<{^_^}>
[nixpkgs] @samueldr pushed commit from @FlorianFranzen to release-19.03 « grub: Support 32bit EFI on 64bit x86 platforms »: https://git.io/fjkMM
kleisli has joined #nixos
abueide has joined #nixos
abueide has quit [Remote host closed the connection]
<colemickens>
So, this is weird. This reminds me of the nscd issues I was having a while back.
<colemickens>
I got my laptop out of my bag. Firefox can't resolve anything. If I go to Google's IP it bounces me to google.com and fails.
<colemickens>
I was going to say that Firefox still doesn't work, even after enabling DoH, but it still has to resolve the initial CF DoH server, which is probably failing.
<colemickens>
Maybe this is still just NSCD being annoyed?
gratinb has quit [Remote host closed the connection]
gratinb has joined #nixos
justan0theruser has joined #nixos
justanotheruser has quit [Ping timeout: 245 seconds]
<{^_^}>
[nixpkgs] @worldofpeace pushed 2 commits to master: https://git.io/fjkDM
<Ashy>
i had never heard of nscd until using nixos
gratinb has quit [Ping timeout: 246 seconds]
evil-olive has joined #nixos
drakonis1 has quit [Quit: WeeChat 2.3]
justan0theruser is now known as justanotheruser
tomberek has quit [Ping timeout: 245 seconds]
kvda has joined #nixos
evil-olive has quit [Quit: leaving]
<rhitakorrr>
Anyone know why I might be getting this error? "/nix/store/[...]-ruby-2.6.1/lib/ruby/2.6.0/bundler/spec_set.rb:91:in `block in materialize': Could not find rake-12.3.2 in any of the sources (Bundler::GemNotFound)"
<mightybyte>
Hmmm, I blew away /nix and reinstalled and now it works. I must have messed something up when I installed nix-darwin or home-manager before.
<rhitakorrr>
It happens when I try to run my app (`ruby app.rb`)
evil-olive has joined #nixos
<rhitakorrr>
If I run `which rake` or `rake --version`, the proper version seems to be installed as expected.
<slack1256>
rhitakorrr: did you set up the environment that the ruby interpret see correctly? If rake is a library is should be specified at construction time
<rhitakorrr>
slack1256: I used bundix to create a gemset.nix from the Gemfile.lock, then used that with bundlerEnv, which used to work until I just updated to Ruby 2.6.
gagbo has quit [Ping timeout: 245 seconds]
tomberek has joined #nixos
<{^_^}>
[nixpkgs] @dhl opened pull request #58644 → wasm-pack: init at 0.7.0 → https://git.io/fjkye
InFlames has quit [Quit: Konversation terminated!]
InFlames has joined #nixos
<colemickens>
monokrome: there's an draft PR that builds a bootable image
<CMCDragonkai>
In CLI land, I can use `nix-store --query --requisites $(nix-instantiate ./default.nix)` to get the list of all closed dependencies of a derivation. Is there a way to do this within the nix language?
<monokrome>
colemickens: oya? :O
* slack1256
reads builtins commands
adetokunbo has joined #nixos
devalot has quit [Remote host closed the connection]
ddellacosta has quit [Ping timeout: 245 seconds]
ddellacosta has joined #nixos
<clever>
CMCDragonkai: one of the new string context things may allow it
<CMCDragonkai>
clever: What do you mean?
<CMCDragonkai>
Also I found this thing called `exportReferencesGraph`
<CMCDragonkai>
But I need to know what it looks like. So I've been looking for a way to run `nix-build` but leave the build directory untouched.
<CMCDragonkai>
I would normally use `nix-build -K`, but that only preserves the directory if the build failed.
sgronblo has quit [Ping timeout: 250 seconds]
<CMCDragonkai>
Ok so I forced a a failure with `nix-build -K`. So it produced the reference graph file. It sort of gives me what I want, but I cannot use it properly because I actually need as a nix variable, not as something that is accessible by build scripts.
<elvishjerricco>
CMCDragonkai: You can write a derivation that puts those files in the output then `builtins.readFile` them. `pkgs.closureInfo` is such a derivation that outputs JSON files, so you can do readFile and fromJSON to get a Nix attr set
<{^_^}>
[nixpkgs] @worldofpeace pushed 2 commits to master: https://git.io/fjky7
sgronblo has joined #nixos
vk3wtf has joined #nixos
<{^_^}>
[nixpkgs] @Ericson2314 pushed 361 commits to staging: https://git.io/fjkyF
tunawithbacon has joined #nixos
appleclusters has joined #nixos
Rusty1 has quit [Quit: Konversation terminated!]
<tunawithbacon>
If a C program has '#include <editline/readline.h>' what should I put in my nix-shell default.nix to be able to compile it?
<acham>
tunawithbacon: you need to add buildInputs = [ readline ]; I suppose
<acham>
or perhaps
<acham>
buildInputs = [ pkgconfig readline ];
<appleclusters>
Running into a ridiculous issue. My campus WiFi is currently hanging on authentication. When I reboot, NetworkManager blocks the systemd processes because it keeps trying to connect. This never happens if there's no network available. It's happening because the network is malfunctioning. I literally can't boot into any generation with wifi enabled
wfranzini has quit [Quit: wfranzini]
<appleclusters>
Worse, I can't just comment out networking because nixos still attempts to install packages via the network
<appleclusters>
How do I even start fixing this?
<acham>
appleclusters: maybe you can boot into single user mode or some other mode without networking
<appleclusters>
Oddly enough, I commented out NetworkManager.enable and switched into the latest generation and now I can't authenticate into gnome with root or otherwise, yet the wifi icon shows it's working (!)
<appleclusters>
Also, how do I prevent network manager from attempting to connect on boot?
domogled has joined #nixos
<appleclusters>
Is there a command to switch from window manager to shell?
gerschtli1 has quit [Quit: WeeChat 2.2]
ym555 has quit [Ping timeout: 255 seconds]
gerschtli has joined #nixos
<slack1256>
Control-Alt-F5
domogled has quit [Ping timeout: 255 seconds]
<slack1256>
will throw you on a tty
tunawithbacon has quit [Quit: Page closed]
ninjin has quit [Ping timeout: 256 seconds]
<appleclusters>
Thanks!
<slack1256>
appleclusters: Also a tip, if X sometime grabs your keyboards and doesn't let you change tty via Control-Alt-F5, your can "reclaim" your keyboard from X via Alt-PrintScrn-R , which is a low level linux command for that
<slack1256>
then you can change your tty as before
<appleclusters>
@slack1256: one more question--my username shows up in /etc/passwd but after doing rebuild switch on a very early generation, I can no longer log in with the credentials I set
<appleclusters>
(on the latest generation)
<appleclusters>
I can log in as root in tty, but not in gnome3
palo1 has joined #nixos
<slack1256>
you did a switch on a live system right?
palo has quit [Ping timeout: 255 seconds]
palo1 is now known as palo
<appleclusters>
Yeah. And I fixed it, just had to re-run passwd
<appleclusters>
For whatever reason the declarative process messed it up I guess
Myhlamaeus has quit [Ping timeout: 258 seconds]
kvda has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<monokrome>
Yo! Does anyone know of good documentation for setting up kubernetes on nixos?
* monokrome
can only find single-node examples :/
<monokrome>
ls
<monokrome>
oops sorry lol
* monokrome
is still getting used to her new keyboard >.<
<jackdk>
monokrome: is that on the purism? did you get nixos going on it?
<manveru>
monokrome: Check Out kubenix. Also I think zimbatm wrote something about it
<Arahael>
I'm still figuring out this shell.nix thing. What I need is to have a particular python dependency (amongst other non-python dependencies), so I have done a 'let myPython = python3.withPackages(ps: [ps.jinja2]); in mkShell ...', however, I require a specific version of jinja2 for some reason. How do I do this?
<monokrome>
manveru: I did
<monokrome>
also how did you know I was using it?
<monokrome>
oops sorry, jackdk
<Arahael>
I think I need to write a custom 'jinja2_2.7.nix' file or something, but then, how do I refer to it from shell.nix?
<manveru>
rhitakorrr: you'll need to use the wrappedRuby that your bundlerEnv gives you
<monokrome>
jackdk: Setup was a bit funky w the keyboard, but it'll be fixed once systemd is updated to latest it'll be working fine :D
slack1256 has quit [Remote host closed the connection]
<monokrome>
jackdk: or are you wondering if I want to run k8s on the purism? If so, no, I'm trying to run it on DO
<monokrome>
but I'll probably just use swarm for now
<monokrome>
kubenix seems maybe cool but no docs afaict
reinhardt has joined #nixos
<jackdk>
you asked about it here a while back. I'm getting close to upgrade o'clock on my current machine, and if the librem13 is sound then maybe I should start voting with my wallet
reinhardt has quit [Client Quit]
adetokunbo has quit [Quit: This computer has gone to sleep]
<monokrome>
The librem13 is mostly pretty cool
<monokrome>
The wifi feels weird when wpa_suppliant isn't configured, but works great once you get wpa_supplicant.conf set up
<monokrome>
The keys are really rigid but hopefully they wear a bit with use
<monokrome>
Pressing '!' is rough sometimes :D
<monokrome>
jackdk: lol, I thought you were like sitting behind me at Zeitgeist in SF or something and I was like "small world wtf" lol
reinhardt has joined #nixos
ddellacosta has quit [Ping timeout: 268 seconds]
reinhardt has quit [Quit: Leaving]
endformationage has quit [Quit: WeeChat 2.4]
vk3wtf has joined #nixos
linarcx has joined #nixos
<CMCDragonkai>
elvishjerricco: But the problem is that `builtins.readFile` runs at Nix expression evaluation time, and the time to get the closure outputs is at buildPhase time.
