
dmj`: that is true. haskell for high level stuff is unbeatable. wonderful language and runtime system.
<jonge>
but for low level stuff, i would always select C++ because the compiler is much better in a lot of things than humans will ever be. we wrote our own hypervisor in c++ and are glad to have its functionality sanitized with as much type safety as possible.
<jonge>
the lack of type safety in C is simply disturbing. it's asking for bugs.
<dmj`>
in some cases C is more typesafe than java
<dmj`>
in others, not so much :)
<jonge>
java is a whole different thing that i like to avoid as much as possible
<dmj`>
you wrote a hypervisor?
<dmj`>
like xen?
<jonge>
not i myself. but my company is a hypervisor startup and we wrote that thing from scratch, yes
<dmj`>
why not use an existing one
<dmj`>
geez man
<dmj`>
that’s a cool company
<jonge>
the problem is that the existing ones are monolithic and we do microkernel stuff.
<jonge>
awesome interesting technology, yes.
<__monty__>
How about C++ vs Rust? You think Rust might be a better C++ one day?
<dmj`>
jonge: I used to work on the haskell virtual machine, it was basically a statically linked haskell file with a think C layer that was bootable by xen, drivers written in haskell, like the network stack
<dmj`>
s/think/thin
<__monty__>
Would've been cool if the microkernel hypervisor was added to kvm though.
<jonge>
hm i had a look at rust and it follows some really cool ideas. but i think it has no real advantages over c++ (talking about the latest c++ standard). at the same time a lot of the advantages in rust come from reducing the programmer's freedom, which is what c++ never does.
<jonge>
and i think that the c++ type system with its meta programming capabilities is not matched by anything in rust
<__monty__>
Haskell's advantages come from restricting the programmer too though.
Ariakenom has joined #nixos
<jonge>
__monty__: that is true, but the abstractions are so generic and versatile that it's never disturbing
<jonge>
c++ helps me writing typesafe code that is partly even functional - and at any point in my code i can jump back to raw imperative shit-code if i find that there is no feasible alternative.
<{^_^}>
[nixpkgs] @hisandeepsingh opened pull request #48770 → Darwin sandbox → https://git.io/fxKlS
<dmj`>
I’ve seen people take haskell too far and basically ruin a company
jtojnar has joined #nixos
<__monty__>
Can't you do that in Rust? It's not like you can't use `unsafe`.
<dmj`>
Like where everything is all about category theory
<jonge>
dmj`: yeah i like the bazaar vs. git discussion... so bazaar is so much nicer and cooler but git just works as fast as people need it, lol
<__monty__>
What company, dmj`?
<__monty__>
I can imagine that making it hard to hire though.
<dmj`>
I should write a blog post, “when types hurt"
<dmj`>
__monty__: that’s the problem, you can’t ramp people up if you’re just making the ceiling higher
<jonge>
although i would not weigh one language against the other by referencing extreme failures that some individual companies did
<__monty__>
I actually don't think over-engineered is the right word, java has plenty of over-engineered abstractions. It's just the level of abstraction that's the problem probably.
<jonge>
c++ and haskell give you plenty of ways to build real abstractions. so people will build abstractions and sometimes build stupid/costly abstractions. this won't happen in C too often because this language does hardly allow for any abstractions.
<jonge>
and the few abstractions in C, like e.g. structs with functions pointers that i would call "polymorphic objects for the poor" are horrible.
Dedalo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<jonge>
i was once asked to debug a kvm crash that in the end was caused by a bad function pointer in such a pointer struct. such things would not have happened in c++ because you can't possibly fuck up a virtual object this much.
<jonge>
in c++ nobody forces you to use all the abstractions. you can keep it to a minimum. but the abstractions you get if you use any are of high quality. in C you just get none and need to do everything by hand - forever.
<jonge>
ok i have to pack my things and go. thank you very much for the interesting discussion and dmj` for the hydra/github/netrc inspiration. have a nice day!
<dmj`>
jonge: Likewise and thank you for your help with patching hydra. Cheers!
