dywedir has quit [(Remote host closed the connection)]
gnuhurd has quit [(Remote host closed the connection)]
<aupiff>
I am on darwin using nix-shell. I installed clang within the nix-shell. Any c programs I compile with nix's clang (which is significantly older than the default clang on OSX) won't run; the programs exit with the message "Killed: 9".
gnuhurd has joined #nixos
<aupiff>
why is that?
yegods has joined #nixos
yegods_ has joined #nixos
yegods_ has quit [(Remote host closed the connection)]
arximboldi has joined #nixos
Infinisil has quit [(Ping timeout: 240 seconds)]
yegods has quit [(Ping timeout: 255 seconds)]
Infinisil has joined #nixos
AndChat|114816 has joined #nixos
ison111 has joined #nixos
AndChat114816 has quit [(Ping timeout: 240 seconds)]
sigmundv__ has quit [(Ping timeout: 268 seconds)]
sigmundv__ has joined #nixos
kiloreux has quit [(Ping timeout: 248 seconds)]
hellrazor has quit [(Ping timeout: 268 seconds)]
filterfish has quit [(Read error: Connection reset by peer)]
aupiff has quit [(Quit: leaving)]
hellrazor has joined #nixos
kwork has quit [(Ping timeout: 276 seconds)]
kwork has joined #nixos
kiloreux has joined #nixos
AndChat|114816 has quit [(Read error: Connection reset by peer)]
AndChat114816 has joined #nixos
filterfish has joined #nixos
eacameron has joined #nixos
gnuhurd has quit [(Remote host closed the connection)]
elvishjerricco has quit [(Ping timeout: 276 seconds)]
akaWolf has quit [(Ping timeout: 260 seconds)]
eacameron has quit [(Ping timeout: 255 seconds)]
akaWolf has joined #nixos
lstanley has quit [(Ping timeout: 255 seconds)]
[0__0] has quit [(Ping timeout: 276 seconds)]
avn has quit [(Ping timeout: 276 seconds)]
sigmundv__ has quit [(Ping timeout: 255 seconds)]
avn has joined #nixos
[0__0] has joined #nixos
sigmundv__ has joined #nixos
lstanley has joined #nixos
jgertm has quit [(Ping timeout: 255 seconds)]
mbrgm has quit [(Ping timeout: 260 seconds)]
mbrgm has joined #nixos
sigmundv__ has quit [(Ping timeout: 268 seconds)]
sigmundv__ has joined #nixos
takle has quit [(Remote host closed the connection)]
eschnett has joined #nixos
takle has joined #nixos
pie_ has quit [(Remote host closed the connection)]
<adisbladis>
timclassic: The rust overlay is actually downloading the hash, might also be an option
<adisbladis>
It would make things non-reproducible but might make sense for a nightly package
_Vi has quit [(Ping timeout: 258 seconds)]
_Vi has joined #nixos
astronavt has joined #nixos
<MP2E>
When packaging a closed source binary for nixpkgs, is there a way to make a library available to that binary at runtime if it isn't in the dynamic section of the elf?
<MP2E>
basically patchelf is working for everything that I see under 'dynamic section' if I view the executable in the nix store with 'objdump -x exename'
<MP2E>
but at runtime it needs another shared object
<MP2E>
and despite putting that in the libPath to be patched in, it isn't finding it
<MP2E>
and objdump shows no trace of it either
<MP2E>
hm.
<clever>
MP2E: by default, the stdenv will strip things from the rpath if ldd says they arent needed
<clever>
that can remove paths you have added for use by dlopen
<MP2E>
that makes sense.
<clever>
3 options i can think of to fix it
sigmundv__ has quit [(Ping timeout: 240 seconds)]
<clever>
#1 patchelf --add-needed LIBRARY
sigmundv__ has joined #nixos
<clever>
MP2E: this will just mess with the ELF headers to make the library load way ahead of time (might break things in weird ways)
arximboldi has quit [(Quit: arximboldi)]
Supersonic112 has quit [(Ping timeout: 240 seconds)]
<clever>
#2 set dontPatchELF = true; in the derivation
<adelbertc>
i'm a bit confused.. i have a nix expr that does `with import <nixpkgs> {}; { inherit (pkgs) ... }` which as i understand means there is a `pkgs` attribute that is a set in `nixpkgs`
ebzzry has joined #nixos
fresheyeball has joined #nixos
<fresheyeball>
Hey out there
<clever>
adelbertc: yeah, and the pkgs attribute just contains the entire package set, including itself
<adelbertc>
from there i'd like to get the proper path to `git`. i see a `gitAndTools.gitFull` and a `gitMinimal` when I do `nix-env -qaP git` and i want to inherit the precise path in the aforementioned nix expr
<fresheyeball>
Anyone got MATE desktop working?
lambdael has joined #nixos
<clever>
adelbertc: gitMinimal and gitAndTools.gitFull are attributes from the root of nixpkgs
<clever>
adelbertc: i dont think inherit works with foo.bar type paths
<adelbertc>
bah
<adelbertc>
where can i see where `git` is defined then
<adelbertc>
"yeah, and the pkgs attribute just contains the entire package set, including itself" <-- hmm doesn't this mean i should be able to just write `{ pkgs ? import <nixpkgs> {} }: { inherit (pkgs) git }` instead of `{ pkgs ? import <nixpkgs> {} }: with pkgs; { inherit (pkgs) git }`
<adelbertc>
oh wait
<adelbertc>
i misread, the `pkgs` itself contains everything including itself
<clever>
prefixing an attr path with pkgs. has no impact at all, once your in the set
<adelbertc>
but i do need at least one `with pkgs` it seems
<clever>
or with import <nixpkgs> {};
<adelbertc>
i cant do that first line you entered for instance
<clever>
which runs with directly on the entire set
<adelbertc>
right
<clever>
yeah
<adelbertc>
cool beans
<clever>
another option, create a foo.nix that properly does { stdenv, foo, bar }:
<clever>
and then make a default.nix that does { pkgs ? import <nixpkgs>{} }: { foo = pkgs.callPackage ./foo.nix {}; }
<adelbertc>
ah
<adelbertc>
in this case i'm trying to write a nix expr that lets me single-command install a bunch of stuff (for whenever i get a new computer)
<clever>
ah
<clever>
i do that in ~/.nixpkgs/config.nix
<clever>
previously i made a single derivation in packageOverrides that used buildEnv to merge things
<adelbertc>
hm how does .nixpkgs/config.nix work?
<clever>
but i have since heard a new trick, just make a new set in packageOverrides like { packageOverrides = pkgs: with pkgs; { group1 = { inherit gitFull; }; }; }
<clever>
adelbertc: nixpkgs will load config.nix on startup, and that lets you set package overrides
<adelbertc>
woaa
<adelbertc>
oh i see
<adelbertc>
so youd have a set in config.nix
<adelbertc>
and then you can do nix-env -iA nixpkgs.whatever
<clever>
and that set is automatically available in every nixpkgs you import as that user
<adelbertc>
an itll just get picked up
<clever>
including the one nix-env loads
zeus__ has quit [(Remote host closed the connection)]
<clever>
line 11 is the new trick, 14 is the old trick
endforma1 has joined #nixos
<adelbertc>
buildEnv installs all the packages as one unit right
<clever>
yeah
<clever>
which makes it difficult to update just one item
<clever>
you have to update all of them as one unit
<adelbertc>
yeah im using buildEnv too right now and have hit issues a couple times, hence me trying to migrate to a different scheme
endformationage has quit [(Ping timeout: 248 seconds)]
<adelbertc>
ok cool
<adelbertc>
this was extremely instructive
<adelbertc>
thank you
<clever>
another related issue
<clever>
collision between ‘/nix/store/g1f383mjbgv7wnqxgmgdymmc26abk5ry-mystuff/bin/as10k1’ and ‘/nix/store/fca9kqx6czwgcbp2k39c1bwzpc4ag03a-alsa-tools-1.1.3/bin/as10k1’; use ‘nix-env --set-flag priority NUMBER PKGNAME’ to change the priority of one of the conflicting packages
<adelbertc>
onoes
qtness[m] has joined #nixos
<adelbertc>
yeah thats what i run into
<clever>
the alsa-tools in newstuff conflicts with the alsa-tools in mystuff
<clever>
so i have to modify mystuff, and reinstall it (which upgrades ALL of mystuff)
<adelbertc>
yep
<adelbertc>
ive had to do that a couple times
<clever>
before i can switch away from the system that forces mass-updating
<clever>
thats why i havent been maintaining mystuff properly, and my nix-env -q has group out of control, lol
<adelbertc>
lol
<clever>
[clever@amd-nixos:~]$ nix-env -iA nixos.newstuff -A nixos.mystuff
<clever>
replacing old ‘mystuff’
<clever>
installing ‘alsa-tools-1.1.3’
zeus__ has quit [(Ping timeout: 258 seconds)]
<clever>
you can also install 2 things at once as an atomic operation
elvishjerricco has joined #nixos
astronavt has quit [(Ping timeout: 240 seconds)]
<adelbertc>
right
<clever>
adelbertc: only downside i can see with this new trick, there is no easy way to see what is missing from newstuff, since its no longer showing as a single unit
<adelbertc>
right
<clever>
but its now far easier to add things without a mass-rebuild
<clever>
just add it, and nix-env -iA nixos.foo
<adelbertc>
i dont think i need that convenience for now... once i do i will revisit
<clever>
with the buildenv, id have to uninstall it because it would have conflicted in the future
<adelbertc>
yep
<maurer>
So, something in the nixos wrapper for gcc is preventing a plt from being generated in some cases
<clever>
maurer: plt?
<maurer>
program linkage table
<maurer>
it's normally how dynamic symbol resolution gets done
rnhmjoj[m] has joined #nixos
<clever>
ah
<maurer>
It looks like it's some attempt at optimization, or something to do with how we're binding to a particular library
<clever>
that definitely sounds important, when exactly is it breaking?
<maurer>
It's not actually breaking anything most people would care about
<maurer>
It's breaking objdump -d's ability to resolve the target of library calls in a program
<clever>
ah
<clever>
it might be the automatic pass of "strip" over all binaries
<clever>
dontStrip = true; can prevent that on a per-derivation basis
<maurer>
No, this isn't in a derivation
<maurer>
I'm literally running
<maurer>
gcc foo.c -o foo
<maurer>
objdump -d foo
<maurer>
in a shell
<clever>
ah, and its already missing even there
<clever>
then it hasnt hit the strip yet
<maurer>
if I do the same thing, same version of gcc on debian, plt entries are created and calls resolved
<clever>
maurer: try setting the NIX_DEBUG variable to anything
filterfish has quit [(Ping timeout: 255 seconds)]
<maurer>
It's not necessarily incorrect behavior for nixos to be doing this btw - I just need to be able to disable it
<dtzWill>
I think the nox review is mostly to automate "oh I had no idea *random package* depended on this due to an override in some other package...."
<dtzWill>
grantwu: "wip" only works if you ahve unstaged changes, lol
<dtzWill>
AFAIK
<grantwu>
ah
<dtzWill>
I got.. a success re:nox on your PR, and definitely more output than what you got xD
zeus__ has quit [(Remote host closed the connection)]
zeus__ has joined #nixos
<grantwu>
o
<grantwu>
That's fast
<grantwu>
Mine is still fetching, sigh
<dtzWill>
curiously running the same nox thing a few times ... it keeps doing things. Why is it not cached, I thought this was nix-land
<dtzWill>
yeah it seemed to be mostly fetching a lot and then a short blip with some python thing :D
<grantwu>
I got a "subprocess.CalledProcessError: Command '['git', 'merge', '61e149502a3c49e151f69c584440729f129d9b05', '-qm', 'Nox automatic merge']' returned non-zero exit status 128."