<elvishjerricco>
CMCDragonkai: I'm pretty sure you can readFile a derivation, and it'll just build that derivation before continuing evaluation
reinhardt has joined #nixos
<CMCDragonkai>
elvishjerricco: That doesn't make sense to me. This is the export references graph.
<CMCDragonkai>
elvishjerricco: How can a nix expression read a file, when that file doesn't exist when the nix expression is being evaluated?
* Arahael
wonders if that readFile is the correct way to refer to such a jinja2_2.7.nix module?
<srhb>
CMCDragonkai: Nix is pretty good at dependency graphs. :)
<elvishjerricco>
CMCDragonkai: It builds the derivation, that derivation completes building, and evaluation continues
<CMCDragonkai>
elvishjerricco: can you give an example of a derivation that acquires its own closure of dependencies, searches through that to get the version of a particular store output, and encodes that as env variable within the derivation?
<elvishjerricco>
CMCDragonkai: No, because that sounds like 15 minutes of work :P
<elvishjerricco>
probably more
<CMCDragonkai>
elvishjerricco: I don't think it's possible right now based on what you told me, and if it is, it appears highly convoluted, basically sort of running a derivation twice.
<elvishjerricco>
CMCDragonkai: No, you have a completely different derivation, unrelated to whatever you're trying to do
<elvishjerricco>
That derivation copies the files to $out
<elvishjerricco>
and you use readFile to read them, do your Nix evaluation, and produce whatever derivation you care about
<elvishjerricco>
CMCDragonkai: This gives you a nix expression containing the contents of the references graph hello file. You do this to acquire whatever info you need, and you evaluate a derivation you care about based on that
<elvishjerricco>
Maybe I just need to know more about what you're actually trying to do
<CMCDragonkai>
elvishjerricco: seems like you're saying that I should create a utility derivation that outputs the `exportReferencesGraph` and then my main derivation uses that utility derivation to get to use the graph
<elvishjerricco>
CMCDragonkai: Exactly
<CMCDragonkai>
elvishjerricco: I see how it is possible, but I don't like the implementation... this is mainly so that I can get the version of CUDA and NVIDIA driver so I can add it as a constraint
<CMCDragonkai>
The code complexity of doing this outweighs the benefit :(
<elvishjerricco>
CMCDragonkai: Is that info not present in those derivations? like cuda.version or something?
<elvishjerricco>
because yea this would definitely be overkill for that :P
<elvishjerricco>
unless they're transient dependencies and you need to *find* them
<Arahael>
Hmm, it turns out that if I copy the 'jinja2' directory from nixpkgs, which consists of just it's default.nix, and then try to import it, it doesn't seem to actually read the file. (CHanging teh version makes no difference. Neither does changing the sha hash...)
* Arahael
tries readFIle instead.
invokesus has joined #nixos
kleisli has quit [Ping timeout: 250 seconds]
linarcx has quit [Ping timeout: 245 seconds]
Myhlamaeus has joined #nixos
linarcx has joined #nixos
<Arahael>
Ok, I'm perplexed. readFile doesn't seem to have any effect, it had an error unless I used the *full* path to the default.nix file I wanted, but then nix-shell magically works. Unfortunately, it persists in using an older python3 environment to what I want.
<elvishjerricco>
Arahael: Something like that might work, depending on what that nix file looks like
bemeurer_ has joined #nixos
<monokrome>
I added authToken under deployment.digitalOcean, but it doesn't seem to work?
sgronblo has quit [Ping timeout: 255 seconds]
<Arahael>
elvishjerricco: That actually managed to get it to read the default.nix file - I deliberately put in a typo there. However, how did it "work" before?
<monokrome>
oh it says use an env variable, so maybe the option is just broken
<elvishjerricco>
Arahael: how did what work?
<Arahael>
elvishjerricco: The gist I gave "worked".
<monokrome>
hooooray it worked
<elvishjerricco>
Arahael: Oh, but I assume the build failed catastrophically?
<Arahael>
elvishjerricco: I could type in 'nix-shell', and I would have a python3 that was in the nix store, and I'd even be able to 'import jinja2', but it was the wrong version, and shouldn't have been possible since my default.nix file didn't even have valid expressions.
bemeurer has quit [Ping timeout: 250 seconds]
<elvishjerricco>
oh
<elvishjerricco>
yea that makes no sense at all :P
<Arahael>
Yeah... I just took out my deliberate corruption of the defualt.nix file. :( Now, it's gone back to "working". :(
adetokunbo has joined #nixos
<Arahael>
It's as if it's reading - parsing - and then ignoring the jinja module.
<Arahael>
Also, as if the 'python35.withPackages' is somehow cached and not using the ps argument.
<Arahael>
It seems that if I type in 'nix-shell', it might use a previously-invoked nix-shell without re-reading the shell.nix
<elvishjerricco>
Arahael: That shouldn't be the case
<elvishjerricco>
Arahael: Do you have both a shell.nix and a default.nix in the current working directory? shell.nix takes precedent
<Arahael>
What I did was: write shell.nix; try using it with 'nix-shell', and discover it wasn't right, so *whilst in that shell*, I did 'vim shell.nix', and thereafter tested the changes by running a sub-shell in vim.
zupo has joined #nixos
<Arahael>
ie, I was recursively running nix-shell.
<Arahael>
Seems only the first invocation actually changes the environment.
<monokrome>
that usually works though
<monokrome>
ya it willl
<elvishjerricco>
Hm I've experienced very strange behavior trying to nest nix-shell in the past. I generally avoid doing it
<Arahael>
Hmm, I'll have to remember this.
<ArdaXi>
hiya, I've got a bit of a weird issue with a derivation I'm developing… the call to mkDerivation has only name and buildCommand args, but changing the buildCommand does not trigger a rebuild
Fare has quit [Ping timeout: 245 seconds]
fusion809 has joined #nixos
<Arahael>
elvishjerricco: The 'throw' works. :) (Not in a nested nix-shell anymore, though)
<Arahael>
New python has 'myJinja2', though it seems to be an empty module for some reason.
<Arahael>
I added a line: foo = throw "should throw;" to myJinja2/default.nix, it does NOT throw. :(
Fare has joined #nixos
wfranzini has joined #nixos
<Arahael>
Even `pname = throw "should throw";` has no errors.
<Arahael>
Just did a nix-collect-garbage, I had to re-download a bunch of stuff, but no throws(!)
<immae>
Arahael: for "foo" it’s expected, if that field is not used in your derivation
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<Arahael>
immae: What about the pname variable?
<immae>
can you try with `name = throw "should throw!";` ? Not every derivatino has pname, does it?
<immae>
(Arahael: you wrote quicker than me :p )
<Arahael>
immae: I could add a name, but I think it uses pname.
<immae>
pname should throw if it is used
<Arahael>
immae: Im not convinsed the file is used, at all. I mean, if I put in something like `1hjhfoalionr1lon;1rij;kj1r23`, it fails. But if it parses fine, or if I use readFIle, it magically "works".
<immae>
maybe it’s ignored in a nix-shell situation?
<Arahael>
Shouldn't be ignored, I mean, throws in the shell.nix itself works.
jbgi has joined #nixos
<immae>
yes, I mean in a nix-shell the pname doesn’t need to be read, does it?
<Arahael>
immae: I dunno, I'm very new to nix, and all I'm trying to do is install a very specific version of jinja2. How would it know what name the module would be otherwise, without the pname?
<sb0>
hi. i've set up nginx with ACME (simply added enableACME = true; and forceSSL = true; to the virtualhost entry). this works fine when connecting with IPv4, but with IPv6 I'm getting obscure SSL errors
<sb0>
curl: (60) SSL: no alternative certificate subject name matches target host name 'git.m-labs.hk'
<sb0>
and wget: ERROR: no certificate subject alternative name matches requested host name ‘git.m-labs.hk’
<immae>
Arahael: yes, definitely. Do you still have a version number with it?
<sb0>
what does that even mean?
<Arahael>
immae: No, as you can see, the default.nix is practically empty. It used to have a version, but not anymore.
<immae>
no, I mean in the result
<Arahael>
immae: (If you're referring to python's __version__, no, that's an attribute error now.
<immae>
you definitely have a name popping out of nowhere, I wonder if you have a version too
<Arahael>
__file__ also results in an attribute error. __name__ is there, though, and it's 'betaJ'.
<Arahael>
) Just weird.
<Arahael>
IT does have a correct path, though, but weirdly, it refers to the location of the blueJ directory.
<Arahael>
It's as if it assumes that the default.nix is in the same location as the rest of the python files for that module.
<sb0>
when looking at the exchanged data with wireshark, it connects then the client sends an "encrypted alert" and disconnects
<Arahael>
I'm looking for the implementation of 'withPackages', by grepping the nix packages repo, maybe it'll give some hints.
adetokunbo has quit [Quit: This computer has gone to sleep]
appleclusters has quit [Quit: Connection closed for inactivity]
sgronblo has quit [Read error: Connection reset by peer]
<Arahael>
immae: Thanks for your time - I think I'll have to look at this another day.
<immae>
ok, good luck :)
Zer000 has joined #nixos
<Arahael>
immae: Thanks - I think if I get a chance to look at this tomorrow, I should switch my channel to a nixpkgs fork, and modify that fork to help narrow things down.
<immae>
I think the end story there is simply that an empty buildPythonPackage ends up in an empty module, there is not much tmore to get
<immae>
not "empty module", just "nothing" actually
<Arahael>
immae: Yes, and that should be an error. Infact, I might raise a bug on that tomorrow?
<immae>
You may if you feel like it
<Arahael>
I should have just used pip instead of nix to install this jinja thing, but then I would have learnt a few things less! (Feels like diving in the deep end, though!)