<teto>
I am doing some changes on the lua support and I got lua5.2 instead of 5.1 pulled at some point I don't know why. I tried `$ nix why-depends 'lua5_1.withPackages(ps: [ ps.lpeg])' lua5_2 `, but it returns `error: don't know what to do with argument 'lua5_1.withPackages(ps: [ ps.lpeg])'
<{^_^}>
[nixpkgs] @arianvp opened pull request #48771 → [WIP] containers: Introduce several tweaks to systemd-nspawn from upstream systemd → https://git.io/fxK4d
smolboye has quit [Ping timeout: 245 seconds]
<infinisil>
teto: Use nix why-depends on store paths instead
<infinisil>
Or maybe the first argument should be wrapped in ()
smolboye has joined #nixos
jasongrossman has joined #nixos
mayhewluke has quit [Ping timeout: 244 seconds]
mayhewluke has joined #nixos
<{^_^}>
[nixpkgs] @hhomar opened pull request #48772 → gnupg: Fix, set current tty in interactive shell → https://git.io/fxKBK
Itkovian has joined #nixos
fragamus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<romildo>
How do I get colored output from grep run by nix-build? Using grep --color in postPatch for instance does not give color.
fragamus has joined #nixos
<tilpner>
What about --color=always?
<romildo>
tilpner, it does not work.
<romildo>
It seems that nix-build somehow disables color for the output of the commands that is run in the process. But I hope there is a way to enable it.
<romildo>
It would help when writing the nix expression.
orivej has joined #nixos
<sphalerite>
romildo: hm well nix isn't suppressing the colour codes
<romildo>
sphalerit, but there is no color code in the output. So what may be happening with it?
fragamus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<romildo>
In fact. I am running nix-build like the following:
<romildo>
nix-build --option sandbox true -K -A deepin.dde-session-ui 2>&1 | tee /var/tmp/nix-build.log
<romildo>
After removing the redirection and piping it works.
<sphalerit>
Aaaah so nix suppresses it when piping
<sphalerit>
romildo: you can use `nix log` instead of saving the logs yourself
Ariakenom has quit [Read error: Connection reset by peer]
<elgoosy>
rawtaz: I don't know how to try to run `emulator -avd Nexus5 -gpu-off from "/home/a/Android/Sdk/emulator/". I can't call `emulate` executable
<rawtaz>
maybe there's a way to tell android studio how to run the emulator, and thereby add the argument
<rawtaz>
it makes sense that there's a setting for it
<elgoosy>
i am looking at that, but no success yet
<teto>
infinisil: thanks but can't make it working. It's driving me crazy. I've reworked the lua support so that it works like haskell (with a generated part) but at some point it compiles with lua5.2 instead of lua5.1 and I can't find where this lua5.2 comes from.
<grokkingStuff>
Hi there! Trying to get wifi up and running on a nixos live cd (on surface 2017)
<grokkingStuff>
Can't seem to get wpa_supplicant to not fail after 10 seconds
<grokkingStuff>
anyone have this issue before?
<grokkingStuff>
on a related note, nixos works pretty well with my laptop (excluding the wifi issues). Good job, whoever worked their magic on the livecd
<ivegotasthma>
I'm testing out nixos in a qemu vm on arch linux, I start the VM with -display sdl. When I start X I cannot move the cursor, is this a known issue or am I doing something wrong?
EarlDeLaWarr has joined #nixos
<hyper_ch>
you're doing it wrong I guess
FRidh has quit [Quit: Konversation terminated!]
crmlt has quit [Remote host closed the connection]
<samueldr>
ivegotasthma: 500MB for the minimal installer image is being worked on; the way nix works makes it really easy to accidentally bring in more into the system closure than desired
jD91mZM2 has quit [Ping timeout: 246 seconds]
crmlt has joined #nixos
<samueldr>
though, as far as the cursor is concerned, unless you're using a 4.18 series kernel which happens to exhibit the issue, it may be another issue
<hyper_ch>
I thnk 500MB aint much
crmlt has quit [Remote host closed the connection]
<samueldr>
hyper_ch: it's all relative :)
<samueldr>
ain't much, but ain't the best we could do
<hyper_ch>
smaller than a cd
<hyper_ch>
smaller than any usb stick you can buy nowadays
<sir_guy_carleton>
it kind annoys me how you can't buy anything keys less than 8 GB, so you end up wasting all that space ...
