hiciu has quit [Quit: Connection closed for inactivity]
<lb5tr>
hey, i have this problem with repl where imports are not reloaded. For example if if have file deafult.nix then i import it using "import ./default.nix", then change the file and import again, nix doesn't reload the file (i.e. whatever changes i've made to deafault.nix are not reflected)
<lb5tr>
is nix lang _this_ lazy :)?
<lb5tr>
s/this/that
<infinisil>
lb5tr: You need to :l
<clever>
lb5tr: nix uses the path to a filename to cache the result of parsing things, you need :r to clear that cache
<infinisil>
Oh :r right
EmoSpice has joined #nixos
<lb5tr>
ack, thanks!
orivej has quit [Ping timeout: 256 seconds]
martin__ has quit [Quit: Connection closed for inactivity]
orivej has joined #nixos
johnnyfive has quit [Quit: I quit now.]
johnnyfive has joined #nixos
codygman has quit [Read error: Connection reset by peer]
_______________ is now known as parsnip
parsnip is now known as parsnip0
codygman has joined #nixos
afreak has quit [Ping timeout: 240 seconds]
EmoSpice has quit [Ping timeout: 272 seconds]
EmoSpice has joined #nixos
codygman has quit [Read error: Connection reset by peer]
Rusty1 has joined #nixos
codygman has joined #nixos
h0m1 has quit [Quit: WeeChat 2.9]
EmoSpice has quit [Ping timeout: 244 seconds]
h0m1 has joined #nixos
nij has joined #nixos
<{^_^}>
[nixpkgs] @aszlig pushed to master « Unmaintain packages I don't use anymore »: https://git.io/JJE3I
orivej has quit [Ping timeout: 260 seconds]
orivej has joined #nixos
<{^_^}>
[nixpkgs] @aszlig opened pull request #94041 → synergy: Add patch to fix CVE-2020-15117 → https://git.io/JJE3Z
<pie_>
adisbladis: if i have something already build by mkPythonApplication, can I add that to mkPoetryApplication in a way that I get access to the library that is the application?
<pie_>
actually mkpoetryenv, but probably doesnt really matter
EmoSpice has joined #nixos
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/JJE3i
EmoSpice has quit [Ping timeout: 272 seconds]
proofofkeags has joined #nixos
noogie has quit [Remote host closed the connection]
sigmundv has quit [Ping timeout: 240 seconds]
proofofkeags has quit [Ping timeout: 256 seconds]
Heirlung has quit [Ping timeout: 272 seconds]
<{^_^}>
[nixpkgs] @aanderse opened pull request #94043 → nixos/zabbix*: replace extraConfig option with settings option → https://git.io/JJE3N
<{^_^}>
[nixpkgs] @orivej-nixos pushed commit from @orivej to master « sooperlooper: 2019-09-30 -> 1.7.4 »: https://git.io/JJEnL
<nij>
Q: I learn that `nix-env -iA <pkg>` is a dirty way to do things in nixos. Is there always a better way in general? Or there are some cases when one indeed needs to do that?
<clever>
nij: add things to environment.systemPackages in configuration.nix
<nij>
clever: one can get everything from doing that as one can get from `nix-env -iA`?
<nij>
If that's the case, then `nix-env -iA` seems to be a way to quickly testing out something.. and it seems that the downloaded stuff should be removed eventually for the system to be clean :D?
<clever>
nij: if you just want to test something, use nix-shell or `nix run`
<nij>
clever: ah! And even 'comma'! As I recalled from a tutorial vid.
<clever>
nix-env requires you to uninstall it with -e and then delete generations, before it can be GC'd
<nij>
clever: what then would be valuable with `nix-env`?
<nij>
`nix-env -iA`*?
<clever>
even if you remove something with `nix-env -e foo`, the generations and rollbacks keep it around
<clever>
so you can undo the removal
growpotkin has joined #nixos
orivej has quit [Ping timeout: 256 seconds]
<{^_^}>
[nixpkgs] @CajuM opened pull request #94049 → Update Vulkan, SPIR-V, glslang and shaderc → https://git.io/JJEnH
orivej has joined #nixos
<DigitalKiwi>
frequently i use nix-env for things i want to evaluate/use longer than a nix-shell is convenient but don't want to nixos-rebuild switch at that time for reasons
<{^_^}>
[nixpkgs] @peterhoeg opened pull request #94050 → dict: look for config in /etc → https://git.io/JJECk
<{^_^}>
[nixpkgs] @peterhoeg pushed to f/dict « dict: look for config in /etc »: https://git.io/JJECI
<colemickens>
Is there a world where Nix would evolve differently if it didn't support Darwin? I wonder if Guix will especially with some of their community's focus on rootless building.
<colemickens>
(I was thinking of userns-like stuff?)
EmoSpice has joined #nixos
lord| has quit [Read error: Connection reset by peer]
<{^_^}>
[mobile-nixos] @samueldr pushed 4 commits to master: https://git.io/JJE8V
<Church->
Oh there's a nixfmt?
<Church->
Now that's something I can use
<DigitalKiwi>
there's also nixpkgs-fmt
<Church->
Nifty
<DigitalKiwi>
i use both, depending on what i am working on
<Graypup_>
huh, interesting, my statically linked haskell program is pulling in ghc-8.6.5-doc
<DigitalKiwi>
more precisely; nixpkgs-fmt if it's going into nixpkgs and nixfmt for most other things
<Graypup_>
due to NIX_GHC_DOCDIR getting stuffed into the executable somewhere
<Graypup_>
it was also pulling in every library package that had binaries and their respective dependencies (not staticly linking those)
<jtojnar>
cole-h: yeah "GNOME session X11 services is not active" again
<jtojnar>
so it failed to start earlier
<dxb[m]>
anyone using pywal for system colors? i'm not too sure how to set that up with home-manager
<benley>
hm. If I use conf-mode for .nix files instead of nix-mode, emacs is lightning quick with all-packages.nix
<DigitalKiwi>
jtojnar: https://github.com/NixOS/nix/issues/3799 i hit this *a lot* it's basically made the aarch64.nixos.community not useful, anything i can do to help?
