<ivanivan>
LnL: ah, right. I read through it while back but it didn't sink in. I'll take another look at it
greymalkin has quit [Ping timeout: 256 seconds]
ivanivan has quit [Quit: WeeChat 1.9.1]
greymalkin has joined #nixos
coot has joined #nixos
dbe_ has quit [Ping timeout: 240 seconds]
coot has quit [Ping timeout: 240 seconds]
dan_b has joined #nixos
kier has quit [Ping timeout: 268 seconds]
swayf has joined #nixos
simukis has quit [Quit: simukis]
ma27 has quit [Ping timeout: 256 seconds]
kier has joined #nixos
goibhniu has quit [Ping timeout: 264 seconds]
Drakonis has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
nixpkgs/master 58ff18b Michael Weiss: pythonPackages.raven: 6.4.0 -> 6.5.0
NixOS_GitHub has left #nixos [#nixos]
<NixOS_GitHub>
[nixpkgs] primeos pushed 1 new commit to master: https://git.io/vNgz6
ssmike1 has joined #nixos
NixOS_GitHub has joined #nixos
ssmike has quit [Ping timeout: 240 seconds]
<NixOS_GitHub>
[nixpkgs] shlevy pushed 1 new commit to master: https://git.io/vNgge
<NixOS_GitHub>
nixpkgs/master 69d4ade Shea Levy: idris: Move library setup to the setup hook....
NixOS_GitHub has left #nixos [#nixos]
ssmike1 is now known as ssmike
Nazral has quit [Quit: Changing server]
civodul has quit [Quit: ERC (IRC client for Emacs 25.3.1)]
jtojnar has quit [Ping timeout: 268 seconds]
<ennui>
ixxie: yeah, my mistake, sorry. test inputs are only used during building, so they should by added to nativeBuildInputs (the manual says buildInputs in the python example. that's functionally equivalent, but nativeBuildInputs signals the intent better)
tomberek has joined #nixos
jensens has quit [Ping timeout: 256 seconds]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] primeos pushed 1 new commit to master: https://git.io/vNggC
<NixOS_GitHub>
nixpkgs/master 02f6827 Michael Weiss: gns3Packages.{server,gui}{Stable,Preview}: 2.1.2 -> 2.1.3
nuncanada has quit [Read error: Connection reset by peer]
orivej has quit [Ping timeout: 240 seconds]
tf777 has joined #nixos
Gohla has quit [Ping timeout: 264 seconds]
<tf777>
I'm using nix on MacOS; I'm getting this error message "name collision in input Nix expressions, skipping ‘/Users/thisMac/.nix-defexpr/channels_root/nixpkgs"
<tf777>
I found a candidate solution by Eelco Dolstra while searching but it didn't get rid of the error unfortunately.
Gohla has joined #nixos
Acou_Bass has quit [Quit: byeeeeeeeeeeeeeee]
Acou_Bass has joined #nixos
<tf777>
Ok nevermind, rolling back nix-channel to generation 1 fixed it I think.
hellrazor has joined #nixos
hellrazo1 has quit [Ping timeout: 264 seconds]
orivej has joined #nixos
threepointone has quit [Quit: Page closed]
ennui has left #nixos ["ERC (IRC client for Emacs 24.5.1)"]
mbrgm_ has joined #nixos
mbrgm has quit [Ping timeout: 268 seconds]
mbrgm_ is now known as mbrgm
tf777 has quit [Ping timeout: 260 seconds]
la_putin has quit [Remote host closed the connection]
la_putin has joined #nixos
martinklepsch has quit [Read error: Connection reset by peer]
pkill9 has quit [Ping timeout: 248 seconds]
Supersonic112 has quit [Disconnected by services]
dag_ has joined #nixos
Supersonic112_ has joined #nixos
Supersonic112_ is now known as Supersonic112
martinklepsch has joined #nixos
kranium has quit [Read error: Connection reset by peer]
dag__ has quit [Ping timeout: 264 seconds]
MP2E has joined #nixos
ison111 has quit [Ping timeout: 265 seconds]
pkill9 has joined #nixos
orivej has quit [Ping timeout: 246 seconds]
markus1189 has joined #nixos
s33se has joined #nixos
iyzsong has joined #nixos
markus1199 has quit [Ping timeout: 256 seconds]
s33se_ has quit [Ping timeout: 248 seconds]
iqubic has joined #nixos
<iqubic>
Does nixos have ispell in the repos?
<iqubic>
Or some other kind of spellchecker application, like hunspell or aspell?
iqubic has left #nixos ["ERC (IRC client for Emacs 25.3.1)"]
xcmw has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
MP2E has quit [Remote host closed the connection]
MP2E has joined #nixos
hakujin has joined #nixos
yegortimoshenko has quit [Remote host closed the connection]
yegortimoshenko has joined #nixos
asuryawanshi has joined #nixos
hakujin has quit [Ping timeout: 265 seconds]
<samueldr>
__red__: nix-env's "-A" attribute expects an attribute path from the "top-level" expression; in this case, it is either nixos or nixpkgs, the name used for the channel
<samueldr>
(nixpkgs being the default for other distros, and nixos for nixos)
<samueldr>
so in your case, it would probably have been nix-env -iA nixos.virt-manager
<samueldr>
(just like it's pkgs.virt-manager in the configuration)
xcmw has joined #nixos
<samueldr>
hm, and I may have goofed in saying pkgs.virt-manager and nixos.virt-manager, it's *.virtmanager, sorry
xcmw has quit [Client Quit]
<samueldr>
with nix-env -qa, adding -P (-qaP) will add the attribute path to the results
asuryawanshi has quit [Remote host closed the connection]
hellrazor has quit [Ping timeout: 268 seconds]
ryantm has joined #nixos
* ryantm
Who knows about the llvmPackages setup?
<ryantm>
Sorry. Who knows about the llvmPackages setup?
verite has quit [Ping timeout: 240 seconds]
Drakonis has quit [Read error: Connection reset by peer]
<tnks>
When using a Haskell environment, I'd expect my ~/.cabal to stay pretty bare... just the downloaded index.
<tnks>
because all the dependencies should come from a ghc-with-packages provided by Nix.
<samueldr>
Guest82411: why?
<tnks>
but I'm noticing that my ~/.cabal has some dependencies in it like alex, c2hs, and syb.
<zimbatm>
Guest82411: not mine
<samueldr>
(and yes, zimbatm)
<samueldr>
ah, not his sorry
<tnks>
not many... but still, I was expecting to not see anything built outside Nix beyond my own projects.
<zimbatm>
it was already taken when I started nixbox
<samueldr>
I must have been confused I thought the nixos box was built using nixbox
<adisbladis>
__red__: The difference here is that "virt-manager-x.y.z" is the name attribute of the derivation and "virtmanager" is the attribute name in the package set
<adisbladis>
__red__: If you search for a package using "nix-env -qaP virt-manager" you will get a result split into two columns, the left one is the attribute name :)
justanotheruser has joined #nixos
<Guest82411>
samueldr: I would like to use Vagrant, but I am targeting a newer version than the one there
<Guest82411>
Unrelated, but I must be missing something: error: selector ‘make’ matches no derivations
<Guest82411>
ah, gnumake
babyflakes has joined #nixos
MP2E has quit [Remote host closed the connection]
datakurre has quit [Ping timeout: 276 seconds]
Rusty1_ has quit [Quit: Konversation terminated!]
MP2E has joined #nixos
Dezgeg has quit [Ping timeout: 276 seconds]
endformationage has quit [Quit: WeeChat 1.9.1]
`_ has joined #nixos
<Ralith>
hmm
<Ralith>
dockerTools.buildImage screws up my terminal
<Ralith>
contrapumpkin: were you the one discussing it previously?
ent has quit [Ping timeout: 246 seconds]
seequ has quit [Ping timeout: 255 seconds]
<hyper_ch>
samueldr: you mean stuff like thousands of entries in the hosts file and adblock plus, ublock origin, umatrix etc?
<hyper_ch>
and deactivation of loading stuff from third party sites by default?
<samueldr>
specially µmatrix
<samueldr>
and yes, those kind of things :)
<samueldr>
though I personally do them too
<hyper_ch>
but who still uses a web browser without that?
<samueldr>
many people!
<hyper_ch>
impossible!!!
<samueldr>
but it's why I took care in *detecting* most fail states
<samueldr>
so it's not simply a white page with a silent error in the browser console, but something a bit more noisy
<hyper_ch>
gsc.io - very shady sounding name
<samueldr>
the page is (temporarily) hosted on my own server, but the websocket is on gchristensen's infra, so right there that's something content blockers love to block :)
<hyper_ch>
so you're telling me that gsc.io belongs to gchrisensen?
<samueldr>
the whois records does say that :)
<samueldr>
(and I was 99.9% sure before looking)
<hyper_ch>
of course whois records are always true :)
<hyper_ch>
samueldr: didn't think of umatrix last night :)
<samueldr>
when I wrote the error message, I was debating whether to mention content blockers or not, since the tool already targets more technologically conscious individuals
<hyper_ch>
samueldr: I'm just a normal internet user who barely knows how to turn on and off the computer
<samueldr>
and looking at live build process logs? :)
<hyper_ch>
it's like the matrix :)
<hyper_ch>
you know, the woman in red
<Guest82411>
Apparently there's some code rot in nixbox :\ I *think* it'll work with my changes, but of course it takes 20 min to find out there's a new wrinkle at the end of the process
* Guest82411
crosses fingers
<samueldr>
I haven't done much with the box once produced, but what's rotting?
<Guest82411>
The Atlas stuff is deprecated and needs to be rewritten as a 'vagrant-cloud' postprocessor
<hyper_ch>
I've heard about bit rot but not code rot
<Guest82411>
code is made of bits
<samueldr>
oh, I removed the atlas bit entirely though, since I wasn't publishing the image
<hyper_ch>
Guest82411: that's what they want to make you believe
<MP2E>
I thought code rot referred to unmaintained codebases and how they tend to become hard to compile as compilers/depencies/build tools update
<Guest82411>
samueldr: there's also a change in the Ruby JSON module in gen_template
<MP2E>
idk though honestly
<samueldr>
code rot mostly meaning unmaintained code isn't updated to match changes in updated libraries or services
<Guest82411>
but that's it
<MP2E>
dependencies*
<samueldr>
MP2E: that too
<hyper_ch>
but then I think code rot is a wrong term... as rotting implies change of the item itself
<Guest82411>
If I'd removed the atlas thing entirely instead of trying to fix it I'd be enjoying my new vagrant box now >_>
<samueldr>
probably
<hyper_ch>
but then language isn't always logical: sleep meds -> give sleep; so pain meds give pain, right?
<hyper_ch>
lets see if I can make zfs panic again
la_putin has quit [Quit: Konversation terminated!]
el_putin has joined #nixos
hakujin has joined #nixos
Itkovian has joined #nixos
el_putin has quit [Read error: Connection reset by peer]
el_putin has joined #nixos
hakujin has quit [Ping timeout: 246 seconds]
Itkovian has quit [Client Quit]
<hyper_ch>
did I kill the channel?
* hyper_ch
pokes samueldr: "Still alive?"
bongsun has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] scode opened pull request #34096: (flameshot): init at 0.5.0 (master...scode/flameshot) https://git.io/vNgXJ
NixOS_GitHub has left #nixos [#nixos]
Guest82411 is now known as josePhoenix
josePhoenix has quit [Changing host]
josePhoenix has joined #nixos
kim0 has joined #nixos
<Ralith>
any way to use NixOS modules in a docker image?
ryantm has quit [Quit: Connection closed for inactivity]
<typetetris>
So I worked through all the nix pills, what should I read next? nixpkgs manual?
<josePhoenix>
samueldr: rats, no /vagrant mount in the final built box :(
jluttine has joined #nixos
hellrazor has quit [Ping timeout: 268 seconds]
hellrazor has joined #nixos
Arcaelyx has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
asuryawanshi has joined #nixos
el_putin has quit [Remote host closed the connection]
el_putin has joined #nixos
verite has quit [Quit: Live long and prosper.]
asuryawanshi has quit [Ping timeout: 240 seconds]
Itkovian has joined #nixos
<adisbladis>
kim0: You probably installed electrum and electrum-ltc with nix-env at different times. If you upgrade both of them you should end up having the same hash for PyQt
Itkovian has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
alex`` has joined #nixos
fragamus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<hyper_ch>
electrum - is that a new crypto currency? :)
ma27 has joined #nixos
<kim0>
adisbladis: I did install them at different times yes .. but I also did "nix-env -u" yesterday (multiple times) .. and it's not settling on a single version
<kim0>
I am worried about disk space really ;) so I want one version to disappear
<kim0>
hyper_ch: electrum is the wallet for bitcoin and (-ltc) for litecoin
<hyper_ch>
all those new confusing things :)
<josePhoenix>
arghhh why is virtualbox so janky
<kim0>
hhhh
<hyper_ch>
josePhoenix: I switched to kvm :)
<josePhoenix>
I was hoping xhyve would get adopted by Vagrant/HashiCorp
<josePhoenix>
Instead it's become the core of some Docker abomination
<hyper_ch>
Q: can Nixos be booted in read only mode?
<typetetris>
How do I find the file a certain nixpkgs derivation was declared in?
<hyper_ch>
typetetris: not quite sure what you mean
<typetetris>
somewhere in the nixpkgs repo is a file, which contains the stuff to build for example `st`. How do I find this nix file?
<hyper_ch>
I'd try find /nix/store -iname *st -type d
Mateon3 has joined #nixos
Mateon1 has quit [Ping timeout: 248 seconds]
Mateon3 is now known as Mateon1
<typetetris>
I found what I needed, but I wondered wether there is a command to print the path of the nix file for an nixpkgs attribute.
linuxdaemon has joined #nixos
linuxdaemon has quit [Max SendQ exceeded]
<hyper_ch>
pretty sure there is
<hyper_ch>
but it's advanced magic far beyond my comprehension
<typetetris>
thank you anyway
<josePhoenix>
I can pass in an extra folder that gets added to NIX_PATH
<josePhoenix>
But that isn't enough to make my "imports = [ ./relative/path.nix ];" resolve
<josePhoenix>
... it's as easy as `<path.nix>`?
