zeus_ has quit [(Remote host closed the connection)]
zeus_ has joined #nixos
zeus_ has quit [(Remote host closed the connection)]
zeus_ has joined #nixos
tomberek has quit [(Ping timeout: 260 seconds)]
michas has quit [(Ping timeout: 248 seconds)]
mkoenig has quit [(Remote host closed the connection)]
Supersonic112 has quit [(Disconnected by services)]
Supersonic112_ has joined #nixos
Supersonic112_ is now known as Supersonic112
darlan has joined #nixos
Wizek_ has quit [(Ping timeout: 260 seconds)]
darlan has quit [(Client Quit)]
jgertm has quit [(Ping timeout: 260 seconds)]
joehh has quit [(Ping timeout: 255 seconds)]
joehh has joined #nixos
lambdamu has joined #nixos
Sonarpulse has quit [(Ping timeout: 255 seconds)]
lambdamu_ has quit [(Ping timeout: 240 seconds)]
johnramsden has joined #nixos
slack1256 has joined #nixos
<johnramsden>
Does anyone have ethminer or any type of etherium miner running on nixos? I have unsuccessfully tried to install ethminer and was wondering if anyone else had gotten it working.
<johnramsden>
*ethereum
joehh has quit [(Ping timeout: 260 seconds)]
joehh has joined #nixos
b has quit [(Quit: Lost terminal)]
sary has joined #nixos
s33se_ has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] peterhoeg pushed 3 new commits to master: https://git.io/v7Lm6
<NixOS_GitHub>
nixpkgs/master dcd6467 Peter Hoeg: xxd: init at 8.0.0442
<NixOS_GitHub>
nixpkgs/master 0345aed Peter Hoeg: mfcj6510dwlpr: use new xxd package instead of vim
<NixOS_GitHub>
nixpkgs/master 45248be Peter Hoeg: cryptominisat: use new xxd package instead of vim
NixOS_GitHub has left #nixos []
darlan has joined #nixos
s33se has quit [(Ping timeout: 248 seconds)]
mbrgm has quit [(Ping timeout: 255 seconds)]
darlan has quit [(Client Quit)]
mbrgm has joined #nixos
<clever>
johnramsden: what error did you have when you tried?
slack1256 has quit [(Remote host closed the connection)]
jgertm has joined #nixos
<johnramsden>
clever: I was getting a stream of 'no version information available' messages. I don't know if I need to set something up for my graphics card, but ethminer https://github.com/ethereum-mining/ethminer didn't mention anything.
<clever>
ah
<clever>
no idea about that error
RayNbow`TU has quit [(Read error: Connection reset by peer)]
RayNbow`TU has joined #nixos
<johnramsden>
clever: Are you mining?
<clever>
not currently
<clever>
but i used to mine btc many years ago
<clever>
was able to get about 0.2 via cpu mining before i cooked a server
<Infinisil>
rycee[m]: It wasn't available in 17.03, so I included -I nixpkgs=path/to/nixpkgs, my local version is using nixos-unstable which worked
<Infinisil>
rycee[m]: I just did a (WIP) PR for htop :D
<srhb>
When exactly does cache.nixos.org *start* caching a package for, say, nixos-unstable? When the tests pass (ie. when the channel is advanced anyway?) Or when the package is first built from master (whether or not the tests succeed and the channels advanced?)
<Infinisil>
I too would like to know that
<rycee[m]>
Infinisil: Cool! I'll try to have a look at it later today :-)
reinzelmann has joined #nixos
silver_hook has quit [(Ping timeout: 240 seconds)]
<LnL>
when it's built
<srhb>
LnL: Then I suppose the next question is: When is it removed?
zeus_ has quit [(Remote host closed the connection)]
<srhb>
I would sort of expect the channels to be the ones to keep cached builds alive
<LnL>
not at all at the moment AFAIK
<srhb>
Oh.
<rycee[m]>
Infinisil: About the htop fields. It's not possible to do exactly what you want (as far as I know). But you could set the type to something like `listOf (enum (attrNames htopFields))` and that should check the values, including giving nice error messages...
junaidali has quit [(Read error: Connection reset by peer)]
junaidali has joined #nixos
silver_hook has joined #nixos
<dhess>
Anyone here know how I can make sure a certain firmware file is included in the generated initrd for NixOS?
<dhess>
I need xbin.usb from non-free firmware
<et4te>
i've been trying to get Hydra to build something, everytime it tries to build gcc.. Is that normal? I've tried setting NIX_PATH and binaryCaches but nothing seems to change it
<et4te>
using: git://github.com/NixOS/nixpkgs-channels.git as the jobset input
<et4te>
(unstable)
<et4te>
and nixpkgs=http://nixos.org/channels/nixos-unstable/nixexprs.tar.xz
<ikwildrpepper>
et4te: think you need to set : use-substitutes = 1
<ikwildrpepper>
in hydra.conf
wkennington has quit [(Quit: Leaving)]
<ikwildrpepper>
ah, there is also a nixos option for it in hydra-module.nix
<methylsalicylate>
does setting the "home" value (let's say, "/home/user") in users.extraUsers.user deletes "/home/user" in case it already exists?
<rycee[m]>
But I guess these are not enough to make it available in initrd.
<bachp>
Does someone know if, when I enable autoUpgrade in my nixos config, the machine automatically reboots if necessary (e.g. kernel update?)
<LnL>
I don't think so
jgertm has quit [(Ping timeout: 246 seconds)]
freusque has joined #nixos
<bachp>
LnL: Would it be possible to detect if a restart wad required?
<srhb>
methylsalicylate: Not to my knowledge.
<srhb>
dhess: You might want to open an issue on this, if you haven't already. At the very least it's nonobvious how to achieve what you're doing.
<LnL>
bachp: maybe, diff the kernel version in /run/current-system with uname or something
silver_hook has quit [(Ping timeout: 240 seconds)]
<bachp>
LnL: I think kernel upgrades would be the only reason a restart would be required, but I could be wrong
lukego has joined #nixos
<dhess>
srhb: thanks. I opened an issue that covers it, along with a couple of other problems I'm having on this NVIDIA Tegra TX1 board.
<bachp>
I'm thinking about implementing something similar to coreos, where each machine get a lock from e.g. etcd or cobsul and then reboots. This way only one machine at the time would reboot.
nslqqq has quit [(Ping timeout: 240 seconds)]
FRidh has joined #nixos
nslqqq has joined #nixos
hamishmack has joined #nixos
kugelblitz has joined #nixos
kugelblitz has quit [(Client Quit)]
indi_ has quit [(Remote host closed the connection)]
indi_ has joined #nixos
<dhess>
Hmm looks like I want boot.initrd.prepend
<dhess>
so, like, boot.initrd.prepend = [ "${firmware}/nvidia/tegra210/xusb.bin" ];
<dhess>
or something like that
<dhess>
actually I have to cpio it first
<dhess>
maybe not. Anyway, nixos/modules/hardware/cpu/intel-microcode.nix does something similar to what I want.
indi_ has quit [(Remote host closed the connection)]
indi_ has joined #nixos
ThatDocsLady has joined #nixos
<dhess>
yeah, cpio then add the resulting image to boot.initrd.prepend
<dhess>
I've actually got a stable system up and running. Main problem is it only sees 1 of the CPUs at the moment. But it's functional.
takle has joined #nixos
erictapen has joined #nixos
freusque has quit [(Quit: WeeChat 1.9)]
freusque has joined #nixos
sary has joined #nixos
<et4te>
ikwildrpepper: what does that do? useSubstitutes i mean?
mbrgm_ has joined #nixos
mbrgm has quit [(Ping timeout: 240 seconds)]
mbrgm_ is now known as mbrgm
<et4te>
nvm i found it
bennofs has joined #nixos
<lukego>
Hey dhess :)
<dhess>
lukego! Hi buddy! :)
<dhess>
lukego: look, you've turned me into an inveterate NixOS user :)
<lukego>
Cool :) yeah I'm hooked too. On the server at least. I switched to Chromebook+NixOS for a while but back on the Mac now :)
civodul has joined #nixos
<dhess>
lukego: Yeah, I went through my Linux-on-the-desktop phase. Too old for that shit now :)
bennofs has quit [(Quit: WeeChat 1.9)]
<dhess>
lukego: Congrats on the Snabb stuff. It looks amazing. I look forward to my next switch running it.
<lukego>
Thanks. It's fun. This year I am working on tooling. Nix-based IDE. Have to see if anybody else ever wants to run it :)
<dhess>
IDE for Snabb?
<lukego>
Yeah. Or really more like a bunch of diagnostic tools including deep JIT visualizations. Solving that problem that anybody can write a Snabb/LuaJIT application but very few people can consistently write a fast one.
<dhess>
Is this what you're using Pharo for?
rpifan has joined #nixos
<lukego>
Yeah. So the UI will be a homogeneous Pharo application, where you can hack Smalltalk code to add new visualizations etc, and the backend will be a heterogeneous zoo of tools/languages/etc managed by Nix.
Fare has quit [(Ping timeout: 248 seconds)]
<dhess>
Wow, sounds ambitious. That will be interesting to see.
Wizek_ has joined #nixos
erictapen has quit [(Ping timeout: 240 seconds)]
<lukego>
I'm imagining that there will be 10x more hacking on the backend, with special things for each application, while the frontend should be a bit more stable e.g. a small number of visualizations can go a long way (e.g. Nix produces a directory full of SVG files and a generic viewer browses those.)
<rpifan>
hello
<dhess>
lukego: Why Pharo rather than something Lua-based, since you presume the user can already write Lua for the Snabb side?