<{^_^}>
nix#3799 (by jtojnar, 2 weeks ago, open): hash '' has wrong length for hash type 'md5'
<benley>
as soon as I tell it to use nix-mode it slows waayyyyyyyy down again
growpotkin has quit [Ping timeout: 264 seconds]
endformationage has quit [Quit: WeeChat 2.7.1]
ddellacosta has quit [Ping timeout: 260 seconds]
<DigitalKiwi>
benley: i guess just open it once and never close it again ;D
EmoSpice has joined #nixos
<Graypup_>
hmmm really weird: I put in my nix file `executableSystemDepends = [ self.diagrams-builder ];` ie not the haskell-packages one, and it is not in the nix-store -q --requisites list??
stoile has joined #nixos
domogled has joined #nixos
<jtojnar>
DigitalKiwi: you could try figuring out what prints that error, setting up a gdb breakpoint…
<jtojnar>
if it comes from the box itself, it will be much harder
<DigitalKiwi>
i think we should git bisect the box
<DigitalKiwi>
lol the git blame has a lot of "Fix and cleanup hash parser" a month ago ... i wonder if it did not fix it quite enough or if it's not in nixUnstable yet... (not sure what nixUnstable tracks??)
<DigitalKiwi>
Ericson2314: ohhi
<{^_^}>
[nixpkgs] @cript0nauta opened pull request #94054 → Add nplusone and pytest-pythonpath Python packages → https://git.io/JJE4l
domogled has quit [Remote host closed the connection]
o1lo01ol1o has joined #nixos
domogled has joined #nixos
pkral has joined #nixos
o1lo01ol1o has quit [Ping timeout: 256 seconds]
noudle has joined #nixos
gthm has joined #nixos
<{^_^}>
[nixpkgs] @DieGoldeneEnte opened pull request #94061 → llvmPackages_{8..10}: comment out broken manpages → https://git.io/JJEzY
cjpbirkbeck has joined #nixos
alp has joined #nixos
evernite has quit [Ping timeout: 265 seconds]
evernite has joined #nixos
fendor has joined #nixos
growpotkin has quit [Ping timeout: 240 seconds]
<bpye>
Hey, is there any example of how to override the src input for a package? I want to try the latest git commit to see if it fixes a bug I see. I’ve tried an overlay with overrideAttrs but it seems to have no effect, I can change the input but even when the fetchFromGitHub is clearly bogus it doesn’t error
<srhb>
bpye: overrideAttrs is usually right, what did you try specifically?
<Church->
Overlays always work
<Church->
What I do when testing new versions out.
<bpye>
I later reference pkgs.brlaserGit, I get a renamed package but I do not seem to be trying to pull down the source from GitHub because that would certainly fail
<graf_blutwurst>
Good Morning. Anyone else got issue with the newest slack package saying it has no connection?
<srhb>
graf_blutwurst: Yes, I have an upgrade ready, but haven't had time to PR it yet.. One moment..
<srhb>
bpye: You're using the brlaser attribute in your nixos config?
<xps>
I'm trying to use an overlay to add rav1e support to ffmpeg-full - the build fails claiming that rav1e could not be found using pkg-config. I don't know where to go from here, any ideas? ttps://privatebin.net/?3592175d5580b3c2#D9wDtGZkKYVJ2fmyNSJfmM823KWtBNttpFaoeYwaCz39
meh` has joined #nixos
<{^_^}>
[nixpkgs] @Mic92 merged pull request #93581 → add github action to wait for ofborg → https://git.io/JJCdA
<{^_^}>
[hydra] @edolstra pushed to no-nar-accessors « Keep the SHA-1 column in existing installations »: https://git.io/JJErH
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
laerling has joined #nixos
laerling has quit [Changing host]
<typetetris>
I am really confused with a nixos module test I made. When I do `nix-build -A driver test.nix` and execute `result/bin/nixos-test-driver` the machine looks like the test already did run, but it shouldn't have, as there is no systemd-unit defined in it for it. Just the `testScript` thing.
<typetetris>
Does it impurely get the disk from somewhere else?
hoijui has quit [Ping timeout: 260 seconds]
o1lo01ol1o has joined #nixos
<lejonet>
How does nix handle having nix-like syntax inside of a multiline string and how can I tell it to leave it be? I have this conf: https://dpaste.com/BA7SBSW3N where I'm trying to write a rsyslog config, which includes the very nix-like if () then {} construct, which is making nixos-rebuild erroring out on me with "error: syntax error, unexpected ')', expecting ';', at /etc/nixos/configuration.nix:60:43" where
<lejonet>
line 60 is the row with the if () then {} in it
<lejonet>
is it because its enclosed in a {} that makes the parser think its a nix construct, even tho its neither prepended with $ nor is the {} anywhere near the if
zupo has joined #nixos
<lejonet>
if I remove the three lines that compose the if () then {}, it doesn't whine at all
EmoSpice has quit [Ping timeout: 246 seconds]
<typetetris>
Okay. it definitely stored the machine somewhere, in systemd logs are multiple reboots ...
<siraben>
Hi all, I've been trying to get this to build but it fails to find imagemagick
ap29600 has quit [Read error: Connection reset by peer]
ap29600 has joined #nixos
itshaydendev has joined #nixos
<srk>
also nix edit
ap29600 has quit [Client Quit]
ap29600 has joined #nixos
<lejonet>
siraben: sorry for the absence, but you'd have to send in an older version to the package, perhaps with packageOverrides
<{^_^}>
[nix] @edolstra pushed to master « Bump version to 3.0 »: https://git.io/JJEXM
asheshambasta has quit [Remote host closed the connection]
<lejonet>
However if you're making the derivation yourself, then "all" you need to do is modify the callPackage call to also include an older version in the call, i.e. callPackage /path/to/package { asciidoc = asciidoc-older; } instead of just simply callPackage /path/to/package { }
<lejonet>
However, you'd have to create the asciidoc-older package yourself, to my memory we don't have any versioning of it in nixpkgs, like we do with like boost and similar
<fergie>
I am trying to use rstan, an R package. Part of what it does is turn a model specification into C++ which (I think) is then compiled. On first use, it does some checks and I get the following error:
<sdisavona>
*note* I'm using just nix and not nixos. Programs installed via nix have huge .desktop files, like 250 lines, because the repeat many lines for each language.
<sdisavona>
Is there any way to change this behaviour and have .desktop in English only?