<josePhoenix>
how nice
max3raza has quit [Quit: ZNC 1.6.3+deb1 - http://znc.in]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
nixpkgs/staging af32623 David McFarland: mesa: set sysconfdir to /run/opengl-driver/etc...
<NixOS_GitHub>
[nixpkgs] vcunat pushed 2 new commits to staging: https://git.io/vNgDU
<NixOS_GitHub>
nixpkgs/staging a0f8e3a Vladimír Čunát: Merge #33590: mesa: make it find drirc
<hyper_ch>
srhb: the fun will happen if it builds and if I reboot server and evreytghing is RO and I need to create new encrypte zfs datasets and migrate everything :)
<infinisil>
hyper_ch: oh wait you're patching zfs yourself without having any idea about zfs internals and using that patch on your working machine???
<hyper_ch>
infinisil: yes :)
<infinisil>
oh boy, well i wouldn't do that
<hyper_ch>
why not?
<hyper_ch>
well, it's not patches I wrote :)
alex`` has quit [Quit: WeeChat 1.9]
<hyper_ch>
applying the zfs stability patch #6864 and using zfs master for it
<hyper_ch>
but using zfs master broke mic92s patch since some stuff seems to have been moved around or something... so I applied that manually :)
<hyper_ch>
I mean what could possibly happen?
<pmeunier>
gchristensen: I can help with carnix now, or later today (sorry for the late answer)
<pmeunier>
Hi! does anyone know how to build/install NixOps from source?
<pmeunier>
I keep getting a weird error when building with the latest nixpkgs
<pmeunier>
cannot coerce a set to a string, at /home/pe/Projets/nixpkgs/pkgs/stdenv/generic/make-derivation.nix:148:11
<srhb>
avn: Essentially you'd be replacing mirror-nixos-branch with whatever script you need to use to deploy your thing (eg some combination of nixops, nix-copy-closure, switch-to-configuration, ...)
<infinisil>
pmeunier: what command are you running?
<pmeunier>
infinisil: being unfamiliar with python packaging, I was simply running `nix-build release.nix -A build`, whereas the correct command was `nix-build release.nix -A build.x86_64-linux`. Adding that to the README *right now*.
Dezgeg has joined #nixos
<infinisil>
pmeunier: hmm, -A build works for me
Dezgeg has quit [Client Quit]
datakurre has quit [Quit: all work no play makes jack a dull boy]
<pmeunier>
infinisil: wow, that's interesting, what versions of nix and nixpkgs are you using?
<infinisil>
but i feel like it would build all 3 archs like that
shabius has joined #nixos
<avn>
srhb: ty, looks interesting. Even simple `rsync $out /var/www/site.new && mv /var/www/site.new /var/www/site` should works.
<srhb>
avn: Something like that..
<infinisil>
latest master nixpkgs, and nix version 1.12pre5732_fd10f6f2
<infinisil>
pmeunier: ^^
<infinisil>
nixops master too
<pmeunier>
alright, I have a slightly older nix (1.11.16)
<pmeunier>
but same master nixpkgs and nixops
<srhb>
That's interesting.
<srhb>
pmeunier: I certainly cannot build nixops master with nixpkgs master and nix 1.12 without supplying the system
<srhb>
"error: a 'x86_64-darwin' is required to build '/nix/store/hmcivgxcx7g1884733m31l3sjjz27smz-libvirt-3.10.0.drv', but I am a 'x86_64-linux'"
<srhb>
I feel like infinisil is doing something secret :-)
<LnL>
huh, how are you building that?
<srhb>
nix-build release.nix -A build
<LnL>
yeah
<pmeunier>
so, I get a "cannot coerce set to string", you get a "darwin required", and infinisil can build fine. I don't really understand.
<infinisil>
pmeunier: Hmm I'm actually not sure why i have nix 1.12
<LnL>
nix-build release.nix -A build.x86_64-linux
<srhb>
Which is behavng as expected to me.
<srhb>
LnL: Yeah, I know. just wondering why it's working for infinisil.
<LnL>
the other will build for all supported platforms, including darwin :)
<srhb>
pmeunier: !FUN!
<pmeunier>
LnL: not on my machine (cannot coerce set to string, which sounds like a type-ish error), and not on srhb's (who gets a "darwin required")
<pmeunier>
srhb: indeed…
<LnL>
with x86_64-linux at the end?
<srhb>
No, that works as expected to.. The conundrum is why a) it work without the system for infinisil 2) why pmeunier and I are getting different errors. :)
<srhb>
too*
<srhb>
pmeunier: I just tried with nix 1.11 and I still get the (expected) darwin message.
<srhb>
I guess I'm looking forward to regnat's next talk ;-)
<ixxie>
Does anybody know where are some of the missing talks from NixCon?
el_putin has quit [Ping timeout: 264 seconds]
<ixxie>
there were some talks scheduled which are not on youtube
<srhb>
ixxie: Which ones are missing?
<srhb>
ixxie: Are they in the full stream? That should still be online, I think.
<infinisil>
Having tried out hnix myself, it's not really done, there's functionality missing and some bugs, also it's a pain to use imo (maybe that's due to my low haskell experince though)
<ixxie>
srhb: oh, I guess it used to be on the program but no more; I saw in the past something about containers / Nix in relation to Docker
<ixxie>
but I can't find it anymore
<Aleksejs>
hello, what drivers should I use for Intel HD 620?
<infinisil>
srhb: pmeunier: And this type system for nix business didn't really impress me at all tbh. I doubt this would get used at all, because it's so unergonomic and weird, it just doesn't feel right
<infinisil>
and you'd need to annotate a whole lot of nixpkgs to make it worthwhile
<srhb>
infinisil: I think that's an unavoidable consequence of gradually typing Nix.
<srhb>
And throwing everything out seems like a non-option.
<infinisil>
yeah probably
<infinisil>
i don't have a better solution
<srhb>
Well, maybe it does become an option once we're an army of maintainers :-)
<pmeunier>
I'm pretty sure many people would be willing to contribute to such an effort
ison111 has quit [Ping timeout: 264 seconds]
<srhb>
Right now I think it's too huge.
<srhb>
It's a scary amount of SLOCs. :P
<srhb>
I imagine a transitional, gradually typed phase that will be sort of ugly as a phase 1.
<infinisil>
I think other things would profit more from our effort, things like e.g. closing PR's (over 600 btw), or cleaning up nixpkgs in general, it's a mess
<srhb>
Ah well, I can dream... :-)
<Mic92>
13008 lines of code in lib/
<srhb>
infinisil: On it.. :-P
Dezgeg has joined #nixos
<infinisil>
btw, does anybody have some read-to-go command for testing PR's?
<infinisil>
maybe just something to get to the PR's head
<srhb>
Sure..
<woffs>
Hi. It might have been discussed before: rustc is broken on 17.09, hydra has no binary substitute anymore, it does not compile locally, and I therefore have to remove firefox-esr and nix-index from my config. I do not understand what's happening here.
<deltasquared>
right, so, is my understanding of how fetchgit and fetchgit local work correct, in that they create a separate derivation with the checked out clone and return the store path as a string?
<MP2E>
github's interface leaves a lot to be desired. I tend to run this command on a local branch 'curl -L https://github.com/NixOS/nixpkgs/34092.patch | git am -'. You just replace the number in the github URL with the pull request number you want to check out, then cherry-pick over to master when you're done testing, squashing, or otherwise altering the commits
<MP2E>
then it's just a simple push
<srhb>
MP2E: I have an alias for checking out PRs, which works fine
<srhb>
MP2E: My problem is merely psychological, I know I will ultimately fuck up nixpkgs if I have a push remote set up for it :P
<fearlessKim[m]>
I install git extra and do $ git pr #PR_NUMBER
<srhb>
fearlessKim[m]: Oh, this is already packaged? I rolled my own.
<deltasquared>
srhb: couldn't you ask someone to roll it back if you fucked up?
<srhb>
nice to know!
<fearlessKim[m]>
shrb yes
<srhb>
deltasquared: Sure. :P It's just embarrasing.
<deltasquared>
srhb: you're going to get embarrassed at some point... "it is inevitable" :P
<srhb>
True, true...
phaebz has joined #nixos
<deltasquared>
thankfully the biggest goof I've made thus far was on a minetest mod.
ssmike has quit [Ping timeout: 246 seconds]
ssmike1 is now known as ssmike
<deltasquared>
not quite as important as a distro's "build recipes" methinks
<srhb>
Hehe.
<deltasquared>
does anyone else get reminded of the BSD ports collection when they think about how nixpkgs works?
<srhb>
Yes, an Portage by reference.
<srhb>
Mostly for the ease of modification.
<ixxie>
how do I nix-env install an attribute from a local close of nixpkgs?
<deltasquared>
so yeah, after reading fetchgit and fetchgitlocal for a bit, I'm under the impression that they effectively create another derivation (containing the sources... I think) and return it's path, I'm hoping I'm reading the relevant nix expressions correctly
<ixxie>
I tried nix-env -i -I . -A my.attribute
<MP2E>
I once broke GHC and the haskell package set. that was pretty embarassing
<srhb>
MP2E: :D
<MP2E>
since then it's been fine though, only a false alarm when I was really tired and thought I merged the master branch over a release branch
<MP2E>
thankfully that did *not* happen
<MP2E>
I was just tired :P
<srhb>
ixxie: nix-env -f path/to/checkout -iA my.attribute should work
<deltasquared>
MP2E: at least you didn't drink and nixpkgs as it were
<srhb>
MP2E: That makes me feel a bit better, thanks. ;-)
<MP2E>
haha, that's good
<MP2E>
and yes I have yet to drink and nixpkgs :P
<ixxie>
cheers srhb :)
<deltasquared>
“remember kids, don't drink and dd”... or code, or anything semi-important
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] srhb pushed 2 new commits to master: https://git.io/vNg79
NixOS_GitHub has left #nixos [#nixos]
<NixOS_GitHub>
nixpkgs/master 8c25e0d Sarah Brofeldt: Merge pull request #33717 from matthiasbeyer/fix-taskwarrior-completion...
phaebz has quit [Quit: ERC (IRC client for Emacs 25.3.1)]
<deltasquared>
well, fetchgitlocal, that's the one I think I'll be using for the moment
<srhb>
deltasquared: Sure,w hat about it?
<deltasquared>
srhb: if you read my above, my understanding is that it effectively plonks the clone of the repository into another derivation, is that correct?
<deltasquared>
given it's use of runCommand which in turn calls mkDerivation under the hood
<MichaelRaskin>
Well, force-pushes are forbidden, you cannot really really mess up stuff…
<deltasquared>
and then the return result of fetchgitlocal is the store path of that "checkout derivation"
<srhb>
If you mean that fetchgitlocal {...} will ultimately result in a store path of the (cleaned) checkout, then yes
phaebz has joined #nixos
<deltasquared>
srhb: oh cool, nice to know. I can start with that then in a nix expression and work from there
<srhb>
deltasquared: That's what a lot of src attributes are. :)
<deltasquared>
srhb: it looked straightforward but there are always caveats learning a new language
<srhb>
I haven't actually used fetchgitlocal, but if it works like fetchgit, that's what it does.
<deltasquared>
srhb: I've looked at the two default.nix files relevant to each one. fetchgitlocal is a bit more primitive, it only takes a path and doesn't do anything particurlarly fancy
<deltasquared>
(one can't specify an exact commit for instance)
<deltasquared>
in fact it appears to require a working tree.
<srhb>
Ah.
pie_ has quit [Ping timeout: 240 seconds]
phaebz has quit [Client Quit]
<deltasquared>
not to mention doing strange things to the git index
phaebz has joined #nixos
<deltasquared>
at least I can feel like I'm starting to grok this functional-ish programming going on... it's pleasingly simple not to see any imperative constructs anywhere
<deltasquared>
it makes me have to squint a bit more though, I'm kinda used to seeing "function x(a, b, ...)" to give my brain a hint of what's going on
<srhb>
deltasquared: You'll eventually see a lot of {a, b, c, ...}: ... when that's more relevant than having a curried function.
<srhb>
deltasquared: But being able to partially apply trivially is so nice.
<deltasquared>
onto my next question then: seeing as most nix expressions require input dependencies, does one have to do anything particularly special with nix-build in order to "import" those dependencies for an expression I wrote?
<deltasquared>
the nixpkgs quick start guide is suggesting I locally modify the nixpkgs (in particular the top-level expression), which I'd rather not right now.
<srhb>
deltasquared: Then you can just use callPackage from nixpkgs to fill it in
<ixxie>
steveeJ: have you tried bumping GDM to unstable?
<deltasquared>
srhb: right, and then the "with" bit makes the keys of the resulting set become part of the scope of the callPackage bit afterwards
<deltasquared>
... I think
<steveeJ>
ixxie: no. I'm staying away from unstable for boot components
<srhb>
deltasquared: Right, or we could have used 'let pkgs = import <nixpkgs> {}; in pkgs.callPackage ...'
kim0 has quit [Quit: Connection closed for inactivity]
<deltasquared>
damnit this is really hard, even this relatively domain specific language has so much information in a small space
la_putin has joined #nixos
<ixxie>
fair enough steveeJ
<srhb>
deltasquared: It's quite terse, we can do a lot of things in very little space. :)
<srhb>
deltasquared: Once you start thinking of nixpkgs as a library (or a set of libraries) for programming nix expressions, it all expands a bit.
<deltasquared>
srhb: oh, I got it could be considered a library, I was just getting to the part where I understood how it was pulled in and glued together
* srhb
nods
ambro718 has quit [Quit: Konversation terminated!]
<deltasquared>
even in a functional language dependencies have to come from somewhere, if not magically from the void
<infinisil>
import is the magical function
<srhb>
I think it's often more explicit, since it's usually just function parameters.
quam6246 has joined #nixos
<infinisil>
import takes a path to a file and evaluates the nix expression in it
<deltasquared>
infinisil: I guess, but that's reasonable magic :P
<deltasquared>
and <nixpkgs> is just an expansion in this case?