<dhess>
just looking for a reason to play with Pharo? ;)
isHavvy has joined #nixos
Wizek has joined #nixos
<lukego>
Could be so :). Here is my rationalization though. The tools should be much more general than only for Snabb - something you can extend to any application, like Emacs. So Lua will not be a prereq for other people. Then that using Lua for GUI seems very low level e.g. I will be taking a Tk/GTK/Qt/etc binding and doing everything from scratch.
rigelk has joined #nixos
rigelk has quit [(Changing host)]
rigelk has joined #nixos
mudri has joined #nixos
<lukego>
Pharo is nice because there are some excellent high-level frameworks from the university of Bern (just around the corner from me) that I can build on. Researchers working on how to quickly and easily make visualizations of application data structures
<dhess>
right.
<lukego>
The crowd-pleaser would be to do something web-based, with d3.js etc, but personally I'd rather put my hand in a blender :)
Havvy has quit [(Ping timeout: 246 seconds)]
andymandias_ has joined #nixos
<dhess>
lukego: I'll have to look into that, too, then. I'm interested in visualizing code and the workings of a VM as well.
andymandias has quit [(Ping timeout: 240 seconds)]
<lukego>
There is a backend/frontend split where Nix preprocesses the data to reduce the amount of code that is needed in the frontend. e.g. nix converts the funky DWARF debug information into JSON so that I can slurp it in and directly understand C objects (like GDB does) without having to put a DWARF parser or libdwarf dependency into the GUI app.
<dhess>
lukego: that looks really interesting and possibly relevant to my own plans. Thanks for the pointer.
<lukego>
http://github.com/studio/studio is the IDE project homepage. still early days without *quite* having a runnable GUI yet.
<lukego>
"watch this space" :)
<lukego>
I need to find a better term than "IDE" btw. This is really on-the-side tooling like Intel Vtune, etc, and not supposed to be handling basic stuff like editing source code etc. Everybody already has tools that they like for that stuff.
<dhess>
It's like a process visualizer.
ThatDocsLady has quit [(Ping timeout: 255 seconds)]
thc202 has joined #nixos
<lukego>
I'll add that one to the candidate list :-)
ThatDocsLady has joined #nixos
MP2E has quit [(Quit: leaving)]
<dhess>
it's not very catchy :)
Tucky has joined #nixos
seppellll has quit [(Ping timeout: 248 seconds)]
<mudri>
manveru, schoppenhauer: did you work out the Ruby runtime dependencies issue yesterday?
lambdael has joined #nixos
<Infinisil>
rycee[m]: I'm polishing it up now
<Infinisil>
rycee[m]: I may have overengineered it a bit, but considering htop crashes when something is even slightly wrong I'd consider it ok :)
<Infinisil>
rycee[m]: I'd say it's done now :)
<rycee[m]>
Infinisil: Sweet :-)
<rycee[m]>
Do you know what happens if you change the settings inside htop? Will it overwrite the link that home manager created?
<manveru>
mudri: nope
<manveru>
mudri: using path dependencies in Gemfile is quite hard...
<mudri>
Is bundlerEnv a recent addition to nixpkgs or something?
<Infinisil>
rycee[m]: It won't do anything (because it's ro)
<manveru>
nah, it's been in for quite a while
<manveru>
just doesn't support all possible gemfiles
<mudri>
Wait, what's the actual problem?
<rycee[m]>
Infinisil: Cool, that's fine. Having a look at the PR now. An ambitious module! :-)
<manveru>
those kinds of dependencies aren't handled
<Infinisil>
rycee[m]: htop applies the settings on quit, so you can actually change it while running, but it will be the same when restarting again
<Infinisil>
rycee[m]: Indeed :D
<manveru>
because ../something is relative, and bundler even converts absolute paths to relative ones
<manveru>
which doesn't work in the nix store
<rycee[m]>
Infinisil: Yeah, I think that behavior is what I would find intuitive. Many thanks for the contribution!
<manveru>
i'll try doing a sed tonight to replace those relative paths with absolute ones in the Gemfile and Gemfile.lock, but maybe i have to patch bundler itself
<Infinisil>
rycee[m]: The fact that it only keeps the changes for a run is actually perfect. Because everytime you change even a single option (which I do a lot), the file changes, which gets really annoying with source control
<Infinisil>
rycee[m]: Was fun!
<manveru>
mudri: why are you interested?
<mudri>
manveru: so just gem 'something' from a repository should work fine?H
rpifan has quit [(Quit: Leaving)]
<manveru>
sure
<manveru>
that's been working for over a year now
<Infinisil>
rycee[m]: I didn't change the modes 1-4 into actual words because that would only make the config longer, I just included the description of the 4 modes in the description attr
<Infinisil>
LnL: Eh, the points are pretty lame imo
<GlennS>
I have a pull request to Nixpkgs that I need to change. Is it better to add a second commit, or to replace the existing commit with a corrected version?
ThatDocsLady has quit [(Quit: Arma-geddin-outta-here!)]
<Infinisil>
LnL: mkEnableOption is probably the least of our concerns on making nixpkgs easier. And it's such a small thing, you can learn in 10 seconds what it des
<LnL>
Infinisil: it's the same as adding mkStringOption imo, you can do it but you don't gain anything by doing that
<Infinisil>
GlennS: I think usually you'd keep pushing more and more commits and then before it gets merged you'd squash them together into one
justelex_ has quit [(Quit: Konversation terminated!)]
justelex_ has joined #nixos
<Infinisil>
LnL: Well sure, but now we already have mkEnableOption. The discussions about mkEnableOption vs mkOption probably wasted already more energy than what it's worcth
<LnL>
yeah like I said it's probably not going away :)
<LnL>
just misremembered
<Infinisil>
:)
koserge has joined #nixos
justelex_ has quit [(Read error: Connection reset by peer)]
seppellll has quit [(Ping timeout: 260 seconds)]
justelex_ has joined #nixos
<LnL>
GlennS: it's fine to amend changes, unless it's something nontrivial then it can be useful to see what changed
indi_ has quit [(Remote host closed the connection)]
<GlennS>
ok, thanks Infinisil and LnL
<GlennS>
In this case, it's simple so I'll amend it and add a comment to the pull request saying what I changed
ThatDocsLady has joined #nixos
simukis has joined #nixos
indi_ has joined #nixos
nh2 has joined #nixos
kiloreux has quit [(Ping timeout: 246 seconds)]
michas has joined #nixos
michas_ has quit [(Ping timeout: 255 seconds)]
tmaekawa has joined #nixos
indi_ has quit [(Remote host closed the connection)]
ThatDocsLady has quit [(Ping timeout: 240 seconds)]
<kuznero>
While building it gives me the error: Setup: Encountered missing dependencies: process >=1.1.0.1 && <1.5
<kuznero>
I guess while configuring Cabal-1.24.2.0...
indi_ has joined #nixos
erictapen has quit [(Ping timeout: 255 seconds)]
indi_ has quit [(Remote host closed the connection)]
erictapen has joined #nixos
indi_ has joined #nixos
gm152 has quit [(Quit: Lost terminal)]
indi_ has quit [(Remote host closed the connection)]
indi_ has joined #nixos
phreedom has quit [(Ping timeout: 240 seconds)]
aristid has quit [(Ping timeout: 246 seconds)]
aristid has joined #nixos
justelex_ has quit [(Read error: No route to host)]
justelex_ has joined #nixos
phinxy has quit [(Read error: Connection reset by peer)]
phinxy has joined #nixos
erictapen has quit [(Ping timeout: 255 seconds)]
MoreTea has joined #nixos
<nh2>
domenkozar clever: I discovered that ld-wrapper still takes 10 seconds overhead per executable in bash for large Haskell projects, even on master
ThatDocsLady has quit [(Ping timeout: 240 seconds)]
ottidmes has joined #nixos
takle_ has joined #nixos
takle has quit [(Read error: Connection reset by peer)]
<niksnut>
wow
<hodapp>
O_O
<niksnut>
shell considered harmful
<symphorien>
what is ld-wrapper ?
catch22_ has quit [(Remote host closed the connection)]
justelex_ has quit [(Ping timeout: 260 seconds)]
erictapen has joined #nixos
<domenkozar>
nh2: let's open a separate issue
MoreTea has quit [(Ping timeout: 276 seconds)]
<nh2>
niksnut: by now I honestly think that the reliance on bash is nix's biggest technical problem
<nh2>
domenkozar: will do
<domenkozar>
nh2: you mean like anything we do with computers nowadays :D
<LnL>
let's change it all to perl :p
<niksnut>
that would actually be a significant improvement
<nh2>
domenkozar: outside nix I can at least run away from it
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] edolstra pushed 4 new commits to staging: https://git.io/v7Lbc
<NixOS_GitHub>
nixpkgs/staging 8cfe774 Eelco Dolstra: gnumake40: Remove unused version
<NixOS_GitHub>
nixpkgs/staging 9f345ce Eelco Dolstra: gnumake41: Remove unused version
NixOS_GitHub has left #nixos []
<domenkozar>
I'll never forget when niksnut/shlevy ported perl to C++ in Nix
<domenkozar>
someone said
<niksnut>
but then, *any* language would be an improvement
<domenkozar>
"wait, you changes a memory safe language for non-safe one?"
aminechikhaoui has quit [(Quit: Lost terminal)]
<domenkozar>
changed*
<niksnut>
domenkozar: actually a fair point
<domenkozar>
yes
aminechikhaoui has joined #nixos
<LnL>
true
<dbe>
it's also about the only fair point you can make about that
<domenkozar>
yes that's why pragmatism won
leat has quit [(Ping timeout: 260 seconds)]
erictapen has quit [(Ping timeout: 248 seconds)]
<nh2>
is there a technical reason why nix doens't use Python in the places where it now uses bash, like in all those wrapper scripts?