<CMCDragonkai>
Given a derivation you can find its outputs. Just look at its output paths. But given an output (its output path), can you find what derivation produced that output?
<CMCDragonkai>
"output provenance"
erictapen has joined #nixos
zupo has joined #nixos
<Arahael>
CMCDragonkai: I've gone home, so I am no-longer at the affected machine, however, that's an excellent tip, I should have looked for that. However, I don't know how this works w.r.t. python's packages, which seem a bit /different/.
<Arahael>
I'll try that tomorrow, though - any other tips?
thc202 has joined #nixos
pie__ has joined #nixos
Xal has quit [Ping timeout: 250 seconds]
sgronblo has joined #nixos
InFlames has quit [Ping timeout: 250 seconds]
periklis has joined #nixos
Xal has joined #nixos
polman has quit [Ping timeout: 255 seconds]
<sb0>
wtf? setting enableIPv6 to false now causes the same bug for both IPv4 and IPv6
fendor has joined #nixos
InFlames has joined #nixos
pietranera has joined #nixos
polman has joined #nixos
kvda has joined #nixos
<hyper_ch2>
hmmmm, I was wondering why my homeserver is so "slow" and has high load displayed.... it's 1st of the month - scrubbing time
<johanot>
hyper_ch2: it's the 1st of the month. Buy a new server? $$$ :P j/k
<hyper_ch2>
wasn't easy to attach 10 hdd to that big tower homeserver
<johanot>
pff.. hardware.. so complicated
barrel has quit [Ping timeout: 245 seconds]
<johanot>
but fair point :-)
<hyper_ch2>
unfortuantely they don't sell 100TB SSDs for $ 99.99 yet
justanotheruser has quit [Ping timeout: 250 seconds]
CcxWrk has quit [Killed (verne.freenode.net (Nickname regained by services))]
CcxWrk has joined #nixos
<fendor>
any recommendations on backup software? kind of afraid of rsync.
<tilpner>
I use duplicity, but my script isn't very smart about when to schedule full backups
<jasongrossman>
fendor: A lot of people here use ZFS, which makes backups almost trivial. I know it's not feasible for everyone to reformat their disks, though.
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<hyper_ch2>
fendor: zfs send/recv :)
<fendor>
jasongrossman, yeah, i am not gonna reformat the disk a third time this week
<jasongrossman>
fendor: Made me laugh - sorry!
<tilpner>
I plan to try zfs for backups too, but it requires a running host and doesn't really have encryption yet
<elvishjerricco>
tilpner: Soon TM
<fendor>
jasongrossman, no worries :D
<hyper_ch2>
there's native zfs encryption in unstable
<elvishjerricco>
But actually. Should be released extremely soon
<tilpner>
I know, but I can't really justify using an pre-release filesystem
<hyper_ch2>
also, backups can be sent to just a normal server somewhere with the raw flag, so it will store it in encrypted format as fail
<tilpner>
as fail?
<tilpner>
or file?
<hyper_ch2>
zfs master even has SSD trim support now :)
<hyper_ch2>
s/fail/file/
<elvishjerricco>
yea the TRIM support was a nice surprise
<elvishjerricco>
I've heard it should make it into 0.8 since releases and candidates are cut straight from master. Not sure if that's true though
<hyper_ch2>
one of the reasons 0.8 isn't released yet was trim
<tilpner>
duplicity works fine with cloud services, so it supports a different usecase
<jasongrossman>
elvishjerricco: I'm sure it is true. (See, I just jinxed it.)
<hyper_ch2>
there's also one more thing they want to include but I forgot what it is
<hyper_ch2>
there's also Borg backup... looks like it works great but never used it
barrel has joined #nixos
<tilpner>
There was FUD about their encryption, and it didn't support S3 backends
<hyper_ch2>
but I've been using rsync just for years
gal_bolle has joined #nixos
<Taneb>
Is there a convenient way to fetch and untar a tarball, as a derivation?
<tilpner>
fetchTarball?
Jetien has joined #nixos
<fendor>
tilpner, duplicity is only a command line utility?
<hyper_ch2>
welcome to linux :)
<tilpner>
Yes. I think there are GUIs, but those can't be automated and I want my backups to run at least daily
<tilpner>
(They can be, but it's much more pain, so why would you do that)
<tilpner>
fendor: Yes, duplicity supports a large selection of remote backup targets. I use it with Wasabi for $5/TB/month (with a minimum of $5/month), and my laptop and server backup to that
<sphalerite>
I like both, depending. I prefer pushing them from my laptop to my backup server, but pulling them from my various-stuff-server to my backup server
<fendor>
currently, i do not have a backup server, but your approach seems cool!
<hyper_ch2>
I just have the mindset: "The live server shouldn't dictate how many backups are kept etc... a malicious introduer on the live server could remove them all"
<Arahael>
hyper_ch2: Have you considered just rsync?
fendor has quit [Remote host closed the connection]
<jasongrossman>
hyper_ch2: I'd never thought of that. Good thinking.
<Arahael>
Erm...
<Arahael>
tilpner: Have you considered just rsync? :
<sphalerite>
but zzzzffffffsssss
<tilpner>
Arahael: I have, and it's not suitable for my needs at all. No versioning, no compression, no deduplication, no encryption
<tilpner>
Arahael: You may have meant to ping fendor though, who asked about backup solutions originally
fendor has joined #nixos
<hyper_ch2>
I used rsync and hardlinking in the past
<Arahael>
tilpner: Ah, well, rsync and cp, I mean. :)
<{^_^}>
[nixpkgs] @lihop opened pull request #58665 → zzuf: init at 0.15 → https://git.io/fjkQ4
<tilpner>
Arahael: I also don't have a backup server, so where would I rsync to?
<tilpner>
Arahael: And I can't just plug a USB drive into a Pi, Wasabi most likely has more redundancy than that
<d10n-work>
fendor: I switched from duplicity to borg because borg lets you delete old backups
<tilpner>
duplicity lets you delete old backups too
barrel has quit [Quit: leaving]
<Arahael>
tilpner: Ah, once I get a decent internet connection, I'll probably sync to amazon s3.
<tilpner>
Arahael: For which you can't use rsync IINM
<d10n-work>
unless something changed in the last year, incremental duplicity backups rely on the most recent full backup, so you can't delete that and keep incremental backups since then
<tilpner>
d10n-work: That hasn't changed. But you can still delete old backup chains
<d10n-work>
tilpner: every borg backup is incremental and you can delete any arbitrary old backup without affecting any other backup. I guess I wasn't specific enough to be fair, but it's a possible 'gotcha' for duplicity
<tilpner>
Yes, definitely. duplicity isn't meant for incremental-forever, but then what is?
<tilpner>
If you kept 5 years of incremental backups on S3, the restore costs would hurt a lot if you need to fetch all of those to restore a single file
<tilpner>
(I don't know if borg requires that)
<fendor>
so, if i want incremental backups, what is to prefer? or recommended?
<d10n-work>
borg is incremental-forever
<d10n-work>
borg lets you mount any backup which lets you restore single files
<tilpner>
But what files does it need locally do let you mount the backup?
<tilpner>
s/do/to/
Fare has quit [Ping timeout: 246 seconds]
<d10n-work>
I'm checking now, but I think either everything or nothing, depending on whether you consider mountable cloud drives to be local
<tilpner>
d10n-work: When you have 10TB of backups in S3, does it need to download that locally to let you mount it? A mountable cloud drive just hides what's being downloaded, though it would help with local disk storage
<Dagger>
IIRC it uses some sort of rolling content checksum to split your data into blocks which are stored in a key/value store, so logically it should only need to download the data referenced by the backup you want to mount
<d10n-work>
tilpner: I am 90% sure that it would only stream what it needs but I am double-checking now with my own backups
__monty__ has joined #nixos
<Dagger>
disclaimer: I haven't so much as looked at my borg install in $too_long; I'm not exactly an expert at it
<d10n-work>
side note: one borg gotcha I ran into is that the max_segment_size determines the max size of the archive. I have 1 big archive, so I increased it
random_yanek has quit [Ping timeout: 268 seconds]
jbgi has quit [Ping timeout: 268 seconds]
adetokunbo has joined #nixos
Jetien has quit [Remote host closed the connection]
<jomik>
betaboon: Well, I want to wrap it, so that it gets passed automatically. I don't want to write `kitty --config /nix/store/...` ^^
iyzsong has joined #nixos
<acham>
jomik: you could wrap it again lol
<betaboon>
jomik: couldnt you just rely on kitty automatically loading the config form eg `/etc/xdg/kitty/kitty.conf` or setting `KITTY_CONFIG_DIRECTORY`-env-variable. and then do `environment.etc."kitty/kitty.conf".text = ''...'';` and `environment.KITTY_CONFIG_DIRECTORY = "/etc/kitty";` ?
<jomik>
betaboon: How would I manage to set the KITTY_CONFIG_DIRECTORY environment variable without wrapping? :P
<betaboon>
jomik: are you running nixos ?
<jomik>
acham: Yeah.. I did that, I went down the rabbit hole :D I have a `kitty`, `.kitty-wrapped` and `..kitty-wrapped-wrapped` :D
<jomik>
betaboon: Yeah. I don't want to set a environment variable on my entire system though - it is only needed by kitty :P
adetokunbo has joined #nixos
<betaboon>
jomik: but thats what env-vars are for ? so you want to modify the package-definition to take in a config and add it to the wrapped binary? i would realy try to avoid that, as that would cause a rebuild of the package every time you change your config and you wont be able to use the build from the binary-cache.
<betaboon>
jomik: maybe a better option would be to create a new derivation that takes in the kitty-package and your config.