<infinisil>
<foo> looks up "foo" in the NIX_PATH env var
<infinisil>
and it has to be a path
<deltasquared>
infinisil: ... hah, I had forgotten that had even been added to my profile
<deltasquared>
(my shell profile I mean)
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] dezgeg pushed 2 new commits to master: https://git.io/vNg5d
<NixOS_GitHub>
nixpkgs/master 44dc31b Ben Wolsieffer: perf: apply patch to fix build of 4.9
<NixOS_GitHub>
nixpkgs/master 94276da Ben Wolsieffer: perf: inherit makeFlags from kernel
NixOS_GitHub has left #nixos [#nixos]
<infinisil>
NIX_PATH is the gateway to nixpkgs :)
<infinisil>
All of nixpkgs uses references to <nixpkgs>, so if that's missing from your NIX_PATH you can't do much
* deltasquared
is trying to find where in the documentation <> expansion is described just to keep his sanity
<joko>
Anyone using system.autoUpgrade in NixOS? How do you deal with reboots for kernel upgrades?
<LnL>
I'm not sure if it's documented anywhere
<deltasquared>
LnL: ಠ_ಠ ok welp I guess I'll just have to remember that particular bit of magic sauce
<srhb>
joko: Manually!
<LnL>
it's kind of like PATH but you can also add named entries
<deltasquared>
LnL: I get the concept ish, just wondering why it's not written down anywhere
<LnL>
it probably should :)
<deltasquared>
seems to be a recurrent problem in my journey thus far. :P
<infinisil>
hmm yeah i can't find it in the manual
<deltasquared>
I had to go learn how fetchgit works by staring at the nix expression for a bit
<joko>
srhb: ok, are you at least aware that the machine needs a reboot?
<srhb>
joko: I usually do it if booted system kernel and running system kernel don't match.
<deltasquared>
srhb: ooooh, ok. you'd have thought it'd be mentioned within the syntax part of the manual, being as <> looks like it would reserve those characters in a sense
<infinisil>
srhb: Ohh, that's pretty hidden
<srhb>
joko: That might produce some false positives though
<infinisil>
deltasquared: yeah it does, should probably be in the syntax part
<joko>
srhb: ok, I see the pattern
* deltasquared
lets all this new information float around in his head for a bit
<joko>
And any ideas for combining nixops with system.autoUpgrade or sth else equivalent?
<srhb>
Not really..
<joko>
So with nixops one cannot perform unattended upgrades if I understand well
<srhb>
I think nixops has its own method of signalling that a deployment needs reboot.
<srhb>
But I don't use any advanced features from it
<infinisil>
joko: you can still just do an automatic periodic deploy (which is what autoUpgrade does with nixos-rebuild anyways)
<deltasquared>
"string doesn't represent an absolute path" d'oh.
phaebz has quit [Read error: Connection reset by peer]
<joko>
infinisil: how to do so?
<infinisil>
joko: look at the source of autoUpgrade, with a systemd timer/service
<joko>
infinisil: so you mean, on my PC to have a systemd timer/service which updates the nixpkgs channel I am going to use and runs nixops deploy?
<infinisil>
deltasquared: `import ./foo` works (because ./foo expands to $PWD/foo, which is absolute), `import "./foo"` does not (because "./foo" is not absolute), but `import (./. + "/foo")` does (because `./.` = $PWD, `./. + "/foo"` = $PWD/foo)
phaebz has joined #nixos
<srhb>
Hrm, how does one enable bash completion in nix-shell (bash) when my user shell is zsh? I tried enabling programs.bash.enableCompletion, but that doesn't seem to suffice
<infinisil>
srhb: you could just switch to zsh upon entering the nix-shell
<srhb>
infinisil: Unfortunately I'm trying to test out whether a bash completion PR works or does not work on 17.09 :-)
<infinisil>
Ah
<srhb>
So I just want to be able to spawn a bash with completion enabled.
<joko>
I also run zsh right after running nix-shell, any way to automate this?
<deltasquared>
infinisil: btw nix-pills documentation isn't present in that repository you suggested I clone earlier, is it in another perhaps?
<srhb>
joko: I have an alias for nix-shell --command "zsh"
<srhb>
deltasquared: I already linked the nix-pills repo, or at least intended to.
<infinisil>
deltasquared: huuuuuuhhh, weird
<joko>
srhb: bingo
<infinisil>
i have the nix-pills folder in nixos-homepage/nixos
<infinisil>
oh, it's in untracked files, i must've messed something up with git, and it probably got removed from there
<deltasquared>
oh, docbook again? DX
<deltasquared>
*engineering language*
<steveeJ>
ixxie: my gdm process has the environment variable GDM_SESSIONS_DIR=/nix/store/wmynl6cm9758s1x8aj4ygs8nnmd6ma8r-desktops; this directory contains valid session files. it looks like the button/menu is just hidden from the GUI
ssmike1 has joined #nixos
ssmike has quit [Ping timeout: 256 seconds]
ssmike1 is now known as ssmike
<deltasquared>
“anonymous function at /h/user/doc/nix-pills/default.nix:1:1 called without required argument ‘revCount’” aww hell. just when I thought I was getting my head around this... now what are those values meaning here... hmm
<deltasquared>
... oh, that's what the release.nix is for
<deltasquared>
... yas, that worked!
<deltasquared>
I don't suppose there's any shortcut for that "with import <nixpkgs> ..." incantation? I might be using that a lot
<NixOS_GitHub>
[nixpkgs] nmattia opened pull request #34098: gpp: init at 2.25 and add @nmattia as maintainer (master...nm/add-gpp) https://git.io/vNgFY
fragamus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<srhb>
deltasquared: Here's a fun game, always call them result and at the end of the week, count how many result links you have in your homedir. :-)
<srhb>
Bonus points if you somehow made some outside of your homedir.
<deltasquared>
the concept of "homedir" isn't that well defined on my system anyway
<srhb>
I'm somehow unsurprised ;-)
<deltasquared>
srhb: what, that I have a bunch of arcane hacks?
<infinisil>
JosW: Not sure why you're asking me specifically, but they seem to have build instructions you could make a nix package with
<srhb>
deltasquared: I'd phrase it differently, but... :-)
<deltasquared>
srhb: don't worry about it, even my fellow archers toss me the occasional "why the fuck"
<srhb>
hehehe.
jtojnar has joined #nixos
<infinisil>
deltasquared: i was surprised by your /h/user path, what's up with your home directory?
<JosW>
infinisil: sorry for being direct, saw that you are one of the maintainers for the x2go stuff
<infinisil>
JosW: I am not, not sure where you got that from haha
<infinisil>
i did once do a mass refactor that touched a whole lot of files in nixpkgs, so maybe from that
<deltasquared>
infinisil: my actual home directory as reported by $HOME is /h/user/dot. I have files above that which I actually want to keep (e.g. /h/user/files) and the "dot" directory is for apps that literally litter dotfiles that I don't care about
<deltasquared>
anything important is symlinked
<JosW>
arghhh contributor :)
<infinisil>
deltasquared: ahh nice
<ixxie>
are there tricks to dealing with memory limits?
<ixxie>
nix-env -i python3 runs out of memory on my 1GM droplet
<deltasquared>
funnily enough, my home dir is probably candidate for some nix-like treatment.
<infinisil>
deltasquared: i should probably do that too. But why /h/user and not /home/deltasquared ?
<deltasquared>
infinisil: because on my system users are named after roles, not people
swayf has quit [Remote host closed the connection]
<deltasquared>
"user" was just the generic everyday one
swayf has joined #nixos
<infinisil>
hmmm i see
<tilpner>
ixxie - Try -iA python3
<infinisil>
deltasquared: and you just chose h instead of home
jtojnar_ has joined #nixos
<deltasquared>
infinisil: oh, that's because /h is specifically for "user" and it's encrypted. it has e.g /h/uniwork which I need to protect.
<infinisil>
bit confused by that
<deltasquared>
infinisil: there's no particular reason I couldn't make it /home/user/dot but there would be a few more hoops
<deltasquared>
(it's the latter on my desktop tower for instance)
quam6246 has quit [Quit: My computer has gone to sleep. ZZZzzz…]
<Aleksejs>
I'm trying to install mesa from unstable and I got this error: http://vpaste.net/Hom27. Any ideas how to install it?
<ixxie>
tilpner: it doesn't find the attribute in the selection path
<7IZAAOQYG>
nixpkgs/release-17.09 510a474 Sarah Brofeldt: Merge pull request #34099 from srhb/nixos-17.09...
<deltasquared>
erm, why does nix-daemon not have any --help output or anything useful in it's man page
ma27 has joined #nixos
<deltasquared>
(just curious about running it on a non-nixos system)
<MichaelRaskin>
Just run it!
<deltasquared>
MichaelRaskin: yeah, but does it fork to the background? what would I need to know to throw it at a systemd service etc.
<MichaelRaskin>
Ah you want systemd
<deltasquared>
MichaelRaskin: the only sane way to run services these days ;)
* deltasquared
dodges the sysvinit mob
<MichaelRaskin>
deltasquared: No, it is still not sane
<deltasquared>
ok, it's the sanest we have.
<MichaelRaskin>
I mean, the uselessd attempted fork was
<MichaelRaskin>
systemd is worse than just running a list of runner scripts sysv-style (which is what I chose)
<MichaelRaskin>
I think nix-daemon stays in foreground
<deltasquared>
I generally find I can't live without systemd's ability to express dependencies and ordering.
<deltasquared>
especially with some of the obscure stunts I've pulled on my system >_>
<infinisil>
+1 to that
<MichaelRaskin>
When systemd just wanted journald to be used (which is the wrongest logging I have seen), I was OK with it, but its opinions on management on basically every resource are too much for me.
ssmike has quit [Ping timeout: 240 seconds]
<deltasquared>
right, fair enough. for the most part I've been fine to let systemd handle the minutia of bootup
<infinisil>
I don't use a lot of systemd features i think, but it has everything i need and works nicely for me. And having it integrated into nixos helps too
<MichaelRaskin>
Well, the very idea that systemd breaks screen and only later starts negotiating with screen upstream…
<deltasquared>
don't a few things in nixos lean on systemd anyway? like nix-shell?
<deltasquared>
MichaelRaskin: wait wat
<MichaelRaskin>
Nope, nix-shell works fine on my systemd-less system
<MichaelRaskin>
(I do have the libraries, just not a running instance)
<infinisil>
Systemd is deeply integrated in nixos
<MichaelRaskin>
I think some container stuff might be relying on systemd.
<deltasquared>
it figures, though the ability to have alternative spawn backends would be nice.
<MichaelRaskin>
NixOS core is. But it is not of much value to me as long as I can just export runner scripts
<adisbladis>
deltasquared: systemd implemented killing off all processes after a session was closed (which change the behaviour people expected from screen and tmux)
<MichaelRaskin>
As far as I understand, this is because Gnome failed to implement anything resembling process management and needed that behaviour on session exit.
<deltasquared>
adisbladis: oh, that. I happen to be of the opinion to kill such rogue processes as I don't use screen myself, but again I can see why it would be annoying
<deltasquared>
MichaelRaskin: that sounds about right, actually >_>
<adisbladis>
deltasquared: Yeah I agree.. The behaviour of not killing off processes not associated with login sessions/services was always broken imho
<deltasquared>
"but we use systemd! we'll just let that do it!"
<deltasquared>
in all fairness, systemd *is* in a better position in that case to be able to cleanly contain and whack remaining processes, much better than gnome could hope to.
<MichaelRaskin>
I mean, couldn't they run a user service that spawned all the unmanageable processes, and then spawn the user processes from something outside that cgroup?
<deltasquared>
that said, "libcgroupsession" would be a useful thing to have, so one could accurately track and kill things.
<MichaelRaskin>
I also actually want to be able to manage vts by _my_ code.
<MichaelRaskin>
It is OK if it hs to run as root.
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] jokogr opened pull request #34100: xmonad-log: init at 0.1.0 (master...pkg/xmonad-log) https://git.io/vNgbM
NixOS_GitHub has left #nixos [#nixos]
abcrawf has quit [Remote host closed the connection]
<deltasquared>
MichaelRaskin: interesting, so do you use the traditional setuid xorg wrapper?
<MichaelRaskin>
It is not OK if I need to track APIs of people who cannot even get logging right.
<MichaelRaskin>
By now, even weirder.
<deltasquared>
oh yeah journald's logging format, blegh
<MichaelRaskin>
I have a daemon in Lisp that manages various policies.
<deltasquared>
he's a lisper! RUN! :P
<fearlessKim[m]>
when generating lua packages from luarocks specs, what should I do if I can't map the luarocks license to a nix license ? is licenceless package a thing ?
amir has quit [Ping timeout: 276 seconds]
<MichaelRaskin>
I can ask it to spawn _an_ X session (not _the_ X session, you know), it will check if the _physically present user_ agrees (you know, me-over-ssh should have higher barriers to doing some things), then do What I ACtually Encoded In MY Policy
<deltasquared>
huh, I didn't know there was a thing for turning rock specs into nix expressions. nifty.
<MichaelRaskin>
It is possible to not set a licence in a Nix package.
<MichaelRaskin>
It will somewhat work.
<MichaelRaskin>
Not sure if we have reached the coverage where it will be considered unfree and require explicit opt-in.
abcrawf has joined #nixos
<gchristensen>
ohh that is a good idea
<MichaelRaskin>
A bit disrupting at the point of the switch, though.
<deltasquared>
adisbladis: and that's fine as well
<deltasquared>
I'm just not protective of the code I write.
dywedir has quit [Read error: Connection reset by peer]
<deltasquared>
if it ends up benefitting humanity, great. if it doesn't because $CORPORATE then I already have my money so...
<deltasquared>
all my personal projects are public domain, that's for sure.