<dtzWill>
>.<
<grantwu>
pretty lulzy
<dtzWill>
so please understand I'm a nox noob and never really looked into it much, but... that seems to happen sometimes w/nox
<dtzWill>
might be .. "encouraged" by my control-C'ing it mid operation
<grantwu>
I'm also a Nox noob
<grantwu>
although I do know that you get something mildly amusing if you type a non-integer into the nix UI
<grantwu>
*nox
<dtzWill>
IIRC I just nuked ~/.nox (or whatever the path is) and then it did a fresh clone and it worked again. YMMV
<dtzWill>
what do you mean "hello is not a valid integer", how _dare_ you
zeus__ has quit [(Ping timeout: 276 seconds)]
<grantwu>
kek?
<grantwu>
Incidentally, anyone here know much about packaging Gtk packages
<maurer>
clever: OK, well, I found the problem - it's adding "-z now"
<maurer>
clever: This is probably good behavior for normal builds, but I need to be able to invoke gcc without it
<grantwu>
I got a whole bunch of output and then at the end it says "Ran 0 tests in 0.000s"
<dtzWill>
in the derivation you wrote or are looking at?
<dtzWill>
missed earlier context :3
<clever>
maurer: oh, i was looking at this file earlier, and didnt think -z had anything to do with plt
<clever>
maurer: its an env variable
<clever>
from a shell, it would be hardeningDisable=bindnow
<clever>
+ export
<maurer>
dtzWill: The context is that I'm doing some program analysis stuff, and as part of it I am statically following calls in built code across the library boundary
reinzelmann has quit [(Quit: Leaving)]
<maurer>
OK, if it'll take shell input that's good enough
<grantwu>
Is there a way to make a particular package depend explicitly on a particular version of a dependency?
sellout- has joined #nixos
<maurer>
dtzWill: In a final version of this I'd make it support gcc's new got inlining stuff that you're turning on there, but htat's just engineering work for the moment
<grantwu>
apparently some of these deluge RuntimeWarnings are due to a libtorrent mismatch
<maurer>
grantwu: Make it depend on the dependency in general as a function, then in the invocation to add it to the package set, pick the correct version manually
<dtzWill>
maurer: GOT inlining?? (!) can you point me to any info on that? :)
<maurer>
sorry, that's probably a bad description
<maurer>
but look at gcc's -fno-plt flag
<dtzWill>
maurer: this is very interesting, I would like to subscribe to your newsletter
<grantwu>
what the hell, deluge depends on libtorrent >= 1.1.1 and nixpkgs has 0.13.6
<grantwu>
I am mildly shocked deluge manages to run
<maurer>
dtzWill: Basically, if you are using -z now and or similar, gcc will attempt to not produce chunks of the plt
thblt has quit [(Ping timeout: 258 seconds)]
<grantwu>
libtorrent is more than 3 years out of date!?
<dtzWill>
excellent! makes sense, honestly, I jsut always assumed it "only" set the BINDNOW flag or w/e changing behavior at start
<maurer>
grantwu: Well then update it, and use nox to check if there was a reason for that
<dtzWill>
^
<maurer>
*nox-review
<dtzWill>
also ty, chasing down "-fno-plt" and related does the trick :)
<maurer>
In any case, the -z now behavior is probably correct for normal nix operation
<maurer>
it's just annoying in my weird case, so I'm glad I can flip it off
<dtzWill>
yeah
<dtzWill>
maurer: I too do compiler and program analysis research... I can't tell you how many times I've forgotten that nix-provided compilers might "sneak" in flags if I'm not careful about my env or in using a cc-wrapper with taht stuff perma-disabled
sellout- has quit [(Quit: Leaving.)]
<dtzWill>
sigh, so much of shared libraries and related are just madness
<NixOS_GitHub>
[nixpkgs] mt-caret opened pull request #28060: json-refs: init at 3.0.0 (master...json-refs) https://git.io/v7XqF
NixOS_GitHub has left #nixos []
<grantwu>
Is there any sort of formal policy around who is listed under maintainers?
Ralith_ has quit [(Ping timeout: 260 seconds)]
<adisbladis>
grantwu: Hehe just 3 years out of date ;)
<adelbertc>
for a package?
<adisbladis>
I just found some package that is 6 years out of date
<grantwu>
Yeah
<grantwu>
adisbladis: :C
<adelbertc>
not to my knowledge. you just might get pinged when someone goes to update the expression
<adisbladis>
grantwu: Some packages really don't get enough love
Ralith_ has joined #nixos
oida_ has joined #nixos
LinArcX has joined #nixos
LinArcX has quit [(Remote host closed the connection)]
<grantwu>
dash: Didn't seem to work
<dash>
what'd you get
<grantwu>
Same thing
<grantwu>
It appears that the build is actually failing within ./configure
oida has quit [(Ping timeout: 255 seconds)]
<dash>
mmh, likely
<grantwu>
so the bjam route doesn't seem to involve configure
<grantwu>
or autotool.sh
silver_hook has joined #nixos
silver_hook has quit [(Changing host)]
silver_hook has joined #nixos
<pierron_>
timclassic: the problem is finding out about the date without having to keep track of it manually. Honestly, I think if files are out-of-date, this is a problem which should be fixed upstream.
Ralith_ has quit [(Ping timeout: 260 seconds)]
<pierron_>
timclassic: also, what do you mean by out-of-date, how do you check that?
Ralith_ has joined #nixos
<grantwu>
dash: Huh. Literally, copying the luabind stuff, and commenting out the "./autotool.sh" step is causing it to actually build things
endforma1 has quit [(Quit: WeeChat 1.9)]
nslqqq has quit [(Ping timeout: 240 seconds)]
nslqqq has joined #nixos
<grantwu>
it built! \o/
<dash>
yay
<adisbladis>
pierron: I check it with git blame and checking the package upstream
<pierron_>
adisbladis: Unless you are talking about firefox nightly being out-of-date, we are not talking about the same thing.
Ralith_ has quit [(Ping timeout: 240 seconds)]
mbrgm has joined #nixos
<grantwu>
Does the nox-review thing not work if you can't fast-forward merge from nixpkgs master to your PR?
<grantwu>
Because that's been my experience
reinzelmann has joined #nixos
hellrazor has quit [(Ping timeout: 260 seconds)]
<grantwu>
okay, so nox-review tells me that my changes are breaking rtorrent
Ivanych has joined #nixos
Ralith_ has joined #nixos
gm152 has quit [(Quit: Lost terminal)]
<pierron_>
timclassic: by the way, the latest directory is supposed to target the latest finished build. It might be that the build was not complete when you did it.
_Vi has joined #nixos
<grantwu>
http://ix.io/z0w this is from rtorrent. Anyone know how to approach fixing this?
<pierron_>
timclassic: the checksum file is checked with a signature file, and the checksum file, once verified is used to check that the download from nightly is done correctly. So maybe you found an issue in one of our CDN which does not distribute the same version.
<pierron_>
timclassic: downloading from date directory should give the same verification process.
<grantwu>
Is there any particular reason why rtorrent is under tools/networking/p2p while every other torrent client is under applications/networking/p2p
FRidh has joined #nixos
<adisbladis>
grantwu: Looks like other distros have 2 separate versions of libtorrent packaged
<adisbladis>
I'm guessing rtorrent doesn't build with an up to date libtorrent
<grantwu>
so nixpkgs is conflating the two.
<grantwu>
and my package is all wrong
<grantwu>
wait, so how did deluge every work???
_Vi has quit [(Ping timeout: 276 seconds)]
<MP2E>
well, imo, this would be an instance where packaging 2 versions of libtorrent might be in order. libtorrent_0_13_6 set to version 0.13.6, and libtorrent set to the latest, with rtorrent now depending on libtorrent_0_13_6
<grantwu>
no, no, that's not right.
<hyper_ch>
probably yes
<grantwu>
There are two versions of libtorrent - libtorrent-rakshasa and libtorrent-rasterbar
<grantwu>
It's unclear to me whether or not one is a fork of another or if they have no relation
<grantwu>
But rtorrent uses libtorrent-rakshasa and deluge, qbittorent use libtorrent-rasterbar
goibhniu has quit [(Ping timeout: 240 seconds)]
<grantwu>
A G H there's already a libtorrentRasterbar
<grantwu>
So basically, deluge was somehow set to use the wrong _fork_ of libtorrent
<simpson>
grantwu: They have zero relation. Rakshasa's libtorrent is for rtorrent and I don't think anybody else uses it.
<grantwu>
simpson: Okay.
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Mic92 pushed 1 new commit to master: https://git.io/v7XYP
<NixOS_GitHub>
nixpkgs/master 531a5b1 Joe Hermaszewski: git-fame: init at 2.5.2
<sphalerite>
Someone mentioned recently that it's a (n unfortunate) tradition to treat libc as special and not just another library. Does anyone know why this actually happened in the first place?
georges-duperon has quit [(Ping timeout: 255 seconds)]
<linarcx>
i want to change sddm theme.set: "services.xserver.displayManager.sddm.theme".but when rebuild configuration it say:"error: The unique option `services.xserver.displayManager.sddm.theme' is defined multiple times, in `/nix/var/nix/profiles/per-user/root/channels/nixos/nixpkgs/nixos/modules/services/x11/desktop-managers/plasma5.nix' and `/etc/nixos/configuration.nix'.(use ‘--show-trace’ to show detailed location information)"
ertes-w has joined #nixos
<symphorien>
plasma5 seems to have already set a theme for you
<symphorien>
use option = lib.mkForce value
mudri has joined #nixos
hyper_ch has joined #nixos
StevenTian has quit [(Quit: Connection closed for inactivity)]
eacameron has quit [(Ping timeout: 240 seconds)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] FRidh pushed 1 new commit to master: https://git.io/v7XRp
<NixOS_GitHub>
nixpkgs/master 345b35c Frederik Rietdijk: Python: add buildPythonPackage.overridePythonPackage method....
<NixOS_GitHub>
[nixpkgs] FRidh pushed 1 new commit to staging: https://git.io/v7XEZ
<NixOS_GitHub>
nixpkgs/staging 7a6d4f4 Frederik Rietdijk: Merge remote-tracking branch 'upstream/master' into HEAD
NixOS_GitHub has left #nixos []
takle has quit [(Ping timeout: 255 seconds)]
roberth has quit [(Ping timeout: 255 seconds)]
bitchecker has joined #nixos
mightybyte has joined #nixos
nwuensche has joined #nixos
freusque has quit [(Quit: WeeChat 1.9)]
Fuuzetsu has joined #nixos
oahong has joined #nixos
oahong has quit [(Changing host)]
oahong has joined #nixos
mk-fg has joined #nixos
mk-fg has quit [(Changing host)]
mk-fg has joined #nixos
periklis has quit [(Ping timeout: 240 seconds)]
leothrix has quit [(Ping timeout: 240 seconds)]
mpcsh has quit [(Ping timeout: 240 seconds)]
pietranera has joined #nixos
justan0theruser has joined #nixos
[0__0] has joined #nixos
mpcsh has joined #nixos
Bane^ has quit [(Ping timeout: 260 seconds)]
<nwuensche>
Hello everybody! I'm trying out NixOS right now, and I have a question concerning package installs. I want to set up NixOS only for my personal computer, so I don't have to do much profile isolation. But I want to have a nice configuration.nix file, that I can use to reinstall my system. What is the best practice to install new packages for me? I don't like nix-env -i, because the packages aren't added in the
<nwuensche>
config file. However, I also don't know if it's best practice to add all packages in environment.systemPackages and just do nixos-rebuild switch everytime I added a new package. Is there a better way to handle pacakges?
<srhb>
nwuensche: Tastes vary a lot on this topic
Bane^ has joined #nixos
leothrix has joined #nixos
<srhb>
nwuensche: I keep things that "i might want to remove later" in nix-env, keep all my "normal/required" packages in configuration.nix, and use nix-shell for all development dependencies etc.
rumble is now known as grumble
<adisbladis>
I never use nix-env. For all my development environment stuff I use nix-shell, other than that packages are mostly in configuration.nix
<nwuensche>
Thank you for the answers. However, if I take out a package in environment.systemPackages and nixos-rebuild the system, will this package also be deleted from my computer?