<jomik>
betaboon: That's how the vim derivation on Nix works, as an example.
<jomik>
So I think it is a pattern we follow ? Since we want to isolate dependencies.
<jomik>
It also makes it easier to test, because I don't depend on changing a file on a set path in my system, I can just run kitty pointing at the generated config.
<betaboon>
jomik: well if you want it that way you would have to modify the derivation.
vk3wtf has quit [Quit: WeeChat 2.4]
vk3wtf has joined #nixos
silver has joined #nixos
jomik has quit [Quit: WeeChat 2.4]
andymandias_ has quit [Remote host closed the connection]
<over7head>
im interested in nix and im reading doc's, but all this looks strange to me
<over7head>
this is not user friendly system, isnt it?
<Yaniel>
depends on your definition of what makes a system user friendly
<over7head>
can i use NixOS in tradiucional Linux way ?
<over7head>
for example, i dont want kde...
<Yaniel>
of course you can
<Yaniel>
it's just another linux distribution
<Yaniel>
then don't enable kde in your configuration file
<over7head>
so, just get minimal and do some "install xorg-server" "install awesome-em" and so..
<Yaniel>
you don'
<Yaniel>
you don't do "apt install <stuff>" style things on nixos (usually, you can do that in your user profile if you want)
<over7head>
thats what looks strange to me.... enable, disable in configuration
<Yaniel>
instead you have a file that specifies all things that should be installed
<Yaniel>
configuration for system-level stuff is also specified in that file
<over7head>
hm... so i need to configure this file and then run some reconfigure and it will download and install...
<Yaniel>
why strange?
<Yaniel>
people use docker for the exact same reason
<over7head>
usally i would install minimal and just apt get stuffs... thats why it looks strange
<gchristensen>
it is definitely different than other linuxes
<gchristensen>
but in pretty cool ways
<over7head>
thats why im interested, because its different
<gchristensen>
on debian I would have no hope of successfully ridding my system of Plasma if I installed it, but with NixOS you can enable it, decide you don't care for it, and then just remove it from the configuration file
<betawaffle>
is it possible to use nix-shell with fish?
<etu>
betawaffle: nix-shell --run fish -p hello
<betawaffle>
cool, thanks
<etu>
(I also have this abbr: "abbr -a -U -- nsh 'nix-shell --run fish -p'" so when I type "nsh" and press space I can start typing package names :)
<etu>
)
<betawaffle>
nice
<betaboon>
over7head: i came from gentoo to nixos, always aiming as the smallest possible install. would finaly pushed me over the edge was having to move my installation from one system to another. with other linux distros that ended up taking some time. remember what i installed, reinstalling all the things, setting everything up. with nixos everything is described in nix-expressions. so recreating a system is just a
<betaboon>
matter of running that same expression on a new system. and everything is as it was before. you can forget nothing as everything that makes your system is described in those expressions. nothing is installed by hand
<gchristensen>
betaboon++
<{^_^}>
betaboon's karma got increased to 1
<betawaffle>
what about /home and /etc though?
<symphorien>
/etc is managed by NixOS for the most part
<betawaffle>
it'd be nice if we could segregate all state from generated things
<symphorien>
for /home there is home-manager if you want
<symphorien>
betawaffle: find /etc -type f is a good approximation
gal_bolle has quit [Ping timeout: 245 seconds]
<etu>
betawaffle: Some people run their / on tmpfs, that way you have a "fresh" system on each reboot except from things that use a persistent disk as backing :p
<betaboon>
betawaffle: i guess that takes some discipline. i now have /home on a seperate partition that i can just rsync to a new system. i dont ever touch anything by hand that is outside of home. so I'm just regarding everything outside of /home or /var/lib as dispensible
<betaboon>
over7head: correct. when i move to different hardware i just install the bare minimum (with network support), clone my git-repo (that contains my whole system-configration), apply that and end up with the system that i expect
<betawaffle>
how does that... bootstrap itself?
<gchristensen>
Nix can boot with nothing but a /nix/store/
<over7head>
but i will need linux-firmware first to get my neetwork up
<gchristensen>
the kernel's `init` parameter points to /nix/store/hash/bin/init - and this init script "activates" nixos including updating (or creating) etc files
<gchristensen>
I do the same, except by erasing / on boot, instead of using a tmpfs
<betawaffle>
oh, so is the value of nix.nixPath baked into something then?
<etu>
betawaffle: It sets the global environment variable $NIX_PATH
<gchristensen>
oh, bootstrapping. you can pass a specific config: sudo nixos-rebuild switch -I nixos-config=$(pwd)/configuration.nix
<betawaffle>
nice. well this has been enlightening
<betaboon>
over7head: maybe you just read up on the install guide to get a feeling how that works. and grab the nixos-virtualbox-image to have a running system to play around with. just to get a feeling ;)
<betawaffle>
is there a wiki page for this stuff? would be a good addition
<gchristensen>
that is a cool idea
kreisys has joined #nixos
<etu>
betawaffle: Yeah, first setup you need to specify the config location. After that you'd have it in the environment already :)
<gchristensen>
I have a friend whose ~ is read-only but that seems too annoying
<betawaffle>
i've just hit my first kernel re-compilation
<betaboon>
is anyone using the refind-bootmanager ?
gal_bolle has joined #nixos
<etu>
gchristensen: yeah... I don't do that. I just put all the home directories within /nix/persistent :p
<betawaffle>
does nixos have any way to cache kernel build products to speed up future recompilations?
rauno has quit [Remote host closed the connection]
<srhb>
betawaffle: No, not without introducing an impurity.
<gchristensen>
betawaffle: what have you done to cause a kernel build?
<manveru>
domenkozar: trying to build hnix-lsp, but seems like hpack can't find libyaml
<monokrome>
srhb: was getting an example
<monokrome>
I thought `overrideAttrs` was on everything
<monokrome>
or mkDerivation
<monokrome>
but neither are there
<srhb>
monokrome: makeOverridable is responsible for creating that.
<srhb>
Or is that override...
dlfjklgjelkrtj has quit [Killed (Sigyn (Spam is off topic on freenode.))]
<monokrome>
oh maybe I can just union them w/ //?
<srhb>
monokrome: Sure.
<monokrome>
thnx <3
reinhardt has quit [Quit: Leaving]
<srhb>
The reason it's usually done via some fixpoint-like mechanism is for compositional reasons, mostly.
<monokrome>
weeee it worked
<monokrome>
:D
bathos has quit [Ping timeout: 244 seconds]
<srhb>
monokrome: It looks like you're getting close to something that resembles the module system though. You may want to reuse that instead.
<betawaffle>
does anyone know why slack might not be working on sway? it was (partially) working for me before, but links were never opening in a browser
<betawaffle>
and now even the sign-in button isn't working
<srhb>
monokrome: (That is, it looks like you basically have some expressions that looks like { pkgs, config, options }: { ... })
<gchristensen>
betawaffle: it works for me, but I logged in from X11 and then moved to sway.
<gchristensen>
betawaffle: but also, slack looks bad under sway if you scale your display, people have recommended I just switch to a browser
<gchristensen>
oh, betawaffle I just realized, the reason it doesn't work for you is probably the same breakage for me: I can't click links in Slack and have them open in a browser. I have to copy the link and paste in a browser
<{^_^}>
[nixpkgs] @marius851000 opened pull request #58682 → pioneer: update to 20190203 → https://git.io/fjkbj
jabranham has joined #nixos
sumTimes has joined #nixos
<{^_^}>
[nixpkgs] @NeQuissimus pushed to master « linux: Fix kernel-testing test »: https://git.io/fjkNt
<monokrome>
Is there a variable for getting the name of a node? O_o
<gchristensen>
what do you mean?
<gchristensen>
networking.hostName may be what you want?
<monokrome>
x = { networking.hostName = "${referToXhere}.example.com"; }
<monokrome>
in nixops
<gchristensen>
hrm I don't know
<monokrome>
hmm
domogled has quit [Ping timeout: 246 seconds]
zupo has joined #nixos
sigmundv has joined #nixos
<betaboon>
monokrome: cant you just set `networking.domain = "example.com";` ?
<monokrome>
betaboon: Hmm, maybe! That'd be cool :D thanks
<betaboon>
monokrome: just out of curiousity why do you need the hostname to contain the domain ?
<monokrome>
that's how hostnames work
<monokrome>
hostname is intended to be FQDN
<monokrome>
` hostname -s ` gets the short name
<betaboon>
i never ever set a hostname to a FQDN oO
* monokrome
shrugs
<monokrome>
It's supposed to work that way
<srhb>
Be advised that the shortname/fqdn stuff is broken in NixOS in Interesting^ways
<monokrome>
hmm
<monokrome>
#WorksForMe lol
<betaboon>
monokrome: i always just set a hostname and a domain which combined makes up the fqdn tho
<monokrome>
betaboon: If you run the `hostname` command, does a FQDN come back?
<srhb>
monokrome: Oh, maybe it was finally fixed.
<srhb>
I don't have a fqdn right now, but iirc hostname --fqdn never worked.
<monokrome>
that's probably why it didn't work
<monokrome>
"aaa: SSH Key is already in use on your account"
<monokrome>
wtf DigitalOcean lol
<betaboon>
monokrome: nope, neither `hostname` nor `hostname --fqdn` returns a FQDN even with `networking.domain` set.
<betaboon>
still. i never had the need to do that tho
<monokrome>
is networking.domain just for resolv.conf?
<srhb>
betaboon: Ah, thanks for checking. Yeah, I'm not sure anyone ever figured out why it's broken.
<betaboon>
srhb: I'm not quite sure what is considered 'broken' here. what would you expect ?