<nh2>
(I should say "why nixpkgs ...")
<niksnut>
to be pedantic, Nix does not rely on bash
<niksnut>
Nixpkgs does
<nh2>
yes I just corrected
<joepie91>
specifically, the default builder, I believe
<niksnut>
sorry
<domenkozar>
nh2: it's just the question of the person who wrote those scripts to pick what's his/her go-to scripting language
<NickHu>
I'm trying to call callCabal2nix on a src obtained from fetchFromGitHub, but the problem is the github repo actually has the haskell module in a subdirectory - is it possible to change to that directory before passing the src to callCabal2nix?
<domenkozar>
actually having shellcheck for stdenv in Nix would be a huge improvement
<nh2>
domenkozar: I was wonering if there's an argument of portability or minimum dependencies
<nh2>
wondering
<domenkozar>
NickHu: --subpath was added to cabal2nix so you'll have to teach callCabal2nix function about it
<LnL>
I think that's definitively one reason
<domenkozar>
nh2: it does make boostrapping easier
<niksnut>
the one big advantage of bash is that it's the one language that everybody knows
<nh2>
domenkozar: even shellcheck wouldn't help with the fact that bash doesn't make it easy to write non-quadratic appends. Python at least has sane defaults and easy to use constructs for that
<nh2>
niksnut: I'd argue that 99% of people I know "don't know bash" (me included) because they make heavy mistakes all the time
<nh2>
nixpkgs included (see 200-second overhead per link we had before ryantrinkle's patch; at least now its "only" 10 seconds)
<NickHu>
actually if anyone could point me to where fetchFromGitHub is defined that would be great; I see fetchgit has a postFetch parameter, which I think I could just put a "cd directory" in
<dbe>
nh2: I guess that setting up packages is enough of a crowd-sourced thing that everyone does their own thing there; as long as it's reproducible then I guess it's cool. Many packages use scons and whatnot and you have to deal with those too, so, you know...
<NickHu>
I wonder if fetchFromGitHub is the same
<domenkozar>
bash wins at stockholm syndrome
<dbe>
nh2: err, well, maybe you're talking about the internal scrips that set up the stages, not the per-package stuff
<domenkozar>
NickHu: git grep "fetchFromGitHub ="
<LnL>
I think it's in all-packages.nix
Ivanych has quit [(Ping timeout: 255 seconds)]
<nh2>
dbe: I'm talking about things at the level of `ld-wrapper` specifically currently. domenkozar, do you know if that level is already past the bootstrapping stage where bash may give an actual technical benefit?
<LnL>
since it's just a "simple" function
<NickHu>
domenkozar: I tried that but I only got two results... Turns out I had typo'd to fetchFromGithub lol
<NickHu>
Yeah, you're right LnL
<domenkozar>
nh2: afaik it's part of stdenv
<niksnut>
afaik, quadratic behaviour in bash is fairly easy to prevent by using array rather than string concatenation
<domenkozar>
but not sure at what stage
<niksnut>
i.e. args+=("--arg")
<LnL>
yeah, I think the ld-wrapper mainly appends strings
kuznero has quit [(Quit: Page closed)]
<niksnut>
also, foo+=" bla" is much faster than foo="$foo bla"
<domenkozar>
nh2: that's why guix exists, Scheme all the way down
<domenkozar>
but then people will argue if Scheme is a good choice :)
<LnL>
the rpath stuff uses "$rpath $1"
justelex_ has joined #nixos
<nh2>
niksnut: I couln't find documentation if += on bash guarantees an exponential buffer growth or not
MoreTea has joined #nixos
<hyper_ch>
nh2: why would one use python over bash? :)
mudri has joined #nixos
<hyper_ch>
niksnut also, foo+=" bla" is much faster than foo="$foo bla" -- is is much faster?
<nh2>
(I assume from the smiley that this is a rethorical question)
<niksnut>
x= ; time for ((i = 0; i < 10000; i++)); do x="$x $i"; done => 3.7s
indi_ has quit [(Remote host closed the connection)]
<niksnut>
x= ; time for ((i = 0; i < 10000; i++)); do x+=" $i"; done => 0.07s
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] edolstra pushed 1 new commit to staging: https://git.io/v7LAm
<NixOS_GitHub>
nixpkgs/staging aa4a92d Eelco Dolstra: cc-wrapper/ld-wrapper: Minor speedup in string concatenation...
NixOS_GitHub has left #nixos []
<LnL>
that's a bigger difference then I was expecting
<dbe>
hey, yeah, that's pretty good.
<hyper_ch>
niksnut: hmmm, interesting
<NickHu>
Okay so I modified haskellSrc2nix to understand the --subpath argument, but I'm looking for a non-destructive way to change callCabal2nix (i.e. not touch its interface)
<hyper_ch>
since both were bash internals I wouldn't have thought it matters much
<NickHu>
Currently, it takes two arguments: name and src
<domenkozar>
NickHu: make a new function and share the internals between the two
<domenkozar>
where the new function has an attribute set of parameters passed to cabal2nix
<domenkozar>
so it can be extended in the future
<NickHu>
domenkozar: Something like callCabal2nixWithSubpath?
<domenkozar>
more like callCabal2nixWithOpts
<NickHu>
So I guess it ought to take arguments in the same order, except with an additional set of options as the last argument..?
<domenkozar>
yup
tvon has quit [(Quit: System is sleeping...)]
<domenkozar>
did clever figure out why are our tests freezing?
<domenkozar>
self.haskellSrc2nix { inherit src name ops; }
<NickHu>
ah I see
<domenkozar>
opts*
<NickHu>
I need it to call haskellSrc2nix itself
<ottidmes>
I thought I saw a tool mentioned to bootstrap NixOS in an existing system. I do have an example of using kexec by clever, but I was curious about this other tool as well, that is, if it exists.
<ottidmes>
I will bookmark them, so I will not forget about them again, thanks!
erictapen has joined #nixos
<NickHu>
What is the last argument to callCabal2nix supposed to be anyway? I always see it called with the empty set passed in, but I can't see it in the function signature
<NickHu>
It seems without passing in the empty set at the end the expression doesn't evaluate or something
tmaekawa has quit [(Ping timeout: 240 seconds)]
<NickHu>
Also, it seems I can pass overrides in it, but I don't know where this argument is defined
<domenkozar>
I recommend first learning the basics of the language
<FRidh>
So now we're having a mass-rebuild on master because of b087618ac06256cb3c06278e7eaba45841f4ea91 but this hasn't been pushed/merged to staging. Staging kind of finished building, but there are still several odd failures. How about merging simply merging the two at this point? Both branches are partially broken anyway.
<gchristensen>
FRidh: _ouch_
takle_ has quit [(Remote host closed the connection)]
<domenkozar>
FRidh: didn't our most awesome ttuegel fix staging?
<gchristensen>
(ttuegel is quite stressed right now, by the way, due to breakage on his system)
<FRidh>
domenkozar: most of it yes. go doesn't build though, and I've had some packages that just fail to build without giving any reason.
IITaudio_ has joined #nixos
<NickHu>
Oh I see it's a curried argument to mkDerivation
<NickHu>
I don't have a problem with the language, but the nixos API documentation is horrible/nonexistent
teh[m] has quit [(Quit: Bridged client killed)]
<domenkozar>
we need a proper CI
<domenkozar>
and a proper git branching workflow
<domenkozar>
then all this will be history
Criena[m] has quit [(Quit: Bridged client killed)]
* domenkozar
sips more coffee
* gchristensen
sips coffee and thinks about a brave new world where we "proper CI"
jascot[m] has quit [(Quit: Bridged client killed)]
<domenkozar>
FRidh: /nix/store/xilxav9i6xkrra3vclmlm7zwbbvr46qv-go-1.8.3/share/go/pkg/tool/linux_amd64/link: too many errors
<domenkozar>
this gives some clues
jesper has quit [(Quit: Bridged client killed)]
lfont[m] has quit [(Quit: Bridged client killed)]
<gchristensen>
a major blocker that I found w.r.t. proper CI is how do you compare 2 builds, where you're building 40,000 pieces of software? if you break 3 packages, is that the updater's fault? what about security / time-sensitive patches?
<domenkozar>
well that has to do with branching workflow
<MoreTea>
gchristensen, the build system should drop everything else if a time-sensitive thing comes along
<LnL>
I think it depends
tvon has joined #nixos
jensens has joined #nixos
leat has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] volth closed pull request #26839: fetchMavenArtifact: prevent leaking nix hash to jar name (master...fetchmavenartifact-do-not-leak-hash) https://git.io/vQObr
NixOS_GitHub has left #nixos []
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] volth closed pull request #27514: Set virtualisation.libvirtd.qemuOvmf to false (master...patch-2) https://git.io/v7e50
NixOS_GitHub has left #nixos []
<gchristensen>
two feature requests for hydra that I'd like to propose but can't yet: (1) there should be a higher level "bump to front of queue" which says this is most important no matter what anyone else says, restricted to certain users. this way, a critical security patch rebuild can be bumped to the front of the queue and not displaced by another eager tester
<gchristensen>
(2) much less important, a way to trick the autoscaler in to scaling up for a bit :) see also #1
reinzelmann has quit [(Ping timeout: 260 seconds)]
<gchristensen>
this feeds in to my organizational bug report: a private copy of nixpkgs on github or something, which should come with a private bug tracker for security issues, set up in Hydra as a private job. this way we can pre-build and pre-test security issues without having them leak publicly, but when they do leak they'll already be in the cache
<gchristensen>
LnL: done
<LnL>
thanks, no point in running that since I can't compare with anything properly
kiloreux has quit [(Ping timeout: 255 seconds)]
koserge has quit [(Read error: Connection reset by peer)]
<ikwildrpepper>
gchristensen: I don't think we can have private repos on the current nixos org account
takle has joined #nixos
<gchristensen>
because it is unpaid?