<gchristensen>
:)
<deltasquared>
mostly so if someone comes along and asks me "can I use this for ..." I can preemptively yell YES then strategically run away
<adisbladis>
deltasquared: Just saying I would have trouble convincing an employer such a loose thing as "public domain" is a good licensing strategy :)
amir has joined #nixos
<deltasquared>
adisbladis: I'm not saying the world should adopt this view; partly because I'd have to have a very good argument for that, which I do not. :P
dywedir has joined #nixos
<fearlessKim[m]>
as a first step , I will use nix-instantiate --eval -E 'with import <nixpkgs> { }; stdenv.lib.licenses.ROCKSPEC_LICENSE.shortName' and if that fails there will be no license
bpa has quit [Read error: Connection reset by peer]
<gchristensen>
deltasquared: so fetchgit is considered ok because you're specifying the sha256 that will come out. it is perhaps not repeatable overtime though because indeed the HEAD will change making the hash wrong making it unbuildable
bpa has joined #nixos
<deltasquared>
gchristensen: well actually I'm not specifying the hash at all right now
<gchristensen>
oh you're using builtins.fetchgit then?
<deltasquared>
gchristensen: no, I'm pretty sure it's pkgs.fetchgit
<deltasquared>
... wait, how the faken am I getting EACCESS to rm -rf that store path when everything nix does is under my user
<MichaelRaskin>
Well, read-only permissions do mean that you need to chmod firts
<MichaelRaskin>
first
<deltasquared>
MichaelRaskin: firstly it was -rf which usually just plows through that, and secondly I did chmod it anyway and it still gave me EACCESS
<deltasquared>
there aren't any funky attributes set on it AFAICT
<MichaelRaskin>
And stat confirms that is under your user?
bpa has quit [Read error: Connection reset by peer]
<MichaelRaskin>
Haven't used non-root installations for quite some time…
<deltasquared>
MichaelRaskin: yep. it couldn't be anything else anyway, it was just made by my user.
bpa has joined #nixos
<gchristensen>
deltasquared: you're on nix-not-nixos?
<deltasquared>
gchristensen: yep
<deltasquared>
for the moment anyway.
<MichaelRaskin>
Could you pastebin stat of a file inside the path, and a strace of removing that file?
<deltasquared>
one second
ssmike1 has joined #nixos
<typetetris>
is it possible to use a package override or overlay in a single build environment instead of per user?
<gchristensen>
well anyway you probably have sandboxing off which allows for certain impurities, like local filesystem access
<deltasquared>
gchristensen: true, though that doesn't explain how nix-build ended up making a file that my own user can't remove :P
<gchristensen>
I dunno, that is completely normal for me ;)
<MichaelRaskin>
typetetris: you can just evaluate an override of the expression that passes a different argument
<gchristensen>
why are you rm -rf'ing in the nix store? you should not do that
ssmike has quit [Ping timeout: 256 seconds]
ssmike1 is now known as ssmike
<deltasquared>
gchristensen: nothing else is going to depend on this, this was just me trying to show you what happens with fetchgit on my system, the thing you said couldn't happen
<MichaelRaskin>
NIX_PATH can be used to set nixpkgs-overlays to something different
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
nixpkgs/master d1a04ef Michael Weiss: tdesktop: 1.1.23 -> 1.2.1
NixOS_GitHub has left #nixos [#nixos]
<NixOS_GitHub>
[nixpkgs] primeos pushed 1 new commit to master: https://git.io/vNgjc
<MichaelRaskin>
nix-store --delete is better than rm -fr
<gchristensen>
though even with sandboxing you could do fetchgit { url=/usr/local/packages/nftfirewall; } and it'd be no problem
<deltasquared>
gchristensen: I was only using file:// to allow shallow clone to work correctly.
<MichaelRaskin>
Although this would be a bit inefficient
bpa has quit [Read error: Connection reset by peer]
<deltasquared>
no point making another complete duplicate when the .git just gets thrown away afterwards...
bpa has joined #nixos
rihards has quit [Quit: rihards]
<gchristensen>
sure
<deltasquared>
in any case, now I can think about what I want to do about the impurity of HEAD. I'll probably just have to slap a tag on it.
<burgdorfer>
Hi everyone, I'm just getting started using nix for haskell development with reflex and am getting the following error: Configuring heist-1.0.1.0... Setup: Encountered missing dependencies: aeson >=0.6 && <1.2. How would I go about resolving this issue? I'm not even sure nix is supposed to build my dependencies from source, I'm just following this guide: https://github.com/ElvishJerricco/reflex-project-skeleton
<gchristensen>
deltasquared: you can specify an exact rev
<burgdorfer>
My app depends on snap which depends on heist which cannot seem to be built from source by my current setup.
<deltasquared>
gchristensen: huh, my reading of the comments in the nix expression for fetchgit was that this was not possible due to git's security mechanisms
<gchristensen>
eh?
<deltasquared>
unless I wasn't paying attention to whatever clever hack it used to get around this
<deltasquared>
gchristensen: the remote commit has to be a ref. the remote server won't let you specify an exact sha1
<deltasquared>
“The workaround is getting the last n commits so that it's likely that they still contain the hash we want.”
<gchristensen>
fetchgit { url = /home/grahamc/projects/ofborg/ofborg/test-srcs/bare-repo; rev = "947318524af9e5f11b7aca5fb17439f2911af9be"; } this seems to work for me
<deltasquared>
gchristensen: that's not a URL though; that's a path. internally it's handled differently
<gchristensen>
I see
<deltasquared>
let me see what happens if I try to specify a commit id
<gchristensen>
gotta go
<deltasquared>
... wait, file:// works that way too
<deltasquared>
hrm
<deltasquared>
is it only for proper git remotes I wonder
bpa has quit [Read error: Connection reset by peer]
bpa has joined #nixos
<deltasquared>
wait, no, that works as well, so how many commits is it fetching for that "workaround" I wonder
el_putin has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] mnacamura opened pull request #34107: cataclysm-dda{,-git}: build on Darwin (master...cdda-darwin) https://git.io/vN2eg
NixOS_GitHub has left #nixos [#nixos]
Ross has joined #nixos
la_putin has quit [Read error: Connection reset by peer]
Xorkle has quit [Ping timeout: 240 seconds]
ixxie has quit [Ping timeout: 248 seconds]
<jtojnar>
steveeJ: there are actually two codepaths in GDM that contain the path
<jtojnar>
I recall patching them when I was working on wayland
<jtojnar>
though I cannot find the patch now
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] dywedir opened pull request #34108: dunst: 1.2.0 -> 1.3.0 (master...dunst) https://git.io/vN2e5
<contrapumpkin>
Ralith: I was, but am not sure why it would screw up your terminal
ixxie has joined #nixos
phaebz has joined #nixos
<lejonet>
srhb: I've now come to the point where I'm starting to test the service file :D
marek has joined #nixos
<lejonet>
However I'm getting it whining at having an unexpected IF when I'm trying to do: if cfg.mon.enable then map (mon: systemd.services."ceph-mon@${mon}".enable = true ) cfg.mon.daemons; (where cfg just is a shortcut for config.services.ceph)
<lejonet>
What is the recommended way of doing that map conditionally in the implementation of a service?
periklis has joined #nixos
<Rovanion>
So I've got this teamviewer.deb or teamviewer.tar.xz that I would like to install onto my nixos, how would I go about it? Is this where I write my first nix package definition?
<deltasquared>
Rovanion: have you not seen if there exists an expression already
<lejonet>
so you just have to enable using unfree software and you should be able to install it
* lejonet
just uses find, on linux, in a checkout of the nixpkgs git repo :P
asuryawanshi has joined #nixos
<MichaelRaskin>
Let me see if nix-index handles unfree packages in the expected way…
<infinisil>
Rovanion: nix-env -qaP
<Rovanion>
nix-env -qaP '.*teamviewer.*' isn't quite as catchy though as apt search teamviewer...
<infinisil>
Rovanion: Use `nix-env -qaP | fzf`
<lejonet>
Its heck a lot more flexible than apt search tho... :P
<Rovanion>
Perhaps, but a user new to the ecosystem needs usable more than flexible.
vidbina has joined #nixos
<lejonet>
That I do agree, but then you compare the young age of this distro, and available resources, compared to debian/ubuntu :P
<infinisil>
lejonet: young age?
<infinisil>
since 15 years?
<infinisil>
s/since//
asuryawanshi has quit [Ping timeout: 248 seconds]
<lejonet>
infinisil: isn't that just for nix tho? The actual distro is younger iirc
<adisbladis>
Rovanion: In nix 1.12 we have "nix search ..." :)
<Rovanion>
On the other hand, nixos has the advantage of hindsight. There's no need for a usability study to figure out that the most used commands such as install and search should probably have aliases in some way, like nix search (which was added to the discussion while I was writing this).
<lejonet>
ah no, I'm speaking out of my ass, my apologise :)
<MichaelRaskin>
I think I have used Nix for more than 10 years, and NixOS was already a few years old
<infinisil>
lejonet: wikipedia says 2003 was the initial nixos release
<lejonet>
infinisil: yep, just found that, scratch my comments :)
<lejonet>
sure, debian got ~10 years on that, and a larger community
<lejonet>
But then apt doesn't support downgrading stuff anyway
<Rovanion>
"The program ‘nix’ is currently not installed. You can install it by typing..."
<adisbladis>
Rovanion: Its in nixUnstable. Not final release yet.
<Rovanion>
How do I know which version of nixos I'm on? lsb_release doesn't seem to be part of the base install.
<infinisil>
Rovanion: nixos-version
<Rovanion>
1.12 doesn't seem to match 17.09.2683 in a good way.
<infinisil>
i don't think that should matter much
<typetetris>
Are there some haskellers here, which could explain to me, what the point of using nix for haskell development is? (I got the impression, it just replaces stackage snapshots with nix channels.)
<infinisil>
nix version is mostly independent of nixos version
endformationage has joined #nixos
<ixxie>
can nix-shell -p access packages installed by nix-env?
<infinisil>
ixxie: i dont' think so
<ixxie>
hmmm
<ixxie>
I wanna nix-shell into an environment with some newly packaged stuff which is my local clone of nixpkgs
<ixxie>
and it seems there isn't really a -I flag on nix-shell either
<infinisil>
ixxie: there is
<ixxie>
oh
<ixxie>
well in that case it isn't documented as far as I can tell
<ixxie>
oh it is
<ixxie>
just not in the synopsis
<ixxie>
alright cool
<infinisil>
yeah
<infinisil>
most nix commands share these arguments
ciil has joined #nixos
<ixxie>
infinisil: so how would it go? I tried nix-shell -I ./nixpkgs -p mypackage without luck
<ixxie>
also -A
<ixxie>
I still haven't completely for the hang of namespaces
<infinisil>
-I nixpkgs=path/to/nixpkgs
<ixxie>
ah
<ixxie>
infinisil: I think something is still missing
<ixxie>
nix-shell -I nixpkgs=./nixpkgs -p mypackage doesn't find the package
<infinisil>
ixxie: it won't expand relative paths
<infinisil>
use nixpkgs=$PWD/nixpkgs instead
<infinisil>
a few gotchas :)
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Ericson2314 closed pull request #33110: bintools-wrapper: symlink all programs that do not need wrapping (staging...bintools-wrapper) https://git.io/vbdwq
<hyper_ch>
hi Mic92, I tried to build zfs master with the encryption stability patch but I failed :(
<Rovanion>
samueldr: Thank you, I'll try to bookmark the new one but Mozilla seems keen on moving/removing features.
<Rovanion>
Ah, found it!
<infinisil>
hyper_ch: what does the patch do btw?
spear2 has joined #nixos
jmeredith has joined #nixos
<sphalerite>
so, my chromebook's chipset includes a codec for various audio and video encodings… is there a way to make use of this hardware acceleration on nixos?
fragamus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<infinisil>
hyper_ch: oh nice
<hyper_ch>
that sounded sarcastic?
<infinisil>
haha no
<infinisil>
raw sends are nice
<hyper_ch>
in theory yes, in practice they don't work currently properly as it can't be mounted again
<infinisil>
hyper_ch: is this the thing that didn't work for encrypted znapzend?
<hyper_ch>
if encrypted
<hyper_ch>
infinisil: that's something else
* infinisil
wants znapzend to work without problems for encryption
<hyper_ch>
I wrote my own backup script meanwhile
bpa has quit [Read error: Connection reset by peer]
bpa has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
nixpkgs/master 26f86b4 Michael Weiss: tdesktop: 1.2.1 -> 1.2.6
NixOS_GitHub has left #nixos [#nixos]
<NixOS_GitHub>
[nixpkgs] primeos pushed 1 new commit to master: https://git.io/vN2TM
<hyper_ch>
but I'd really love to test the stability patch
nschoe has joined #nixos
bpa has quit [Read error: Connection reset by peer]
bpa has joined #nixos
pie_ has joined #nixos
<nschoe>
Hi everyone
* deltasquared
waves
<nschoe>
I have a quite embarassing problem: I have had nixOS and Win10 on dual boot (UEFI) for, like 3 years. Worked flawlessly. And I regularly reboot and dual boot. So it worked. This morning, however I cannot boot nixOS anymore. It fails at stage 1. 2 things to be noted: 1) I use (and have always used) LUKS encrypted partitions for nixOS and 2) this morning there was a quite big Win10 upgrade.
<deltasquared>
nschoe: did win10 clobber your UEFI boot entries?
bpa has quit [Read error: Connection reset by peer]
<nschoe>
It is since this Win10 upgrade that my nixOS fails to boot, although I don know if this is related. This is partition Win10 should not be concerned about.
bpa has joined #nixos
<nschoe>
deltasquared, I first thought this, but I don think so: when I boot the computer I still have the bootloader selection (gummiboot) and I can still chose between Windows and nixOs
<nschoe>
deltasquared, the problem is that it now says ¨¨/dev/sda5 is not a valid LUKS device¨
<deltasquared>
nschoe: have you validated this from some external media?
<deltasquared>
(one really should have a rescue disk for this kind of thing)
<hyper_ch>
s/disk/usb stick/
<LnL>
sda5? if something changed in the partition scheme that might be the wrong device name
<nschoe>
deltasquared, not yet, no.
<nschoe>
deltasquared, but yeah Iĺl have a rescue dis from now on
<LnL>
it's generally not a good idea to use sd* names directly
<deltasquared>
nschoe: LnL has a point, why were you using /dev/sda5 instead of e.g. /dev/disk/by-uuid/xxx...
<deltasquared>
or even partuuid, if it's GPT.