<monokrome>
betaboon: some scripts will use it expecting a FQDN when possible, and it will help prevent excessive outbound traffic is all
<srhb>
betaboon: hostname --fqdn should return a fqdn. :)
<monokrome>
`hostname` and `hostname --fqdn` should be equivalent
<monokrome>
`hostname -s` gets the short name
<{^_^}>
[nixpkgs] @dotlambda opened pull request #58684 → acpilight: use default python3 → https://git.io/fjkNi
<betaboon>
srhb monokrome: what i find curisous is that `hostname -y` and `domainname` returns the proper domain
<srhb>
monokrome: I don't think that's correct?
selfsymmetric-pa has quit [Remote host closed the connection]
<monokrome>
betaboon: I think it's resolving it's own host w DNS if you do that?
<monokrome>
domainname maybe uses resolv.conf, not hostname? Not sure on that one
<srhb>
I mean, it's true if you literally set the hostname to the fqdn, but only then
<monokrome>
that's what it's intended to be though
<betaboon>
monokrome: doesnt seem like it
<srhb>
Whereas, if you set the shortname and the domain, `hostname (--short)` will give you a shortname (gethostname) and hostname --fqdn will give you a fqdn (gethostbyname)
<betaboon>
monokrome: I've been doing networking stuff for quite a while and have never seen the hostname set to a fqdn
<betaboon>
monokrome: and in my mind that doesnt make sense. as FQDN = $host.$domain. therefore assuming $host=fqdn would create an infinitly recursive definition
<monokrome>
srhb: No, because --fqdn is the default behavioor
<monokrome>
-s is a non-default behavior / a flag
<srhb>
monokrome: Where do you see this?
<srhb>
monokrome: I can find no reference of that being the default.
IRCsome has joined #nixos
<samueldr>
as they discourage use of dots, makes me think FDO also recommends not using a FQDN, but I'm unsure /etc/hostname actually sets the hostname for FQDN
<monokrome>
srhb: fix your hostname and run `hostname`, even on Debian
<monokrome>
you'll get the FQDN
Glider_IRC has quit [Ping timeout: 246 seconds]
<monokrome>
if your hostname is not set to a FQDN, you won't
<monokrome>
(even with --fqdn)
IRCsum has quit [Read error: Connection reset by peer]
<srhb>
monokrome: I may be just missing your argument entirely, so bear with me: What I hear you saying is that if I set a FQDN as /etc/hostname, then hostname will print the FQDN, and thus --fqdn is the default behaviour. This seems like circular reasoning, because if the behaviour of hostname --short is to print /etc/hostname (which shouldn't have dots), then that looks like --fqdn. So your reasoning seems circular
<srhb>
to me?
<srhb>
Again, I may just be missing understanding here :)
wfranzini has joined #nixos
Anton-Latukha has quit [Ping timeout: 245 seconds]
<betawaffle>
what happens if i run nixos-rebuild twice by accident? like i forgot it was still running in another terminal
<monokrome>
srhb: `hostname` should be equivalent to `hostname --fqdn` which should be equivalent to `uanme -n`
<srhb>
betawaffle: One will inevitably wait on the other.
<betawaffle>
ok
<monokrome>
/etc/hostname should be, for instane, mymachine.example.cmo
<slabity>
betawaffle: You'll create two new profiles. They shouldn't collide
Anton-Latukha has joined #nixos
<srhb>
monokrome: If you look above, this is expressly against debian guidelines.
<monokrome>
If that is the case, `hostname --short` will come back with "mymachine" but `hostname` comes with "mymachine.example.com"
<srhb>
monokrome: Oh, I see.
<srhb>
monokrome: Time to dig out the source of this beast... It seems no matter what you do, one or the other will be incorrect :P
<monokrome>
srhb: Debian is probably trying to support old packages with bad expectations, but one can assert that the tools expect the other method because "hostname" and "hostname --fqdn" to be the FQDN because it's broken if you set /etc/hostname to the short name
<betaboon>
monokrome: yeah i understand your argument :) but i disagree on your assumption that [[ $(hostname) == $(hostname --fqdn) ]] should be true
fendor has quit [Read error: Connection reset by peer]
<betaboon>
monokrome: i disabgree just on a feeling and am just trying to figure out what actually should be true
<monokrome>
betaboon: Then --fqdn will be broken
fendor has joined #nixos
<monokrome>
that's the part that is objective about it
<betaboon>
monokrome: yes the --fqdn is defenitly broken. i completly agree on that part
<monokrome>
If they aren't true, --fqdn doesn't match uname -n
<monokrome>
My bet is that Debian recommends the short name because they deal with old packages that assume it's the short name and Debian takes ages to fix these things
<samueldr>
--fqdn is documented to be what gethostbyname2() will return when given the value of gethostname(); seems that whether it is a FQDN or not already, both situations should be fine?
<betaboon>
monokrome: does uname -n return the FQDN ?
<monokrome>
samueldr: If you're okay with --fqdn not returning a FQDN, sure
<monokrome>
betaboon: yep
<monokrome>
Only if it can figure out out, I think.
<samueldr>
monokrome: the FQDN should be configured in a "it depends" way depending on your setup, according to the assumptions that hostname has, so yeah, if --fqdn doesn't return a FQDN. and you need a FQDN things will need to be changed according to your setup
<betaboon>
monokrome: that's weird, as man(2) of gethostname says that GNU-C implements `gethostname` as just resolving to what `uname` tells it
<monokrome>
The only difference really is that if your hostname isn't an FQDN, you'll be sending extra unnnecessary traffic to your router when using the FQDN in your software
<monokrome>
not a huge deal
<samueldr>
monokrome: from the manpage, it depends, though yeah it possible it will. They seem to recommend using /etc/hosts to handle it without sending traffic outside
<samueldr>
>> Note: It is not specified if the hostname displayed will be a fully qualified domain name. Applications requiring a particular format of hostname should check the output and take appropriate action.
<betaboon>
monokrome: and i just tried running `sudo domainname foobar` and still `hostname --fqdn` did not return a FQDN WTF
<betaboon>
samueldr: even better xD
<betaboon>
damn this is a mess
<samueldr>
I think the LSB has the answer, paraphrased "¯\_(ツ)_/¯"
<srhb>
I am rebuilding hostname.c with fprint everywhere now... This is annoying :P
<monokrome>
betaboon: Yeah, it'll only get a FQDN if /etc/hostname has a FQDN in it even though $(uname -n) will work
<samueldr>
srhb: "not needed" since LSB said "¯\_(ツ)_/¯" :)
<monokrome>
even though the docs imply otherwise
<srhb>
haha
<srhb>
The default is neither short nor fqdn, at least...
hyper_ch2 has quit [Quit: Page closed]
<monokrome>
srhb: To be clear, I said the default in terms of user expectation - not what the code does. I haven't read the code.
<srhb>
Oh, hang on, I did that wrong..
<srhb>
monokrome: Sure, I'm not out to disprove you or anything, I'm just completely baffled by this seemingly core feature being indecipherably weird. :P
<monokrome>
EG, if $(hostname) isn't an FQDN then $(hostname --fqdn) isn't either - which feels wrong if some software wants to use --fqdn
<monokrome>
srhb: ya, for sure, I just feel like I'm not being clear maybe
<monokrome>
:)
<srhb>
monokrome: There's a better chance I'm just not very knowledgable on networking and failing to comprehend ;-)
<monokrome>
At the end of the day, it only matters if you want traffic to not hit your router unneecssarily and you don't want to change /etc/hosts to have a FQDN in it
<jabranham>
Does anyone run nixos without "networking.networkmanager.enable = true;" for a wired connection and experience slow startup? dhcpcd.service takes nearly 10 seconds to get going according to "systemd-analyze blame"
ixxie has joined #nixos
<monokrome>
jabranham: I've seen that behavior but never knew why
<monokrome>
I think it was maybe an overworked router because it fixed itself maybe around the same time I upgraded
<betaboon>
monokrome: i just found this: `domainname` takes the domain name as set in the uname structs `domainname`. `hostname --fqdn` does not. and `uname -n` just displays the `nodename` in that struct
<jabranham>
monokrome: hm... definitely not an overworked router here. I'm the only one on it :-)
<monokrome>
like, I put a switch w/ storm control in front of it and it fixed the issue
<monokrome>
oh hmm
<jabranham>
monokrome: did you switch to using NetworkManager too?
<monokrome>
jabranham: naw, I can't stand networkmanager
<monokrome>
betaboon: Yeah, it's a big mess and the only way to keep things consistent (that I've found) is to set /etc/hostname to FQDN which fixes every case
<betaboon>
monokrome: sounds reasonable :D
<jabranham>
hmmm... interesting. I thought since I see this on my desktop (without networkmanger) but not laptop (with networkmanager) that was perhaps to blame. Maybe something else is going on though.
<monokrome>
jabranham: it could be, my router was def overheating - so maybe your issue is different but w/ similar symptoms
<betaboon>
monokrome: on a machine with `networking.domain` and `networking.hostName` set "$(hostname).$(domainname)" actually resolves to the FQDN just using the uname-struct XD
<monokrome>
jabranham: When my router was overheating, it wasn't due to too many devices as much as that it was sending messages iirc
<monokrome>
so the switch maybe didn't even fix that particular issue
dejanr has joined #nixos
ixxie has quit [Ping timeout: 245 seconds]
<monokrome>
betaboon: not sure I follow, you updated hotname to be the FQDN?