<ikwildrpepper>
and switching it to paid account is probably also a bad idea, as we'd need to pay per user in the org
Fare has joined #nixos
<gchristensen>
ouuch
<ikwildrpepper>
we can make separate org for it, or use e.g. gitlab
<ikwildrpepper>
gitlab has free private repos, if I remember correctly
<gchristensen>
yeah they do
<TimePath>
or ask for a freebie?
<gchristensen>
however, I think if we're seeking entry in to the oss-security list, having our private bug tracker not even be paid-for would be maybe frowned upon
<gchristensen>
(much less our own hosted infra)
<gchristensen>
but I don't know what they really expect to be honest :)
kyren has joined #nixos
mkoenig has quit [(Remote host closed the connection)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] NeQuissimus pushed 1 new commit to master: https://git.io/v7tkr
<NixOS_GitHub>
nixpkgs/master 1dd6e7d Tim Steinbach: linux: 4.13-rc1 -> 4.13-rc2
NixOS_GitHub has left #nixos []
rauno has joined #nixos
kyren_ has joined #nixos
kyren has quit [(Ping timeout: 255 seconds)]
Rotaerk has quit [(Quit: Leaving)]
kyren has joined #nixos
<domenkozar>
hmm
kyren_ has quit [(Ping timeout: 246 seconds)]
reinzelmann has quit [(Quit: Leaving)]
<joepie91>
gchristensen: ikwildrpepper: I have no idea whether this is desirable or not, but fwiw, feel free to use https://git.cryto.net/ :P
<joepie91>
since that's running anyway
<gchristensen>
I think if we're using some otherwise not popular hosting system we'll host ourselves
<gchristensen>
they're rightfully quite paranoid about the issues
kyren has quit [(Ping timeout: 240 seconds)]
<joepie91>
sure, that works too :P
fikse has joined #nixos
kyren has joined #nixos
dannyg has joined #nixos
MercurialAlchemi has quit [(Ping timeout: 240 seconds)]
<nh2>
it just claims that it can't fetch the package, even though it's there on the other machine. And it works when the command is run from a different machine. I'm wondering if it has something to do with the ssh-agent socket shown in the strace
kyren has quit [(Ping timeout: 260 seconds)]
kyren has joined #nixos
jtojnar has joined #nixos
Piece_Maker has joined #nixos
ryanartecona has joined #nixos
kiloreux has quit [(Ping timeout: 255 seconds)]
Acou_Bass has quit [(Ping timeout: 258 seconds)]
Piece_Maker is now known as Acou_Bass
jtojnar has quit [(Client Quit)]
pie_ has quit [(Read error: Connection reset by peer)]
pie_ has joined #nixos
pie__ has joined #nixos
pie__ has quit [(Remote host closed the connection)]
<niksnut>
copumpkin: even more radical: get rid of .drv
<joepie91>
copumpkin: if JS tooling is undesirable for some reason, should be fairly trivial to port that parser to a PEG generator in any other language as well :)
<Sonarpulse>
I'd like to a) make everything propagated-* related accept \n and ' ' as deliminator when reading b) make nixpkgs just use '\n' when reading, c) bump the min version of nixpkgs
<Sonarpulse>
*min version of Nix in nixpkgs
<Sonarpulse>
then there is no breaking change
<niksnut>
MoreTea: the alternative is to simply not materialize derivations on disk, but to just send an in-memory representation of derivations to the daemon (basically BasicDerivation and buildDerivation() already do this)
<Sonarpulse>
OTOH, accepting '\0' or ' ' when reading would just be weird
<gchristensen>
niksnut: would that make embedding the nixpkgs revision in to fetchurl's user agent a reasonable thing to do?
<niksnut>
Sonarpulse: you mean "when writing"?
<Sonarpulse>
niksnut: un-revert my printLines change
<copumpkin>
ixxie: I like how that page list RHEL as inactive
<Sonarpulse>
so nixpkgs unstable and 17.09 only write newline-delimited files
<copumpkin>
oh, it's listing RHL as inactive, not RHEL
<Sonarpulse>
niksnut: but since those nixpkgs can read with either deliminator, mixing packages from older nixpkgs still works
<niksnut>
gchristensen: not sure. Having a virtual drvPath that doesn't change too much might still be useful
rigelk has joined #nixos
civodul has quit [(Quit: ERC (IRC client for Emacs 25.2.1))]
Tucky has quit [(Quit: WeeChat 1.9)]
<copumpkin>
what's the best intermediate solution to this nix-shell screwup in master?
<copumpkin>
assume I can't just nix-env -iA bashInteractive before running nix-shell
<copumpkin>
is there something I can do from inside the expression?
<copumpkin>
Sonarpulse?
<Sonarpulse>
copumpkin: hmm
<Sonarpulse>
well I backported the PR
<LnL>
we could check $BASH_VERSION and recommend people to run nix-env -iA nixpkgs.bashInteractive
<Sonarpulse>
to Nix
<copumpkin>
Sonarpulse: I still don't have a release that contains it :)
<LnL>
but that's not ideal ofcorse
<copumpkin>
I have a lot of build infrastructure that's failing with the new version
<copumpkin>
because I use nix-shell all over the place
<copumpkin>
expecting it to work without jumping through hoops
<Sonarpulse>
copumpkin: I could revert it two
<copumpkin>
there isn't something I can do inside the expression to make it work?
<Sonarpulse>
*too
<copumpkin>
like put in an explicit reference to bashInteractive from there?
<Sonarpulse>
copumpkin: oh
<Sonarpulse>
probably in nativeBuildInputs
<Sonarpulse>
err wait, nah that's too late in the pipeline
<Sonarpulse>
just minor disagreement on whether to special-case PATH or not
<copumpkin>
ah okay
<Sonarpulse>
copumpkin: also if you look at nix-shell
<Sonarpulse>
the initial script is all bad
<copumpkin>
so there's nothing I can do easily right now without finding a way to sandwich nix-env -iA bashInteractive before I ever run nix-shell?
<Sonarpulse>
you could just edit nix shell a la my backport instead
<copumpkin>
this is on a ton of automated build infrastructure
<copumpkin>
"just edit X" or "just run X" isn't super feasible
<Sonarpulse>
copumpkin: seeing that a revert still causes a mass-rebuild, yes I suppose there is nothing instantaneous and easy :/ sorry
<copumpkin>
I'll just figure out how to get bashInteractive into scope before it
<copumpkin>
(this is the CentOS case btw, not macOS)
<Sonarpulse>
ah
<Sonarpulse>
yeah installing system wide could be risky
<Sonarpulse>
if some jank old script relies on weird set -e for example
<copumpkin>
yay impurity
<Sonarpulse>
(bash no understand Either monad, keep on changing definition of `set -e`)
<copumpkin>
how painful would it be to revert?
ThatDocsLady has quit [(Ping timeout: 246 seconds)]
<Sonarpulse>
copumpkin: well at this point staging and master already won't cleanly merge
<copumpkin>
because I'm basically going to have to tell everyone my team to run a command when this change hits them, or reinstall nix
<copumpkin>
both of which feel kinda shitty
<Sonarpulse>
and it might be an easy revert on master
<Sonarpulse>
it will be a difficult revert on staging
<Sonarpulse>
I rather just revert on master
<copumpkin>
can we just do a minimal change of local -n to the eval thing you don't like
<copumpkin>
retaining the rest of the functionality?
<copumpkin>
just reverting on master will push the pain down the road to the staging merge though, right?
<Sonarpulse>
well, I hope we can release a new nix version by then
<Sonarpulse>
writing the eval stuff correctly isn't easy
<copumpkin>
but the new nix version is still an awkward solution
<Sonarpulse>
I forgot about lib/minver
<Sonarpulse>
it's not with that
<copumpkin>
like I'm still going to have to tell everyone on my team (and all my build infra) to update to a new version of nix
<Sonarpulse>
that's going to happen sooner or later with 1.12 anyways?
<copumpkin>
I dunno, I'd rather stick with the most compatible thing until we have reasonable expectation that most users are on a version that doesn't rely on impurities
<Sonarpulse>
with minver you can let Nixpkgs tell them?
<Sonarpulse>
I rather revert than eval
<Sonarpulse>
how long do you suspect that "most users" will be?
ertes-w has quit [(Ping timeout: 255 seconds)]
<copumpkin>
I'd expect lots of folks to update to 1.12 :) just seems crazy to me to force a ton of folks to update to get basic functionality because of what should've been a minor refactor
<copumpkin>
like lots of wasted time for lots of people everywhere
<copumpkin>
I've already wasted several hours on this thing
<copumpkin>
that's before I'm asking everyone else to do the same thing. I have nixpkgs pinned to a version before this change went in right now
<copumpkin>
for the entire team
<Sonarpulse>
I don't need to immediately land this stuff
<copumpkin>
but I also need some more recent stuff and now I'm stuck making a choice between spreading the time wastage to everyone else or not getting the new things I need
<Sonarpulse>
I assume somebody can make my release.nix hydra job work
<Sonarpulse>
copumpkin: yeah I have all the creds I need
<copumpkin>
no idea why it's been pending for so long
<MoreTea>
Sonarpulse, cute
<LnL>
Sonarpulse: are you sure you can use system as an imput?
<Sonarpulse>
do I just need to bump number of shares?