<{^_^}>
[hydra] @edolstra pushed to no-nar-accessors « Copy paths in the right order »: https://git.io/JJEM0
<simpson>
sdisavona: Currently looking at how they're generated. Is this a serious space problem?
pmy has quit [Ping timeout: 240 seconds]
<infinisil>
I doubt that's specific to nixpkgs
<sdisavona>
It's not about space in my case, it's about making a script to search through .desktop
<sdisavona>
Does it depend on the source?
evanjs has quit [Read error: Connection reset by peer]
<infinisil>
Yeah I think these desktop files are usually just copied from the source
<simpson>
I think so. We don't have our own l10n structure for building those files.
<infinisil>
Sometimes they're generated with Nix, but then they only include the very basic stuff
evanjs has joined #nixos
pmy has joined #nixos
dermetfan has quit [Ping timeout: 240 seconds]
orivej has quit [Ping timeout: 240 seconds]
<srxl>
Got some software that I'm not quite sure how to make a derivation for. It's comprised of mostly binaries, but also has some python scripts that expect to be in a virtualenv. Any tips?
<Guest79073>
What is the general approach in nixos to setting environment variables like these, that define paths to plugins common to many packages (some of which can be installed system-wide, some to user profiles)? https://github.com/NixOS/nixpkgs/issues/45663#issuecomment-665003249
ap29600 has quit [Read error: Connection reset by peer]
ap29600 has joined #nixos
ap29600 has quit [Read error: Connection reset by peer]
ap29600 has joined #nixos
jlv has left #nixos ["User left"]
jlv has joined #nixos
<inquisitiv3>
I've installed an application on my Fedora machine using Home Manager by adding the application to the `home.packages` list. If I would like to report a bug, should I do that against Home Manager xor Nixpkgs?
EmoSpice has joined #nixos
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<makefu>
inquisitiv3: depends on the issue you encounter, if it is related to the package then nixpkgs would be correct
<nbp>
infinisil: Who needs builtins.fromTOML, really?
<makefu>
if it is related to the location of certain files, maybe home-manager
<infinisil>
nbp: poetry2nix for one, I think the mozilla-nixpkgs rust overlay too
orivej has quit [Ping timeout: 246 seconds]
orivej has joined #nixos
<infinisil>
Oh and probably all other <rust>2nix tools
<nbp>
infinisil: hum … I wonder who made the mozilla-nixpkgs rust overlay :P
<inquisitiv3>
<makefu "inquisitiv3: depends on the issu"> Thanks!
<inquisitiv3>
@makef
<aml>
Hi! I've just started using NixOS and there are some cases where I don't want to start services automatically but rather run systemctl start service. How is that usually handled? My specific use cases are for docker and wireguard.
<nbp>
aml: use `systemd.services.<name>.enable`, IIRC
<nbp>
^ apparently no
<Taneb>
Is there a way to query what signatures a path has been signed with?
<infinisil>
aml: I think `systemd.services.<name>.wantedBy = lib.mkForce []` (this removes the standard multi-user.target wantedBy)
fergie has quit [Remote host closed the connection]
<srhb>
Taneb: timestamp on the system symlink maybe?
<srhb>
Only works for the last one though.
bqv has joined #nixos
o1lo01ol1o has quit [Remote host closed the connection]
domogled has quit [Ping timeout: 260 seconds]
pmy has quit [Ping timeout: 265 seconds]
<immae>
Taneb: `sudo nix-env -p /nix/var/nix/profiles/system --list-generations` could give an approximation (although I think it gives build time, not switch time)
<immae>
if you never do "only build" it might be accurate
<antifuchs>
(not sure if my previous message made it here, my client let me send it but I wasn't joined? I'll resend, sorry for the dup if it did)
<antifuchs>
I'm getting started with nixos/pkgs/home-manager, and am wondering if anyone has produced a Dash/zealdocs docset of nix projects' documentation yet - it's how I access most of my docs, and would love to do the same for nix (:
<ScottHDev>
Hello, I'm trying to build a docker image with nix, I tried with this expression based on alpine linux : https://termbin.com/yhal. But I get run-as-root.sh: line 5: apk: command not found.
karolus has quit [Client Quit]
cosimone has joined #nixos
Kk2 has joined #nixos
<{^_^}>
[nixpkgs] @ehmry pushed to master « acme: build for aarch64 »: https://git.io/JJEFc
proofofkeags has quit [Remote host closed the connection]
proofofkeags has joined #nixos
<{^_^}>
[nixpkgs] @Mic92 merged pull request #94079 → github-action/ofborg: only run in main repo → https://git.io/JJE50
<{^_^}>
[nixpkgs] @jonringer pushed commit from @ruuda to master « python3Packages.sentry-sdk: fix trytond dependency »: https://git.io/JJue0
<{^_^}>
[nixpkgs] @teto merged pull request #94050 → dict: look for config in /etc → https://git.io/JJECk
<{^_^}>
[nixpkgs] @teto pushed commit from @peterhoeg to master « dict: look for config in /etc (#94050) »: https://git.io/JJueK
wnklmnn has quit [Remote host closed the connection]
philr_ has quit [Ping timeout: 240 seconds]
ap29600 has joined #nixos
<{^_^}>
[nixos-homepage] @github-actions[bot] pushed commit from GitHub Actions to master « Update flake.lock and blogs.xml [ci skip] »: https://git.io/JJueQ
domogled has quit [Ping timeout: 260 seconds]
domogled has joined #nixos
<neonfuz2>
Hey just wanted to ask if this is a bug or intended behaviour: I'm cloning a repo with nix-prefetch-git, and it outputs fetchSubmodules: false, but looking in the output directory the submodules were fetched
<neonfuz2>
and if I don't set fetchSubmodules to true manually, the hash is different
<neonfuz2>
so nix-prefetch-git is fetching submodules but returning fetchSubmodules: false
<clever>
yorick: ip6tnl0 is an unrelated interface
<yorick>
maybe kernel 4.9.220 is too old for rpi3
<clever>
yorick: what is the vid/pid under usbether@1?, what does lsusb report?
<evanjs>
matthiaskrgr: sorry about the lack of updates. I've been having issues building with offline_tests and I'm going to see if #93128 changes things
<clever>
you can also use `hexdump -C` to dump any binary files
<evanjs>
but I'd rather handle as we can (within reason) in the nix expression before we start adding OS checks into the code (yuck lol)
<infinisil>
neonfuz2: Oh damn, didn't know or expect that either
<infinisil>
fetchFromGitHub doesn't fetch submodules by default (in which case it doesn't use fetchgit but fetchzip)
<clever>
yorick: ah, i want compatible, not vid/pid
<yorick>
clever: usb424,ec00
m4ts has quit [Quit: bye]
<evanjs>
speaking of which... I realized that fetchGit is still not used like anywhere
m4ts has joined #nixos
<clever>
yorick: yeah, everything looks normal, all i can guess is that the kernel/drivers arent right
<evanjs>
is there a point at which it will be preferred over fetchgit? Maybe after recurse submodules is available in a stable release?