<betaboon>
monokrome: no. assuming `sudo hostname FOO` sets the `nodename` in the uname-struct and `sudo domainname "bar.com"` sets the `domainname` in the uname-struct -> `hostname` returns the nodename, `domainname` returns the domainname -> "$(hostname).$(domainname)" returns the FQDN
<betaboon>
monokrome: still `hostname --fqdn` still does not return it
i1nfusion has quit [Remote host closed the connection]
i1nfusion1 has quit [Remote host closed the connection]
<betaboon>
anyhow. I'll stop disturbing xD
<monokrome>
betaboon: Did you reboot? I don't think nixos-rebuild will call `sethostname`
civodul has quit [Quit: ERC (IRC client for Emacs 26.1)]
WhittlesJr has joined #nixos
knupfer has joined #nixos
<nh2>
where's a good place to insert a GCC flag so that it's used in all package builds? I want to add `-ffunction-sections` and see how much smaller it makes executables
<johanot>
srhb: yeah. that one exactly. Sry for being imprecise :P
<srhb>
Just checking. :P
InFlames has joined #nixos
Ariakenom has joined #nixos
<johanot>
currently trying to find a programmatic way of repro'ing. Don't want "git bisect run selenium" or something :P
Jetien has quit [Ping timeout: 245 seconds]
<srhb>
Oh dear. What's the issue?
wfranzini has quit [Remote host closed the connection]
<johanot>
UI gets out of sync with git (repos). Usually, when that happens, there's some sort of error msg in gitaly or sidekiq logs, but alas. It happend after upgrading to 11.9.1
<srhb>
Ew. Guess I'll refrain from updating.
erictapen has quit [Ping timeout: 246 seconds]
<johanot>
srhb: good call! if only the gitlab test didn't need >15mins to complete :(
<samueldr>
srhb: weirdly enough, master uefIUsb workd fine here?
<flokli>
that's mostly due to the 9p access being very slow, and ruby reading waaaay to many files
<srhb>
samueldr: Maybe a commit snuck past me somehow?
erasmas has joined #nixos
<samueldr>
I'm building against the revision that failed within hydra
jbgi has quit [Ping timeout: 245 seconds]
<samueldr>
I assumed it was the "usual qemu weirdness"
<srhb>
samueldr: I doubt it is. I am able to reproduce it perfectly at least. but with filesystem ordering being involved, ??????
<srhb>
samueldr: Which revision exactly?
<samueldr>
yeah, since you repro'd I assume it itsn't
<samueldr>
from the "Inputs" tab in hydra, 005fb92d32c9d5b636846ce30ae01aab81454161
<srhb>
Checking that then.. Just in case.
<srhb>
How did you reproduce?
<samueldr>
used the command you listed
<johanot>
flokli: I think the issue I'm facing is due to the repo post-receive hook not being executed for some reason.
<samueldr>
srhb: nix-build '<nixpkgs/nixos/tests/boot.nix>' -A uefiUsb
<srhb>
samueldr: Stupid question, but does your <nixpkgs> point to your checkout? :)
<betaboon>
gchristensen: I'm just realy stuck in getting this dotnet application to build, as that stuff wants to fetch dependencies from the network iI
<srhb>
betaboon: Are you sure you can't get by with fixed-output drvs though? Allowing network access without fixed-output is almost certainly impure.
<srhb>
Ouch. :)
<betaboon>
srhb: after dumping two days into this i realy dont care about impure for now, just need to get this working as it is blocking me from getting actual work done xD
<samueldr>
I have a hunch, waiting on build of master (just in case)
<srhb>
samueldr: My immediate suspect was -type f -- yours too? :-P
xorkle has quit [Read error: Connection reset by peer]
<betaboon>
gchristensen: thanks for the pointer. but i dont feel like setting sandbox to relaxed xD
<samueldr>
I was thinking more along the ways of "not everything is at the right place", which might be what you saw
<srhb>
samueldr: Ah..
<gchristensen>
betaboon: I've never brought myself to turn it on either :D
argo has joined #nixos
drakonis1 has quit [Quit: WeeChat 2.3]
<srhb>
betaboon: Any chance that's it's just completely reproducible and you can just do it the fixed-output way at some checkpoint? I realize this is probably not super likely with .net, but... You never know :P
<srhb>
(I had this pain with scala's sbt. Yikes...)
<betaboon>
srhb: i realy dont have a clue about dotnet xD
Ariakenom has quit [Quit: Leaving]
<gchristensen>
yeah, maybe you can do a "fetch deps" only phase and make that fixed output
<srhb>
gchristensen: Where's your spirit of adventure, just do the _whole_ build and mark it fixed :-P
<srhb>
(not serious, just in case...)
<gchristensen>
now you're cooking with gas
<srhb>
:D
<betaboon>
i know that i can run `dotnet restore` which populates `$HOME/.nuget/packages` by downloading stuff. then when building the package in question in can run `dotnet restore --source $SOMEPATH` to prevent it from downloading again.
<gchristensen>
btw did y'all see the nixos 4/1/2019 joke commit
<johanot>
flokli: that worked! a lot.. :) did you backport this ? hehe
<{^_^}>
[nixpkgs] @Ekleog opened pull request #58694 → [WIP] iso-image: make reproducible by not relying on mcopy's readdir → https://git.io/fjkpm
ddellacosta has quit [Ping timeout: 240 seconds]
<johanot>
we should really try and incorporate gitaly testing in the nixos test (on another note)
<srhb>
johanot: The interesting thing about the speed of the NixOS test is that we now know that you deployed directly to prod. :-P
argo has quit [Quit: argo]
<johanot>
srhb: well .. it had a tiny stopover at a staging box, but you're basically right :P
<srhb>
johanot: But agreed! I'm pretty sure it has enough json support that we don't need selenium hackery.
<srhb>
Maybe even the diff endpoints...
ddellacosta has joined #nixos
sigmundv_ has joined #nixos
dejanr has quit [Ping timeout: 250 seconds]
<das-g[m]>
Apropos selenium: I'd like to write a NixPkgs test that checks whether selenium version, Firefox version and geckodriver version are compatible to each other. Would someone be willing to hand-hold me through that?
<das_j>
Hey, is there a way to make Hydra use binfmt to build for other architectures?
<infinisil>
gchristensen: Haha nice
<infinisil>
aszlig++
<{^_^}>
aszlig's karma got increased to -666
Ariakenom has quit [Remote host closed the connection]
Ariakenom has joined #nixos
<srhb>
das-g[m]: Can you give some short pointers on what that test would logically require?
dejanr has joined #nixos
<srhb>
das-g[m]: (Like, is a simple command enough to check this?)
sigmundv has quit [Ping timeout: 246 seconds]
<das-g[m]>
Kinda, but that command needs to be able to remote-control Firefox.
<das-g[m]>
(Thus requirering a GUI, but no manual interaction with it.)
<srhb>
das-g[m]: And how do you verify it's working? (we have some OCR in the chromium test, iirc, but I'm guessing that's not needed with selenium?)
<das-g[m]>
It'd just be a smoke test, so if the return code is `0`, I'd consider that a pass.
<srhb>
But I think it would still be a good starting point, and it's not too long a read :)
<srhb>
Ah, even easier.
jasongro` has joined #nixos
<srhb>
das-g[m]: I would definitely just copy the chromium test to start out with and return here for any questions :)
<das-g[m]>
Wait a moment. I'll dig up the (not-yet-working) test I tried to write a while back.
jasongrossman has quit [Ping timeout: 246 seconds]
ng0 has joined #nixos
johanot has quit [Quit: WeeChat 2.4]
pietranera has quit [Quit: Leaving.]
<WilliamHamilton[>
infinisil: thanks, in the end I managed to do that without changing the compiler version; I have another question: I have a package that has a dependency hidden behind a flag (that's activated by me). So when starting a nix shell, it complains that it cannot find it (although it is in nixpkgs). How could I modify the overlay to add that dependency explicitly?
<srhb>
das-g[m]: Yeah but your start is fine. :) I guess you just want to do something more robust with the python script and figure out why it fails?
sigmundv__ has quit [Ping timeout: 268 seconds]
<das-g[m]>
Actually, if that would run successfully, it'd be the whole test. If the version of any of the 3 packages is incompatible to the other two, I'd expect that to fail.
domogled has joined #nixos
<das-g[m]>
But currently, it fails for some other reason. (I forgot which. How again do I run a test?)
<srhb>
das-g[m]: You should be able to just run nix-build on it, if it lives in the nixos/tests tree
sumTimes_ has quit [Ping timeout: 256 seconds]
<das-g[m]>
So `nix-build nixos/test/geckodriver.nix` from the nixpkgs repo root?
<srhb>
das-g[m]: Your cwd doesn't matter, but yes.
<das-g[m]>
that results in a stack overflow
<srhb>
um. :)
<das-g[m]>
(which wasn't the problem I encountered back then)
<manveru>
you might wanna wrap that in another function for the default arguments, but i'll leave that up to you
<flokli>
srhb: huh, johanot is gone.
<flokli>
I did backport it, together with other important backports
<srhb>
flokli: He doesn't use a bouncer. Feel free to point it out repeatedly, each time I link him to the logs of someone saying "where did johanot go"
<srhb>
:-)
<srhb>
flokli: (Also thanks! :D)
<manveru>
WhittlesJr: oh, i just realized an ip might have more hostnames... that won't work :(
<srhb>
das-g[m]: I'm no python expert, but as far as I could surmise, the error came from using -c and some weirdness therein.
<srhb>
das-g[m]: So I just went with what I know: Write a file and run that.
<srhb>
So, I guess I prefer it because it works :-P
<WhittlesJr>
manveru: Yep, that works perfectly! I started to get close with my own implementation but yours is already looking more elegant than mine, the way things were going. Thank you so much for taking the time to do that! This will help me for future projects sure, no doubt
Ariakenom has joined #nixos
drakonis has quit [Ping timeout: 268 seconds]
<WhittlesJr>
er, grammar
orivej has quit [Ping timeout: 250 seconds]
<das-g[m]>
Ah right, that was the error. (I didn't see it, yet, as nix-build is still compiling stuff.)