ma27 has quit [Quit: WeeChat 2.2]
drakonis has joined #nixos
drakonis_ has quit [Ping timeout: 252 seconds]
crmlt has quit [Remote host closed the connection]
drakonis has quit [Read error: Connection reset by peer]
<samueldr>
might be interesting to make a new installer image type that works like the sd-image for aarch64; which is writable, and has its rootfs expanded to the size of the drive
<samueldr>
allowing a test-drive before an install
drakonis has joined #nixos
<samueldr>
or allowing live customization for things like wireless drivers
<samueldr>
permanent customization of the media*
ma27 has joined #nixos
drakonis1 has quit [Ping timeout: 252 seconds]
drakonis has quit [Read error: Connection reset by peer]
Mr_Keyser_Soze has quit [Ping timeout: 264 seconds]
wpcarro has joined #nixos
<wpcarro>
I keep losing my wireless connection after my laptop sleeps or hibernates (I'm not sure I know the difference). I can reproduce this behavior by closing my laptop and reopening it. I'm using a Macbook Pro and haven't been able to debug this for awhile
drakonis has quit [Read error: Connection reset by peer]
drakonis_ has joined #nixos
<wpcarro>
Someone told me to switch to kernel 4.18, which I did, but that hasn't fixed the issue
orivej has quit [Ping timeout: 240 seconds]
<wpcarro>
I've seen some solutions online for Ubuntu and other OSs, but I usually cannot follow them command-for-command because some files are in different locations on my NixOS than the answer's OS
<wpcarro>
For now, I'm just going to set `system.logind.lidSwitch = "ignore"` because the system is tough to use when the network continuously drops. I'd prefer to have a more elegant solution, but it seems like I'll need to dig around for longer until I find that
dbmikus__ has quit [Ping timeout: 252 seconds]
sanscoeur has joined #nixos
sanscoeur has quit [Remote host closed the connection]
wpcarro has quit [Remote host closed the connection]
<samueldr>
,tell wpcarro [about wifi on resume] with a link to said instructions, many more experienced users could probably help figure out how to port them to nixos :)
<arianvp>
I made a small patch to the systemd package in NixOS
o1lo01ol1o has joined #nixos
<arianvp>
but that makes almost the entire world rebuild
<arianvp>
is there a way to trick it to not rebuild everything? And if not,is there a way to check how much will be rebuilt ? because i've been building for 2 hours now
<arianvp>
and wondering how many packages are left
<ivan>
heh, I have the same problem but I resigned to setting up a hydra to build everything
<{^_^}>
[nixpkgs] @matthewbauer closed pull request #48286 → avr: use new cross compilation infrastructure → https://git.io/fxBq4
<{^_^}>
[nixpkgs] @matthewbauer reopened pull request #48286 → avr: use new cross compilation infrastructure → https://git.io/fxBq4
freeman42x has joined #nixos
fragamus has joined #nixos
pie__ has joined #nixos
freeman42y has quit [Ping timeout: 260 seconds]
dbmikus_ has joined #nixos
<aleph->
Is there a way to create directories in environments as well? Ideally I'd like to create the logrotate.conf.d dir and all the files it needs as well.
<mightybyte>
I'd like to do this so I can use nix to test that the cabal solver works with my packages.
fragamus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<mightybyte>
...and to also automatically run "cabal install --allow-newer" every so often and notify me about failures with new versions of dependencies.
<mightybyte>
When I'm writing apps I don't really care about this. But for my hackage packages I do.
<infinisil>
mightybyte: With recent nixpkgs versions you should be able to do `myhaskellpackage.getBuildInputs.systemBuildInputs`
<infinisil>
Oh, and that's available through `(haskell.lib.getBuildInputs myhaskellpackage).systemBuildInputs` for all nixpkgs versions
<infinisil>
> deepEval (map (x: if isNull x then null else x.name) haskellPackages.xmonad.buildInputs)
<{^_^}>
[ ]
<infinisil>
Well I tried
<mightybyte>
No luck?
<mightybyte>
I suppose you could just set up the full environment with all the Haskell dependencies and then just delete the GHC package repository. :P
shabius has joined #nixos
shiver has quit [Remote host closed the connection]
<infinisil>
mightybyte: What do you by "an environment for a haskell package without any of the haskell deps" anyways?
<elvishjerricco>
What would be a good way to measure the performance impact of a change to callPackage on the evaluation time? Is there a particular expression I could `time nix eval`?
<mightybyte>
infinisil: That's not sufficient though because some Haskell packages need additional non-haskell dependencies. For instance, zlib is a common one.
<mightybyte>
Sure I can do this manually on a per-project basis, but I'd like to have something completely automatic.