<evanjs>
wasn't sure how it would compare usage-wise, at least in terms of best practices and etc
<yorick>
clever: I also have a Revision: 0000 in my /proc/cpuinfo
<clever>
yorick: is the ethernet driver built into the kernel or a module in `lsmod` ?
<clever>
revision shouldnt be 0
<yorick>
builtin, I think
<infinisil>
evanjs: I think it maybe shouldn't be used in nixpkgs, because it slows down (I think) evaluation
<clever>
yorick: can you add `lsmod` to the gist?
<yorick>
clever: done
<clever>
yorick: and also `hexdump -C /proc/device-tree/system/linux,revision` ?
<yorick>
clever: there's no /proc/device-tree/system
<evanjs>
infinisil: to clarify, do you mean fetchGit, or the recursion? I vaguely remember dolstra maybe bringing that up, and that it's defaulted to false or something, idk
<clever>
yorick: poke around , is system in another dir?
<infinisil>
evanjs: builtins.fetchGit
<yorick>
no
<clever>
yorick: thats not right, that implies something isnt right with device-tree
<infinisil>
evanjs: What recursion?
<{^_^}>
[nixpkgs] @kolaente opened pull request #94092 → flutter beta (1.19.0-4.3.pre -> 1.20.0-7.2.pre), dev (1.20.0-3.0.pre -> 1.21.0-1.0.pre) update → https://git.io/JJukv
<yorick>
clever: I did need to add bcm2837-rpi-3-b.dtb to make this work at all
<evanjs>
infinisil: sorry, fetchSubmodules. the branch name is fetchgit-recursive so idk lol
<{^_^}>
[nixos-homepage] @github-actions[bot] pushed commit from GitHub Actions to master « Update flake.lock and blogs.xml [ci skip] »: https://git.io/JJukU
<clever>
yorick: sounds like the firmware in /boot/ is mixed up some
<yorick>
(I think, it was in the repo already)
Neo--- has joined #nixos
<infinisil>
evanjs: Ah. I think even just fetchGit without submodules slows down eval, since the git fetch is done at eval time. pkgs.fetchgit in comparison happens at build time
<infinisil>
Also, it's a bit impure, since it relies on your git in PATH
<yorick>
clever: ah yes, the firmware
<evanjs>
infinisil: great, thank you! that's what I was wonderingggg
<evanjs>
but yeah, would be super helpful for closed-source stuff at work and any private instances in general
<yorick>
yeah, there's only bootcode.bin
<clever>
yorick: start.elf and fixup.dat are a matched pair and must always be together, the rest is fairly flexible
<evanjs>
I remember trying to fight with fetchgit and, well... bleh
<edef>
infinisil: submodules make sense for data with different privacy properties
<edef>
infinisil: they have a *significant* cost but git doesn't really have any other tools that fit
maier has joined #nixos
<edef>
infinisil: i dislike "just encrypt it and check it in publicly", since that means you need to nervously guard that key for the rest of your life, which probably means it should be on an HSM of some kind
<chiiba>
Is it possible to overlay nginx service configuration from another package? Not in upstream, but for personal use. I want to add a global location declaration.
<edef>
infinisil: (because otherwise you can never prove key destruction with any certainty)
<edef>
infinisil: (and the ACLs are not part of the version-controlled history there, and rotating secret material within GOOG is generally quite easy)
<infinisil>
edef: I haven't played around with git subtrees, how do you commit and push changes to subtrees? (can you even do that or do you need a separate checkout?)
<edef>
infinisil: you just commit
<Guest10762>
NixOS was name dropped in the current #hope2020 talk
<edef>
infinisil: like, it's not anything special, it's just git merges that happen to rename all the files to have a common prefix
bqv has quit [Ping timeout: 240 seconds]
bkv is now known as bqv
<porphyrogenetos>
I recently switched from a single user to a multi-user install of nix (on a non-nixos distro) and now I cant build docker/podman images with nix. Its failing at the point that it tries to boot into the image with KVM to run the runAsRoot section, with the error "qemu-system-x86_64: failed to initialize kvm: Permission denied"
<edef>
infinisil: so you can just behave as if it's one big repository
<edef>
infinisil: (because it is)
<porphyrogenetos>
Im guessing its because the nixbuilders dont have the right group membership but im trying to figure out the nixy way to fix that
<infinisil>
edef: So if you have some changes in the main tree and some in the subtree and you `git add -A && git commit`, would it do two separate commits for the subtree and the main tree?
<infinisil>
I should really just play around with it shouldn't I
<edef>
infinisil: no
<edef>
infinisil: it's one commit
<edef>
infinisil: imagine you just copied the other repository over and committed
<edef>
infinisil: you have one git tree now, right?
<infinisil>
Yeah
<edef>
infinisil: now imagine that, but you give that commit another parent — namely the commit whose contents you just merged
cosimone has joined #nixos
mattlevan has quit [Quit: WeeChat 2.7.1]
<edef>
infinisil: restarting and approaching from the other direction — imagine you went to your nixpkgs tree and did `git mv * nixpkgs/`
<edef>
infinisil: and committed that
<edef>
infinisil: and then kept working
<infinisil>
Hm I get that I think. But then my question is: Can you push the changes you did to the subtree to the original remote you imported from?