<samueldr>
I may be wrong, but I have a feeling I remember reading that a windows update would add partition(s) soon or maybe not soon
<nschoe>
LnL, thatś the weird thing, the screen at nixOS boots says: ¨¨ Device /dev/sda5 is not a valid LUKS device¨ but then it says ¨waiting for /dev/disk/by-uuid/xxx-yyy-... to appear¨ and then it fails
<nschoe>
LnL, so... it seems part of it uses the uuid denomination, part of it doesn
<samueldr>
it's in two parts, the second part is the uuid of the unencrypted disk
<LnL>
sounds like samueldr kind of confirmed my suspicion :)
<samueldr>
well, de-encrypted, not un-encrypted
<nschoe>
samueldr, that´s possible indeed. But then how / where can I change the partition name where it looks for the LUKS device?
* samueldr
is thinking
<deltasquared>
I never did trust windows to not at some point decide it'd use a partition it didn't recognise for it's own purposes
<LnL>
nschoe: maybe the mount is setup correctly with a uuid but not the lurks part
<nschoe>
deltasquared, wait what? I thought samueldr meant Windows would **add** a new psrition not take over one of mine!
<samueldr>
nschoe: one thing is sure, download a nixos iso and burn it to usb drive as you may as well need it
<deltasquared>
nschoe: I have no idea, but I would not be suprised either way
<nschoe>
LnL, from what samueldr said, I think it´s the case. How / where can I change the partition of the LUKS device?
<samueldr>
let's leave this to the .01% probablities, that would be so rude of microsoft to do such a thing
<nschoe>
samueldr, I will do this for sure!
<nschoe>
samueldr, indeed.
<hyper_ch>
nschoe: boot live cd
<hyper_ch>
nschoe: unlock encrypted disk
<hyper_ch>
mount everything as you should in /mnt
<hyper_ch>
and make reinstall
<nschoe>
Okay.
bpa has quit [Read error: Connection reset by peer]
<nschoe>
¨make reinstall¨?
<samueldr>
(the ESP will be important!)
<hyper_ch>
and then make sure that uuid is being used in the hardware configuration
<MichaelRaskin>
samueldr: just as rude as upgrade from W8 to W10 when the user closes the nag popup using Alt-F4 instead of Cancel? Oh wait.
<nschoe>
hyper_ch, yeah I thought I had to compile smth manually. But I just have to change the luks device in configuration.nix and nixos-rebuild rebuild, right?
<hyper_ch>
nothing confusing about that
<samueldr>
hyper_ch: isn't it generate-config that overwrites hardware config?
<hyper_ch>
nschoe: just make a backup copy of your hardware-configuration.nix in case you have added/altered stuff there as well
<nschoe>
hyper_ch, yep.
<nschoe>
So.. no way to do it ¨live¨, right?
<deltasquared>
nschoe: AFAIK nixos-rebuild does not work correctly from the iso. you'd need nixos-install, and *after* running nix-channel --update (the ISO will likely have out-of-date nixpkgs)
<hyper_ch>
pretty sure there's a way to do it directly on your system but it seems a lot more complicated
<deltasquared>
assuming your configs are in place there shouldn't be a difference between first install and "rebuild" as it were.
* samueldr
checks if nixos supports single user rescue boot
<nschoe>
deltasquared, hum okay. I am just very confused by the nixos-install
<deltasquared>
nschoe: it installs your system. consider that the output of the operation will always be the same for a given configuration.nix, therefore you should be able to use the first install tool to get back up and going
<samueldr>
try adding boot.shell_on_fail in your boot command line
<hyper_ch>
nschoe: nixos-install will make a chroot from /mnt and then install nixos according to your configuration.... at the end it will prompt you for new root password
<deltasquared>
it's what I use when I needed to change the main system from iso.
<samueldr>
but yes, as others point out, using nixos-install on existing system works fine
<deltasquared>
just don't forget the nix-channel --update first.
<nschoe>
Okay sure
<nschoe>
Well I´m going to test this right now
<deltasquared>
good luck
<nschoe>
Well after iḿ done burning the ISO on the USB.
<hyper_ch>
only thing I haven't found out yet is how to set nix-channel to unstable and directly use unstable in the live iso
<samueldr>
if you want to do it live, boot.shell_on_fail will add options for interactive shell on failure to mount
<nschoe>
Thanks! I am on the laptop so I´ll have to log out.
<deltasquared>
hyper_ch: hmm yeah...
<nschoe>
samueldr, I´ll try that first, and then the USB stick if this doesn work
<samueldr>
which should allow you to mount it manually, and identify partitions (e.g. lsblk) and mount manually
<hyper_ch>
so from live ISO I just a very basic installation
<nschoe>
samueldr, but if the shell doesn´t have luks installed I won´t be able to do much
<samueldr>
and iirc, that rescue shell once logged out continues the boot process
<infinisil>
hyper_ch: you mean build an iso from unstable nixpkgs!
<infinisil>
?*
<samueldr>
nschoe: the shell is the stage1, the same thing that boots the system
<hyper_ch>
then reboot into new system and then restore configuration.nix and stuff, switch channel and then to rest install
<nschoe>
samueldr, oh! So yes. cool.
<hyper_ch>
infinisil: no, I mean I don't know how to make the live cd use unstable channel for installation
<nschoe>
So I´m logging out and trying. I´ll hopefully be back soon. Thanks a lot guys!
nschoe has quit [Quit: Leaving]
<infinisil>
hyper_ch: I'd think just -I nixpkgs=path/to/unstable
<infinisil>
For nixos-install
<hyper_ch>
infinisil: no idea how that is supposed to work
<NixOS_GitHub>
[nixpkgs] Mic92 pushed 2 new commits to master: https://git.io/vN2I5
<NixOS_GitHub>
nixpkgs/master 457bf6d Luke Adams: phantomjs2: set QT_QPA_PLATFORM to allow use in daemons...
phaebz has quit [Ping timeout: 246 seconds]
fragamus has joined #nixos
bpa has quit [Read error: Connection reset by peer]
bpa has joined #nixos
hotfuzz_ has joined #nixos
GiGa has quit [Remote host closed the connection]
hotfuzz has quit [Ping timeout: 264 seconds]
<jtojnar>
steveeJ: it does not make any sense
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Baughn opened pull request #34112: initrd-network: Document the need for modules (master...nix-docs) https://git.io/vN2L8
NixOS_GitHub has left #nixos [#nixos]
<josePhoenix>
I'd like to define my nginx config in a general way for a dev and prod configuration. My best idea of how to do this is to have an importable file containing a function that returns the config based on a hostname argument (e.g. "dev.foo" vs "foo")
<josePhoenix>
I don't really understand how names become available in the scope of an imported nix file. e.g. if I start with "{}:", will I have things like "services.nginx" in scope?
<steveeJ>
jtojnar: are you debugging gdm?
<Baughn>
josePhoenix: Sounds like a job for `nixops` (tm).
<Baughn>
josePhoenix: Though I don't have time to talk about it, here's a halfway converted example config: http://ix.io/EsN
<josePhoenix>
Is "imports = [ ... ]" treated specially? It appears in the default nixos configuration.nix to import hardware config, but I can't find it in the nix language manual
ssmike has quit [Ping timeout: 256 seconds]
ssmike1 is now known as ssmike
<vcunat>
josePhoenix: it's not a nix feature
<vcunat>
it's a nixos feature
phaebz` has quit [Quit: ERC (IRC client for Emacs 25.3.1)]
<myguidingstar>
actually my laptop's built-in wifi is broken so I have to buy that new wireless usb
<myguidingstar>
i did re-generate hardware.nix
fragamus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<hyper_ch>
myguidingstar: nixos is installed?
<myguidingstar>
but seems not to work
<myguidingstar>
yup, it's installed
<hyper_ch>
stable or untable=
<hyper_ch>
?
<MichaelRaskin>
And did you enable anything about firmware in the configuration?
<myguidingstar>
I've been using nixos happily for about a year
<steveeJ>
jtojnar: I mean a race within gdm
<myguidingstar>
stable, with some unstable package
<MichaelRaskin>
Then WiFi died, then USB WiFi doesn't work, right?
<steveeJ>
although I don't think it's very likely, since we just add a directory to the search path
<myguidingstar>
right
xcmw has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<MichaelRaskin>
Do you have any wired connection options for rebuild? I would try just rebuilding with hardware.enableAllFirmware = true; and rebooting.
<hyper_ch>
so 8188eu isn't in kernel yet?
<MichaelRaskin>
It is in 4.14
<MichaelRaskin>
In staging
<hyper_ch>
what does "staging" mean there?
<MichaelRaskin>
Quite likely it requires firmware
<MichaelRaskin>
Kernel driver staging area
<hyper_ch>
if he sets his configuration.nix to latest kernel should he get it?
<myguidingstar>
I have `hardware.enableRedistributableFirmware = true`
<MichaelRaskin>
A.k.a. we have no idea how this chip behaves in reality, but we have a driver that works in this specific room and we want more testing.
<hyper_ch>
modprobe: FATAL: Module 8188eu not found in directory /run/current-system/kernel-modules/lib/modules/4.14.14
<hyper_ch>
:(
<MichaelRaskin>
I would try to enable more firmware just in case. I am not sure that the corresponding firmware has a sane license.
<MichaelRaskin>
hyper_ch: it is r8188eu
burgdorfer has joined #nixos
<MichaelRaskin>
Hm
<hyper_ch>
ok, that works fine
<MichaelRaskin>
I am wrong
<MichaelRaskin>
NixOS non-redistributable firmware is just Broadcom
<hyper_ch>
MichaelRaskin: so adding boot.kernelPackages = pkgs.linuxPackages_latest; # use latest kernel to his configuration.nix should give him 4.14 kernel and the r8188eu mod
<MichaelRaskin>
Yes
<hyper_ch>
myguidingstar: add boot.kernelPackages = pkgs.linuxPackages_latest; # use latest kernel do you configuration.nix
<jtojnar>
steveeJ: asking on the GDM mailing list will probably the best option
<hyper_ch>
and then nixos-rebuild boot --upgrade (I guess)
<MichaelRaskin>
Ideally, one needs to create a package that adds that file to $out/lib/firmware/rtlwifi/ and then add this package to hardware.firmware NixOS option (which is a list of packages)
<hyper_ch>
myguidingstar: got it?
<myguidingstar>
hmm, I don't get it
<hyper_ch>
myguidingstar: edit your configuration.nix
bpa has quit [Read error: Connection reset by peer]
<hyper_ch>
myguidingstar: add to it this line: boot.kernelPackages = pkgs.linuxPackages_latest; # use latest kernel
<MichaelRaskin>
Simple way: put the file into /home/user/firmware/r8188eu/lib/firmware/rtlwifi and add ("" + /home/user/firmware/r8188eu) to hardware.firmware
<hyper_ch>
myguidingstar: then upgrade your nixos: nixos-rebuild boot --upgrade
bpa has joined #nixos
<myguidingstar>
I did have that line
deltasquared has joined #nixos
<hyper_ch>
oh, you are on 4.14 already?
<MichaelRaskin>
Wait what weird.
<myguidingstar>
Yes, since meltdown :)
<hyper_ch>
myguidingstar: uname -a
<hyper_ch>
what's the output?
<myguidingstar>
i'm currently in Ubuntu live usb for the internet
<hyper_ch>
can't you just plugin network cable?
<MichaelRaskin>
firmwareLinuxNonfree already comes with that firmware.
<myguidingstar>
no :')
<hyper_ch>
why can't you plugin network cable?
<MichaelRaskin>
If you have enableRedistributableFirmware set to true, you should have firmware already…
<MichaelRaskin>
I probably want a dmesg…
bpa has quit [Read error: Connection reset by peer]
bpa has joined #nixos
<MichaelRaskin>
hyper_ch: not everyone has even a spare ethernet cable lying around
<rcshm>
thanks @simpson
<MichaelRaskin>
Not to mention where physically the WiFi AP can be…
phaebz` has quit [Ping timeout: 264 seconds]
<hyper_ch>
doesn't need to be a spare one
<hyper_ch>
just got to a computer and use that one's :)
<deltasquared>
even so, some people just plain don't have the option of a wired setup.
<myguidingstar>
MichaelRaskin: so do i have to download the .bin file from github to my home dir?
<samueldr>
some laptops don't come with RJ45
<MichaelRaskin>
I think no.
<myguidingstar>
or is it available in nixstore?
<hyper_ch>
usb ethernet adapter :)
<samueldr>
sure, when you have one :)
<MichaelRaskin>
hyper_ch: which some people by some mistake happen not to have.
<deltasquared>
hyper_ch: can you magic one out of thin air for me? that would be nice :P
<MichaelRaskin>
samueldr: not really, when you have _any_
<MichaelRaskin>
Having five is not worse, trust me.
<samueldr>
but then, you need a switch that can accomodate 5 + upstream !
<samueldr>
/s
<hyper_ch>
deltasquared: that magic is called "3d printer" :)
<deltasquared>
omg a 3D printer that can make ASICs I want one
<MichaelRaskin>
hyper_ch: you have a 3D printer that can print the adapter part?
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] das-g opened pull request #34115: tablib: 0.10.0 -> 0.12.1, re-enable on Python 3 (master...tablib-0.12.1-cherrypicked-to-master) https://git.io/vN2YU
NixOS_GitHub has left #nixos [#nixos]
bpa has quit [Read error: Connection reset by peer]
<hyper_ch>
no :(
<MichaelRaskin>
OK, now _I_ am interested.
<MichaelRaskin>
Pity
bpa has joined #nixos
<MichaelRaskin>
myguidingstar: if there is no misunderstanding, you have the driver and the firmware already installed correctly
<hyper_ch>
but if you can get me to the 24th century I can brng back something much bette - a replicator and awesome borg algorithms :)
<vcunat>
is that not why we have printed circuit boards and 3D trasistors?
KingOfThePirates has quit [Ping timeout: 260 seconds]
<deltasquared>
hyper_ch: *sigh* bring back the ARG while you're at it
<MichaelRaskin>
Whether the driver gets loaded at all is a good question (you can try to modprobe it manually)
<rcshm>
simpson, i am specify haskell packages by using haskellPackages.ghcWithPackages. how do i specify the version with this?