<adisbladis>
nwuensche: Yes it will
<nwuensche>
Ok, cool
<avn>
it will remove from $PATH, and physically removed on next GC cycle
<adisbladis>
Technically it will still be around in your nix store until you garbage collect
<nwuensche>
I also found a way of editing default.nix and add a packageOverrides command there. Should this be used to install packages in my case?
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] chrisburr opened pull request #28066: Fix download URL of libtiff for 17.03 release (release-17.03...fix-lib-tiff-url) https://git.io/v7Xz8
NixOS_GitHub has left #nixos []
<pietranera>
Hello, a bunch of cURL security advisories has been made public, upgrading to cURL 7.55.0 is recommended... is a cURL upgrade/patch in the works?
<07EABFYFG>
[nixpkgs] globin pushed 1 new commit to staging: https://git.io/v7XaP
<07EABFYFG>
nixpkgs/staging 0ff782e Frederik Rietdijk: makeWrapper: fix regression introduced in #24944...
07EABFYFG has left #nixos []
<globin>
peti: fyi just started a new staging eval with FRidh's fix, would you like to join #nixos-dev, to make co-ordinating staging merges easier? :)
<catern>
I want to make a derivation which does a simple thing: it looks at an arbitrary path, and generates a file containing a Nix expression which is a fixed-output derivation to get that path into the store
<catern>
is there a utility for this?
queiw has joined #nixos
<catern>
both of the original derivation and the resulting derivation are super impure, of course
<catern>
since they look at a path on the host machine
<catern>
(but I know in this case that that path will never change)
Ivanych has quit [(Ping timeout: 258 seconds)]
linarcx has joined #nixos
<linarcx>
linarx
<tsmeets>
catern: are you looking for a function that takes a path as an argument and returns a derivation?
<catern>
tsmeets: basically, yes
<linarcx>
Infinisil: thanks. its work:)
<tsmeets>
catern: that should be possible.
<catern>
tsmeets: yes but I'm hoping one already exists in nixpkgs that I can use ;)
<catern>
or somewhere
<catern>
that I can just copy
<linarcx>
i want to set another theme for sddm. in the file:"/nix/var/nix/profiles/per-user/root/channels/nixos/nixpkgs/nixos/modules/services/x11/desktop-managers/plasma5.nix ", there is an option for this purpose.it is called:"services.xserver.displayManager.sddm "and set the theme atribute to "breeze".i want to set it to another theme.how find another name?where "breeze" store?i mean in which file?
<catern>
because I don't trust that I will do it with good style
<tsmeets>
catern: but what are you trying to do with it? I think you can just pass a path expression, and nix will copy it to the store.
<catern>
tsmeets: yeah, but I need to get a fixed-output derivation instead of a path, because I want to allow use of a binary cache
<catern>
hmm
<catern>
I suppose a path may also allow use of a binary cache, but to rephrase: I want to be able to use the binary cache without having to rescan/rehash the path each time
<catern>
on different hosts
newhoggy has quit [(Remote host closed the connection)]
newhoggy has joined #nixos
<goibhniu>
catern: it might help to explain the big picture
dywedir has quit [(Remote host closed the connection)]
Ivanych has joined #nixos
erictapen has quit [(Ping timeout: 240 seconds)]
<tsmeets>
catern: there is `pkgs.copyPathToStore`
Rotaerk has quit [(Quit: Leaving)]
<tsmeets>
catern: this is what you need I think
<tsmeets>
catern: it takes a path as an argument and returns the path to the nix store
<catern>
no, that isn't what I need
<tsmeets>
catern: oh, i see. it is the same as using the path
<catern>
right
<catern>
I need to separate "hash the path" and "copy the path into the story", because the latter is expensive, but I only actually need to do the former (to use a binary cache for builds from that path)
<catern>
er
gnuhurd has quit [(Remote host closed the connection)]
<catern>
(by "hash the path" I guess I mean "obtain somehow a hash of the path", which could also be from a cache)
<tsmeets>
catern: you could use something like: `path: runCommand "" {} "cp -a ${path} $out"`
ixxie has joined #nixos
<catern>
tsmeets: that's also equivalent to using the path, though, I think
detran has joined #nixos
<catern>
the core thing is that I need *two* derivations: one impure derivation which generates a fixed-output derivation
reinhardt has joined #nixos
goibhniu1 has joined #nixos
newhoggy has quit [(Remote host closed the connection)]
newhoggy has joined #nixos
goibhniu has quit [(Ping timeout: 240 seconds)]
goibhniu1 is now known as goibhniu
reinzelmann has quit [(Ping timeout: 255 seconds)]
tvon has quit [(Quit: System is sleeping...)]
hlavaty has joined #nixos
eacameron has joined #nixos
<Infinisil>
catern: Would just using nix-instantiate work?
<catern>
Infinisil: inside the impure derivation? or what do you mean?
<Infinisil>
catern: nix-instantiate generates just the derivation path, it doesn't run the builder
<catern>
interesting...
reinhardt has quit [(Quit: Leaving)]
<Infinisil>
then nix-store -r <path to derivation> builds it
newhoggy has quit [(Ping timeout: 260 seconds)]
<catern>
Infinisil: and necessarily nix-instantiate would need to read and hash the impure path I passed it, so that it could generate the derivation path, right?
<Infinisil>
catern: man nix-store is very interesting
eacameron has quit [(Ping timeout: 260 seconds)]
<catern>
Infinisil: hmmm wait a second though
ThatDocsLady_nom is now known as ThatDocsLady
Wizek has joined #nixos
<catern>
if I passed the derivation the path as a string (which I would need to do to avoid the path being copied immediately into the store)
<catern>
then nix-instantiate would just hash the *path string*, not the contents of the path
<LnL>
err
<Infinisil>
catern: Then it's become a run-time dependency
<Infinisil>
i don't know if that's really what you want
<catern>
Infinisil: unfortunately it is what I want
<catern>
I see it as similar to giving network access to fixed-output derivations
<catern>
this is a special network filesystem
<Infinisil>
catern: Ahhh, well yeah
<Infinisil>
No need to mess with derivations, a string is enough then :P
<catern>
oh hmm interesting point
alx741 has joined #nixos
<catern>
are you suggesting that it's sufficient to hash the string?
alx741 has quit [(Client Quit)]
<Infinisil>
catern: Do you want to make sure that the file doesn't change?
<catern>
yes
alx741 has joined #nixos
<catern>
(well, I already know it isn't going to change)
<Infinisil>
catern: Then you should probably write a bash script or so which runs at runtime and checks it against a known hash of it
<catern>
I don't really like hashing the string because, I would like to be able to pass paths that *will* change
<catern>
which would necessitate re-scanning
nevermind has joined #nixos
<ixxie>
anybody have this thing with GDM where it doesn't detect users so you have to type in the name yourself?
tvon has joined #nixos
<Infinisil>
catern: Huh? How do you want to make sure it doesn't change without rehashing?
<catern>
hmm I guess that's at a higher level though... I would replace the fixed-output derivation I end up with, with a path
<catern>
Infinisil: yeah sorry I was confused and confusing there for a second
nevermind has quit [(Client Quit)]
* Infinisil
is also kinda confused
gnuhurd has joined #nixos
<catern>
Infinisil: I have a network filesystem which contains immutable sources, and also sources local to my system. I want to be able to build either of them, uniformly, but I also want to be able to fetch the immutable sources with a fixed-output derivation, so that I can get build caching from a remote build cache
nevermind has joined #nixos
<catern>
(that is, get the cached build output without actually scanning the immutable sources)
<Infinisil>
catern: Ahhh, how about builtins.fetchurl then
newhoggy has joined #nixos
<Infinisil>
wait
<Infinisil>
maybe not
eacameron has joined #nixos
<catern>
(the network filesystem is mounted locally everywhere, at /opt/some/path)
<Infinisil>
catern: By sources you mean actual source code you want to use as the src attribute in a build?
<catern>
Yes
<catern>
(I will be passing the sources in as an argument)
<Infinisil>
catern: How do you intend to have a cache? What caches it?
newhoggy has quit [(Ping timeout: 276 seconds)]
newhoggy has joined #nixos
erasmas has joined #nixos
peacememories has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
MercurialAlchemi has quit [(Ping timeout: 276 seconds)]
<disasm>
If anyone hanging out in here is near Central PA area, I'm doing a talk for State College DevOps on nix, in particular, the language using nix-repl, basics of setting up nix on any linux distribution, creating derivations, and how to use nix-shell to run build steps manually. Food is provided. https://www.meetup.com/devops_statecollege/events/242005997/
newhoggy has quit [(Ping timeout: 276 seconds)]
<dash>
nice
v0|d has joined #nixos
iyzsong has quit [(Quit: ZNC 1.6.5 - http://znc.in)]
newhoggy has joined #nixos
Infinisil has quit [(Quit: leaving)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] copumpkin pushed 1 new commit to master: https://git.io/v7XpJ
<NixOS_GitHub>
nixpkgs/master ed55bdb Dan Peebles: lkl: 2017-06-27 -> 2017-08-09...
NixOS_GitHub has left #nixos []
newhoggy has quit [(Ping timeout: 260 seconds)]
<catern>
Infinisil: you are gone, but, Hydra will build it and I'll download from there
linarcx has quit [(Remote host closed the connection)]
<NixOS_GitHub>
[nixpkgs] vcunat pushed 1 new commit to master: https://git.io/v7Xjw
<NixOS_GitHub>
nixpkgs/master 10bcf08 Vladimír Čunát: knot-resolver: security 1.3.2 -> 1.3.3...
NixOS_GitHub has left #nixos []
gnuhurd has quit [(Remote host closed the connection)]
gnuhurd has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] vcunat pushed 1 new commit to release-17.03: https://git.io/v71eL
<NixOS_GitHub>
nixpkgs/release-17.03 9349886 Vladimír Čunát: knot-resolver: security 1.3.2 -> 1.3.3...
NixOS_GitHub has left #nixos []
newhoggy has quit [(Ping timeout: 240 seconds)]
vandenoever has quit [(Ping timeout: 260 seconds)]
frankpf has quit [(Ping timeout: 276 seconds)]
peacememories has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
<FRidh>
nix-prefetch-url supports file:///some/path where /some/path is an absolute path. Can it also handle relative paths? Or will I have to use realpath?
<linarcx>
Hi. i want to override some lines in /etc/sddm.conf.i use services.xserver.displayManager.sddm.extraConfig option and pass my lines to it.but after rebuild configurations, it append my lines to end of that file. i want to replace those lines.how achieve this goal?
Ralith_ has quit [(Ping timeout: 260 seconds)]
newhoggy has quit [(Ping timeout: 258 seconds)]
<avn>
linarcx: can you explain bit widely, what you try to do?
<linarcx>
i want to use another theme.
Ralith_ has joined #nixos
<bkchr>
linarcx: why not displayManager.sddm.theme?
<avn>
linarcx: services.xserver.displayManager.sddm.theme is not enough?
<linarcx>
yes. i try it.but it is limited to 3 themes reside in nix/store/"hashcode"/sddm/themes
<linarcx>
i want to override it and add more themes
<avn>
just add package with your theme, and add "sdda/themes" to pathsToLink
<avn>
It looks for themes into ThemeDir=/run/current-system/sw/share/sddm/themes
<avn>
so mix in own packages, which provide $out/share/sddm/themes
<clever>
linarcx: it gets the sddm path from services.xserver.displayManager.sddm.package
gnuhurd has quit [(Remote host closed the connection)]
<clever>
linarcx: so you can create an override on sddm that adds themes to it
<tnias>
is there an easy way to generate the nixos options page locally?