<edef>
infinisil: and you can still merge upstream because, well, the only divergence is the path prefix
michiel__ has quit [Remote host closed the connection]
<edef>
infinisil: re: upstreaming patches, i just `git format-patch --relative`
michiel__ has joined #nixos
<edef>
infinisil: and that generates patches i can apply to upstream
<edef>
infinisil: (with git am)
<infinisil>
Hm I see. So there's nothing builtin for `git push` ing subtrees. It's pull & merge only for subtree remotes
<edef>
infinisil: which is effectively a cherry-pick (cherry-pick is just format-patch + am)
<edef>
infinisil: yes, correct
<infinisil>
Got it
<edef>
infinisil: both cherry-pick and merge work in the other direction
bqv has quit [Read error: No route to host]
bqv has joined #nixos
griff__ has quit [Quit: griff__]
<infinisil>
So if my main tree is private and the subtree is public, this might be pretty annoying to push public updates without some good git wrappers
<edef>
infinisil: actually, i *think* cherry-pick actually works both ways
bqv has quit [Read error: Connection reset by peer]
bkv has joined #nixos
<edef>
infinisil: my git alternate setup just doesn't *quite* work with it
<infinisil>
And the subtree repo would have to be checked out separately somewhere else for me to apply and push the patches
<edef>
yes
<edef>
although if you're emailing patches you don't really need to
<infinisil>
I see
mattlevan has joined #nixos
<yorick>
clever: ... oh
orivej has quit [Ping timeout: 246 seconds]
<edef>
i have a perl script that lets me go "hey gimme upstream, except with this mailable patch"
<edef>
it means i have to touch github as little as possible
orivej has joined #nixos
<infinisil>
edef: I conclude that git subtrees could be pretty nice if most changes you do are in the main tree
<edef>
(i consider github's UI at best a creative form of torture)
<clever>
yorick: (facepalm)
<edef>
infinisil: most changes i do marinate in my tree for a while anyway
<dtz>
Lmao edef
<infinisil>
edef: Hehe "marinate", very fitting
<infinisil>
Some of my nixpkgs changes do too
<edef>
infinisil: generally they see a few revisions before i upstream them
<infinisil>
(though I mostly forget about them really)
<dtz>
*nods*
<edef>
infinisil: although, conversely, sometimes i pick something off one of my trees and decide it needs to be a touch cleaner to be acceptable for upstream
<edef>
infinisil: and then i do a revert+cherry-pick to fix up my tree
<edef>
infinisil: i have two or three trees worth caring about (work, personal, and an infrastructure project with friends)
<infinisil>
I see
<edef>
infinisil: personal tree uses full nixpkgs history, the others use squashed history, with git replaces used to graft full history in locally and transparently
<edef>
infinisil: flokli and i upstream stuff from the work tree with some regularity
<{^_^}>
[nixpkgs] @worldofpeace pushed 2 commits to master: https://git.io/JJutZ
<yorick>
clever: still I learned a lot about device trees
<Ke>
in the configuration,nix
alp has joined #nixos
<Ke>
or whatever you build your system with, you can't change it at runtime, at least if you don't even have it installed
<Ke>
I guess there might be some integration to grub menus for previous generations that have not been purged
<ScottHDev>
Hello, I'm running a jekyll server in a docker container. From the logs I know that the server is running accessible on http://127.0.0.1:4000 so I just remapped the port with -p 8080:4000 and I tried to connect to http://127.0.0.1:8080 but can't reach the server. So I tried opening port 8080 with the firewall (networking.firewall.allowedTCPPorts)
<ScottHDev>
but still not working. Do you have any ideas what's going on?
<sshow>
any thoughts on nixops vs morph vs krops?
<numkem>
I'm getting crazy high memory usage trying to run nix-index on unstable-small
artemist has quit [Ping timeout: 240 seconds]
rajivr has quit [Quit: Connection closed for inactivity]
<lassulus>
sshow: krops is when you deploy from multiple computers or want to be able to debug the target system manually with nixos-rebuild switch. haven't used the other ones though
<ScottHDev>
nix-index is huge
<ScottHDev>
But you just have to run it once
griff_ has joined #nixos
<numkem>
42GB of ram almost
<infinisil>
sshow: I wasn't enticed by any of them, so I started making my own :P https://github.com/Infinisil/nixus (very experimental though, not recommended for production, also not much docs for now)
<Profpatsch>
kalbasit: I’m not a darwin user, so I did not notice.
<evanjs>
missed opportunity
<Profpatsch>
There was a test failing when we updated, but the build itself worked.
<evanjs>
just saw bqv's comment on sops-nix lol
<Profpatsch>
I think it was the protobuf test
<{^_^}>
[nixpkgs] @svend opened pull request #94097 → dual-function-keys: init at 1.1.0 → https://git.io/JJutH
<evanjs>
like "Hey, I know that guy" :P
artemist has joined #nixos
<evanjs>
I also need to check to make sure I'm not scrolled back before replying lol
<kalbasit>
Profpatsch: The Bazel 3 commit works, it's the latest master that's broken.
<Profpatsch>
kalbasit: Ah, ok
<Profpatsch>
Then it’s not my fault :P
<kalbasit>
Profpatsch: I believe it's related to all the C stack changes that went in the repo recently
fendor_ has quit [Ping timeout: 240 seconds]
<kalbasit>
Profpatsch: I did not say it was :)
<Profpatsch>
We unfortunately don’t have reverse dependency tests for everything in ofborg, because wow that would be expensive.
<Profpatsch>
But maybe bazel is complex enough to be tested on changes like that.
<Profpatsch>
I don’t know how we could make it so.
<Profpatsch>
gchristensen: maybe ^
<Profpatsch>
(revdep test for bazel on stdenv changes)
<Profpatsch>
(on Linux and Darwin)
<kalbasit>
That would be great, for sure!
<Profpatsch>
kalbasit: also, Ericson is in #nixos-dev iirc
<Profpatsch>
yes, he is
<Profpatsch>
So maybe we can ask him directly
<kalbasit>
sure I'll try over there
<Profpatsch>
He might have a clue on what is going on, after all it looks like he changed the toolchain a bit
<kalbasit>
Yea he did, and I think it's related because clang is not finding libstdc++ anymore. I'm trying a few things myself, but it's pretty much a change/try without much knowledge on it
fendor has joined #nixos
<Profpatsch>
Who is the person that did much of the darwin work? Maybe they can help.
<infinisil>
The "1.46.1 -> 1.47.2" commit should be "1.46.0 -> 1.47.2" and the "1.46.1 -> 1.47.3" one should be "1.46.2 -> 1.47.3"
Mateon1 has quit [Read error: Connection reset by peer]
Mateon2 is now known as Mateon1
<kuznero>
infinisil: according to the previous commits to vscode, vscodium it was the same. Do you mean that I should make it 2 separate PRs instead for vscode and vscodium separately?
<kuznero>
I see now.
<kuznero>
I can squash commits, should that fix it?