<Sonarpulse>
LnL: hmmmm
<Sonarpulse>
I just cargo culted that
<Sonarpulse>
but perhaps from the wrong job
<LnL>
try supportedSystems [ "x86_64-linux" ]
<niksnut>
tbh, whether we use newlines or spaces seems kind of a bikeshed issue
<LnL>
inputs are similar to --arg AFAIK
<niksnut>
yes, newlines may be slightly nicer, but not really worth the breakage
<copumpkin>
it feels like another issue like the widespread user expectation thing
<niksnut>
so I'm fine with adding support for newlines to Nix, but at the same time it would be best not to rely on it in nixpkgs in the near future
<copumpkin>
like, get a change into nix, wait until it's widely adopted
<copumpkin>
then make the change in nixpkgs that relies on that change
<copumpkin>
(bringing it back to the local -n thing)
<copumpkin>
it's not ideal but the cost/benefit of the current break and fix and hope folks upgrade feels all wrong
<Sonarpulse>
well, let's revert all the things
<Sonarpulse>
some of ttuegle's changes are seperate
<Sonarpulse>
so will need to filter carefully
<Sonarpulse>
domenkozar niksnut: can we revert on master directly? cancel the eval last night?
<nh2>
niksnut: the bash change you pushed to staging, is that currently being built? I'm not quite sure how to determine that rom the hydra site / where to look
<gchristensen>
Sonarpulse: that is certainly a down side to working so close to core
<copumpkin>
Sonarpulse: I hear that there are hacks to get ccache into nix builds
<copumpkin>
and yes when I was developing the darwin stdenv it was miserable
<Sonarpulse>
gchristensen: indeed, and it is unavoidable
<copumpkin>
until we slimmed down the stdenv, it was 3 llvm+clang builds per stdenv
<Sonarpulse>
OTOH the propagated- change at least is orthogonal
<LnL>
copumpkin: oh man that's annoying
<Sonarpulse>
I had no idea it was even a breakging change as stdenv/setup.nix already slurped the entire file
<Sonarpulse>
niksnut: I know everyone always asks you about 1.12 and it must be annoying, but I haven't followed that stuff much and have no idea even roughtly when to expect it: 17.09? 18.03? 18.09?
<Sonarpulse>
the propagated change is not important
<copumpkin>
LnL: you should've seen when I decided to kill bootstrap-tools references in the final stdenv :P big clumps of torn out hair all over the house :P
<Sonarpulse>
but the non-old bash stuff is easy
<Sonarpulse>
as working with eval just makes my debug cycles even longer
<Sonarpulse>
s/easy/necessary/
s33se_ has quit [(Ping timeout: 260 seconds)]
<niksnut>
17.09 seems unlikely
<niksnut>
Sonarpulse: ^
<Sonarpulse>
right, makes sense to me too
<Sonarpulse>
OTOH if we release a 1.11.14 now, can we require it for 17.09?
<Sonarpulse>
is that enough delay?
s33se has joined #nixos
seppellll has quit [(Ping timeout: 255 seconds)]
<LnL>
I have a small nix-profile change if 1.11.14 is going to be released soon
<Sonarpulse>
:) if we have a enough small non-breaking changes, requiring 1.11.14 becomes more reasonable
<copumpkin>
Sonarpulse: I still don't get the local -n vs. eval
<copumpkin>
it looks from your diff like the meat of it is replacing
<copumpkin>
eval "local -a dummy=("${$var[@]}")"
<copumpkin>
which isn't pretty
<copumpkin>
but I don't see the awfulness
<copumpkin>
we'd been doing that for years
<Sonarpulse>
the merged diff?
<copumpkin>
don't get me wrong, I'd rather not :) but in the ranking of "rather nots" it's lower than the alternative
<niksnut>
Sonarpulse: it should also be possible to upgrade from the previous NixOS release
<niksnut>
so requiring 1.11.14 is problematic
<Sonarpulse>
niksnut: doesn't upgrading already download new Nix and restart if necessary?
<copumpkin>
that's NixOS
<copumpkin>
we have a lot of non-NixOS users :(
<niksnut>
Sonarpulse: yes, but that's the fallback-of-last-resort :-)
<niksnut>
(iirc)
<niksnut>
I'm worried about these "Too many heap section" errors on hydra...
<Sonarpulse>
copumpkin: maybe I've overblown it, but keeping track of bash's word splitting takes a fair amount of mental engery without eval
<copumpkin>
it's just all a trade-off, right? like, is it more mental energy than disentangling the reverts in master? or forcing all users to jump through hoops for basic nix functionality on a tiny minor upgrade?
ixxie has quit [(Ping timeout: 248 seconds)]
<copumpkin>
like I'm not saying we should never switch away from local -n
<copumpkin>
just that we phase that after the nix change has been in for a few months
<copumpkin>
I mean swithc away from eval, to local -n :)
<Sonarpulse>
:) well, I'll take a look at adding back eval then
<copumpkin>
I mean, if you think it's hopeless, I'll trust you on that, but of the three options it seems like the smallest to my uneducated eye
<copumpkin>
can even leave the local hook in place
<joepie91>
error: infinite recursion encountered, at undefined position
<joepie91>
whee
<copumpkin>
it's just s/var/dummy/ and replacing the local -n line with the old eval line
<copumpkin>
joepie91: my favorite error message
<joepie91>
tremendously helpful, too :P
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Ericson2314 opened pull request #27614: stdenv-setup: Add quotes that don't do anything for consistency. (staging...stdenv-quotes) https://git.io/v7tiv
NixOS_GitHub has left #nixos []
fikse has joined #nixos
aminechikhaoui has quit [(Quit: Lost terminal)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] NickHu opened pull request #27615: fix haskellPackages.cuda and some tools dependent on cudatoolkit (master...cuda) https://git.io/v7tir
NixOS_GitHub has left #nixos []
takle has quit [(Remote host closed the connection)]
<et4te>
guys i have the following release.nix + default.nix: https://gist.github.com/et4te/9c33ff2c21005d263ee3dbf88a0b8719, keep getting 'nixpkgs not found in default.nix line 2'. how do I actually pass that through? the package seem to build local
<et4te>
i guess the question is confusion over how inherit etc works
<copumpkin>
cat `which echo` | /lib64/ld-linux-x86-64.so.2 /dev/fd/1 hi
<niksnut>
copumpkin: never really thought about it, but yeah, it would be possible to turn the installer into a self-contained self-extracting executable
<Sonarpulse>
niksnut: recall I did that associate array thing?
<Sonarpulse>
for findInputs memoization?
<niksnut>
copumpkin: but of course, it wouldn't be possible
ixxie has joined #nixos
<Sonarpulse>
if ${var[$pkg]}; then continue; fi # already seen
<niksnut>
*portable
<Sonarpulse>
basically
<Sonarpulse>
that will also break on Darwin
<hyphon81>
Hello. In my environment, matplotlib crashed with segmentation fault in libopenblasp-r0.2.19.so. "nix-shell -p python3Packages.numpy -p python3Packages.matplotlib -p qt4 -p python3Packages.pyqt4"
jgertm has joined #nixos
<Sonarpulse>
It shouldn't break on CentOS, however
<hyphon81>
Running plt.show() and obtainning segmentation fault in libopenblasp-r0.2.19.so.
tmaekawa has joined #nixos
<copumpkin>
niksnut: unless we came up with a pure bash (without local -n :P) decompressor :D
<copumpkin>
of course it's only as portable as the underlying closure tarballs anyway, so don't really need to go overboard
aborsu has joined #nixos
aborsu has quit [(Remote host closed the connection)]
<aristid>
why is the channel stuck? :(
<copumpkin>
the tubes are clogged
ryanartecona has joined #nixos
jensens has quit [(Ping timeout: 255 seconds)]
<copumpkin>
aristid: which channel?
aborsu has joined #nixos
alx741 has quit [(Quit: alx741)]
<aristid>
copumpkin: unstable-small
<joepie91>
well, unstable has been stuck for a whilke
<joepie91>
while*
<copumpkin>
ah, nixos-unstable :)
<aristid>
i don't use any of that fancy stable stuff
<hyper_ch>
unstable for 20 days... unstable small for 12 days or so
<aristid>
unstable forevah!
xadi has quit [(Quit: Leaving.)]
<hyper_ch>
aristid: I keep bugging peopler here as well about this
<aristid>
i mean, wkennington hasn't broken booting for quite some time, even in unstable! so there's no reason not to use it :)
tokudan has joined #nixos
<copumpkin>
aristid: he just gave up on nixpkgs/nixos :P
<aborsu>
Hello, trying to create a package using fetchFromGithub, how can I get the hash from the revision (other than making a bad hash and watching the logs when nix crashes)?
<aristid>
gchristensen: why did you break hydra? :(
<copumpkin>
that's more like it
<gchristensen>
copumpkin: what have you done! :o
<aristid>
hmm
<aristid>
i'm using stable nix
<aristid>
maybe that's why?