<avn>
clever: I feel, mixing themes to sw/share/.... is easier, that modify sddm
<avn>
especially for newbie
gnuhurd has quit [(Remote host closed the connection)]
<linarcx>
very very thanks dudes.i'll try it.
gnuhurd has joined #nixos
<FRidh>
clever: thanks. For what I'm working on I prefer to stick with nix-prefetch-url. Using realpath or in python os.path.abspath isn't that big of an issue.
<avn>
linarcx: environment.pathsToLink = [ "/share/sddm/themes" ]; and add package which will have theme in $out/share/sddm/themes
<tnias>
nvm. just found appendix a in the manual
gnuhurd has quit [(Remote host closed the connection)]
gnuhurd has joined #nixos
newhoggy has joined #nixos
newhoggy has quit [(Ping timeout: 240 seconds)]
<bkchr>
Were there some changes to how buildInput etc are handled in the last time?
<Infinisil>
bkchr: What's the problem you're having?
<clever>
emmanuelr: yeah, i'm not sure where the patches would get applied
<emmanuelr>
clever, yeah I was looking at that file and I checked bundler-common and when mkDerivation is called it doesn't pass along @args. So I don't see a way to apply patches. Damn it.
vandenoever has quit [(Ping timeout: 248 seconds)]
<clever>
emmanuelr: overrideAttrs applies things between the bundler stuff and mkDerivation
jonte_ has joined #nixos
<clever>
so that should let you squeeze things in
<clever>
but you need to apply it to the mkDerivation return value, and there are at least 2 buildEnv
<clever>
's in the way
vandenoever has joined #nixos
vandenoever has quit [(Changing host)]
vandenoever has joined #nixos
<rsa>
any optirun + steam users?
<emmanuelr>
clever: oh goodness. I may need to PhD to pull this off. LOL!
<clever>
Returns a status of 0 or 1 depending on the evaluation of the conditional
tdc has quit [(Read error: Connection reset by peer)]
tdc has joined #nixos
<sphalerite>
clever: yes, but there are two [[s on the same line
<sphalerite>
with only one ]]
<sphalerite>
GNUmakefile[[
<sphalerite>
is that a typo perhaps?
<nliadm>
it seems like it's looking for a file with that name
* sphalerite
git blame
<nliadm>
looks like a typo
<clever>
ah, i see the second one, bash will probably treat it as part of the filename
newhoggy has quit [(Ping timeout: 240 seconds)]
<sphalerite>
yeah
<nliadm>
'GNUmakefile` isn't widely used, it seems
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] lheckemann opened pull request #28079: stdenv: fix typo in setup.sh (staging...setup-typo-fix) https://git.io/v71cu
NixOS_GitHub has left #nixos []
Camdar has quit [(Ping timeout: 260 seconds)]
endformationage has joined #nixos
bkchr has quit [(Ping timeout: 255 seconds)]
simendsjo has joined #nixos
roberth has quit [(Ping timeout: 240 seconds)]
newhoggy has joined #nixos
zeus_ has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] cocreature opened pull request #28080: Fix GHC 7.10 config of vector and tasty-ant-xml (master...ghc-7.10) https://git.io/v71WG
NixOS_GitHub has left #nixos []
goibhniu has quit [(Ping timeout: 255 seconds)]
<Naughtmare[m]>
How do I get the qt frontend for transmission installed? I have installed the 'transmission' package, but it doesn't include 'transmission-qt'. The only other related package is 'transmission_gtk` package, but I want the qt frontend. Am I missing some configuration option or do I have to make a custom package myself?
<clever>
Naughtmare[m]: the normal and gtk versions are the same package, one of them just has enableGTK3 set to true
newhoggy has quit [(Ping timeout: 240 seconds)]
<clever>
Naughtmare[m]: but i dont see a QT option, so that would have to be added in pkgs/applications/networking/p2p/transmission/default.nix
mudri has quit [(Ping timeout: 240 seconds)]
<NickHu>
Does anyone know how to get ghc compiled without -dynamic? I'm trying to profile some haskell code but it's giving me this error http://lpaste.net/357548
Wizek_ has joined #nixos
newhoggy has joined #nixos
ixxie has quit [(Quit: Lost terminal)]
leat has quit [(Ping timeout: 248 seconds)]
newhoggy has quit [(Ping timeout: 255 seconds)]
pietranera has quit [(Quit: Leaving.)]
grw has quit [(Ping timeout: 240 seconds)]
newhoggy has joined #nixos
_rvl has joined #nixos
zraexy has joined #nixos
rauno has joined #nixos
newhoggy has quit [(Ping timeout: 240 seconds)]
bkchr has joined #nixos
<eacameron>
So it's about time I figure this out: How do I create my own custom nixos module? I have a bunch of derivations for doing daily backups but putting them into my config is awkward
ixxie has joined #nixos
ylwghst has joined #nixos
<ylwghst>
Hello
mudri has joined #nixos
<ylwghst>
I imported nix store closure into my nixos.
<ylwghst>
How can I install it now?
<srhb>
ylwghst: Just like you would if you hadn't copied the store closure, presumably?
edude03 has joined #nixos
<ylwghst>
Its a closure of network-manager-1.4.4
<ylwghst>
Im totally noob with nix
<ylwghst>
I ve succesfully installed and booted nixos on my machine.
<ylwghst>
but without network connection\
<srhb>
ylwghst: I haven't actually done this before, but I assume you just install it like you did on the machine you copied the closure *from*
<srhb>
Since the store paths are now all available, I would expect it to not require network connectivity.
deltasquared has joined #nixos
<ylwghst>
i installed from live cd there is nmtui i have used to connect
<clever>
ylwghst: simplest option, enable it in configuration.nix, then boot from the installer, re-mount the filesystems to /mnt, and re-run nixos-install
<clever>
and that will upgrade the install
newhoggy has joined #nixos
<ylwghst>
but after I imported the closure it should work someting like nix-env -iA nixos.pkgs.linuxPackages.networkmanager
<clever>
networkmanager isnt under linuxPackages, and the pkgs. is redundant
<srhb>
ylwghst: Yes. Or simply point nix-env -i to the store path
<ylwghst>
where is?
periklis has joined #nixos
<clever>
and it needs system wide config via configuration.nix, so it wont work right with nix-env
<ylwghst>
srhb: ok ill try
<ylwghst>
clever: hm
<clever>
just set the right settings in configuration.nix and nixos-rebuild switch
dpino has joined #nixos
* srhb
nods
<ylwghst>
ok
<dpino>
hi
<dpino>
after upgrading nixos when I use nix-env to search to a package I got the following error:
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] tnias opened pull request #28081: nixos/tor: add tor hidden service options (master...tor_service) https://git.io/v71BA
NixOS_GitHub has left #nixos []
<dpino>
$ nix-env -qaP nix-env
<dpino>
error: cannot coerce a set to a string, at /nix/store/1jc7annw6k8c34a8l6md1z7zb9hwnhxh-nixpkgs-17.09pre112229.e619ace733/nixpkgs/pkgs/development/interpreters/ruby/default.nix:57:17
<clever>
ylwghst: if you have a /boot you must also mount it to /mnt/boot/
<clever>
dpino: then you have 3 channels
<dpino>
clever: interesting, that explains why the path is nixpkgs-17.09pre...
<srhb>
ylwghst: And last I checked, using rebuild in the chroot was broken, but nixos-install worked.
<clever>
dpino: and 15.09 is really old
<dpino>
clever: I'm thinking of removing all the channels bu 16.09, then slowly upgrade to newer channels
newhoggy has quit [(Ping timeout: 276 seconds)]
<clever>
srhb: "rebuild switch" will definitely fail in a chroot, because it tries to mess with the active systemd, you want "rebuild boot"
georges-duperon has quit [(Ping timeout: 255 seconds)]
<srhb>
clever: Oh, that would explain.
<ylwghst>
clever: so I need run nixos-install
<dpino>
clever: does it sound like a good plan?
<srhb>
clever: Thank you :)
<clever>
dpino: the main thing is that nixos-rebuild expects a chanenl called nixos on root
<dpino>
clever: ok, didn't know that
<clever>
ylwghst: yeah
willprice has joined #nixos
<dpino>
I will name 16.09 nixos then
<deltasquared>
ok, stupid question time. I'm aware that nixos's "functional-ness" extends to it's configuration, so I would assume somewhere there is something generating /etc or other config files. I'm not sure what part of the documentation I should look at to see how this process works under the hood for etcfiles
<clever>
deltasquared: etc.nix defines the nixos options for it, make-etc.sh converts those options into a derivation, and then setup-etc.pl runs on bootup to re-sync /etc with the derivation
<deltasquared>
oooooh nice, so there is a generator script
<clever>
yeah
<deltasquared>
clever: I haven't full-on embraced nixos yet but I figured the concept would be useful for forcing me to keep my /etc under control, as it's not easy to determine which files I've modified
<clever>
deltasquared: setup-etc.pl manages it by symlinking everything via /etc/static, and keeping track of which files it has created, so it can remove them later on
<clever>
then it can atomicly swap most of /etc by updating just /etc/static
<deltasquared>
hmm, I guess the rename() operation would be atomic from the point of view of most other processes
<clever>
yeah
<deltasquared>
though if I were to start quibbling the precise ordering semantics of system calls and the VFS I'd get lost in kernel land real fast.
kiloreux has quit [(Ping timeout: 276 seconds)]
<deltasquared>
"stdenvNoCC"... I assume means "no C compiler"
<clever>
yep
mounty has quit [(Ping timeout: 260 seconds)]
<deltasquared>
clever: so, if for some reason I wanted to directly provide an /etc tree to override "defaults" from packages with, could I do that?
<clever>
that must be done from nixos modules
<deltasquared>
I kinda need to take baby steps from wanting complete control over my config files
<clever>
not from the packages
<clever>
at its core, nixos is a module framework to combine option and config definitions from many modules, and allow the modules to set eachothers config
mounty has joined #nixos
<deltasquared>
something is yelling "conflict resolution!?" in the back of my head
<clever>
and in the end, you have a single derivation, that directly (or indirectly) depends on all of the config and everything it gnerates
<clever>
and 105 builds the top-level derivation using the builder defined on 30
<dpino>
clever: I noticed the error happens when as soon as I switch to 17.03. If my only channel is 16.09, I can use nix-env normally. But under a configuration where my only channel is 17.03 then a search with nix-env returns an error in a ruby/default.nix
<clever>
deltasquared: and in the case of /etc, line 122 refers to the sum of all activation scripts, which are defined in another module
<dpino>
any idea of what it could be?
<dpino>
or any work around I could try?
<clever>
dpino: i generally use nix-repl to do all of my searching, but the command you previous gave fails with a different error here, error: selector ‘nix-env’ matches no derivations
newhoggy has quit [(Ping timeout: 260 seconds)]
<dpino>
I think the error I posted earlier was $ nix-env -qaP ruby
<dpino>
error: cannot coerce a set to a string, at /nix/store/v1zvii5pzvx76dgkb8ifdsyp8fwj6ng6-nixos-17.03.1661.1f7114aec3/nixos/pkgs/development/interpreters/ruby/default.nix:57:17
<clever>
yeah, getting a different error here
<clever>
[root@amd-nixos:~]# nix-channel --list
<Ralith>
cstrahan: it's really difficult to overstate how frustrating it is for your debug info to be silently clobbered
<dpino>
clever: I tried to edit that nix file, and amend that line but it's read-only (even for root). Is there any way I can edit and modify that file?
kiloreux has joined #nixos
<clever>
dpino: you must never manualy modify anything in /nix/store
<clever>
that can seriously break the entire system
<clever>
nixos automatically mounts it read-only to prevent that
<ylwghst>
clever: if I enable networking.networkingmanager will the nixos get also bc43 driver?
<deltasquared>
clever: so I could have / readonly most of the time huh? intradesting
newhoggy has joined #nixos
<deltasquared>
I wish arch had a feature to do that when doing pacman stuff.