<infinisil>
Yeah squashing the commits would work too
<kuznero>
infinisil: done, please merge if possible. Btw, was that the reason it was not merged before?
fendor_ has quit [Ping timeout: 240 seconds]
<kuznero>
infinisil: please do NOT merge yet!
<kuznero>
Slight mistake in the rebase -i ;(
<infinisil>
kuznero: The ratio of (open PRs) to (people who have merge rights and are actively reviewing PRs) is just too high at the moment, meaning everything is rather slow
<lassulus>
it was probably just not seen. reaching out is helpful as there a lot of PRs and we are just a couple of humans :D
<infinisil>
kuznero: We should wait until ofborg is happy anyways :)
<kuznero>
infinisil: yes, sure. Now it is one single commit.
<infinisil>
Looking good
<evanjs>
when one computer hands on a derivation that another builds just fine >_>
<evanjs>
hangs*
<kuznero>
lassulus: :+1: I see, I was wondering if it is an option to get merge right for a subtree in git somehow (just to be able to merge in this particular package(s))?
<infinisil>
kuznero: If I forget about it, feel free to ping me again when ofborg is ready, I'll merge it then
alp has quit [Ping timeout: 256 seconds]
<kuznero>
infinisil: thanks, will do
zupo has joined #nixos
<lassulus>
not yet sadly. there was a RFC or something for that if I remember correctly
<kuznero>
that would have solved the issue with active reviewers with merge right. I can imagine that there will be way more people who's a lot more comfortable with specific packages (like me, e.g.)
<infinisil>
Do I need to implement retry for API requests now when they return 404
<infinisil>
(i run the {^_^} code that should link PR references for context)
proofofkeags has quit [Remote host closed the connection]
proofofkeags has joined #nixos
<duairc>
Is it possible for a closure to include a .drv file, but not the result of building that drv file?
proofofkeags has quit [Remote host closed the connection]
<pbogdan>
anyone else having trouble with `nix-index` not being able to find / index packages?
proofofkeags has joined #nixos
<pbogdan>
just reran it (am on unstable) a lot of stuff that's definitely listed in the cache is missing
o1lo01ol1o has quit [Remote host closed the connection]
malook has joined #nixos
<infinisil>
duairc: I doubt it, what do you intend to use this for?
<numkem>
pbogdan: won't work with me either. It's taking 42GB of ram
<pbogdan>
it doesn't seem to have a memory issue for me, it's just not finding stuff when querying the cache
<pbogdan>
like for example it doesn't have ffmpeg in the database..
hspak has quit [Ping timeout: 246 seconds]
<duairc>
infinisil: I'm trying to find a way to write a derivation describing how to build a large VM image (basically download and then run through qemu-img convert), but to have it build on the machine I'm deploying to (with NixOps)
<duairc>
infinisil: I was trying to be clever and only copy a script that contains the drvPath but not the outPath, and then nix-store --realize is called at runtime when the VM starts up
c0c0 has quit [Quit: WeeChat 2.7]
<duairc>
But it seems to want to build the derivation anyway
<mrosenbe>
I'm writing a nixexpr for a project that I'm working on, and I'd like to use callPackages passing in a path into the nix tree. Is this something I can do?
<mrosenbe>
for reference, I'm trying to install idea-ultimate on darwin, and I can't because all-packages.nix always forces it to use the jetbrains-jdk, which doesn't work on darwin.
<mrosenbe>
and I'd like to just create a jetbrains object that *doesn't* override jdk.
iyzsong has joined #nixos
justanotheruser has joined #nixos
<duairc>
Is there even some way to reconstruct a derivation from the contents of a drv file? Or to run nix-store --realize on a drv file that isn't in /nix/store?
<evanjs>
mrosenbe: IIRC it's something like `callPackage (<nixpkgs/path/to/thing/expression.nix>)`
<infinisil>
duairc: The .drv file *is* a derivation really
cosimone has joined #nixos
<infinisil>
But not in the Nix expression way I guess
<mrosenbe>
ahh, that looks like an extension of syntax that I have seen before.
stephaneyfx has joined #nixos
<evanjs>
shouldn't need the parenthesis IIRC
fendor__ has quit [Remote host closed the connection]
<duairc>
infinisil: I guess what I mean is, I can call readFile on the .drv path, and then it doesn't get added to the closure (which is what I want). So I have the text of the .drv file. I just need a way to be able to "realise" that then on the remote machine
<slabity>
I deploy all my machines using NixOps and would like to begin switching their configuration over to using flakes
bqv has quit [Ping timeout: 256 seconds]
stephaneyfx has quit [Quit: Leaving]
<duairc>
infinisil: Cool, well thanks anyway, I feel like I'm getting pretty close, hopefully I'll find something that works
<energizer>
i've been delaying switching to flakes until "In the next blog post, we’ll talk about typical uses of flakes, such as managing NixOS system configurations, distributing Nixpkgs overlays and NixOS modules, and CI integration." but the next post came out and it isn't about any of those things
<infinisil>
I guess it is a little differently, but I think that's essentially the same
<infinisil>
duairc: You could use builtins.unsafeDiscardStringContext to get rid of that reference:
<theduke>
mhm, weird issue with Python: I'm using `python3.withPacakges [ scipy pandas opencv ...]`
<mrosenbe>
right. next question: how do I access callPackages?
<duairc>
infinisil: Oh wow, I have no idea what the hell that is, but that seems like exactly the kind of undocumented evilness I'm looking for
<theduke>
most dependencies work, excpet pandas
stephaneyfx has joined #nixos
<infinisil>
But even if you manage to copy the .drv file itself over (which I don't think is possible), you'd still have to get the whole closure of it too. Because .drv files depend on a whole lot of other .drv files for them to work
<theduke>
pandas always give me a `ModuleNotFoundError: No module named 'pandas'` error
<duairc>
infinisil: I have managed to copy the *contents* of the drv file over at least
<mrosenbe>
let pkgs = import <nixpkgs> {}; foo = pkgs.lib.callPackages blah; does not work
<infinisil>
duairc: I guess maybe you could run `nix-store --export $(nix-store -qR ${builtins.unsafeDiscardStringContext pkgs.hello.drvPath}) > $out` in a nix-build, then `nix-store --import` and `nix-store -r` that in a separate derivations $out/bin/realise
<evanjs>
I'm still not 100% on when to actually use callPackage vs callPackages
<mrosenbe>
ahh, no lib, it is just pkgs.callPackages
<evanjs>
mrosenbe: oh right, also that :P
<mrosenbe>
evanjs: I'm overriding `jetbrains = (recurseIntoAttrs (callPackages ../applications/editors/jetbrains {`, so I assume I want to use callPackages
<evanjs>
just like callPackage I guess lol
<infinisil>
callPackage <file> for when <file> is of the form `<args>: <derivation>`
<evanjs>
mrosenbe: curious, what are you overriding there?