<gchristensen>
if it is my fault, copumpkin broke it first
<aristid>
gchristensen: i can blame you both equally if it makes you feel better :)
<gchristensen>
it really doesn't
pie_ has quit [(Remote host closed the connection)]
pie_ has joined #nixos
<gchristensen>
we all contribute what we can :P
<aristid>
i think i can reproduce one of the failing tests
<aristid>
at least it's hanging for like a minute now
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nix] LnL7 closed pull request #1085: use /nix/var/tmp as default tmpRoot (master...darwin-nix-tmpdir) https://git.io/vPz4m
NixOS_GitHub has left #nixos []
aborsu has quit [(Remote host closed the connection)]
aborsu has joined #nixos
<Sonarpulse>
copumpkin: what was the cause of that problem?
foxmean has left #nixos ["ERC (IRC client for Emacs 25.2.1)"]
<Sonarpulse>
I am now on 1.11.13 daemon and user
<aristid>
gchristensen: i want the channel to update so everybody can benefit from my fix to the wireguard module to allow you not to store your private keys in the /nix/store! :)
<copumpkin>
Sonarpulse: nixpkgs has a libSystem in it that mentions that library and it doesn't exist in 10.12
<copumpkin>
or something
<aristid>
also, when will 1.12 become the default? :)
<Sonarpulse>
aristid: by 18.03 maybe? :/
<gchristensen>
aristid: we all do
Infinisil has joined #nixos
tmaekawa has quit [(Quit: tmaekawa)]
<copumpkin>
niksnut: if it helps, I've been running nix 1.12 in my EC2 AMIs at work (lots of bootups!) for the past couple of months with no issue
<copumpkin>
haven't done much "user" testing
<LnL>
I've been using it on osx for about a week now
colabeer has joined #nixos
<dshin>
does nix package manager support nvidia opengl on ubuntu?
watersoul has quit [(Ping timeout: 240 seconds)]
<dshin>
im having some driver conflicts
aborsu has quit [(Quit: aborsu)]
<gchristensen>
hard to believe it is almost time for 17.09
<gchristensen>
about this time I switch to unstable to start hunting for problems
<aristid>
gchristensen: ah, maybe the thing to blame you for is that you didn't switch to unstable yet, so that's why there are still problems! :)
obadz has quit [(Ping timeout: 260 seconds)]
<copumpkin>
I thought we agreed it was my fault
<gchristensen>
yeah I mean I don't run unstable for a reason
<copumpkin>
turns out I was wrong about gchristensen
<copumpkin>
you can blame the blame of gchristensen on me
<Infinisil>
gchristensen: The failing test :)
hyphon81_ has joined #nixos
et4te has quit [(Remote host closed the connection)]
<aristid>
gchristensen: i don't know, it's not that bad... most problems i have with unstable is that it doesn't update the channel quite often - i guess the tests do catch most stupid issues
<gchristensen>
yeah, I just don't like running an insecure os locally
<Infinisil>
LnL: Ohh, just noticed that nix-vim is by you :D
<LnL>
yup
<Infinisil>
LnL: Just found a case that doesn't work correctly heh
<aristid>
gchristensen: fair enough :P
<LnL>
Infinisil: feel free to open an issue :)
<LnL>
or contribute if you like viml
<aristid>
so this is weird: the nixos-unstable-small channel appears to be on a newer revision than the last hydra run for it that succeeded
<Infinisil>
LnL: I have never done anything viml, but I'll have a look
<Infinisil>
LnL: It's an expression like this that's failing: "${${foo.bar}}baz" (baz has wrong color)
<LnL>
well don't look at the source if you don't like regexes then :p
<copumpkin>
I'd want to merge to master and then force evaluate nixpkgs trunk
<copumpkin>
on hydra
<copumpkin>
joachifm: ^
FRidh has joined #nixos
Ivanych has quit [(Ping timeout: 260 seconds)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nix] Ericson2314 opened pull request #1485: utils: Disambiguate between normal and malformed EOF in nix::readLine. (master...utils-clarity) https://git.io/v7tjw
NixOS_GitHub has left #nixos []
<FRidh>
pushing more of these mass rebuilds to master will only create larger delays
<copumpkin>
FRidh: you get why I want it though, right?
<copumpkin>
I really just want to revert the channel
bennofs has joined #nixos
<copumpkin>
but I don't have an easy way to do that
<FRidh>
There's currently 4 large sets that are being built
<copumpkin>
I get it
<gchristensen>
cancel the old ones
<FRidh>
copumpkin: I understand, and wish those breaking changes would have been reverted as well, but on staging.
<copumpkin>
FRidh: me too, but wishing won't get us anywhere right now :)
<copumpkin>
I just want a nix install / channel update to not break a bunch of users
endformationage has joined #nixos
luigy has joined #nixos
<domenkozar>
for such cases we might need to rollback the channel
<FRidh>
copumpkin: if you want to get us somewhere you should pause the gcc6 and haskell builds and revert changes on staging.
<copumpkin>
I'd be okay cancelling some evals but I don't think I have such powers
<copumpkin>
can they be paused?
<domenkozar>
we can just bump the builds in a queue
<Sonarpulse>
but I've been busy atoning for my nixpkgs breakage sins on the early side for EU time
<aristid>
gchristensen: shouldn't that be 48 physical cores? :) it says it's a two socket machine
<Sonarpulse>
so when I get back to such things not sure what his workday will be at
ottidmes has joined #nixos
<Infinisil>
But can it run Crysis?
<domenkozar>
I have used weeder to get our dependencies down
<domenkozar>
so we might have a week
<aristid>
gchristensen: ah it's not the big core count xeon :D
<domenkozar>
:D
<aristid>
gchristensen: at work we only have the 2x22-core type xeon machines :D i love seeing my software take up 70+ cores and use them to full utilization
<gchristensen>
aristid: nice, we have a 96 core build box too -- the type 2A -- much slower clock speed though
rigelk has joined #nixos
rigelk has quit [(Changing host)]
rigelk has joined #nixos
darlan has joined #nixos
darlan has quit [(Client Quit)]
<aristid>
gchristensen: yeah it's probably not really optimal to standardise on the huge core count version, but whatever
yegods_ has joined #nixos
<aristid>
gchristensen: 96 core... that's a xeon e7 then?!
<aristid>
gchristensen: who buys e7s?
<gchristensen>
Cavium ThunderX
<aristid>
oh
ixxie has quit [(Quit: Lost terminal)]
<copumpkin>
domenkozar: so you've triggered a new eval and cancelled old builds?
<domenkozar>
yes
<copumpkin>
thanks :)
<copumpkin>
<3 <3
Wizek_ has quit [(Ping timeout: 268 seconds)]
oida_ has joined #nixos
<joepie91>
gchristensen: I can irregularly pitch in when the freelance business is good :P
<gchristensen>
(back in 25 minutes)
tvon has quit [(Quit: System is sleeping...)]
<joepie91>
(which unfortunately is not right now)
<copumpkin>
gchristensen: happy to throw money at it if it's easy to spin up and throw into the mix
Ivanych has joined #nixos
<copumpkin>
I expect the autoscaling EC2 instances things would probably be easier
<gchristensen>
we already do that
<copumpkin>
yeah, I just don't know what triggers it
<gchristensen>
# of jobs
rigelk has quit [(Quit: Lost terminal)]
<copumpkin>
niksnut: need a magic bitcoin address that turns up the # of hydra nodes based on how much money gets sent to it :P
ng0 has quit [(Quit: Alexa, when is the end of world?)]
<gchristensen>
how about our own cryptocurrency called buildcoin (lol)
<xplat|work>
so, like, say i want to use nix to build some containers, but my dev box is not nixos
<joepie91>
copumpkin: that actually sounds pretty neat and doable
<xplat|work>
how would i do this?
rigelk has joined #nixos
rigelk has quit [(Changing host)]
rigelk has joined #nixos
<joepie91>
"builds too slow? throw in some BTC!"
et4te has joined #nixos
<copumpkin>
xplat|work: linux?
oida has quit [(Ping timeout: 255 seconds)]
<taktoa>
nixcoin
<joepie91>
gchristensen: and you mine buildcoins by submitting nixpkgs PRs? :P
<taktoa>
lol
<aristid>
copumpkin: maybe make an ICO, scam uuuh i mean "attract" a lot of buyers, and use all funds to buy hardware for hydra!
<joepie91>
(I am not entirely serious)
<xplat|work>
yeah, it's debian stretch actually
<joepie91>
(this is unfortunately a necessary footnote)
<clever>
aristid: oops, the site was hacked within 5 minutes of launch, the hacker got all of our funds!
<copumpkin>
xplat|work: so dockerTools should just work
<Sonarpulse>
joepie91: bu bu bu we need to build all those PRs!
<joepie91>
clever: that has got to be one of the most shameful tricks I've ever seen
<gchristensen>
you have to _pay_ buildcoin to submit a pr
<joepie91>
lol
<aristid>
domenkozar: btw will the foundation ever make a financial report? :P
<aristid>
gchristensen: i think you should be able to make security fixes for free, but adding a new package is BTCBTCBTC
<xplat|work>
copumpkin: the docs i found on dockerTools were a little sparse. and ideally i'd also run nix in a container so i could isolate the build-deps
<et4te>
does anyone know which key hydra uses to clone repositories? Machine hosting hydra can clone repos with id_rsa.pub within ~/.ssh, however Hydra appears not to use this because build fails with host key verification error
<gchristensen>
host key verification is _NOT_ ~/.ssh/id_rsa.pub
<clever>
et4te: sudo -u hydra -i, then ssh into the target host to populate ~/.ssh/known_hosts
<clever>
et4te: there is also a nixos option to pre-populate it system wide
<aristid>
gchristensen: and not as a donation, just as a part of the lineup.
<aristid>
gchristensen: i wouldn't be a customer though, because i don't need such things privately :P
<gchristensen>
aristid: oh I don't work at packet :)
<aristid>
gchristensen: huh, you seem to have some kind of relationship with them though?
<LnL>
heh :)
<clever>
bbl
<gchristensen>
aristid: I'm quite good at writing proposals to businesses and developing those relationships
<aristid>
anyways, i'm not going to resume my donation to the nixos foundation until it has better (any, really) reporting.
<gchristensen>
Sonarpulse, LnL, adisbladis[m], copumpkin -- looks like we're going to scale up in alternative means for now, I'll let you know later if we're going to go the T2 route
<Sonarpulse>
gchristensen: ok sounds good!
<et4te>
clever: thanks for the info, so its hydra rather than root user whose SSH key matters when cloning repos? I'm not sure what known hosts do here, the scenario i have is simply getting hydra cloning from private repos. So I just added ssh keys (public) to bitbucket expecting it to work as normal.