<infinisil>
Oh yeah so that's what I was going for earlier, but then you didn't mention this
<infinisil>
> systemDeps = pkg: lib.filter (x: ! isNull x) (lib.concatMap (x: if isNull x then [] else (pkgs.haskell.lib.getBuildInputs x).systemBuildInputs) (pkgs.haskell.lib.getBuildInputs pkg).propagatedBuildInputs)
<mightybyte>
I get "error: undefined variable 'lib' at (string):1:7" on that systemDeps line
<nkaretnikov>
acowley: hi, sorry to bother, but perhaps you could help me out a bit. i know you're a mac and nix user, and have some opengl-related packages on github. have you had any issues with opengl when using nix? i'm trying to run a gloss application, but it only shows a black screen, likely because cabal build doesn't add OpenGL.framework to the list of shared libraries. have you seen anything like that? thanks in advance! ps:
<nkaretnikov>
i've tried fixing OpenGLRaw and GLURaw packages to use the framework in my local copy of nixpkgs, but it doesn't seem to help. Or I've messed something up. In any case, i'm pretty much out of ideas at this point, so any tips would be appreciated.
silver has quit [Ping timeout: 252 seconds]
<infinisil>
mightybyte: lib = pkgs.lib
<mightybyte>
Ahh, ok.
<aleph->
Okay adding another backend wouldn't be too hard.
<aleph->
Anybody here use prgmr.com?
<mightybyte>
aleph-: Tried it years ago
<aleph->
Anybody be interested in a nixops target for it?
<aleph->
mightybyte: Oh neat, not a fan?
<mightybyte>
aleph-: It was fine for a small cheap box, but eventually that wasn't what I needed.
<aleph->
Aye fair enough. Talking with the owner about adding a nixops backend for it slowly.
<ivegotasthma>
hello, any help as to why the mouse cursor isn't working inside a qemu vm running a nixos guest on arch linux host?
<mightybyte>
infinisil: So how would I go about integrating what you did above into something that's really easy to add to an existing Haskell package to get that kind of testing?
<mightybyte>
These things are still not obvious to me.
<{^_^}>
[nixpkgs] @Mic92 merged pull request #48797 → corosync: update RDMA support to rdma-core → https://git.io/fxKQj
<infinisil>
Sorry my hands hurt too much already and I'll go to bed now, hopefully somebody else can help you with that. The resulting type of this function is just a list of packages though
<mightybyte>
infinisil: Ok, cool. Thanks a lot!
drakonis has quit [Quit: WeeChat 2.2]
<mpickering>
ap
slack1256 has joined #nixos
civodul has quit [Quit: ERC (IRC client for Emacs 26.1)]
<fiatjaf>
has anyone ever tried to install nixos on a vultr vps?
<akavel>
Hi! Can I run nix-shell with default.nix + some attributes from nixpkgs?
<akavel>
I have a default.nix generated by node2nix, it's not clear to me where would I need to add extra buildInputs in it;
pie__ has quit [Ping timeout: 245 seconds]
<akavel>
Normally, the derivation generated by node2nix can be started with: `nix-shell -A shell`
<akavel>
but when I try to add -p options, it fails:
<akavel>
$ nix-shell -p gifsicle -A shell ## error: attribute 'shell' in selection path 'shell' not found
<jackdk>
akavel: I don't know the patterns for node2nix files, but when I cabal2nix package foo, I usually dump it into a foo.nix, then default.nix imports ./foo.nix and modifies things if necessary.
<akavel>
I'm not yet adding, because I'm not sure how to do that :) hm, ok, so maybe rename the default.nix to foo.nix as you say, and then do as you say further
<akavel>
I'll try now...
<fiatjaf>
jackdk, that would be the answer to my second question :P
<fiatjaf>
jackdk, the first one is: how do I get vultr to stop booting into the installation ISO?
<akavel>
jackdk: thanks for the ideas
<fiatjaf>
akavel, everything is a derivation.