<infinisil>
callPackages <file> for when <file> is of the form `<args>: { foo = <derivation>; bar = <derivation>; ... }`
<evanjs>
so basically call packag... ah lol
<evanjs>
on an attrset of packages
<infinisil>
Yee
<clever>
callPackage will basically // { override = ...; } to the result
<clever>
but if you inherit (callPackage ./many.nix {}) a b c; you loose the override
<clever>
you would need to `many = callPackage ./many.nix {}; (many.override ...).a`
<clever>
callPackage instead adds override to each attr, so it works more normally
<mrosenbe>
that is the expression defining jetbrains. my shell.nix used to override the system jdk with adoptopenjdk and depend on pkgs.jetbrains.idea-ultimate, but that started failing, so now it has jetbrains = pkgs.callPackages <nixpkgs/pkgs/applications/editors/jetbrains> {}; and depends on jebrains.idea-ultimate
<duairc>
infinisil: Your suggestions seem promising; however, when I get this mysterious error when I try to build the derivation that buils the .nar file: error: creating directory '/nix/var': Permission denied
<mrosenbe>
although something is still fishy with this install.
<mrosenbe>
/nix/store/nk1bq9ifz5nhjb5hxlqvqml009ik563g-idea-ultimate-2020.1.3/idea-ultimate-2020.1.3/bin/fsnotifier: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), dynamically linked, interpreter /lib/ld-linux.so.2, BuildID[sha1]=cd2f36e5efe59106ad7c0198d02eb7ae8207927d, for GNU/Linux 3.2.0, not stripped
jonringer has joined #nixos
domogled has joined #nixos
<duairc>
Ah, apparently this is because Recursive Nix isn't supported
<colemickens>
bqv: so does a flake-built sys config contain the flake sources?
maier has quit [Ping timeout: 246 seconds]
<colemickens>
bqv: You mentioned something about a flake system containing the system config, I'm assuming it must stuff slurp up all the inputs to the flake repo including itself?
<manveru>
colemickens: you can basically get `self.outPath`, which is the source of the flake itself
<manveru>
which in turn has all required references to all inputs, thereby giving you everything you need :)
<colemickens>
hm, I'm trying to image a flake-built VM image and how it would be able to use that to rebuild-switch with a user-provided snippet of nix.
<colemickens>
That sounds like it might work.
<{^_^}>
[nixpkgs] @emmanuelrosa opened pull request #94101 → qt5: Add the ability to optionally use NSS 3.44 → https://git.io/JJuOb
<manveru>
yep
<kuznero>
infinisil: thanks!
<mrosenbe>
interesting, I remember looking into this a while back and determining that it actually grabbed the .dmg when installing idea on darwin
<bqy>
It puts it in the store, at least
<mrosenbe>
looks like it just always installs the linux .targz now.
<bqy>
It's not a dependency though
kuznero has quit [Quit: leaving]
incognito9999 has quit [Quit: ZNC 1.7.2+deb3~bpo9+1 - https://znc.in]
bqv has joined #nixos
<mrosenbe>
edwtjo: If I were to do some of the legwork to get idea-ultimate to use the mac build on darwin, would you help get it into the repository?
incognito9999 has joined #nixos
stephaneyfx has quit [Quit: Leaving]
<bqv>
colemickens: you might want to do the trick i do
gthm has joined #nixos
<bqv>
as part of the activation script, link "inputs.self" to /run/current-system/flake
<bqv>
then you can access your system's source via there
<{^_^}>
[nixpkgs] @matthewbauer pushed 2 commits to master: https://git.io/JJu39
alp has joined #nixos
<{^_^}>
[nixpkgs] @matthewbauer merged pull request #94088 → llvm*: remove symlinks to llvm-diff, llvm-as and LLVM IR utilities → https://git.io/JJEh8
<{^_^}>
[nixpkgs] @matthewbauer pushed 2 commits to master: https://git.io/JJu37
o1lo01ol_ has quit [Read error: Connection reset by peer]
<hexagoxel>
is there a reliable way to find the "source" of a derivation? something like "this .drv is the product of a runCommand invocation in foo/bar.nix"
<hexagoxel>
of course foo/bar.nix might no longer exist, or not exist on your system etc.
xps has quit [Quit: Lost terminal]
<hexagoxel>
or maybe the more general question:
<hexagoxel>
suppose you encounter an error, and discover that one .drv does not build on your system, and you figure out a way to fix the contained buildCommand.
<{^_^}>
[nixpkgs] @matthewbauer pushed 2 commits to master: https://git.io/JJusJ
<hexagoxel>
and "build graph" is not even the right category, is it? because I don't want to know deps and reverse deps, but which .nix needs modifying to fix it.
<slabity>
Am I correct in assuming that NixOps does not support deploying NixOS flakes?
<immae>
hexagoxel: builtins.unsafeGetAttrPos "fetchurl" pkgs should lead you to a place where you’d be able to follow the track from there
bqv has quit [Quit: WeeChat 2.9]
<immae>
it won’t give directly the runCommand position, but you should be able to follow the path
crimsonalucard has joined #nixos
stephaneyfx has quit [Read error: Connection reset by peer]
stephaneyfx has joined #nixos
<crimsonalucard>
really new to nix. Installed it on osx and not sure where nix.conf is located? Can anyone help?
<hexagoxel>
immae: ah, interesting. I guess `pkgs` can be replaced by whatever expression triggers the error?
<immae>
hexagoxel: not exactly, it’s builtins.unsafeGetAttrPos "fetchurl" pkgs if you want the position of pkgs.fetchurl
lordcirth has joined #nixos
gthm has quit [Ping timeout: 240 seconds]
<hexagoxel>
I don't understand. I do `nix-build something.nix` and somewhere during that it errors while building fail.drv
<lordcirth>
I see, so mkDerivation returns a derivation, but you need callPackage to take that and return a package?
<dminuoso>
No you dont.