<gchristensen>
but if you want to send me cash anyway ... :P
<copumpkin>
domenkozar: I'm seeing a lot of idle macs and no builds on them from nixpkgs:trunk. Should I be concerned?
<et4te>
honestly i wouldn't be averse to donating some BTC or something
<et4te>
so that documentation is improved
<gchristensen>
et4te: the user's SSH key has nothing to do with this error, as I've told you many times
<copumpkin>
gchristensen: thanks :)
<et4te>
gchristensen: yes but I didn't understand your answer
pie_ has quit [(Ping timeout: 255 seconds)]
<et4te>
gchristensen: hence why I asked the question multiple times
<gchristensen>
why didn't you say so? :P
<xplat|work>
et4te really here you need to be looking at the docs for ssh, rather than for hydra or nixos
<et4te>
its hard to know what you don't understand at first, thats why
<gchristensen>
you said you didn't understand, so you knew you didn't understand, so you should have told me you didn't understand :)
<et4te>
lol
erictapen has quit [(Ping timeout: 246 seconds)]
<et4te>
xplat|work: that may be the case but its with running hydra that the issue arose, and thus i thought it would be ok to ask here
<xplat|work>
it is ok to ask here afaik
<xplat|work>
i'm just telling you which docs have answers
<symphorien>
is there an equivalent of builtins.tryEval but for building instead of evaluating ?
<Sonarpulse>
copumpkin: (forgot to ping back) I am building the tip of staging + one commit
<Sonarpulse>
boootstrapp tools is /nix/store/49g980lqdsy92rji8bw3qlj7qk24a82f-bootstrap-tools/lib/libSystem.B.dylib
<xplat|work>
basically files beginning with id_ are the ones for verifying the identity of the *user* whose .ssh directory it is, and known_hosts is used to verify the identity of the machines they ssh into
<Sonarpulse>
I see the comment in there explaining what happend
<Sonarpulse>
very odd
<Sonarpulse>
oh, time for ppl here to eat --- poor timing on my part my bad
<xplat|work>
copumpkin: so i want to build nix expressions that build containers, but i also want to do the building in a container
<sphalerite[m]>
Anyone know a good self-hostable service that's similar to owncloud but not a PHP nightmare to maintain? Maybe something with an existing NixOS module?
<et4te>
so yea, generating a key to ~/.ssh for id_rsa + id_rsa.pub for the _hydra_ user solved the problem
<LnL>
xplat|work: nix runs fine in a docker container
<et4te>
thank you clever
<xplat|work>
LnL: is there a good image to start with? i found nixos/nix but people have reported that some things from nixpkgs don't build right in it
<aristid>
domenkozar: i guess i'm lucky, the bisect is almost finished
<aristid>
despite staging
<et4te>
gchristensen: so yeah the users SSH key had everything to do with this error btw
<gchristensen>
that is not what host key verification failures are
<copumpkin>
domenkozar: I don't think the latest evaluation is doing anything for darwin :(
<gchristensen>
but perhaps you solved that problem while doing something else
<et4te>
well, generating an SSH key solved the host key verification problem...
erictapen has joined #nixos
<gchristensen>
ok, if you insist, I'm not interested in debating it
<et4te>
nope, i generated an SSH key for the hydra user, and added the id_rsa.pub to bitbucket
<et4te>
i'm just letting you know
<xplat|work>
LnL: i guess the main issue with nixos/nix is it's nix-on-alpine and the version of alpine is outdated enough that a lot of modern ssl certs are unrecognized
<LnL>
an overview of the running steps would also be nice
<copumpkin>
et4te: basically it's a different model for the hashing/dependencies in nix
<LnL>
that makes more sense then a build when rebuilding
<copumpkin>
et4te: not actually implemented in any released version of nix
ottidmes has quit [(Quit: WeeChat 1.7.1)]
<sphalerite[m]>
Is there a simple way to describe the difference between the intensional and extensional models? I've taken a shot at reading the thesis but didn't really understand that part.
<et4te>
sure, extensional is closed world, intensional has an implicit context
<et4te>
as in, as far as i understand it extensionality is simply description of structure, like an n dimensional matrix whereas intensionality is a mapping of possible worlds, but simplified values with arbitrary contexts
<sphalerite[m]>
bash: curl: command not found`
<sphalerite[m]>
\>_>
<aristid>
i don't quite understand why
<aristid>
sphalerite[m]: you need to have curl in your linux install...
<niksnut>
sphalerite[m]: intensional == totally content-addressed nix store (so a store path hash equals the hash of the contents of that path)
<dash>
sphalerite[m]: hash-of-inputs vs hash-of-outputs, basically
<niksnut>
right
<sphalerite[m]>
aristid: yeah I realise that, but I didn't realise that curl isn't installed by default in a debootstrapped debian :p
Ivanych has quit [(Ping timeout: 246 seconds)]
<dash>
sphalerite[m]: wget more likely to be there than curl
jgertm has joined #nixos
<sphalerite[m]>
I see. Sort of.
<sphalerite[m]>
dash: indeed it is there. Oh well, I've installed curl now.
tokudan has quit [(Quit: Leaving)]
tvon has joined #nixos
<sphalerite[m]>
Will the installer script set up a multi-user install if run as root?
<sphalerite[m]>
`sh: sorry, there is no binary distribution of Nix for your platform` :(
<sphalerite[m]>
What does platform mean in this case?
tvon has quit [(Read error: Connection reset by peer)]
tvon has joined #nixos
<sphalerite[m]>
I thought there were binaries available for ARM
<gchristensen>
our installer doesn't handle that yet :/
<LnL>
you can create tar of the closure and copy that to the host to boostrap
FRidh has quit [(Quit: Konversation terminated!)]
hamishmack has quit [(Quit: hamishmack)]
<sphalerite[m]>
ooh ok
<sphalerite[m]>
But where do I get the ARM binaries and their closure?
<LnL>
you can download it from a cache with nix-store -r
pie_ has joined #nixos
<LnL>
eg. nix-store -r /nix/store/36n4gjxwln1v9i660g4nn4z6ibq1xvx7-nix-1.11.13 for aarch64-linux
<sphalerite[m]>
And how do I get the derivation? :p
<LnL>
and then nix-store -qR /nix/store/36n4gjxwln1v9i660g4nn4z6ibq1xvx7-nix-1.11.13 | tar -czf nix-closure.tar.gz -T -
<copumpkin>
the latest nixpkgs:trunk linux build is looking pretty screwy
<aristid>
i really don't understand why fixing the build of re2c would break the nixos NAT test
<LnL>
perl keeps failing
<LnL>
recipe for target 'manifypods' failed
<LnL>
not sure what that's about
<copumpkin>
it isn't failing on my local machine :(
<aristid>
copumpkin: and now the committ that seemed to succeed is failing too
<Sonarpulse>
gchristensen: has anyone mentioned that nix-profile-daemon.sh uses a coreutils-only argument to stat?
<Sonarpulse>
while the topic of the day is my new Bash features, figured I'd mention it :D
<sphalerite[m]>
clever: thanks!
<nh2>
aszlig: I have a situation with nixops where I set up a Hetzner but specified a nonexistent device for a partition. Now it's stuck in `Starting / Obsolete`. I don't seem to be able to tell it to start anew from rescue mode. Do you know what to do in this case?
<copumpkin>
niksnut: yay! any idea what the perl failure is?
<copumpkin>
niksnut: the latest nixpkgs:trunk evaluation is failing basically all of the linux builds because a bootstrap perl fails for a reason I can't reproduce locally
<copumpkin>
but wait, doesn't hydra run with --cores 1?
ebzzry has joined #nixos
<niksnut>
really all makefiles will fail at some point with -j
<niksnut>
no
<copumpkin>
hmm
<copumpkin>
niksnut: so will hydra recover now that the perl build is alive again?
<copumpkin>
or does it need a jolt?
<clever>
may need to go up to the eval and restart all aborted jobs
<copumpkin>
I'm never quite sure when something gets cached as a failure and won't retry
<gchristensen>
if you restart an individually failed job it'll restart
<gchristensen>
if something (A) depends on a job (B) that failed and you restart the failed job (B), you also have to restartA
<copumpkin>
but all the things depending on it and the intermediate build steps will also restart?
<copumpkin>
ah, so we need to restart the entire thing
<clever>
copumpkin: the eval has a dedicated option, "restart all aborted jobs"
<copumpkin>
"Restart all failed builds"?
<copumpkin>
that's all I see
<clever>
that one also works
<gchristensen>
yeah
<clever>
depends on what state they are in
<Infinisil>
Umm, is also for others the case that nixos.org uses HTTP?
* copumpkin
presses that button
<copumpkin>
Infinisil: it does use http unless you ask for https
<copumpkin>
probably good to redirect and HSTS it all
<Infinisil>
copumpkin: Well.. That seems pretty bad..
<copumpkin>
the threat being that someone observes you reading nixos manual or intercepts your conection and convinces you to install something that isn't nix, while making you think it is nix?
<Infinisil>
At least the presented installer script uses https explicitly.. but still, it's 2017
<copumpkin>
I don't deny that it'd be nicer to do https everywhere
<clever>
copumpkin: what about the url's to the installer iso's?
<copumpkin>
there's just a bajillion things to do and limited folks who can administer the site, all of whom have lots of other things to do :)
<gchristensen>
Infinisil: you can send prs to the nixos-org-configuration repo (or similarly named)
<clever>
so hydra cant pick the best one automatically
<wak-work>
clever: is that reference a cross build?