<jackdk>
fiatjaf: do you have the boot.loader.grub settings? I do not recall needing to fiddle boot order in the configuration
<fiatjaf>
jackdk, oh, I see
__monty__ has quit [Quit: leaving]
<fiatjaf>
it's very unexpected for me that the os determines the bootloader order like that, but I guess nix is a different beast
uri-canva has joined #nixos
<jackdk>
I don't think it does. I suspect that boot from CD is the first option, then HD, that fails because your grub didn't install properly, then the next option is PXE
<akavel>
jackdk: mbwahahaha, I managed to squeeze it into a "one-liner", a.k.a. magical incantation: nix-shell -E 'with import <nixpkgs> {}; (import ./default.nix {}).shell.overrideAttrs (a: { buildInputs = a.buildInputs ++ [gifsicle]; })'
<akavel>
Thanks! :)
<akavel>
Now, got some another error, obviously ;)
jtojnar has quit [Quit: jtojnar]
<{^_^}>
[nixpkgs] @matthewbauer opened pull request #48803 → Experiments in pkgsStatic: a fully static overlay → https://git.io/fxKbY
<uri-canva>
Is there a logging bot on the Nix channels so I can search if a particular question has been asked before?
<samueldr>
the logs are even starting to pop up in google search results sometimes :)
<pie___>
if im not using channels, just pointing nixpkgs= to a git repo, to update, i just update the repo and rebuild switch right?
<samueldr>
pie___: right
<pie___>
ok thanks
uri-canva has quit [Ping timeout: 256 seconds]
aanderse has joined #nixos
drakonis has joined #nixos
akavel has quit [Quit: Page closed]
graphene has joined #nixos
camsbury has joined #nixos
camsbury has quit [Ping timeout: 252 seconds]
<pie___>
so, afaict stateversion is basically for not breaking things on an update that generally would?
<slack1256>
Does anybody have problems with TRAMP under nixos?
<pie___>
so when you update you get warnings (do you?) for services that you have an older config for, and then you can go look at a migration guide?
<slack1256>
7w
<slack1256>
sorry
<pie___>
samueldr, ^ (?)
<pie___>
another thing: I looked at my mount list and it had these lines:
<pie___>
/dev/sda1 on / type ext4 (rw,relatime,data=ordered)
<pie___>
/dev/sda1 on /nix/store type ext4 (ro,relatime,data=ordered)
<pie___>
how does it work that the same block device is mounted in two places with different directory structures? or is mount just not showing all the information
<jasongrossman>
slack1256: Tramp as in Emacs? I only use it a bit, but it's been fine for me.
nbardiuk has quit [Ping timeout: 268 seconds]
<slack1256>
jasongrossman: Yes as in Emacs. I just wanted to discard that as as source of error.
<clever>
pie___: bind mounts
<clever>
pie___: you can mount a subdir of anything to another dir
<pie___>
oh ok that makes sense
<pie___>
the information given by mount is kind of misleading then?
<clever>
in this case, nix mounts /nix/store to /nix/store, with an extra read-only mount
<samueldr>
pie___: no warnings, and as of right now, and nothing in the close future, is expected to break at any older stateVersion (AFAIK)
<jasongrossman>
pie___: If you think of the Linux filesystem as a tree then it seems misleading, but it just isn't a tree. Does that help?!
<samueldr>
I'm pretty sure most stateVersion using options are folder locations
<clever>
samueldr: it can also impact the postgresql version, and some ssh hostkey types
<samueldr>
oh right, versions
<samueldr>
but the main thing to remember is that as of today, it's not an issue to keep older stateVersion
<jasongrossman>
pie___: Or do you mean the fact that not the whole of /dev/sda1 is mounted on /nix/store? I agree that the output of mount hides that fact.
<clever>
in the case of sshd, its to prevent the hostkey changing and causing unexpected mitm warnings
Rusty1 has quit [Quit: Konversation terminated!]
<samueldr>
and if you want to update, the safest way is to look at everything using `stateVersion` in <nixos/modules>
<samueldr>
s/want to update/want to change it/
<worldofpeace__>
anyone want to merge/comment on this backport pr #48150 ? (been sitting for over a week)
<pie___>
ugh crap, looks like rebuild switch still fails to restart tinc
sigmundv has joined #nixos
<pie___>
not sure if i just have something configured badly... the errors :https://bpaste.net/show/b3cdceccc098
sigmundv has quit [Remote host closed the connection]
<pie___>
i have an old command laying around that seems to fix it: systemctl start network-interfaces.target network-addresses-enp0s25.service network-local-commands.service network-addresses-tinc.iaacluster.service
sigmundv has joined #nixos
<pie___>
which may or may not make sense. is there some way i could make it so that doesnt fail and i have to physically go and fix the connection?
Lears has quit [Ping timeout: 264 seconds]
fragamus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]