<MichaelRaskin>
Then it might complain in dmesg…
<MichaelRaskin>
Do I understand correctly that the interface doesn't appear?
<rcshm>
it is in my config.nix file.
<myguidingstar>
okay, I'll boot my nixos and try...
<MichaelRaskin>
Get a dmesg, I guess
<MichaelRaskin>
Also cat /sys/class/rfkill/*/name
<hyper_ch>
also check what kernel you're booted into
<MichaelRaskin>
and try modprobe -r and then modprobe on the driver, yes
<MichaelRaskin>
(before dmesg)
<hyper_ch>
but weird, if he uses 1.14 kernel it should be working I think
myguidingstar has quit [Quit: Page closed]
nschoe has joined #nixos
<nschoe>
Hi again / back.
<hyper_ch>
hi nschoe
<MichaelRaskin>
How is your LUKS?
<nschoe>
So, first thanks for helping me earlier. I have confirmed the problem is that the Windows 10 update added a new partition.
<nschoe>
(nice ctahc samueldr for that:P)
<nschoe>
I have also confirmed that my configuration.nix says ´device = /dev/sda5´ for luks.
<hyper_ch>
can you still open the luks partition?
<nschoe>
I have ued the ¨boot.shell_on_fail¨ to give me a shell to be able to do thing. I have successfully opened /dev/sda6 (now ^^) and activated the LVM group on it.
ixxie has quit [Quit: Lost terminal]
<nschoe>
I could mount my files. That´s the good thing.
<nschoe>
The problem is that from that rescue shell, I do not have the ability to run ´nixos-install´, or rebuild or anything.
<hyper_ch>
that's what live iso is for
<MichaelRaskin>
You should be able to mount everything, then just exit, no?
<nschoe>
hyper_ch, yes, but I have tried this, and weidrly enough it also fails on stage 1 -_-
<hyper_ch>
(or what MichaelRaskin says)
<hyper_ch>
nschoe: well, no idea what you did in the live iso
<nschoe>
I do not know what happens. I have downloaded the ISO, booted the USB, selected ¨nixOS xxxxxxx (Installer)¨ and it fails when waiting for device I-don-t-remember-which to appear :/
<MichaelRaskin>
Stage 1 failure of ISO is probably a problem with how exactly it got copied onto USB
<nschoe>
MichaelRaskin, ah? I used Unetbootin
<MichaelRaskin>
Well, it could expect the partition with files to have a specific filesystem label
<MichaelRaskin>
Not sure if unetbootin handles this
<nschoe>
hyper_ch, yes: I have only windwos left at the moment
<nschoe>
MichaelRaskin, and if I go back to my install, open the luks device myself, activate the vg groups, I was wodnering if I could simply go on with stage 2.
<hyper_ch>
I wonder, could you use WSL and dd from there to the usb drive :)
<nschoe>
WSL?
<hyper_ch>
nschoe: get Rufus
<hyper_ch>
it has dd mode to write to usb
<MichaelRaskin>
nschoe you should activate and mount everything
<MichaelRaskin>
What happens when you exit after mounting everything?
<deltasquared>
MichaelRaskin: .... oh, it should start systemd shouldn't it?
<deltasquared>
or whatever init
<MichaelRaskin>
That should hopefully come after the switch.
<nschoe>
MichaelRaskin, well I CTRL + D when I could not find the stage2 init script, and then it rebooted normally.
<MichaelRaskin>
Did you press i or f to get the shell?
<nschoe>
MichaelRaskin, f
<hyper_ch>
nschoe: just select from the ISO dropdown the DD
<nschoe>
hyper_ch, I am DLing it, yes.
<nschoe>
hyper_ch, thanks.
<MichaelRaskin>
Try with i
<hyper_ch>
nschoe: :)
<MichaelRaskin>
Basically, f means there is nowhere to return but you can launch init yourself
<MichaelRaskin>
i means that stage1 one can continue after you exit the shell, but you must let it launch init
<nschoe>
MichaelRaskin, oh okay.
<hyper_ch>
try first MichaelRaskin's way :)
<MichaelRaskin>
Probably if it just fails LVM, you should unlock/activate and exit
<nschoe>
MichaelRaskin, so I unlock my lucks, activate the vg group and exit? Or should I also mount the root, home, etc partition after activatuing the vg group?
<nschoe>
Iḿ asking because when I mount the root partition on /mnt/root it works by tells me there is a problem with ext 3 and everything is there but the content of /mnt/root/boot .
<samueldr>
(never tried it with nixos, but looking at the source it looked like it)
<MichaelRaskin>
samueldr: i and f are different, after f there is nothing to continue
<nschoe>
MichaelRaskin, I will try this then, thanks.
<MichaelRaskin>
Well, unclean unmount isn't deadly, but there might be some inconsistency
<nschoe>
I´m trying this right now, rebooting. See you in a few (hopefully ^^)
nschoe has quit [Quit: Leaving]
<hyper_ch>
(there's a reason why I never liked lvm)
<deltasquared>
hyper_ch: why, pita to get going from a rescue shell?
phaebz has quit [Remote host closed the connection]
<hyper_ch>
deltasquared: my old setup was mdadm raid1 -> luks/dm-crypt -> ext4 .... lvm would just have added another layer of complexity.... now I just use native root encrypted zfs in mirror mode
<deltasquared>
hmm. btrfs on luks on LVM here
<hyper_ch>
I always thought lvm was more hassle than benefit
<hyper_ch>
btrfs... ieeeks
<deltasquared>
interesting, it only in general made my life easier
<hyper_ch>
I manage to crash btrfs so easily :)
<hyper_ch>
and rebalancing nightmares
* deltasquared
probably should rebalance more often
<MichaelRaskin>
It got harder to crash after some time
<deltasquared>
I generally just ensure I have a fair bit of free space on /
<MichaelRaskin>
I just resize /nix/store down and up
<hyper_ch>
usually within 48h of operation under my care, btrfs spits out the first errors
<deltasquared>
hyper_ch: you must have looked at it funny
<deltasquared>
btrfs has been the one FS that's never given me any trouble.
<hyper_ch>
don't worry, it'll also get yu eventually ;)
<deltasquared>
(bar crashes in the middle of updates on an "imperative" distro, but...)
<deltasquared>
in all fairness I had been meaning to migrate to ext4 anyway, btrfs is a tad seek happy
<deltasquared>
not good on less than spectacular spinning rust storage
<MichaelRaskin>
If ext4 had online downsizing and online changing of inode count…
<hyper_ch>
I still think ext4 is a very good fs
zarel has quit [Quit: Leaving]
<deltasquared>
MichaelRaskin: AFAIK btrfs doesn't have visible inode limits anyway
<hyper_ch>
it has invisible ones?
<MichaelRaskin>
Well, there are metadata blocks…
<deltasquared>
hyper_ch: no invisible ones either but don't quote me on that
nschoe has joined #nixos
<nschoe>
Hello again ^^
<MichaelRaskin>
deltasquared: ext4 has static inode limits set on mkfs
<deltasquared>
MichaelRaskin: it's a tree structure everywhere, preallocation of structures like inodes doesn't really make sense.
<deltasquared>
for btrfs anyway
<hyper_ch>
nschoe: now change to uuid in your hardware-configuration and rebuild :)
<nschoe>
MichaelRaskin, so I tried with i instead of d.
<nschoe>
instyead of ¨f¨ sorry.
<deltasquared>
and yes I'm aware of ext4 having it predetermined
<nschoe>
I unlocked luks and activated the vg group then CTRL + D to exit. And it kernal panic-ed
<nschoe>
:/
<hyper_ch>
:(
<deltasquared>
nschoe: "tried to kill init"?
<MichaelRaskin>
Well, you could try f and finding the target init yourself.
<nschoe>
But I tried the USB key burnt with Rufus, and it worked. So I guess I could still do the same thing, then mount on /mnt and run nixos-onstall --no-bootloader.
<MichaelRaskin>
Should be in /nix/var/nix/profiles/system/ somewhere (directly or /bin)
<nschoe>
deltasquared, sorry what?
<deltasquared>
nschoe: the kernel panic, was the message to the effect of "tried to kill init" ?
<nschoe>
deltasquared, hum possibly. I just tried to exit the interactive shell, did not look that closely.
<nschoe>
Hum gotta go for a few minutes. Sorry I´ll be back. Thanks __a lot__ for all the help guys :)
nschoe has quit [Client Quit]
<deltasquared>
in and out this guy...
<hyper_ch>
(1) boot live cd (2) mount everything as you should but in /mnt instead of / [hence, /mnt/boot etc....] (3) make a backup copy of your /mnt/etc/nixos/hardware-configuration.nix (4) run nixos-generate-config --root "/mnt" (5) check the hardware-configuration.nix whether all is fine (6) run nixos-install
<deltasquared>
hyper_ch: he's gone. :/
<deltasquared>
or she. they. whatever applicable.
<gchristensen>
it is tough to debug those problems especially if they have one computer, and they may not be a guy
<hyper_ch>
deltasquared: well, I hide join/part since it's too annoying
<hyper_ch>
deltasquared: or "it"
<deltasquared>
gchristensen: yes I just corrected that :P
<deltasquared>
I was thinking of getting an rpi just for a low power machine I could leave attached to IRC while I rebooted this laptop for nixos things
<samueldr>
looking at the source it f) execs, so exitting definitely kill PID1, but i) should allow exitting to continue
* samueldr
starts a VM
periklis has quit [Ping timeout: 246 seconds]
<hyper_ch>
samueldr: why?
<samueldr>
to look at the behaviour of fail_on_shell
<hyper_ch>
:)
<samueldr>
though, weirdly I thoughg it was a hardware fault on my spare computer, but no, when `debug1` is used it seems I can't choose any option :(
<deltasquared>
well, my efforts to package up my firewall thing continue... https://ptpb.pw/AuDZ is what I have thus far, though I can't seem to get the dependencies to work correctly.
<samueldr>
debug1devices though works
<deltasquared>
(the builder at this point is effectively /bin/false)
<samueldr>
with i), once I ^D (exit) it continues booting at the point it stops MichaelRaskin
<deltasquared>
... oh wait
<deltasquared>
I'm an idiot, I didn't even reference the deps array I made. d'oh
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] zimbatm closed pull request #34098: gpp: init at 2.25 and add @nmattia as maintainer (master...nm/add-gpp) https://git.io/vNgFY
NixOS_GitHub has left #nixos [#nixos]
<samueldr>
ah, re-read what you said and what you said does agree, sorry
<deltasquared>
https://ptpb.pw/S-KF ok *now* I can move onto the next part it seems...
<samueldr>
reading comprehension is hard ;)
<deltasquared>
can confirm
<hyper_ch>
kitty wants something... she patroling around my legs
<deltasquared>
food?
<hyper_ch>
there's plenty
<deltasquared>
"but human, I already had that flavour today!"
ixxie has joined #nixos
<hyper_ch>
:)
<deltasquared>
cats are simultaneosly lovely little things and diabolical little shits
bpa has quit [Ping timeout: 246 seconds]
<Rusty1_>
dogs have masters, cats have subjects
<hyper_ch>
well, kitty thinks the apartment is hers and I'm just tolerated because I provide food and pet her on command
<deltasquared>
ok, I've got this WIP nix expression for that firewall thing thus far and the next bit I need help with is making the systemd units visible.
<deltasquared>
I have no idea how systemd units are "collected" together in a given nix profile.
<hyper_ch>
(usually I just think it's some form of advanced automagic)
<vcunat>
it's a real kitty, right? (not how you call your girlfriend or such :-)
{^_^} has quit [Remote host closed the connection]
{^_^} has joined #nixos
<hyper_ch>
myguidingstar isn't back yet
<deltasquared>
hyper_ch: with regards to the systemd units, just looking at how nix itself does it for it's own expression, and there indeed is no special mention of them in the expression. I would check to see if the make install stage is copying them, but I suspect I won't be able to read autotools.
<hyper_ch>
deltasquared: I create some systemd unit thingies for watch folders
<deltasquared>
hyper_ch: you have that described in a nix expression? I'd like to see if possible
<hyper_ch>
systemd watchfolder notices and runs script for ocr and puts it at the final location
<deltasquared>
hyper_ch: interesting, though not what I need. I already have the systemd units defined in the sources, just need to know how to make them visible to the rest of the system
<hyper_ch>
what do you mean by that?
xcmw has joined #nixos
<hyper_ch>
I used incron before on debian but incron is discontinued
<deltasquared>
hyper_ch: like, I don't understand how I can get the systemd units in the resulting derivation to appear in the "merge" directory for the profile.
<deltasquared>
sure I can copy them there but they'll just be stuck in the store, not visible anywhere else
<NixOS_GitHub>
[nixpkgs] vcunat pushed 1 new commit to release-17.09: https://git.io/vN2Za
jb55 has quit [Ping timeout: 256 seconds]
<Myrl-saki>
vcunat: Stupid hypothetical example. What if I want to go source-based on my distro and use -march=native?
<Myrl-saki>
vcunat: s/distro/machine/
<vcunat>
you can
<Myrl-saki>
vcunat: What flags should I set on my configuration.nix?
<vcunat>
coming from gentoo? :-)
<vcunat>
You want to override the stdenv - that might be problematic, but in the worst case you can just have a simple patch atop nixpkgs.
<vcunat>
Set NIX_ENFORCE_NO_NATIVE, and add -march=native to the default flags.
<vcunat>
but I don't think it's worth doing this system-wide, normally
<Myrl-saki>
vcunat: Used to Gentoo but got tired of it. :P Though, this does have a lot of inspiration from Gentoo.
<vcunat>
Maybe specifically for some package that you use ultra-heavily on CPU.
<Myrl-saki>
vcunat: True. That's pretty much number 1 reason why I no longer use Gentoo. Whole-day rebuilds are not fun.
<MichaelRaskin>
Back when Hydra was smaller, it was multi-day rebuilds for a large NixOS configuration…
<vcunat>
Well, with nix you have "transactional" switches and rollbacks.