<wak-work>
idk how nixpkgs works anymore for cross compiling
<clever>
the above nix expression is a native build
<clever>
everything will be done on an armv7 build slave
<wak-work>
yeah that makes sense then
<wak-work>
idk what else you would want
<clever>
the x86 is probably faster, and hydra could download it without involving any build slaves
<clever>
but mixing the arches right can be tricky
<wak-work>
well yeah if you had a system for seeding fixed-output stuff
<copumpkin>
wak-work: if/when you get around to forking nix itself, you can call it tritonix
<wak-work>
lol
<wak-work>
we have some patches on top
<wak-work>
but im trying to avoid anything major
<copumpkin>
what patches?
<wak-work>
they might not even be necessary anymore
<wak-work>
i've been working on one to actually implement all of the fetchurl functionality in nix
<copumpkin>
what for?
<wak-work>
so you don't have to bootstrap it
<copumpkin>
like the builtins.fetchurl?
<copumpkin>
oh
<wak-work>
yeah
Khetzal has quit [(Ping timeout: 258 seconds)]
<wak-work>
extending builtins.fetchurl to not suck
cpennington has quit [(Ping timeout: 268 seconds)]
<copumpkin>
well, the eval-time one or the <nix/fetchurl> ?
<clever>
<nix/fetchurl> cant unpack a tar, only a nar
<copumpkin>
yeah
<wak-work>
yeah
<wak-work>
fetchurl can't either in trition
<wak-work>
oh wait
<wak-work>
yeah it can
<wak-work>
jkl
<clever>
so you need to provide a nar that contains a pre-built tar
<wak-work>
no
<wak-work>
why are you unpacking in fetchurl typically?
<clever>
if you want to create the bootstrap tools, like the initial gcc for building stdenv
<wak-work>
you mean to unpack the bootstrap tools?
<clever>
yeah
<wak-work>
that's not done in fetchurl though
<clever>
yeah, right now its a busybox and a shell script
<clever>
and the busybox supports ash, xz, and tar
<wak-work>
yep
erictapen has joined #nixos
<catern>
hey #nixos
<gchristensen>
hi catern
<catern>
I want to pass a list of key-value pairs to a builder script
<Infinisil>
catern: Good evening
<clever>
catern: every attribute on a derivation becomes an env variable at build time
<catern>
clever: but
<catern>
I want to build it programmatically
<copumpkin>
catern: if you want something more structured, you can also spit out some JSON into the builder and have something inside it consume it
<catern>
oh, hmm
<copumpkin>
you can also do that
<catern>
I guess I can just use args@{ stdenv, etc. }
<copumpkin>
mkDerivation ({ /* some attrs */ } // some-expression-producing-attrs)
<catern>
then // args
<catern>
(oh i didn't mention that that list of key-value pairs is my arguments)
<catern>
i'll do it
<catern>
thanks #nixos
<clever>
sphalerite[m]: my armv7 gcc has started ubilding again
<catern>
oh right
<catern>
the issue is that I also want the list
<catern>
er, I want both the list, and the names in the list
<catern>
er...
<clever>
yeah, it can be tricky to know which env vars where added
<clever>
json might be simpler
<LnL>
clever: talking about arm builds, did you increase the build timeout in hydra?
<catern>
I have deps "foo", "bar", and "baz", and I want to both have the derivation paths corresponding to them, and the names themselves, because the names are meaningful to my build system
<LnL>
mine keeps aborting
<catern>
hmm, should I extract the name from the derivation path?
<clever>
LnL: ive seen my hydra run a build for 48 hours before, and pass
<LnL>
I get the timed out after 7200 seconds of silence when building with hydra
<clever>
catern: using this, you can turn a set into a series of shell commands
sidpatil has joined #nixos
<clever>
LnL: ah, my arm build of llvm didnt go silent for that long
<clever>
it was chatty the whole 48 hours, lol
<LnL>
:/
matt__ has joined #nixos
<catern>
clever: neat, but scary
<Infinisil>
I wondered about this for a while now. nixos.org/nix /nixpkgs and /nixops all have direct links to the manual at the header, but nixos.org doesn't, you need to click "Get NixOS" and then "Next Step: Manual"
matt__ is now known as lostInTheMatrix
<taktoa>
Infinisil: nixos.org/nixos/manual
<Infinisil>
Very inconvenienc
<Infinisil>
taktoa: Yeah, but it's not in the header of nixos.org as with all others
<taktoa>
oh wait nevermind, I get what you were saying
<taktoa>
yeah
Sonarpulse has quit [(Ping timeout: 246 seconds)]
<lostInTheMatrix>
hi I can't do anything with nox/nix-env since I now always have "Upstream has ceased free support for grsecurity/PaX." though I don't think I enforce grsec. Full log http://paste.ubuntu.com/25165845/ (17.03 upgraded from 16.XX). Googled the pb but couldn't find a solution :'(
<gchristensen>
lostInTheMatrix: hmm can you share your config please?
<gchristensen>
lostInTheMatrix: and how did you see this error? what command did you run? can you paste all of the output?
<lostInTheMatrix>
gchristensen: thanks for the help. The full output was already in http://paste.ubuntu.com/25165845/ . It happens when I do $ nox <program>
python476 has quit [(Ping timeout: 240 seconds)]
<gchristensen>
ohh weird I don't know about nox
<et4te>
gchristensen: you were right earlier. There were two errors, the host verify failure happened as you said because of known_hosts
<gchristensen>
I know :)
<lostInTheMatrix>
hum with nix-env I seem to be able to install (thought It failed too, but that might be because of another problem)
takle has quit [(Remote host closed the connection)]
<et4te>
gchristensen: yeah working beautifully! :D
<gchristensen>
I'm very glad :)
<et4te>
is there a way to pay for documentation to be produced
sigmundv_ has joined #nixos
<lostInTheMatrix>
nice find. I think the problems happened with nix-env too so I didn't check nox tracker but now nix-env seems fine so I can live with it (even if it's weird). thanks for the help
<et4te>
like via BTC or something
<et4te>
maybe 'donation' would be the more appropriate term
<gchristensen>
et4te: unfortunately we're not setup to accept donations in exchange for specific work
<catern>
hey #nixos, I have a strange silly proprietary package manager which I would like to hackily convert into Nix. should I write converter from its package specification into default.nix files? or should I write a Nix expression which reads its package specification and returns a derivation?
<catern>
which would be more idiomatic?
<gchristensen>
catern: the converter route
<gchristensen>
et4te: but we're very happy to help you write documentation ... :)
<catern>
i rescind my last comment about idiomatic, because i definitely know the converter is more idiomatic
<catern>
do you still say converter, if I just care about the best way? :)
<et4te>
gchristensen: ah oke. well sure when I understand things I will try to contribute
<gchristensen>
et4te: you can open issues for underdocumented things
ryanartecona has joined #nixos
<gchristensen>
et4te: from there I guess you could put a Bugbounty on it, but I don't know if many people do that
<gchristensen>
oh I can't even find a bounty thing for it
markus1199 has joined #nixos
lambdael has joined #nixos
<Infinisil>
et4te: Did you encounter anything specific that was underdocumented?
<et4te>
Infinisil: hydra.useSubstitutes was not obvious and non trivial to find out about
<et4te>
the naming confused me because i was looking for something that fixed the binaryCache problem
markus1189 has quit [(Ping timeout: 240 seconds)]
<et4te>
jobsets in hydra, i haven't found a clear description of them anywhere really and how they relate to a package
<NickHu>
Fwiw, the kexec thing doesn't work on scaleway, at least not anymore
<NickHu>
It doesn't find any drives
<catern>
what is the best language for generating Nix code?
<NickHu>
I think scaleway changed recently - all their drives are /dev/nbd rather than /dev/vda now
<Infinisil>
et4te: While I haven't used hydra a lot, it seems that the docs are indeed very sparse
hamishmack_ has joined #nixos
<Infinisil>
catern: No idea, what are you planning to do?
hamishmack has quit [(Ping timeout: 260 seconds)]
hamishmack_ is now known as hamishmack
<copumpkin>
rather than generating nix code, I'd generate json and consume it with nix code
thc202 has quit [(Ping timeout: 255 seconds)]
<copumpkin>
most of the time you have a more restricted data model than what nix can express
<copumpkin>
and json fits better and is easier to generate in other languages
Rotaerk has joined #nixos
<c74d>
Can one access the default value of a NixOS option when setting that option?
<NickHu>
Actually, saying that, it might be because this is the first bare metal server I tried instead of vcloud for the discrepancy
<Infinisil>
c74d: You can use `nixos-option services.openssh.enable`
<c74d>
hm
nslqqq has quit [(Ping timeout: 246 seconds)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] georgewhewell opened pull request #27628: add libcurl to radarr + sonarr wrappers (master...add-libcurl-to-sonarr-radarr-wrapper) https://git.io/v7qwo
NixOS_GitHub has left #nixos []
<c74d>
Infinisil: you mean, I could parse the output of `nixos-option` to get the default value?
<Infinisil>
c74d: Ah you mean programmatically
<c74d>
yes
<Infinisil>
c74d: Well you can evaluate arbitrary nix with nix-instantiate: nix-instantiate --eval -E '(m: m + 2) 3'
<catern>
Infinisil: admittedly that's true but I have lots of experience in other typed functional languages... no, the actual issue is that I need to keep the dependency graph for my generator small
<Infinisil>
catern: Why do you need to keep it small?
<catern>
because it will be a build dep for everything else
<catern>
so, bash it is! :)
<Infinisil>
catern: I would totally use anything other than bash
<Infinisil>
1GB GHC installation, so what, may the people without 1GB of memory not be able to build it then
<catern>
:)
<simpson>
catern: Seriously, do the JSON route; it is *far* easier when you have a common lingua franca for data.