<srhb>
das-g[m]: So, really, you had the whole thing minus that minor launch issue. :)
<WhittlesJr>
manveru++
<{^_^}>
manveru's karma got increased to 7
<das-g[m]>
:-)
__monty__ has joined #nixos
<WilliamHamilton[>
infinisil: so, I finished the overlay I intended to write, and it works with 8.6.4, but I'd still take some advices on how to write it more clearly: https://gist.github.com/59686a42e3206909be71605b9fd52613 what troubles me is that I'd like to write all that function compositions like in haskell with `flip` and `$`. Is there any way to do that?
<WilliamHamilton[>
expecially the reflex-dom clause: I'd like to write `dontCheck . doJailbreak . flip appendConfigureFlag [flags] $ ...`
<srhb>
WilliamHamilton[: Since we don't really have nice composition, I recommend using overrideCabal instead for most things like this.
<infinisil>
WilliamHamilton[: Unfortunately not possible like that in nix, but you can write a function: `compose [ dontCheck doJailbreak (flip appendConfigureFlag [flags]) ] ...`
<WilliamHamilton[>
by the way, why isn't there an operator like `compose` or `flip` already defined in builtins?
<WilliamHamilton[>
or is it elsewhere?
<tilpner>
> lib.flip
<{^_^}>
<LAMBDA>
<WhittlesJr>
manveru: badass! Thank you!
<manveru>
WhittlesJr: np... just spending time while waiting for haskell to build stuff :P
<WilliamHamilton[>
> lib.flip - 2 3
<{^_^}>
attempt to call something which is not a function but an integer, at (string):230:12
<WilliamHamilton[>
> lib.flip `-` 2 3
<{^_^}>
error: syntax error, unexpected $undefined, expecting ')', at (string):230:10
<infinisil>
> lib.flip __sub 2 3
<manveru>
WhittlesJr: - is not really a function you can pass...
<{^_^}>
1
<WilliamHamilton[>
infinisil: very good, thanks
<infinisil>
> let __sub = a: b: "You want these numbers subtracted, eh?"; in 2 - 3
<{^_^}>
"You want these numbers subtracted, eh?"
<WilliamHamilton[>
:D
<infinisil>
(Don't do this pls xD)
<slabity>
> let __sub = a: b: "Eh... Maybe later..."; in 2 - 3
<{^_^}>
"Eh... Maybe later..."
<WilliamHamilton[>
infinisil: I don't need to, lack of static types confuse me enough at it is xD
domogled has quit [Ping timeout: 245 seconds]
<infinisil>
> let s = __sub; in let __sub = a: b: s b a; in 10 - 3
<{^_^}>
-7
<slabity>
> let __sub = __add; in let __add = __sub; in 10 - 3
<{^_^}>
13
Fare has joined #nixos
<infinisil>
> let s = __sub; in let __sub = a: b: if randomUpTo 5 == 0 then s b a else s a b; in 10 - 3
<{^_^}>
undefined variable 'randomUpTo' at (string):230:40
<infinisil>
> let s = __sub; in let __sub = a: b: if randomUpto 5 == 0 then s b a else s a b; in 10 - 3
<{^_^}>
-7
Myhlamaeus has joined #nixos
<infinisil>
Inverts subtraction only occasionally :P
<WilliamHamilton[>
hahahaha bug proof!
bemeurer has joined #nixos
taktoa has quit [Quit: taktoa]
<infinisil>
> let s = __sub; in let __sub = a: b: s (lib.toInt a) (lib.toInt b); in "10" - "3"
<{^_^}>
7
<infinisil>
Dynamic typing!
callmecabman has joined #nixos
<tilpner>
Except it doesn't work for integers anymore c.c
<callmecabman>
has anyone used ganeti on nix?
<infinisil>
> let s = __sub; in let toInt = val: ({ string = lib.toInt; }.${builtins.typeOf val} or lib.id) val; __sub = a: b: s (toInt a) (toInt b); in 10 - "3"
<{^_^}>
7
<infinisil>
tilpner: There you go
<pie__>
does anyone know how to fix broken PS1 on nested tmux
<tilpner>
... yay?
cyberwolf[m][m] has joined #nixos
<srhb>
pie__: I would imagine: Not nesting, not resetting PS1 on nested session, making the PS1 assignment nest-safe?
<pie__>
srhb, no idea how to do the last two, i didnt change any of this stuff, seems like nixos brokenness?
<srhb>
pie__: It's possible the default PS1 isn't safe for nesting, yeah..
* infinisil
plays a bit more in #nix-lang
<pie__>
i get stuff with backslashes and tab completion is broken
<srhb>
pie__: Just tried it in what I believe is the default setup, no issues.
<pie__>
ok maybe its unrelated..
* pie__
goes to try some stuff
<pie__>
ok its broken even if I use it not nested...
<das-g[m]>
I think I've seen backslash weirdness in the PS1 prompt when nesting `nix-shell`s.
<pie__>
yeah i had that ages ago
<pie__>
but im not nesting nix shells
<clever>
das-g[m]: that happens when you run a non-interactive bash
<clever>
pkgs.bash vs pkgs.bashInteractive
<pie__>
clever, any chance my tmux is falling back to the wrong bash somehow?
adetokunbo has quit [Quit: This computer has gone to sleep]
<das-g[m]>
That'd also explain the missing tab completion, wouldn't it?
<clever>
pie__: the wrong one could be in $PATH
<pie__>
clever, i can paste the script in a second, but the rest of my machine works fine
<clever>
nix-shell nesting fails, because nix-shell adds the non-interactive bash to PATH
<clever>
and then the 2nd nix-shell obeys, that
<pie__>
clever, i have tmux running in a nix-shell
<pie__>
is that it?
<clever>
pie__: that sounds likely
<clever>
pie__: any time you run `bash` inside nix-shell, it will fail
<pie__>
ah. doh. needs to be added to "common pitfalls ro something :P wish we could do something about this
<pie__>
clever, so how to I unf*** it?
<clever>
pie__: add bashInteractive to the buildInputs?
<pie__>
hmm sounds reasonable, i was about to try adding bashinteractive to path
<clever>
pie__: is that template build at compile-time or runtime?
<pie__>
clever, i have no clue
<pie__>
clever, there seems to be the same code in like 5 different places
<clever>
pie__: is the "compiled" copy of that script in the nix store?
<pie__>
i have no idea how ruby works
<pie__>
oh, hm, if this is code generation that would explain why theres the same thing all over the place
<pie__>
ill see if i can check
<pie__>
but it might be easier to just try setting shell
adetokunbo has joined #nixos
sumTimes has joined #nixos
<aanderse-work>
if i run nix-collect-garbage -d on a box managed by nixops the whole system isn't going to come to a crash, is it?
bemeurer has quit [Ping timeout: 246 seconds]
<pie__>
great now i have another intermediate problem...will report back when solved...
domogled has joined #nixos
<ryantm>
aanderse-work: No.
<ryantm>
aanderse-work: Or right, it won't.
<aanderse-work>
:)
<aanderse-work>
and if my /boot is getting pretty full on that machine managed by nixops, how do i clean that without making the system come to a crash?
<aanderse-work>
lets say i've already ran nixops delete-generation *all the old ones i don't need anymore*
<aanderse-work>
** sorry, ran that on the nixops host machine, of course
<aanderse-work>
i didn't see these details under the nixops manual, so i just want to make sure i'm not going to cause myself any pain
Neo-- has quit [Ping timeout: 240 seconds]
johanot has joined #nixos
zupo has joined #nixos
<pie__>
clever, im getting this for some reason now "no server running on /home/nixos/stuff/nixapps/nix-learn-git/workdir/socket", it wasnt doing this earlier :/
bemeurer has joined #nixos
* pie__
tries a revert
<srhb>
aanderse-work: when nixops does the switch/boot, it should clean up unused configs as well
<pie__>
clever, yeah so setting the SHELL causes tmux to fail to start...
<srhb>
aanderse-work: Do note that you should never be able to mess that up unless you go on a killing spree in /boot, since /run/booted-system will keep a gc lock on the generation the machine actually booted with.
<aanderse-work>
srhb: good to know
<pie__>
clever, is this not correct?
<pie__>
export SHELL=${bashInteractive}
<pie__>
oh derp...that should have a }/bin/bash shouldnt it
<srhb>
aanderse-work: (This also means that you usually will have at _least_ two generations until you boot into your new profile and then clean up again.)
<aanderse-work>
srhb: i cleaned up on the nixops side (running nix-collect-garbage) and /nix/var/nix/profiles/ looks good there, but after deploying to the target machine the /nix/var/nix/profiles/ directory on the target machine is pretty full
sigmundv_ has quit [Ping timeout: 246 seconds]
<srhb>
aanderse-work: nix-collect-garbage with no further arguments?
<aanderse-work>
--delete-older-than 14d
<srhb>
aanderse-work: If you have system generations in /nix/var/nix/profiles after that, then I guess they're newer than 14d.
<srhb>
(You can confirm by checking the timestamps)
<aanderse-work>
i must be doing something wrong in that case...
<aanderse-work>
Dec 6 17:51 system-96-link
<aanderse-work>
i see system-n-link where n spans from 1 to 207
<monokrome>
it's for users that don't have a $HOME isn't it
<gchristensen>
/homeless-shelter is used during Nix builds because there is no HOME
vendicta has quit [Quit: WeeChat 2.2]
<gchristensen>
there is no renaming it, there is no creating it -- it is exposing a problem in the build
<monokrome>
oic
<monokrome>
can I rename it anywaay
<gchristensen>
to what end?