<dminuoso>
But callPackage is certainly less of a hassle
drakonis_ has joined #nixos
<clever>
lordcirth: callPackage deals with the arguments on line 1 of foundry-vtt.nix
<dminuoso>
Check the nix pill I linked above
<lordcirth>
It seems to be working.. Thanks guys!
ap29600 has joined #nixos
hspak has joined #nixos
drakonis has quit [Ping timeout: 260 seconds]
<hexagoxel>
immae: for one .drv it helped, yes. For another the name is a bit generic and I am still confused.
<hexagoxel>
no need to look closer at this specific problem however. I was just curious if there was a better approach to random guessing this in general.
domogled has joined #nixos
<lordcirth>
clever, is unzipping automatic? I thought that was only .tar
<hexagoxel>
(I also wonder if it might be "good practice" to put a "echo MY_RANDOM_STRING" into buildCommands :D)
<clever>
lordcirth: if you add unzip to the buildInputs, unzip becomes automatic
<lordcirth>
Ah, neat!
<immae>
hexagoxel: ok, then not that I know of, apart from going through all the dependencies of something.nix and recursively
<immae>
(usually if a dependency fails then you can just find it by dichotomy
sonerhalis[m] has joined #nixos
philipp[m]1 has joined #nixos
neonfuz2 has joined #nixos
timclassic has joined #nixos
theotherjimmy[m] has joined #nixos
ndarilek has joined #nixos
easamuilov[m] has joined #nixos
Guest80879 has joined #nixos
rednaZ[m] has joined #nixos
<lordcirth>
What does it take for an executable file in a package to be in $PATH? Just copy it to ./bin?
serrvan has quit [Ping timeout: 256 seconds]
srxl has joined #nixos
dngray has joined #nixos
srid has joined #nixos
Valodim[m] has joined #nixos
jlv has joined #nixos
lwbcdt[m] has joined #nixos
jaewest304[m] has joined #nixos
dhess has quit [Remote host closed the connection]
wlan42[m] has joined #nixos
Cale has quit [Ping timeout: 260 seconds]
sparogy has joined #nixos
dbirks[m] has joined #nixos
dhess has joined #nixos
chrismatheson has joined #nixos
frobenius[m] has joined #nixos
<clever>
lordcirth: yep
delroth[m] has joined #nixos
stites[m] has joined #nixos
adray has joined #nixos
Lars[m]3 has joined #nixos
kekw736[m] has joined #nixos
logan12358[m] has joined #nixos
Markus[m]2 has joined #nixos
mullein[m] has joined #nixos
notmatthew[m] has joined #nixos
HackerFoo[m] has joined #nixos
mt[m] has joined #nixos
<lordcirth>
clever, removing the unpackPhase made it throw: "unpacker produced multiple directories"
Philipp[m]2 has joined #nixos
ptol3my[m] has joined #nixos
sneknek[m] has joined #nixos
AlesHuzik[m]1 has joined #nixos
Guest45779 has joined #nixos
tbenst[m] has joined #nixos
kriteus119[m] has joined #nixos
nikola[m] has joined #nixos
aquarial has joined #nixos
ice7[m] has joined #nixos
drakonis has joined #nixos
eddyb has joined #nixos
LinusSexTips[m] has joined #nixos
<clever>
lordcirth: the zip file should contain a single directory with all of the src, not multiple directories
raju[m] has joined #nixos
hienergy[m] has joined #nixos
domenkozar[m] has joined #nixos
puzzlewolf has joined #nixos
Xamino[m] has joined #nixos
therivercass[m] has joined #nixos
<lordcirth>
Seems like it has everything at the top level
<clever>
then youll need the unpackPhase to keep the mess contained
<lordcirth>
Yeah. Whatever, it's 3 lines of simple boilerplate
shibboleth has quit [Quit: shibboleth]
<clever>
can even be 1 line, doesnt need newlines
turlando has quit [Ping timeout: 256 seconds]
kenji has joined #nixos
wnklmnn has joined #nixos
<{^_^}>
[mobile-nixos] @samueldr opened pull request #190 → initrd: Fix for lvm2 multiple output → https://git.io/JJuGE
<duairc>
infinisil: That seems to be doing something at least!
cosimone has quit [Quit: Quit.]
knupfer has quit [Ping timeout: 260 seconds]
<duairc>
infinisil: Cool, I can at least test your idea from earlier now. It seems recursive nix doesn't allow this kind of thing
cosimone has joined #nixos
<infinisil>
Kind of expected that
<duairc>
I've come up with a much simpler solution though in the mean time: instead of giving a derivation that makes the VM image, I'm just going to give a file containing a nix expression that contains a derivation that makes the VM image, and then I'll just nix-build that on the server when initialising the VM
seku has joined #nixos
plutes has joined #nixos
<infinisil>
Yeah that's probably the best workaround
<{^_^}>
[nixpkgs] @ddelabru opened pull request #94105 → inform6: init at 6.34-6.12.2 → https://git.io/JJunH
orivej has quit [Ping timeout: 260 seconds]
orivej has joined #nixos
<duairc>
I do appreciate you humouring me though :)
meh` has quit [Ping timeout: 240 seconds]
lordcirth_ has joined #nixos
proofofkeags has joined #nixos
maier has joined #nixos
lordcirth__ has quit [Ping timeout: 244 seconds]
o1lo01ol1o has quit [Remote host closed the connection]
marcusr has quit [Remote host closed the connection]
plutes has quit [Ping timeout: 272 seconds]
codygman has quit [Ping timeout: 240 seconds]
marcusr has joined #nixos
codygman has joined #nixos
incognito9999 has joined #nixos
aaronjanse has joined #nixos
dsiypl4_ has joined #nixos
philr_ has joined #nixos
<MarcWeber>
I'd like to use rust and librealsense. However cargo build says that it cannot find the librealense header files. They are at /run/current*/sw/include. NIX_CFLAGS_COMPILE didn't help. cc of rustc is failing. Is there a simple way to pass the -I flags to cc ?
dsiypl4 has quit [Ping timeout: 244 seconds]
<clever>
,libraries MarcWeber
<{^_^}>
MarcWeber: Don't install libraries through nix-env or systemPackages, use nix-shell instead. See https://nixos.wiki/wiki/FAQ/Libraries for details.
<clever>
MarcWeber: only `nix-shell -p library` will setup -I properly
maier has joined #nixos
dsiypl4_ has quit [Remote host closed the connection]