<clever>
deltasquared: only /nix/store/ is read-only
<deltasquared>
clever: true, though considering most of the other bits could be on other FSes, may not be too hard to extend to /
<clever>
deltasquared: and nixos needs a writable / to boot, for a number of things it creates
<clever>
deltasquared: but nixos can boot with a tmpfs on /
<clever>
deltasquared: or a read-only union'd with a tmpfs
<deltasquared>
I am so jelly right now
<clever>
deltasquared: i have booted nixos many times with ONLY /nix/store present, and every other directory just being absent
<clever>
and it fills in all of the blanks on its own
<Infinisil>
clever: /boot?
<deltasquared>
well, what with it regenerating /etc anyway as is possible with functional design of it's kind, it would only really *need* the store
roberth has joined #nixos
michaelpj has joined #nixos
<clever>
Infinisil: i was usually booting it with a special method that didnt rely on /boot
<clever>
it will probably just make an empty /boot
<deltasquared>
oh yeah, what black magic would this be
<michaelpj>
is it possible to link to e.g. the nix manual from the nixpkgs manual?
ison111 has quit [(Ping timeout: 240 seconds)]
<clever>
dpino: do you have anything in your ~/.nixpkgs/config.nix ?
<Infinisil>
deltasquared: I have my wpa_supplicant config in /etc, manually written, but I guess I could move this to a non-default location by passing some option to wpa_supplicant
newhoggy has quit [(Ping timeout: 240 seconds)]
reinzelmann has joined #nixos
<clever>
Infinisil, deltasquared: nixos will leave /etc writable and ignore any file that hasnt been configured in nixos
<clever>
so you can just add /etc/wpa_supplicant.conf manually
<deltasquared>
Infinisil: I have one config per network in my current arch setup, I have no idea how I'd replicate that
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] globin pushed 1 new commit to staging: https://git.io/v71uU
<NixOS_GitHub>
nixpkgs/staging 46a25ea Robin Gloster: linuxPackages.bcc: fix using wrapProgram on .c file
NixOS_GitHub has left #nixos []
<Infinisil>
clever: Yeah I did that for now, but I don't like it
<avn>
clever: I generally like idea of moving /etc/nixos to /nix/etc, and make /etc tmpfs
<NixOS_GitHub>
[nixpkgs] globin pushed 1 new commit to master: https://git.io/v71ub
<NixOS_GitHub>
nixpkgs/master 30d76b8 Simon Lackerbauer: seafile-client: 5.0.7 -> 6.1.0...
NixOS_GitHub has left #nixos []
<ylwghst>
srhb: really cool
newhoggy has quit [(Ping timeout: 255 seconds)]
ixxie has quit [(Quit: Lost terminal)]
* Infinisil
currently is on generation ~560, because he nixos-rebuild switches for everything
<avn>
sphalerite: in theory, /etc/shadow can be moved somewhere to /var
<sphalerite>
yeah, sure
<Infinisil>
avn: Oh, why is it in /etc anyways?
<sphalerite>
I agree it would be nice to separate the mutable and the immutable stuff completely
<sphalerite>
Infinisil: probably because nobody's bothered patching shadow-utils and pam(?) and whatever else needs it yet ;)
<avn>
Infinisil: why not use `nixos-rebuild test` for some modifications, if you know that you will rebuild again in 2-3 mins
<Infinisil>
sphalerite: Ah right, compatibility
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nix] edolstra pushed 1 new commit to master: https://git.io/v71zC
<NixOS_GitHub>
nix/master af765a8 Eelco Dolstra: Use /proc/self/fd to efficiently close all FDs on Linux...
NixOS_GitHub has left #nixos []
<sphalerite>
`grep -r /etc/shadow $(nix-store -qR /run/current-system/)` suggests glibc (libnss), pam, shadow (no surprise there), util-linux, and libcap-ng might need patching
<LnL>
^ nice!
<Infinisil>
Oh nice one indeed
<Infinisil>
brb, testing grep vs rg speed
<sphalerite>
rg is faster, but in a case like this one you'll want to make sure you are going through binary files as well, which rg defaults to not doing
<sphalerite>
iirc.
<avn>
sphalerite: even more, I am sure that some of them would be satisfied with symlink to real shadow (if they only need to read)
<LnL>
was actually referring to the nix commit, but yes grep is slow :)
<Infinisil>
grep: 41sec
<sphalerite>
I just used grep because I knew it would do what I wanted in this case, usually I use rg which I have aliased to rg -S
<cocreature>
hey, hackage-packages.nix is apparently auto-generated but I can’t figure out which file controls what packages are included in it, i.e., which file I actually need to edit. any ideas?
elurin has joined #nixos
<sphalerite>
LnL: and I was actually referring to Infinisil's message ;)
<Infinisil>
sphalerite: Are you sure grep searches binaries by default? Because I used rg -a which searches binaries and it turned up a lot more results than grep
<ylwghst>
srhb: users created using adduser are gone in new build so its neccesary to define users in configuration.nix right? or will users persist after nixos-build switch?
<sphalerite>
Also, if you ran grep before rg, rg will have had a cache advantage
<Infinisil>
ylwghst: By default they persist
<Infinisil>
ylwghst: If you disabled users.mutableUsers, then they won't persist and you'll have to do user management all with configuration.nix
<sphalerite>
But yes, rg is much faster also because it runs multiple threads
ThatDocsLady has quit [(Quit: Arma-geddin-outta-here!)]
<Infinisil>
sphalerite: Damnit, i forgot about that (again)
* Infinisil
mans grep for its parallel flag
<sphalerite>
Having run grep beforehand, meaning the cache should have been warmed up
<avn>
Infinisil: I believe most of us use configuration for user management, except passwords
<sphalerite>
not sure grep has multithread
<Infinisil>
sphalerite: man grep -> /parallel -> Pattern not found
linarcx has joined #nixos
MP2E has quit [(Quit: leaving)]
<ylwghst>
Infinisil: i did not but they are gone after nixos-install
<sphalerite>
yeah
<srhb>
ylwghst: Users will persist after nixos-rebuild switch, boot, whatever, yes.
<sphalerite>
grep is for compatibility, and for when you remember the options and just need to get shit done (and not search a lot of stuff). rg is much awesomer in the general case :)
<Infinisil>
sphalerite: Well I guess if grep can't even compete that's a clear win for rg :P
newhoggy has joined #nixos
<srhb>
Oh, irc lag. Fun.
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] fpletz pushed 2 new commits to master: https://git.io/v71gX
<NixOS_GitHub>
nixpkgs/master 1ed7862 Franz Pletz: searx: cleanup, remove python name prefix
<NixOS_GitHub>
[nixpkgs] michaelpj opened pull request #28082: Nixpkgs manual: expand documentation for overlays (master...overlays-doc) https://git.io/v71w2
NixOS_GitHub has left #nixos []
linarcx has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] dpino opened pull request #28083: Fix ruby built (master...fix-ruby-built) https://git.io/v71w9
NixOS_GitHub has left #nixos []
<Infinisil>
NickHu: I am so confused regarding GHC now
<Infinisil>
I know its self hosted, but how does this work with nixpkgs?
<linarcx>
hi. i want to use unstable channel.so delete stable channel and add unstable.after that update channel with:"nixos-rebuild switch --upgrade".i get this error:"error: file ‘nixpkgs/nixos’ was not found in the Nix search path (add it using $NIX_PATH or -I)".why?
<dpino>
clever: opened a pull-request to fix ruby/default.nix https://github.com/NixOS/nixpkgs/pull/28083 If you're a nixos reviewer would you mind taking a look when you have time? Appreciate
<Infinisil>
linarcx: What is the output of `nix-channel --list`, `sudo nix-channel --list` and `echo $NIX_PATH`?
<NickHu>
But I don't know how to recompile ghc without -dynamic
dpino has quit [(Ping timeout: 240 seconds)]
newhoggy has quit [(Ping timeout: 240 seconds)]
<Infinisil>
linarcx: Seems all well
<Infinisil>
Hmm
<Infinisil>
linarcx: What does `nix-instantiate --eval -E 'import <nixpkgs/nixos> {}' output?
<Infinisil>
linarcx: Does it throw an error?
<ylwghst>
Why this nix-build switch: getting status of /home/user/switch' no such file or directory ?
<linarcx>
yes.one error and one warning:"warning: Nix search path entry ‘/nix/var/nix/profiles/per-user/root/channels/nixos/nixpkgs’ does not exist, ignoring
<linarcx>
error: file ‘nixpkgs/nixos’ was not found in the Nix search path (add it using $NIX_PATH or -I), at (string):1:8
<linarcx>
"
<Infinisil>
linarcx: try sudo nix-channel --update, then run it again
<Infinisil>
ylwghst: nixos-rebuild switch
<cocreature>
Infinisil: iirc nixpkgs uses a binary for some ghc 7.x and then bootstrap from that
<Infinisil>
ylwghst: And you probably also want a sudo
<pbogdan>
linarcx: does it help if you remove the channel and re-add it but using "nixos" as the name, rather than "unstable" ?
revtintin has quit [(Ping timeout: 240 seconds)]
<Infinisil>
cocreature: I thought it would be something like that, imagine having to build the first ever ghc written in haskell, impossible with nixpkgs :P
<Infinisil>
pbogdan: Ohh right, that's it!
<cocreature>
Infinisil: well it’s not really different from gcc or most other compilers
<Infinisil>
Didn't even notice
<Infinisil>
cocreature: Are that many compilers self hosting?
<adelbertc>
how do i get the Git revision of the current nix channel i have on my machine (macOS) ?
<LnL>
yes
<adelbertc>
im in `~/.nix-defexpr/channels/nixpkgs`
<adelbertc>
but that seems to be an SVN thing?
<adelbertc>
not a git repo
<linarcx>
pbogdan:yes. but im thinking that maybe manipulate the $NIX_PATH will fix the problem.what is your idea?
<cocreature>
Infinisil: yeah, compiler writers tend to like the languages that they write compilers for so they want to use them for writing the compiler as well :)
<Infinisil>
cocreature: Yeah, but I thought it was more of an exception than a regularity
<adelbertc>
LnL: nix-env -q? i get "error: selected '/nix/var/nix.......' matches no derivations
<cocreature>
it’s quite common
<NickHu>
Nah, it's pretty normal
<NickHu>
That way they can develop the language in that language too
<NickHu>
e.g. Haskell extensions are written in Haskell etc.
<LnL>
adelbertc: I forgot -p before the path
<cocreature>
it’s also a way to demonstrate your language is mature is enough to write a significant program
<NickHu>
Yeah that too
newhoggy has joined #nixos
<cocreature>
gnah I have no idea how hackage2nix is supposed to work
<NickHu>
Has anyone else used the haskell profiling libraries?
<cocreature>
NickHu: have you seen the section in the manual about that?
<NickHu>
Rather, haskell libraries compiled with profiling enabled
<adelbertc>
ok so i got `nixpkgs-17.09pre111456.d4ef5ac0e9`.. how does that translate into a Git revision on GH? (my actual goal here is to pin nixpkgs in my build)
<Infinisil>
joepie91_to_go_: did you forget to call nixos-generate-config after a filesystem change?
<cocreature>
NickHu: what exactly are you running to get that error?