<jomik>
Yo guys. How do I add fzf as a plugin for vim? I use the `myVimPackage`, built-in vim plugin manager. The issue is that for fzf, I need the plugin part from fzf, to use fzf-vim.
<betaboon>
gchristensen: i now found a way to fetch dotnet dependencies before build :) still not succeeding in building the package in question. i didnt know that dotnet pulls everything from the web .. :(
<wolfman154>
13:20 wolfman154: I added home.nix as a module in configuration.nix and it is throwing an error? https://hastebin.com/ogubigeras.rb
<ryantm>
wolfman154: It looks like you are trying to mix home-manager configuration into nixos configuration, when they are separate.
<wolfman154>
ryantm: if I read correctly I thought you can add home.nix as a sub module?
<ryantm>
wolfman154: Where did you read that?
bemeurer has quit [Quit: WeeChat 2.4]
freeman42y has quit [Read error: Connection reset by peer]
freeman42y has joined #nixos
<wolfman154>
ryantm: towards the top of home-manager news Command, it also explains it somewhere in the middle of the news?
cyphase has quit [Ping timeout: 255 seconds]
<das_j>
wolfman154: add it as channel and do something like: imports = [ <home-manager/nixos> ]; home-manager.users.wolfman154.programs.git.....
<ryantm>
wolfman154: One is the home-manager config files, and one is your configuration of home manager.
eyJhb has quit [Quit: Bye]
<wolfman154>
ryantm: 13:27 wolfman154: 13:20 wolfman154: I added home.nix as a module in configuration.nix and it is throwing an error? https://hastebin.com/ogubigeras.rb
goibhniu2 has joined #nixos
goibhniu1 has quit [Ping timeout: 255 seconds]
<wolfman154>
When I use ~ instead of /home/tjg it thinks I mean root ?
<{^_^}>
[nixpkgs] @dotlambda pushed to master « renderdoc: use default python3 (#58687) »: https://git.io/fjIva
eyJhb has joined #nixos
<wolfman154>
rycee: I already did that so my guess is that it should work without the home.nix import in my configuration.nix, I guess I will remove it: 13:27 wolfman154: 13:20 wolfman154: I added home.nix as a module in configuration.nix and it is throwing an error? https://hastebin.com/ogubigeras.rb
<wolfman154>
infinisil++
<{^_^}>
infinisil's karma got increased to 71
fendor has quit [Ping timeout: 245 seconds]
<rycee>
wolfman154: If /home/tjg/.config/nixpkgs/home.nix is a user's home manager configuration then you can't use it in the NixOS imports.
goibhniu2 has quit [Ping timeout: 246 seconds]
cfricke has quit [Quit: WeeChat 2.4]
<JonReed>
How to configure the docker daemon under nix? Should I just create `/etc/docker/daemon.json` or there is an option to pass that config file from `configuration.nix`? I can't find that option though. The only option I've found is `virtualization.docker.extraOptions`, which means that I'll probably have to set it to `''--config-file ${myJsonConfigFile}''`, if I understand it correctly.
cfricke has joined #nixos
<wolfman154>
rycee: Will my home.nix still work as a sub module to configuration.nix if I remove it?
mconstant_ has joined #nixos
mconstant_ has quit [Client Quit]
<rycee>
wolfman154: You need to do something like `home-manager.users.youruser = import /home/tjg/.config/nixpkgs/home.nix;`
<JonReed>
das-g[m]: I think it doesn't matter much. But one consideration could be that with fetchFromGithub you could more easily update your Python package for nix before updating it for Pypi, or if someone forgot to update it for Pypi you could still update it for Nix with fewer lines changed.
<DigitalKiwi>
ime if you use the pypi version a lot of things won't have tests
rcshm has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @worldofpeace pushed 2 commits to master: https://git.io/fjIJB
<{^_^}>
[nixpkgs] @worldofpeace pushed 0 commits to revert-57700-auto-update/aisleriot: https://git.io/fjIJR
<infinisil>
buckley310: import is a function that takes a file and returns the nix value of that file. imports is an attribute you can use to import nixos modules in configuration.nix: `{ imports = [ ./some-module.nix ]; ... }
<wolfman154>
I’m using rycees instructions: 13:55 rycee: wolfman154: Yeah, looks right to me. This is from a configuration of mine: `home-manager.users.rah = import ../user/lt-ct-rb.rah.nix;`
<infinisil>
buckley310: Oh, you were replying to wolfman154, got it
<wolfman154>
I tried those two ways and it did not work, home-manager news says you can use home-nix as a sub module and it is needed for build-vm command to work?
<rycee>
wolfman154: Change it to `home-manager.users.tjg = import /home/tjg/.config/nixpkgs/home.nix;`;
<{^_^}>
[nixpkgs] @fpletz pushed 6 commits to release-19.03: https://git.io/fjIJw
slack1256 has quit [Remote host closed the connection]
i1nfusion1 has quit [Remote host closed the connection]
drakonis has joined #nixos
Soo_Slow has joined #nixos
<wolfman154>
Finally got it to build, gracias gracias
<wolfman154>
rycee++
<{^_^}>
rycee's karma got increased to 6
jasongrossman has joined #nixos
<wolfman154>
infinisil++
<{^_^}>
infinisil's karma got increased to 73
ericsagnes has quit [Ping timeout: 250 seconds]
erasmus has quit [Quit: reboot]
__monty__ has quit [Quit: leaving]
orivej has quit [Ping timeout: 255 seconds]
orivej_ has joined #nixos
fusion809 has quit [Ping timeout: 240 seconds]
kvda has joined #nixos
rprije has joined #nixos
sgronblo has joined #nixos
goibhniu1 has quit [Ping timeout: 246 seconds]
<sylvie[m]>
hi, i get `error: undefined variable 'google-api-python-client'` trying to install `with import <nixpkgs> {}; python3.withPackages (ps: with ps; [ google-api-python-client ... ])`
goibhniu1 has joined #nixos
erasmus has joined #nixos
sgronblo has quit [Ping timeout: 250 seconds]
<Arahael>
I'm a bit lost, I'm back to trying to install a specific version of jinja2, and I'm starting afresh. I've got a shell.nix, and I think I'll try to write a new package altogether that will represent python with the jinja2 dependency. So I create a directory called 'python', and put in an empty default.nix. I know that default.nix needs to contain a "derivation", however, where can I find out how this is
<Arahael>
As far as I can tell, this is the only documented mention of the primitive built-in function: To build a package with the standard environment, you use the function stdenv.mkDerivation, instead of the primitive built-in function derivation, e.g.
wfranzini has quit [Remote host closed the connection]
wfranzini has joined #nixos
alex`` has quit [Quit: WeeChat 2.4]
m0rphism has quit [Ping timeout: 268 seconds]
ericsagnes has joined #nixos
<wolfman154>
I finished compiling xmonad.hs yesterday and now I found that a you can do this with nix, so do I just place this text in my home-manager and erase ~/dotfiles/xmonad.hs?
<wolfman154>
gchristensen: just to make sure , can I just copy that text to my home-manager home.nix or do I have to create a separate xmonad.nix module?
Anton-Latukha has quit [Ping timeout: 255 seconds]
<wolfman154>
I want to make sure and get this setup correctly
scribbler has joined #nixos
<wolfman154>
Any tips from you guys would be greatly appreciated
<{^_^}>
[nixpkgs] @cpages opened pull request #58719 → calligra: mark broken → https://git.io/fjITI
scribbler has quit [Client Quit]
<wolfman154>
?
jasongro` has joined #nixos
<{^_^}>
[nixpkgs] @astro opened pull request #58720 → openjdk: disable hotspot os version check → https://git.io/fjITq
jasongro` has quit [Remote host closed the connection]
jasongrossman has quit [Ping timeout: 268 seconds]
hlolli has quit [Remote host closed the connection]
orivej_ has quit [Ping timeout: 255 seconds]
Soo_Slow has quit [Remote host closed the connection]
wolfman154 has joined #nixos
<wolfman154>
When is nixos version 19 stable coming out?
<gchristensen>
19.03 will come out "any day now"
samueldr has left #nixos [#nixos]
samueldr has joined #nixos
glenndavy has joined #nixos
kreisys has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<wolfman154>
gchristensen: what it be best to upgrade from scratch(usb) or manually from 18 to 19. From scratch seems easier, and I could just copy and paste my configuration files?
kleisli has joined #nixos
<gchristensen>
people don't typically (ever?) need to do a from-scratch install with NixOS
<gchristensen>
my 15.09 machine has upgraded to 16.03, 16.09, 17.03, 17.09, 18.03, 18.09, 19.03 just fine
<wolfman154>
gchristensen: I’ve already crashed nixos plenty of times and had to install from scratch.
erictapen has quit [Ping timeout: 245 seconds]
<gchristensen>
hah
<gchristensen>
remarkable, that is certainly exceptional
jluttine has quit [Ping timeout: 246 seconds]
shibboleth has joined #nixos
erictapen has joined #nixos
<wolfman154>
gchristensen: I finally removed my desktop manager and sign in window display(gun, sddm) they did not want to work with x server
<gchristensen>
I feel like I get too much karma from them :P
<gchristensen>
more than I've earned, I mean
<Arahael>
In https://nixos.org/nixpkgs/manual/#sec-using-stdenv it states that I must have an 'src', however, what if I don't have a source? (Ie, it's the current directory that default.nix resides in...)
kreisys has joined #nixos
<gchristensen>
then your src is ./.! :) src = ./.;
<glenndavy>
Hi all, I need to install the ruby ffi gem. To do that I need to have ffi.h findable. AFICT, the MacOS_SDK-10.10.10-dev pkg is the correct one (vs libffi ?)