<vcunat>
so it seems not as painful as my old experience with Gentoo.
<MichaelRaskin>
Also isolation
<Myrl-saki>
vcunat: What does transactional here refer to?
<MichaelRaskin>
Which means that unupgradeable combinations are harder to create
<Myrl-saki>
Ah
<vcunat>
you build the whole new system, and then you e.g. reboot into it
<vcunat>
you don't replace libs "in-place", etc.
<Myrl-saki>
A little aside. 2 things I learned this past few weeks are hardeningDisable and NIX_ENFORCE_NO_NATIVE. :P
myguidingstar has joined #nixos
<Myrl-saki>
Thanks. I really appreciate the help.
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
nixpkgs/master 58e7823 Michael Weiss: Merge pull request #33931 from dotlambda/multidict...
<NixOS_GitHub>
[nixpkgs] primeos pushed 2 new commits to master: https://git.io/vN2ZA
<NixOS_GitHub>
nixpkgs/master 2e99541 Robert Schütz: pythonPackages.multidict: 3.3.2 -> 4.0.0
NixOS_GitHub has left #nixos [#nixos]
<MichaelRaskin>
But in NixOS land, we do not even aim for maximum efficiency.
<Myrl-saki>
MichaelRaskin: -O2
<Myrl-saki>
:P
<MichaelRaskin>
Slower and less optimised builds for better reproducibility? We-ell, what are the numbers? Maybe?
<Myrl-saki>
MichaelRaskin: True.
<MichaelRaskin>
Use more HDD to save a bit of thinking? Yes. Just yes.
<Myrl-saki>
Hmmm
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] brown131 opened pull request #34116: Update hy to latest version (master...patch-1) https://git.io/vN2Zj
NixOS_GitHub has left #nixos [#nixos]
<Myrl-saki>
Wow. That just made me realize a lot of things.
<MichaelRaskin>
A typical NixOS system goes through a stage where it needs _twice_ its store footprint, because glibc upgrade.
<MichaelRaskin>
(You just build all the new stuff, then reboot, check everything is fine, then at some point GC the old stuff)
<MichaelRaskin>
This is also a useful fact when planning the partition layout…
<Myrl-saki>
MichaelRaskin: Wouldn't this mean that Gentoo really is effective vs binary-based systems? Take vectorization for example.
<MichaelRaskin>
Well, x86_64 has some vectorisation as the baseline.
<Myrl-saki>
MichaelRaskin: Right, sse2
<vcunat>
yes, IIRC
<MichaelRaskin>
I think a bit more.
<vcunat>
Packages where this matters contain multiple variants anyway and switch during runtime.
<vcunat>
(That's what upstream does already.)
<slyfox>
'gcc -Q --help=target'
<Myrl-saki>
vcunat: Oh. Now it makes sense why gcc has a runtime feature query.
<myguidingstar>
MichaelRaskin, hyper_ch fyi, turns out that network manager messed up with the two wifi devices. All wireless network profiles have been associated with the broken device and they don't show up in network manager's menu
<myguidingstar>
I had to fix them with nmcli
<MichaelRaskin>
Ah.
<myguidingstar>
the reason I thought the driver wasn't loaded properly was that when `lsusb` the new device doesn't have a name
<MichaelRaskin>
Argh.
<MichaelRaskin>
Generally I start with ip link
<MichaelRaskin>
But of course, after that I just launch wpa_supplican directly, that cuts out some possible levels of confusion
leat1 has quit [Quit: WeeChat 1.9.1]
<myguidingstar>
thank you so much for your help
phaebz` has joined #nixos
<MichaelRaskin>
Unfortunately, it was all in the wrong direction.
<Myrl-saki>
Err
<Myrl-saki>
How do I clean gcroots again?
<Myrl-saki>
Do I delete the pointed file, or can I delete the symlinks themselves?
<MichaelRaskin>
You delete symlinks, there is a series of symlinks gcroots → your symlink → actual something, if Nix cannot find the target of indirect gcroot, it just removes the root
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] primeos pushed 1 new commit to master: https://git.io/vN2nb
<NixOS_GitHub>
nixpkgs/master fbe1770 Michael Weiss: gns3Packages.server{Stable,Preview}: Unlock the multidict version
NixOS_GitHub has left #nixos [#nixos]
<Myrl-saki>
MichaelRaskin: Which symlink to delete?
<Myrl-saki>
`for i in $(sudo readlink -f *); do sudo rm -rf $i; done` is what I did
nschoe has joined #nixos
<Myrl-saki>
-r was a bad idea
<nschoe>
Hello again -_-
<MichaelRaskin>
You have built something and have a result (or whatever was passed to -o) symlink, right?
<nschoe>
Sorry to come and bother _again_.
<Myrl-saki>
MichaelRaskin: Yeah
<nschoe>
Concerning my LUKS problem, it seems I keep having bumps in the road.
<MichaelRaskin>
Remove the result symlink, the rest will be done by Nix
<myguidingstar>
actually I've learned useful things from your help. commands in context is easier to get than reading man pages alone :)
<Myrl-saki>
MichaelRaskin: What if I don't know where the result symlinks are?
<Myrl-saki>
MichaelRaskin: readlink the gcroot?
<nschoe>
MichaelRaskin, I tried exiting both from shell started with ¨i¨ and ¨f¨ and the result is the same: a kernal panic with ¨tried to kill init¨.
<MichaelRaskin>
Oh well. Then you can go to gcroot and readlink
<Myrl-saki>
MichaelRaskin: Thanks
<Myrl-saki>
I'll be rebooting brb
<MichaelRaskin>
People recommend something more advanced. debug1devices and i there.
ylwghst has quit [Ping timeout: 246 seconds]
<MichaelRaskin>
At least for some, exiting from the shell continues the boot
<nschoe>
hyper_ch, as for trying with the nixos installer, I can start the installer, I can unlock the LUKS device, I can activate the LVM vg and mount the partitions into /mnt and /mnt/home. That works. BUT when I try to run ¨nixos-install --no-bootloader¨, it immediately fails with ¨/home/nschoe/.config/nginx.conf : file not found¨, because the first ntry in my configureation.nix is how to isntall / configure nginx and it feres t
<nschoe>
o a file that I have placed there.
<MichaelRaskin>
Or maybe you should try switch_root manually (in the f mode)
<Myrl-saki>
MichaelRaskin: Some of my auto gcroots point inside /nix/store, what does that mean?
Itkovian has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<MichaelRaskin>
Do not touch these
<nschoe>
The thing is: I thought ¨nixos-intall¨ was doing a chroot.
<Myrl-saki>
MichaelRaskin: Mkay.
<Myrl-saki>
Brb, rebooting
<nschoe>
MichaelRaskin, ah! it´s good you mentionned ¨switch_root¨ (don´t know if it was for me or not ^^), because when I run ¨/init¨ manually to try and continue stage 1 and stage 2, it fails with a message about ¨switch_root¨.
myguidingstar has quit [Ping timeout: 260 seconds]
<MichaelRaskin>
To you, yes, you are the only one on the channel doing advanced boot debugging right now
<nschoe>
Okay :/
<nschoe>
MichaelRaskin, so what is this ¨switch_root¨ thing?
<MichaelRaskin>
switch_root is a busybox command, which is like chroot and pivot_root, but scarier
<nschoe>
hum ok
<MichaelRaskin>
It removes everything on initramfs (scary part), then basically chroot's but better into the target directory (you want to say /mnt/root here) and then executes the second parameter (by exec, so PID1 is passed)
<nschoe>
MichaelRaskin, because what I tried next in my interactive shell is to change the ¨sda5¨ to ¨sda6¨ in the /init script. So that it would try to unlock the LUKS device on the proper partition. Then I executed the /init script manually. And it works: it does ask for the LUKS passphrase and manages to unlock it, which is good. But then it fails with ¨swithc_root¨. And since I saw something like ¨delete everything in /¨ I
<nschoe>
decided not to try it myself ^^
Myrl-saki has quit [Ping timeout: 256 seconds]
<MichaelRaskin>
yeah, deleting everything in initramfs is what has to be done
<MichaelRaskin>
Not really, but otherwise you have memory occupied by inaccessible stuff
<nschoe>
Okay. Any idea why it fails? I just sed the /nix/store/hash-stage1.sh file to replaced sda5 by sda6. So I expected this to continue booting normally after that.
<MichaelRaskin>
Did you run or exec the file
<nschoe>
MichaelRaskin, run, with ¨./init¨.
<MichaelRaskin>
You need to exec it, because if it doesn't have PID1, it cannot switch_root
<nschoe>
But from your remark earlie rand yoru question... I suppose I should `exec` it. From the f) shell.
<nschoe>
Okay
<nschoe>
Well thanks _again_. I feel like itś going to work this time ^^
<nschoe>
And I learned to double check that I never, ever use /dev/sdaX numbers anymore ^^
<MichaelRaskin>
Yes
<nschoe>
Well I am going to reboot again and see it this works
<nschoe>
Thanks a lot MichaelRaskin for sticking with me and my annoyoing questions :/
nschoe has quit [Quit: Leaving]
Myrl-saki has joined #nixos
<Myrl-saki>
Welps. For the heck of it, I want to use -march=native on my laptop.
<MichaelRaskin>
The best part is that I actually stumble in the dark a bit, as I read the stage-1-init.sh, but I don't use it as I just write my own nowadays…
<samueldr>
ah, missed nschoe
<Myrl-saki>
vcunat: If you're not busy, can you give me a simple runthrough?
<samueldr>
weird that i) KPs on exit... ah, unless they mounted the volumes at unexpected locations...
<samueldr>
(though, haven't tried it in the same conditions, I tried it under controlled and forced conditions in a VM)
nschoe has joined #nixos
Itkovian has joined #nixos
<nschoe>
MichaelRaskin, it worked! I'm on my system now, currently rebuilding nixos with my configuration.nix updated.
<nschoe>
MichaelRaskin, Thanks so much, the 'exec' did it.
<MichaelRaskin>
I wrote just after you had left: The best part is that I actually stumble in the dark a bit, as I read the stage-1-init.sh, but I don't use it as I just write my own nowadays…
<Myrl-saki>
ugh
<Myrl-saki>
changed my mind. I have libreoffice
<nschoe>
MichaelRaskin, what? you write your own stage 1 init files?
<MichaelRaskin>
Myrl-saki: imagine debugging its major updates with slight build system changes!
<MichaelRaskin>
nschoe: I do
<nschoe>
MichaelRaskin, okay, well it's a league I will never be in :/
<nschoe>
MichaelRaskin, I'm curious to know: why?
<samueldr>
MichaelRaskin: curious about your specific use case for a custom stage1
<samueldr>
:D
<nschoe>
samueldr, thanks for helping too by the way. It's still compiling, but I feel like it's a solved problem for me now :)
<MichaelRaskin>
At some level of knowledge of low-level details and also specificity of desires, it turns out to be simpler to write my own stage 1 and stage 2 than prevent systemd from messing everything up.
<samueldr>
(I'm about to look into part of it to allow BOTH keyfile and passsphrase to be used for unlocking LUKS)
<MichaelRaskin>
Also, I want to mkfs /tmp on boot. It is _possible_ but very annoying with NixOS.
<samueldr>
so you're using an on-disk filesystem for /tmp, but always a new one?
<vcunat>
Myrl-saki: that link is for particular packages. For applying it to everything, you can e.g. inspect `git grep hardeningCFlags`
<vcunat>
`
<nschoe>
samueldr, (then you store the keyfile on an external USB stick, right?)
<samueldr>
symphorien: \o/ should've talked about it sooner
<MichaelRaskin>
samueldr: yes
<symphorien>
I just read the conversation :)
<nschoe>
MichaelRaskin, okay, well looks awesome.
<samueldr>
(I mean, I looked about a week ago into decrypting with a key and saw it didn't fallback gracefully)
<Myrl-saki>
vcunat: Thanks!
<MichaelRaskin>
samueldr: I sometimes do large builds. If there is enough RAM, the build effectively happens in the file cache.
<MichaelRaskin>
If there isn't, my system doesn't go to swap completely
<samueldr>
MichaelRaskin: oh, I didn't poo-poo the idea, I just found it interesting to create a FS instead of cleaning in another way (rm mostly)
<MichaelRaskin>
rm is a disaster.
<MichaelRaskin>
(for this use case)
<samueldr>
details?
<MichaelRaskin>
Imagine leftovers from LibreOffice build. And Chromium.
* samueldr
shudders
<MichaelRaskin>
mkfs is actually faster than walking these
<samueldr>
yes I see now :)
<infinisil>
What the hell, why is ping = /run/wrappers/bin/ping for me
ssmike has quit [Remote host closed the connection]
<samueldr>
suid wrapper?
<infinisil>
never seen that before and it has me concerned
nschoe has quit [Remote host closed the connection]
<gchristensen>
wht are you concerned aout?
<MichaelRaskin>
At some point NixOS kept setuid wrappers in /var/setuid-wrappers/, later these moved to /run/wrappers/bin
<infinisil>
it's odd
<infinisil>
oh, setuid stuff, haven't looked into that
<MichaelRaskin>
gchristensen: without _knowing_ that's what NixOS does, it _is_ concerning to have ping in a weird place.
<gchristensen>
yes I understand, just wanted to know specifically what their concern was :)
<infinisil>
my problem is that I wanted to use ${pkgs.inetutils}/bin/ping
<MichaelRaskin>
As I don't use literally NixOS, I actually have _both_ these places symlinkning to the real wrapper location, so I can put off fixing the old scripts indefinitely
<samueldr>
paths in store can't be SUID iirc, which makes it harder for some tools like ping
<infinisil>
but that now doesn't seem to work because "Lacking privilege for raw socket"
<samueldr>
yeah, ping in linux-land has to somehow get elevated privileges since it uses raw sockets
<gchristensen>
infinisil: it is because ping requires root
b has joined #nixos
<infinisil>
So should I just hardcode /run/wrappers/bin/ping ??
jensens has joined #nixos
<gchristensen>
that is common, yes
Itkovian has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<infinisil>
oh god
<gchristensen>
or depend on your PATH
<gchristensen>
or run it as root or a user with the appropriate capabilities
<MichaelRaskin>
Then NixOS changes the wrapper location and you will need to change the hardcoded paths!