<joepie91_to_go_>
nope, no filesystem or hardware changes, but i just switched from 4.4 to 4.9 kernel and back to SDDM
<joepie91_to_go_>
one rebuild failed in a strange way
<joepie91_to_go_>
locked up my system
<joepie91_to_go_>
rebooted, did another rebuild
<joepie91_to_go_>
it appeared to succeed and then killed my display-manager service
<joepie91_to_go_>
did another rebuild from tty1, which succeeded... except it could not switch because of something to do with display-manager.service being masked
newhoggy has quit [(Ping timeout: 240 seconds)]
<Infinisil>
joepie91_to_go_: Sounds nasty
<joepie91_to_go_>
yeah, not entirely sure why it's failing in this way
<joepie91_to_go_>
so I've just rebooted into 4.4 and am now trying another rebuild
<joepie91_to_go_>
and whee, there goes my display-manager again
<joepie91_to_go_>
what the hell :D
<Infinisil>
joepie91_to_go_: rollbacks sure are nice :D
<ylwghst>
hm
<joepie91_to_go_>
bunch of X11 server errors in journalctl
<Infinisil>
joepie91_to_go_: Maybe you could try installing the new version to a stick and boot that instead, to make sure your hardware works with this kernel
<joepie91_to_go_>
right after "switching to system configuration [...]"
<Infinisil>
and your filesystem isn't screwed up
<Infinisil>
or is*
<joepie91_to_go_>
Infinisil: well, I can boot into the 4.4 revision fine
<joepie91_to_go_>
so filesystem should be fine...?
<Infinisil>
joepie91_to_go_: Not sure
<ylwghst>
did nixos-rebuild switch and my user created using useradd again gone
tnks has joined #nixos
newhoggy has joined #nixos
<tnks>
when debugging a Nix expression with nix-shell, do people call the phases explicitly one-by-one?
<Infinisil>
ylwghst: Yeah it doesn't work, why aren't you using configuration.nix for that?
<tnks>
seems like it's easy to forget to run a phase in the right order.
<tnks>
which makes me trust nix-shell less and want to just run through a bunch of nix-build calls instead.
<Infinisil>
tnks: Asking the important questions, I have no idea
<tnks>
AmIDoingItWrong™
thibm has quit [(Quit: WeeChat 1.9)]
<Infinisil>
I believe people are just doing nix-shell> configure -> buildPhase -> buildPhase -> buildPhase ... when doing normal development
<michaelpj>
tnks: I have exactly the same experience. It would be really nice if each phase set $nextPhase or something
<joepie91_to_go_>
okay, I have just produced a generation
<joepie91_to_go_>
let me try to boot into it
<ylwghst>
systemctl start display-manager.service
<joepie91_to_go_>
if it works, then I know what the issue is
<joepie91_to_go_>
:p
<joepie91_to_go_>
or at least, in what direction to investigate
<ylwghst>
nothing happen :/
<joepie91_to_go_>
it works
<joepie91_to_go_>
!
<Infinisil>
joepie91_to_go_: Nice
<Infinisil>
ylwghst: What error?
<ylwghst>
blank line and nothing
<joepie91_to_go_>
Infinisil: okay, so it would appear that the issue is a combination of changing the kernel version and the display manager in the same rebuild; changing the display manager will result in the display-manager.service being stopped (for obvious reasons), but this will break the ongoing rebuild, which apparently somehow breaks the generation of the new kernel image that the new GRUB entry points at
<ylwghst>
already definel all neccesary whats in nixos manual
<ylwghst>
d
pxc has joined #nixos
<joepie91_to_go_>
I have now set my config back to 4.4, and just switched to SDDM, and now it boots fine
<joepie91_to_go_>
although SDDM is still broken but that's likely for unrelated reasons
<pxc>
hey, everyone! does anyone here know off-hand what has changed w/r/t nix-shell syntax in Nix unstable vs current Nix? I've noticed thst some of my old magic shebangs stop working if I use nixUnstable
<Infinisil>
ylwghst: Try hitting Ctrl-Alt-F7
<Infinisil>
or ctrl-F7
linarcx has quit [(Remote host closed the connection)]
<Infinisil>
joepie91_to_go_: And you're updating the kernel separately now?
newhoggy has joined #nixos
<ylwghst>
Infinisil: nothing happens too
<ylwghst>
to
nslqqq has joined #nixos
<Infinisil>
ylwghst: What's your configuration.nix?
<joepie91_to_go_>
Infinisil: that;'s my plan
<joepie91_to_go_>
once SDDM works :)
<joepie91_to_go_>
it seems that display-manager.service is now masked
<joepie91_to_go_>
which is probably not good...
<Infinisil>
joepie91_to_go_: I want to switch to SSDM to some time
<Infinisil>
slim seems to be out of order
jellowj has joined #nixos
<joepie91_to_go_>
Infinisil: I was using lightdm until now for GPU compatibility reasons
<joepie91_to_go_>
which works fine but eh, SDDM is nicer
<joepie91_to_go_>
if I can't get SDDM to get going on this box, I'll just use lightdm again
<Infinisil>
joepie91_to_go_: I honstly don't even know what masked means
<joepie91_to_go_>
Infinisil: 'force-disabled' - can't be activated even manually
<Infinisil>
joepie91_to_go_: From systemd.unit: "If a unit file is empty (i.e. has the file size 0) or is symlinked to /dev/null, its configuration will not be loaded and it appears with a load state
<Infinisil>
of "masked", and cannot be activated."
<joepie91_to_go_>
whee, I have run into a fun bug
<joepie91_to_go_>
for some reason, my config has produced a masked display-manager.service
<Infinisil>
pxc: No idea, what error do you get with what shebangs?
colabeer has joined #nixos
<joepie91_to_go_>
and there are no issues about this on the issue tracker...
<jellowj>
hi, newbie here. i'm trying to make a nix package from a deb file and i constantly get 'libnssutil3.so: cannot opens shared object file'
<pxc>
Infinisil: I don't recall. I think it may have been caused by a change to python.withPackages, since I used that in the shebang. It's not a crucial issue for me. I'm sure I'll find it in the docs when the new release comes out, if it still comes up :-)
<Infinisil>
joepie91_to_go_: You can inspect the final config with nix-repl '<nixpkgs/nixos>', then e.g. systemd Maybe this can help to check the systemd stuff
<jellowj>
i cheked `file` output, same thing on executable and libnssutil3.so
<jellowj>
libnssutil3.so is obviously there
<jellowj>
what can be causing the trouble?
michaelpj has quit [(Ping timeout: 260 seconds)]
<clever>
jellowj: did you patchelf the rpath?
<jellowj>
yes
newhoggy has joined #nixos
<joepie91_to_go_>
Infinisil: meh, I have run out of patience for this for tonight :P
<joepie91_to_go_>
switching back to lightdm for now
<clever>
jellowj: use patchelf to print the rpath out after its built, is the dir still there?
<jellowj>
yes
<joepie91_to_go_>
will investigate more later
<clever>
jellowj: can you gist the output of patchelf, and ldd?
<Infinisil>
joepie91_to_go_: Happens to me too, some problems are just ugh
<joepie91_to_go_>
let's see if the switch to 4.9 works when I don't touch the display manager...
<Infinisil>
joepie91_to_go_: "Just gonna try this last thing.." *5 hours later* "damnit."
fresheyeball has quit [(Quit: WeeChat 1.9)]
<joepie91_to_go_>
hehe, yeuuup
<joepie91_to_go_>
I prefer interesting problems over frustrating problems anyway :)
<Infinisil>
ylwghst: What does `systemctl cat display-manager` on the new installation say?
<Infinisil>
ylwghst: I can't see anything wrong with your configuration.nix
newhoggy has quit [(Ping timeout: 240 seconds)]
<Infinisil>
Although the services.xserver.videoDrivers = [ "nouveau" ]; might be bad
<joepie91_to_go_>
hrm.
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] primeos pushed 1 new commit to master: https://git.io/v711F
<NixOS_GitHub>
nixpkgs/master 6225f1b Michael Weiss: android-studio-preview: 3.0.0.8 -> 3.0.0.9...
NixOS_GitHub has left #nixos []
<Infinisil>
Have no idea about video drivers, but I think normally you don't need to modify this option
<joepie91_to_go_>
"Failed to start cpufreq.service: Unit cpufreq.service is masked."
<joepie91_to_go_>
"warning: error(s) occurred while switching to the new configuration"
<clever>
jellowj: its also available in the nss package, so you could just use that directly
<jellowj>
yes
<jellowj>
i've tried
<jellowj>
same thing
<clever>
jellowj: what does file say about upwork and /nix/store/fxq608q4432gljd6ffadlzycv2cv91k0-upwork/usr/share/upwork/libnssutil3.so ?
<Infinisil>
btw, ylwghst you might want to use https://gist.github.com instead, much better than pastebin (i only said pastebin before because it's kinda common to mean any pastebin service). You can also install the CLI with `nix-env -i gist`, then `gist some-file` and you get an url to it uploaded
<clever>
jellowj: can you also gist the nix expression your using to build it?
newhoggy has joined #nixos
<ylwghst>
Infinisil: nice, definetely, forgot about gist
<joepie91_to_go_>
let's see if it boots
<Infinisil>
ylwghst: Do `gist /nix/store/z96db6hay6jsp0klpyaw8gs3mpkddm7z-unit-script/bin/display-manager-start`
<joepie91_to_go_>
Infinisil: now it fails to boot with the same kernel panic again...
<clever>
jellowj: what callPackage are you using to load the file?
<disasm>
where on github is the install script (that's in the tarball) that's used with the curl bash instructions for installing nix on another distribution?
<ylwghst>
clever: what is callPackage?
<joepie91_to_go_>
aha!
<joepie91_to_go_>
I may have found the issue
<clever>
ylwghst: thats how you load a nix package within nixpkgs
<ylwghst>
Infinisil: Started user manager foor UID 0
k2s has quit [(Remote host closed the connection)]
<disasm>
the install-nix-from-closure.sh is the one I was looking for, just comparing it to the steps listed in pill #2 to see if it's pretty much the same, seems pretty close.
<joepie91_to_go_>
Infinisil: so apparently the way the cpufreq governor is set was changed from 16.09 -> 17.03, and it should now be in hardware-configuration, except I hadn't re-run generate-config after the 17.03 upgrade
<ylwghst-desktop>
clever: rebooted and loged as burim in first tty
<ylwghst-desktop>
clever: and its working
<clever>
ylwghst-desktop: and by "login", so you mean the actual tty, did you never use su at any point?
<ylwghst-desktop>
clever: I use su
<clever>
thats why i asked how you got the shell, heh
<clever>
su breaks a lot of things
<clever>
you must use "sudo -u burium -i" to get a shell as another user
jellowj has joined #nixos
<ylwghst-desktop>
clever: ok thx
_ts_ has quit [(Quit: Leaving)]
gnuhurd has quit [(Ping timeout: 260 seconds)]
ylwghst has joined #nixos
<ylwghst-desktop>
cool
<ylwghst-desktop>
Im in gdm finally
<Infinisil>
\o/
<ylwghst-desktop>
Infinisil: i changed e19 to gnome
<ylwghst-desktop>
Infinisil: gnome is working like a charm
* Infinisil
looks up e19
FRidh has quit [(Quit: Konversation terminated!)]
<ylwghst-desktop>
Infinisil: the xserver is runing I can go sleep 2nite finally haha
<ylwghst-desktop>
I haven't slept since yesterday playing with nixos installation
newhoggy has joined #nixos
<Infinisil>
Heh, it's very addicting indee
<Infinisil>
d
<ylwghst-desktop>
Infinisil: great peace of software really
<Infinisil>
I just set up mpd running on a server, continuously playing my music. I can now listen to it on any machine, and control it on any machine
<Infinisil>
Even works with Sonos!
<Infinisil>
basically my own streaming service, that I fully control
<ylwghst-desktop>
I was using CentOS for years then I have started developing websites and switched to MacOS. Yesterday I was decided to try whats new in linux and fond nixos
newhoggy has quit [(Ping timeout: 240 seconds)]
<ylwghst-desktop>
I decided*
<Infinisil>
I was a mac user before NixOS too, a few months ago I switched
<ylwghst-desktop>
Infinisil: great idea
joepie91_to_go_ has quit [(Changing host)]
joepie91_to_go_ has joined #nixos
<ylwghst-desktop>
Infinisil: Have heard about C.H.I.P. by Nexthing Co.
<ylwghst-desktop>
?
<avn>
ylwghst-desktop: actually is only one distro which have something new in ;)
<Infinisil>
ylwghst-desktop: What the hell is that, an online karaoke thing?
kiloreux has quit [(Ping timeout: 240 seconds)]
<colabeer>
Infinisil: are is your mpd server also running nix?
<Infinisil>
Ah
<Infinisil>
colabeer: Yup :D
<Infinisil>
ylwghst-desktop: Ah I get it now, pretty cool
<adelbertc>
if i want the version of a particular Nix expression on HEAD, do i just copy/pasta it into packageOverrides in ~/.config/nixpkgs/config.nix?
<ylwghst-desktop>
Infinisil: I got Pro kit but havent touched it yet
<adelbertc>
i assume i can dump that in like ~/.config/nixpkgs/docker.nix and then in packageOverrides do dockerTools = import "./docker.nix" ?
<colabeer>
Infinisil: cool, I tried the same and it kinda works, but I wanted it to be configured as a user service in my configuration.nix, but that one did not work out proper
<Infinisil>
colabeer: You probably want to put that file into e.g. ~/.config/nixpkgs/docker.nix and then do environment.systemPackages = [ (pkgs.callPackage ./docker.nix {}) ];
newhoggy has quit [(Ping timeout: 240 seconds)]
<Infinisil>
umm, that's for adelbertc
<adelbertc>
Infinisil: i assume that was meant for me
<adelbertc>
:)
<Infinisil>
adelbertc: Ah right, packageOverrides
<adelbertc>
so what i said sounds right?
<colabeer>
:D
<Infinisil>
colabeer: Did you want to run mpd as a user service?
<Infinisil>
adelbertc: Yup :)
<ylwghst-desktop>
Infinisil: theres another pain thing I need to get work
<Infinisil>
ylwghst-desktop: Whew, not a single clue about any of: drivers, nvidia, GPUs
<Infinisil>
I just know there's some nvidia package or so, but if there's a problem I'm out of ideas
<ylwghst-desktop>
Infinisil: I know how to do it and wheres the problem
<Infinisil>
s/package/option
newhoggy has joined #nixos
<colabeer>
Infinisil: yes is did activate with service.mpd [enable = true; user = "someuser"; .... ] and then I wanted to start it with systemctl --user start mpd.service where it complained that it can not find mpd.service
<ylwghst-desktop>
Infinisil: I first need change pci register via setpci of nvidia PCIE bus
<Infinisil>
colabeer: Ahh yes, you need to put it in systemd.user.services
<colabeer>
Infinisil: hui thx I'll give it a try :)
<Infinisil>
I think
<ylwghst-desktop>
Infinisil: and build the driver on kernel 4.9.. it can be pain too :(
jonte_ has joined #nixos
<ylwghst-desktop>
Infinisil: Isn't there a package in store of nvidia drivers v. 430?
newhoggy has quit [(Ping timeout: 240 seconds)]
<ylwghst-desktop>
340 not 430
<Infinisil>
ylwghst-desktop: Do you know how to search through the repository?
<avn>
Infinisil: git grep is best ;)
<ylwghst-desktop>
Infinisil: not exactly yet, I'm using nixo webpage to list packages.
eschnett has quit [(Quit: eschnett)]
<ylwghst-desktop>
there is services.xserver.videoDrivers = [ "nvidiaLegacy340" ]; in manual
ericsagnes has joined #nixos
<ylwghst-desktop>
looks like thats it
<Infinisil>
ylwghst-desktop: nvidia_x11_legacy340 now
rtjure has quit [(Ping timeout: 255 seconds)]
<ylwghst-desktop>
Infinisil: can be store search trought cli ?
<ylwghst-desktop>
searched
<Infinisil>
ylwghst-desktop: Yes
<Infinisil>
You need a nixpkgs checkout
<Infinisil>
Well you alreday have one actually
<colabeer>
Infinisil: how did you specify the musicDirectory then? it complains that this option does not exist
<Infinisil>
colabeer: Well I'm not using a user service for that
<ylwghst-desktop>
If I change somethign in configuration.nix like nuoeveau to nvidialegacy will be the nouveau files purged after nixos-rebuild switch?
<ylwghst-desktop>
or will stay untouched?
<colabeer>
Infinisil: hm im gonna have to checkout its definiton :D but thx for the hint :)
gnuhurd has joined #nixos
<Infinisil>
colabeer: Why do you want to run it as a user service anyways?
berce has joined #nixos
<Infinisil>
ylwghst-desktop: NixOS will keep around every file it needs for older generations by default, to allow you to revert it back in case something gets messed up (really useful!)
<colabeer>
Infinisil: thought it is cleaner, saver, security and stuff. There is also other stuff running on the machine
<Infinisil>
ylwghst-desktop: You could delete all older generations, but it's not recommended
<jellowj>
how can know path to libudev?
<ylwghst-desktop>
Infinisil: OK
<jellowj>
how can i find out*
<ylwghst-desktop>
good
<Infinisil>
colabeer: I'm not sure, what makes this more secure? the normal mpd service already creates a separate user to run it as, so you can configure access for the "mpd" user specifically
<ylwghst-desktop>
Infinisil: Are there grub2-efi-modules in store?
<Infinisil>
jellowj: Maybe this works: nix-instantiate '<nixpkgs>' --eval -A libudev.outPath
<jellowj>
whould it work in nix expression?
roberth has quit [(Ping timeout: 255 seconds)]
<Infinisil>
ylwghst-desktop: There is grub2 in nixpkgs. grub2 is actually the default
newhoggy has quit [(Remote host closed the connection)]
<ylwghst-desktop>
Infinisil: I know but cant find modules in my installation
<colabeer>
Infinisil: Hm did not know that, I justh thought the less number of priviliges the better and besides I wanted to try out a service configured as user service in my configuration.nix ;)
<ylwghst-desktop>
I actually need setpci.mod grub2 module
<Infinisil>
jellowj: Sure, it's just `pkgs.libudev` in a nix expression, it's at the top level
newhoggy has joined #nixos
<Infinisil>
colabeer: You might wanna check out rycee[m]'s home-manager: https://github.com/rycee/home-manager It's basically a user-level configuration.nix, including user services
<Infinisil>
colabeer: And if you really want to make it super secury, you can use nixos containers
goibhniu has quit [(Ping timeout: 260 seconds)]
<jellowj>
Infinisil: thanks
edude03 has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
<Infinisil>
ylwghst-desktop: Why do you need that?
<Infinisil>
ylwghst-desktop: And what do you need exactly?
<ylwghst-desktop>
Infinisil: I ll try it explain it
<colabeer>
Infinisil: I gues, when I manage it to run as a user service, a container will be the next step. But its just a welcome opportunity to play with it after all.
Cale has joined #nixos
alunduil has joined #nixos
newhoggy has quit [(Ping timeout: 255 seconds)]
edude03 has joined #nixos
<ylwghst-desktop>
Infinisil: There is kind of EFI problem when nvidia driver is in use. The driver cannot get proper PCIE bus ID.
<ylwghst-desktop>
Infinisil: so the X will end on black screen
<Infinisil>
colabeer: I guess :P
<joepie91_to_go_>
Infinisil: I don't believe NixOS supports unprivileged containers yet?
<adelbertc>
hmm do local packageOverrides in ~/.config/nixpkgs/config.nix mess with distributed builds?
<clever>
adelbertc: it gets evalled locally, then the result of that is pushed out
<Infinisil>
ylwghst-desktop: Whew, no idea
<alunduil>
I've recently updated a package on hackage but it's not included in any builds due to failures. Is there anything I can do to get it re-added and building? I've fixed up the issues upstream already.
<adelbertc>
hm alright
<Infinisil>
ylwghst-desktop: Maybe ripgrep a bit through nixpkgs
<ylwghst-desktop>
the proper PCIE BUS and its function can be set using setpci
<ylwghst-desktop>
setpci -s "00:17.00" 3e.b
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] globin pushed 1 new commit to staging: https://git.io/v71h3
<NixOS_GitHub>
nixpkgs/staging 5e89bca Robin Gloster: evolution_data_server: wrap all and only executables
NixOS_GitHub has left #nixos []
<ylwghst-desktop>
setpci -s "00:17.0" 3e.b=8
_rvl has quit [(Ping timeout: 260 seconds)]
<ylwghst-desktop>
but in order to get it work this should be done in very early stage
<Infinisil>
joepie91_to_go_: Could be, yes. But he has root, so a container is possible and would be much more secure
simukis has quit [(Ping timeout: 240 seconds)]
tmaekawa has joined #nixos
<joepie91_to_go_>
Infinisil: well, my point here is that containers aren't "more secure" unless they are unprivileged containers, so if NixOS doesn't support unprivileged containers... :)
<joepie91_to_go_>
privileged containers only protect from accidental contamination
<clever>
irssi.out 11,680 x /nix/store/h1zm9jy0ns014nry2vj8s1vqzg0m7p5k-irssi-1.0.2/etc/irssi.conf
hiratara has joined #nixos
<Infinisil>
awesome
* Infinisil
is running nix-index
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] globin pushed 1 new commit to staging: https://git.io/v7MeT
<NixOS_GitHub>
nixpkgs/staging b1f5305 Robin Gloster: serf: fix darwin build
NixOS_GitHub has left #nixos []
<Infinisil>
oh hold on, isn't that the thing you need for command-not-found to work
<clever>
its the new replacement for command-not-found
<clever>
the old command-not-found used c++ to scan the closure of an entire hydra eval for binaries, and make a programs.sqlite
<Infinisil>
Well it tells me already what packages contain some executable when I try to run it, but it doesn't install it automatically
<clever>
Infinisil: the new nix-index is written in rust, and scans every file in every derivation in the eval
<Infinisil>
Rust \o/
<clever>
there is a command-not-found thing mentioned in the issues on it
newhoggy has joined #nixos
rpifan has joined #nixos
<Infinisil>
This would've saved me a few minutes today: nix-locate mpd.conf
roberth has joined #nixos
<ylwghst_>
Infinisil: what If I want look what contains package that Isn't installed
<ylwghst_>
Infinisil: not in /nix/store
<Infinisil>
ylwghst_: you mean to search for a package?
<ylwghst_>
Infinisil: not exactly
<ylwghst_>
Infinisil: I want to examine what is in package grub-2.x-2015-11-16
<ylwghst_>
Which files it contains
<clever>
ylwghst_: [clever@amd-nixos:~]$ ls $(nix-build '<nixpkgs>' -A grub --no-out-link)
<Infinisil>
Almost had the same thing ^^
<ylwghst_>
clever: what does nix-build '<nixpkgs'> ? How can I understand it?
<clever>
that tells nix-build to load the nix expressions in '<nixpkgs>' and then -A grub tells it to build grub
<clever>
it will then print the path to stdout, and --no-out-link stops it from creating a result symlink
<ylwghst_>
clever: I don't what to build it.
<ylwghst_>
want
<Infinisil>
ylwghst_: It uses the cache
<clever>
it will usually download it from the binary cache
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] volth opened pull request #28088: tinc: allow the daemon to write to files in /etc/tinc/${network}/hosts (master...patch-8) https://git.io/v7MvU
NixOS_GitHub has left #nixos []
<avn>
anyway, grub is not qt, and not chromium, it buitds fast ;)
<avn>
*builds
<ylwghst_>
avn: I don't need build it.
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] LnL7 pushed 2 new commits to master: https://git.io/v7Mvt
<NixOS_GitHub>
nixpkgs/master c7f1414 Matthew Bauer: miniupnpc: fix on darwin...