<infinisil>
exactly ^^
anton__ has quit [Remote host closed the connection]
<MichaelRaskin>
Which is why on my not-quite-NixOS system I have both old and new locations as symlinks to the real place
Itkovian has joined #nixos
<samueldr>
your setup seems interesting MichaelRaskin
<MichaelRaskin>
And I am always iterating it.
<MichaelRaskin>
Plan is to clean up and write up the next version for European Lisp Symposium.
<infinisil>
MichaelRaskin: what's the 'real place' if it can't be in the nix store?
<gchristensen>
samueldr: MichaelRaskin is an interesting fella
Itkovian has quit [Client Quit]
<MichaelRaskin>
It is /var/setuid-wrapper-storage/[hash-of-setuid-program-list]-setuid-programs/
<MichaelRaskin>
Hash adressed, because the system instance has a symlink to _its own_ wrapper set.
<MichaelRaskin>
And /run/wrappers/bin links through current-system
<infinisil>
i don't see any symlinking going on in /run/wrappers/bin
<MichaelRaskin>
The point is: I want _actual_ atomic FS switches. So replacing /var/current-system is the single operation that makes all the new stuff available at once.
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] flokli opened pull request #34117: bpython: fetch from PyPi (master...bpython-pypi) https://git.io/vN2W4
NixOS_GitHub has left #nixos [#nixos]
fyuuri has joined #nixos
<Myrl-saki>
Wow, my /tmp folder is 3.6GB large.
<samueldr>
infinisil: /run/wrappers/bin itself is symlinked
aarvar has quit [Ping timeout: 246 seconds]
<MichaelRaskin>
No, in my case /run/wrappers/bin -> /var/current-system/setuid/wrappers -> /var/setuid-wrapper-storage/fx2yp878306h7dvhry5szpr0gbilwhp1-setuid-programs
<infinisil>
OH in your case, right
<MichaelRaskin>
NixOS does it Obviously Wrong.
<MichaelRaskin>
(because two parts of the PATH get updated one after the other)
<infinisil>
yeah
<infinisil>
that seems definitely something nixos could be improved on
<MichaelRaskin>
Of course, there is a need to do some scripting around resolv.conf
<simpson>
rcshm: Sorry, I don't do PMs and I also don't do much Haskell on Nix. I just wanted to point you to the right corner of the manual.
ma27 has quit [Quit: WeeChat 2.0]
<MichaelRaskin>
I actually want mutable users and read-only /etc proper. That creates an interesting combo, but fortunately _my_ design _assumes_ that I do want to write a complete UID-0 daemon that manages policy, and can update /etc/passwd in a slightly convoluted way if needed.
<MichaelRaskin>
I don't know what would be a good solution for NixOS and /etc; my current solution with FUSE is almost acceptable, but sometimes annoying
<infinisil>
MichaelRaskin: you mean read-only /etc as in a symlink /etc -> /nix/store/<hash>-etc ?
ssmike has joined #nixos
dan_b has joined #nixos
<MichaelRaskin>
Of course, /etc -> /var/current-system/global/etc
<MichaelRaskin>
And that later thing goes into store, yes
<infinisil>
ah right
ssmike1 has joined #nixos
<MichaelRaskin>
Speaking of atomic swiches: it is nice that ln -sfT replaces the symlink with rename and not by deleting the previous one; but using mv -fT is more explicit…
<NixOS_GitHub>
[nixpkgs] grahamc pushed 2 new commits to release-17.09: https://git.io/vN2ll
<NixOS_GitHub>
nixpkgs/release-17.09 013662f Graham Christensen: Merge pull request #34056 from bachp/minio-security2-stable...
NixOS_GitHub has left #nixos [#nixos]
anton__ has joined #nixos
jensens has quit [Ping timeout: 256 seconds]
dywedir has quit [Remote host closed the connection]
dywedir has joined #nixos
ambro718 has joined #nixos
ssmike1 has joined #nixos
alex`` has quit [Ping timeout: 256 seconds]
ssmike has quit [Ping timeout: 246 seconds]
ssmike1 is now known as ssmike
jsgrant_ has joined #nixos
dywedir has quit [Read error: Connection reset by peer]
<Myrl-saki>
Wow. I updated to the latest nixos-17.09
<Myrl-saki>
And 2 GB update.
<Myrl-saki>
Welps
<MichaelRaskin>
Only 2GiB?
* jsgrant_
is building rustc and wowzah it's a biggy.
<Myrl-saki>
Since we're at this point, might as well do the -march=native.
<gchristensen>
fwiw NixOS requires much more recompiling than gentoo
<Myrl-saki>
gchristensen: Why? ; n ;
<gchristensen>
well if python changes in gentoo you just recompile python, here you recompile everything that depends upon python
<Myrl-saki>
Ahhh
<jsgrant_>
So, I set up a 'redox os partition'; Want to mess with building there installer for my sdb -- would it be trivial to chroot to run from there? Want to eventually package some of this stuff.
<jsgrant_>
their installer*
knupfer has quit [Ping timeout: 246 seconds]
<jsgrant_>
Was going to go near-full NixOS; But setting myself up to fail again -- repurposing this old Thinkpad as a dedicated toy/idealistic development box.
goibhniu has joined #nixos
<pmeunier>
How is containerisation handled in hydra? Can I setup an "open hydra server" for an organisation, and let anyone run jobs in there? would that cause security problems?
<MichaelRaskin>
Well, fixed-output derivations have arbitrary network access during the build…
kolb has joined #nixos
<kolb>
is there an option to see a diff between generations?
<jsgrant_>
Looks like I can go full FHS Chroot actually; Seems like the most trivial way to run un-nix'd binaries.
<gchristensen>
pmeunier: run with sandboxing and your only serious concern is what MichaelRaskin's noted, I think
taktoa has joined #nixos
<pmeunier>
MichaelRaskin: that's pretty bad alright, but I guess any CI service (Travis?) would do the same.
<gchristensen>
pmeunier: beyond other risks like Spectre of course :)
<symphorien>
kolb: there is a tool called `nix-diff` to do that
jb55 has joined #nixos
<pmeunier>
gchristensen: ok, thanks. how does sandboxing work internally? how does it prevent jobs to access unwanted parts of the system?
<MichaelRaskin>
pmeunier: Well, all this depends on what you call security problem.
kreetx has quit [Ping timeout: 268 seconds]
<pmeunier>
sure, and thanks a lot for pointing that out :-)
<gchristensen>
it blocks network access and access to the filesystem other than what is declared as an input to the job, iirc
<Myrl-saki>
Well
<Myrl-saki>
Wish me luck, I guess.
<MichaelRaskin>
There (sometimes — these get patched in the Linux kernel relatively soon) are sandboxing escapes via user namespaces, which are allowed on NixOS …
infinisil has quit [Quit: ZNC 1.6.5 - http://znc.in]
<Myrl-saki>
MichaelRaskin: Also, I update only like once every 2 months, so I dun have too many problems... I hope. :x
vcunat has quit [Quit: Leaving.]
infinisil has joined #nixos
<pmeunier>
MichaelRaskin: I wasn't aware of the existence of linux sandboxing, hence my question. Thanks!
<pmeunier>
(reading about it now)
<MichaelRaskin>
Well, a lot of that is based on namespaces, the same as containers only even more granular.
dywedir has joined #nixos
nuncanada2 has joined #nixos
<lejonet>
some throw in some seccomp filtering too
kline has quit [Quit: K-Lined]
<gchristensen>
it does some additional filtering as well, pmeunier, like it is imposible to make setuid binaries
<gchristensen>
but you absolutely should run it with (1) multi-user nix, ie: run NixOS proper, (2) enable sandboxing
<MichaelRaskin>
I would say «e.g. NixOS proper», not «i.e.»
<gchristensen>
sure
<MichaelRaskin>
But multi-user and with sandboxing, and NixOS with nix.useSandbox = true; is the simplest way to get that
nuncanada has quit [Ping timeout: 268 seconds]
<lejonet>
I'm seriously thinking about putting nixos on my laptop too, even tho I hardly use it nowdays, because it would force me to do proper setup with containers and such for all the bs I run :P
<jsgrant_>
lejonet: The more I mess with NixOS, the more I realize how many dirty-hacks I have lying around on the rest of my boxen.
<lejonet>
jsgrant_: that is kindof what I've realized too xD
nuncanada2 has quit [Ping timeout: 264 seconds]
<lejonet>
especially as I do some CTFing and such
cfricke has quit [Quit: WeeChat 2.0]
<jsgrant_>
But yeah, that initial transition friction is hard to overcome & as I've said in here earlier and a few times before ... I've set myself up to fail, a number of times trying to 'go all in'.
<lejonet>
Atleast nowdays I confine them to their own directory, but I should just make a container I do everything in... :P
<lejonet>
I've started to convert my servers firstly, so I can fool around with nixops and learn more of the details of nixops before I go nuts with a desktop/laptop system :P
ambro718 has quit [Ping timeout: 248 seconds]
<jsgrant_>
Messing with flatpak, I do wish there was trivial way (maybe I'm just not aware of) that somehow redirects/forces local program files (configs & share, etc) to a given directory and not cluster /home
<MichaelRaskin>
I think nixos.org needs to create Terms of Service that makes it obligatory for website users to keep Nix expressions nuder version control.
<gchristensen>
there is some lib that does that, but I don't remember the name.
<lejonet>
xD
<gchristensen>
MichaelRaskin: oh?
<MichaelRaskin>
In license or in ToS?
<jsgrant_>
gchristensen: Yeah, I feel like some solution of some-level-of-completeness must exist somewhere.
<gchristensen>
re: jsgrant's comment
nuncanada has joined #nixos
jb55 has quit [Quit: WeeChat 1.9]
jb55 has joined #nixos
<MichaelRaskin>
gchristensen: well, NixOS tries to avoid a few kinds of messes, it also makes versioning the config _easy_, not versioning config is a mess…
<jsgrant_>
Certainly a nice side-effect of flatpak & I assume other containers; That I really did think of prior.
<lejonet>
MichaelRaskin: indeed :)
anton__ has quit [Ping timeout: 246 seconds]
<gchristensen>
aye, yeah
<lejonet>
jsgrant_: that is kindof how I was thinking of using rkt, but I've not had the time to sit with that lately
<jsgrant_>
Does your system-hostname drv that gets generated in every rebuild store your latest config at the time of building?
<lejonet>
seeing as the container for rkt is just a tarball of the rootfs and a json manifest, nothing more special than that
hiratara has quit [Ping timeout: 255 seconds]
anton__ has joined #nixos
<MichaelRaskin>
At least NixOS manual should outright say it as the only sensible option: install some VCS and put the configuration.nix under version control.
ssmike has quit [Quit: ssmike]
kl1n3 has joined #nixos
<kolb>
symphorien: ok nix-diff diffs derivations, and each generation is a derivation?
hiratara has joined #nixos
<kolb>
not sure what to pass as nix-diffs arguments
<symphorien>
kolb: for example: nix-store --query --deriver /nix/var/nix/profiles/system-360-link
<symphorien>
will give you the path to the .drv file you should pass to nix-diff iirc
<Myrl-saki>
Stupid question. How do I use dhcpcd on other networks but a certain IP on a certain SSID?
<symphorien>
probably not the expected answer, but network-manager ?
<Myrl-saki>
symphorien: Let me check.
<Myrl-saki>
symphorien: Is there a way to declarative set the config through Nix?
<symphorien>
not that I know, also it is some ini-style syntax files in /etc
<Myrl-saki>
symphorien: Mkay. Thanks.
hiroshi has quit [Ping timeout: 276 seconds]
<lejonet>
there is always lib.generators.toINI that you can use to generate the config from an nix attrset :D
asuryawanshi has joined #nixos
kline- has joined #nixos
kl1n3 has quit [Ping timeout: 600 seconds]
<Myrl-saki>
lejonet: Interesting. Thanks! I'll give it a look.
hiroshi has joined #nixos
kline- is now known as kline
pukkamustard has joined #nixos
JosW has quit [Quit: Konversation terminated!]
asuryawanshi has quit [Ping timeout: 246 seconds]
pukkamustard has quit [Client Quit]
xcmw has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<Myrl-saki>
I added this +hardeningCFlags=('-march=native')
<NixOS_GitHub>
[nixpkgs] Mic92 pushed 1 new commit to release-17.09: https://git.io/vN2Bb
NixOS_GitHub has left #nixos [#nixos]
<burgdorfer>
I'm using fetchgit to fetch haskell sources from a git repo but nix complains about missing a default.nix which does not exist in that repository. How do I tell nix to build this package just like the other non-nix cabal packages?
simukis has quit [Ping timeout: 240 seconds]
<Myrl-saki>
What's the option for make -j again?
<Myrl-saki>
max-jobs is for Nix, right?
<Myrl-saki>
Oh wait no, I think I mixed it up.
actress has quit [Quit: WeeChat 2.0]
nocoolnametom_ has quit [Ping timeout: 246 seconds]
<Myrl-saki>
-j1 --cores 4?
<MichaelRaskin>
--cores 4 should set make -j4 -l4
babyflakes has quit [Quit: Connection closed for inactivity]
xcmw has joined #nixos
<Myrl-saki>
MichaelRaskin: Right. And -j1 is a Nix option.
<sphalerite>
Myrl-saki: and if you want to set it in nix.conf (or nixos config) it's build-cores iirc
<Myrl-saki>
sphalerite: Ah. Thanks.
<Myrl-saki>
Wait
<Myrl-saki>
A little aside.
<Myrl-saki>
My custom flags are -march=native -ftree-vectorize
<Myrl-saki>
What are the chances I'll kill my system
<Myrl-saki>
(In addition to -O2 + other hardening)
ixxie has quit [Ping timeout: 240 seconds]
<symphorien>
you can always rollback :)
<Myrl-saki>
symphorien: True. :P
<Myrl-saki>
symphorien: With that said, let me nix-collect-garbage
nkpart has joined #nixos
<Myrl-saki>
(jk)
hiratara has quit [Remote host closed the connection]