<sphalerite>
ylwghst_: you wont as long as you're on one of the channels
<ylwghst_>
clever: I don't understand it at all
<sphalerite>
"Build" often means "download"
<Infinisil>
clever: What would be cool is a symlink from /nix/pkgs -> /ipfs/XXXX which would contain all packages of the current profile. which you could check out by just doing cd /nix/pkgs/grub2
tmaekawa has quit [(Quit: tmaekawa)]
<avn>
ylwghst_: it simple, firstly nix calculate hash, if package with that hash exists in cache, it download it (unless package have preferLocalBuild = true), otherwise it build it locally
<ylwghst_>
sphalerite: ok
<clever>
Infinisil: the root problem with anything using nix + ipfs, is that /nix/store/<hash> is a hash over the directions on how to build it, not a hash of the output
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] globin closed pull request #28088: tinc: allow the daemon to write to files in /etc/tinc/${network}/hosts (master...patch-8) https://git.io/v7MvU
NixOS_GitHub has left #nixos []
<avn>
ylwghst_: some packages builds faster than downloads
<clever>
Infinisil: so you need a central index that maps hash(build-script) to hash(output)
<Infinisil>
clever: Yeah yeah, but problems set aside, that would be pretty nifty
<avn>
(latext is one of them)
<Infinisil>
clever: gcroot equivalent in IPFS would just be a pin :O
<ylwghst_>
avn: ok
<avn>
clever: may be hash of .drv salted with hash of builder key can be used as ipfs unique id?
<clever>
avn: the problem is that IPFS is purely a hash(value) = value lookup system, so you need to know the hash of the value
<clever>
avn: and if you have to run gcc on a source file and hash it, you already have the value, and no longer need the cache
Wharncliffe has joined #nixos
<avn>
clever: does it have some indexes?
<Infinisil>
Which also means everything needs to be deterministic i think
<clever>
Infinisil: nope, deterministic just ensures the same input always gives identical output
<ylwghst_>
avn: I acutaly want only to look in package that isn't installed on my system if there is /usr/lib/grub/x86_64-efi/setpci.*
<clever>
that doesnt infer any relation between the hash of the output and the input values
<clever>
avn: the closest thing is IPNS, which is based on publickey = signed(ipfs hash)
<clever>
avn: which points to a file or directory within ipfs
<clever>
[clever@amd-nixos:~]$ ls $(nix-build '<nixpkgs>' -A grub2_efi --no-out-link)/lib/grub/x86_64-efi/setpci* -l
<Infinisil>
clever: But from a.drv == b.drv doesn't follow a.out == b.out which means that people building the same derivation could end up with different hashes -> different IPFS objects, which wouldn't be very nice
<clever>
-r--r--r-- 1 root root 9016 Dec 31 1969 /nix/store/2anxmi1jpn6im0mqwxha2krmdjib2579-grub-2.x-2015-11-16/lib/grub/x86_64-efi/setpci.mod
<clever>
-r-xr-xr-x 1 root root 9560 Dec 31 1969 /nix/store/2anxmi1jpn6im0mqwxha2krmdjib2579-grub-2.x-2015-11-16/lib/grub/x86_64-efi/setpci.module
<clever>
ylwghst_: it is present
<clever>
Infinisil: if a.drv == b.drv, then a.out is always equal to b.out, but the contents within a.out may not match up
<Infinisil>
clever: Yeah that's what I mean, and that's what IPFS does
<clever>
Infinisil: a.out is just a path, computed from the contents of a.drv
<avn>
clever: in ideal case they should, and I hope we sometimes enforce it ;)
<clever>
the path in a.out has nothing to do with the actual value of the files inside a.out
<Infinisil>
IPFS does hashing on every file, therefore if a single bit is different in the output of a derivation, it gets a different hash, that's what I'm talking about
<Infinisil>
Determinism would prevent that
<clever>
yeah
silver_hook has quit [(Ping timeout: 248 seconds)]
dpino has joined #nixos
<clever>
Infinisil: do you know how the narinfo files work?
<Infinisil>
clever: I think you explained it before, but I can't fully remember. *tries to find a nar file*
<clever>
[clever@amd-nixos:~]$ nix-build '<nixpkgs>' -A grub2_efi --no-out-link
<clever>
so determinism never comes into the story
<clever>
its just a different way to download the .nar.xz
<clever>
and all binary-cache stuff continues to work as it does now
<Infinisil>
clever: That would work for a start yes
<clever>
next, using logic from nix-index, you could iterate over every derivation in a given channel release, bulk-download the narinfo's with ipfshash + signature, and setup a mirror
<Infinisil>
But I think that would take up double the space
<clever>
and thats where narfuse and fusenar come into play
<clever>
i wrote a custom fuse layer, that lets you mount a directory full of .nar files at /nix/store
<clever>
and ipfs has said that in the future, they will allow sharing files you have promised to keep immutable, so they dont have to be chunked up and put into ~/.ipfs/blocks/
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] LnL7 pushed 1 new commit to staging: https://git.io/v7Mfy
<NixOS_GitHub>
nixpkgs/staging bf8c125 Daiderd Jordan: cc-wrapper: fix LD_DYLD_PATH on darwin...
NixOS_GitHub has left #nixos []
<Infinisil>
clever: I think it's been implemented already I think
<clever>
with narfuse or fusenar, you can just take a raw directory of 2anxmi1jpn6im0mqwxha2krmdjib2579-grub-2.x-2015-11-16.nar files
<ylwghst_>
need to learn alot about nix
<clever>
and it will magically turn into a working /nix/store/ directory
<clever>
and it doesnt need to talk to ipfs at any point
<clever>
and it wont care if ipfs is sneaking a look and reading the same files to share
<clever>
the only thing really missing, is an RPC that allows adding/removing nar files, and to replace nix's internal "unpack nar to /nix/store" with "copy nar to X and RPC narfuse to reload"
<Infinisil>
clever: So fusenar can provide a /nix/store that has everything magically available like /ipfs. How would this work with ipfs?
<Infinisil>
would fusenar be adopted to enable ipfs support if narinfo's contained an ipfshash
<clever>
a seperate daemon, would need to download the .nar file (http or ipfs), and store it in a path like /nix/nars
<clever>
and ipfs can then optionaly share the content of /nix/nars back to the world
<clever>
and it will keep working even if ipfs is turned off
<mpcsh>
Infinisil: why not? That looks like a perfectly reasonable change!
<Infinisil>
Hmm, but this shouldn't cause it to not display
<mpcsh>
Infinisil: yeah :/ when I run it from the git repo (like if I clone dylanaraps/neofetch and run ./neofetch), it works perfectly
<Infinisil>
mpcsh: How about updating your local neofetch then?
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] volth opened pull request #28089: nixos/tinc: do not tell systemd where is pidfile (master...patch-9) https://git.io/v7MJb
NixOS_GitHub has left #nixos []
<Infinisil>
I mean it's really just a bash script
<mpcsh>
Infinisil: well sure, but I'd like that to be transparent to me
<mpcsh>
Like I'd like it to behave just like any other package
Sonarpulse has quit [(Ping timeout: 260 seconds)]
<clever>
brb
<ylwghst_>
clever: It will download it into /nix/store and create link in users profile right?
digitalmentat has quit [(Quit: Leaving)]
<jellowj>
i get this from `nix-index` attribute ‘webkitgtk24x-gtk2’ missing, at /nix/store/5ikqdj0wg9nyvz31r3n3nf0pxj9zl34k-nixos-17.03.1662.93498869b9/nixos/pkgs/development/haskell-modules/configuration-nix.nix:132:45
<Infinisil>
ylwghst_: Correct
<Infinisil>
jellowj: By just running `nix-index`
<Infinisil>
?
<jellowj>
yes
pxc has quit [(Ping timeout: 260 seconds)]
<Infinisil>
jellowj: does nix-locate work otherwise?
<jellowj>
nope, i need to generate index to use it
dpino has quit [(Ping timeout: 248 seconds)]
willprice has quit [(Ping timeout: 240 seconds)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] globin pushed 1 new commit to staging: https://git.io/v7MUz
<NixOS_GitHub>
nixpkgs/staging cc5480c Robin Gloster: csmith: fix wrapProgram call
<Infinisil>
jellowj: Do you happen to have installed websnap?
<jellowj>
no
<ylwghst_>
ush3d
<clever>
ylwghst_: nix-build creates a symlink called result in the current directory
<jellowj>
not to my knowledge
<ylwghst_>
ylwghst_: and nix-env -iA download into /nix/store and creates links in users profile ?
<Infinisil>
jellowj: What channel are you on?
<jellowj>
unstable
<jellowj>
but i got that error on stable too
<clever>
ylwghst_: yes
newhoggy has joined #nixos
<Infinisil>
jellowj: I get the same error when I switch to that revision, weird
newhoggy has quit [(Remote host closed the connection)]
newhoggy has joined #nixos
<Infinisil>
On master I get: Package ‘webkitgtk-2.4.11’ in /home/infinisil/src/nixpkgs/pkgs/development/libraries/webkitgtk/2.4.nix:21 is marked as insecure, refusing to evaluate.
<Infinisil>
When running nix-instantiate --eval --strict -A pkgs.haskellPackages.websnap
<Infinisil>
Perfect oppurtunity to test git bisect for the first time :D
mizu_no_oto has quit [(Quit: Computer has gone to sleep.)]
Rotaerk has joined #nixos
<ylwghst_>
clever: so it should be run without sudo or links are created in roots profile right?
<clever>
ylwghst_: it will be added it to the ~/.nix-profile of the user running nix-env
mizu_no_oto has joined #nixos
<ylwghst_>
got it
<ylwghst_>
thx
<jellowj>
well, i still don't know how to find libudev.so.0 on my system
<clever>
libudev.lib 0 s /nix/store/qqz3iq8w5cqakdqw2rjbylzabf1pdb5z-systemd-232-lib/lib/libudev.so
<clever>
jellowj: there is a different version in libudev.lib
newhoggy has quit [(Ping timeout: 258 seconds)]
alunduil has quit [(Quit: leaving)]
<ylwghst_>
hm
<ylwghst_>
where is default grub2 configuration?
<ylwghst_>
there is no /etc/grub.d/
<clever>
ylwghst_: it gets generated automatically when you run "nixos-rebuild switch" or "nixos-rebuild boot"
saintcajetan has quit [(Quit: Connection closed for inactivity)]
<Infinisil>
ylwghst_: Yes, it's inserted exactly as it is
<Infinisil>
you should really experiment with the language a bit, you can just write a nix expression that uses this and evaluate to see such things for yourself
<ylwghst_>
Infinisil: can there be also doublequotes " xxx "; will it work same as '' xxx ''; ?
<Infinisil>
ylwghst_: Try it out!
mizu_no_oto has quit [(Quit: Computer has gone to sleep.)]
<Infinisil>
ylwghst_: Then evaluate it with nix-instantiate --eval path/to/file
<jellowj>
thanks everybody
<clever>
ylwghst_: '' based strings will strip the common indentation from every line
<clever>
ylwghst_: so the string in that gist winds up having zero indentation
<adelbertc>
how does packageOverride play with pinning nixpkgs? i have something similar to https://github.com/NixOS/nixpkgs/issues/27994 in my build, but i also have an attribute defined in ~/.config/nixpkgs/config.nix called `customDockerTools`. when i try to call customDockerTools in my build though it complains 'undefined variable 'customDockerTools''
<jellowj>
for a moment there i was considering switching back to arch
<clever>
ylwghst_: which lets you indent the nix file nicely, without spamming the resulting grub config with weird indents that dont line up
<jellowj>
almosl lost hope
<ylwghst_>
it doesnt work :/
<Infinisil>
jellowj: D:
<ylwghst_>
nor with '' string '';
<Infinisil>
jellowj: Glad you're sane \o/
<clever>
ylwghst_: what error, and what is the exact contents of your configuration.nix?
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] joachifm pushed 1 new commit to master: https://git.io/v7MLt
<NixOS_GitHub>
nixpkgs/master 767b2ae Joachim Fasting: nixos/dnscrypt-proxy: default to random upstream resolver
<clever>
ylwghst_: that error doesnt appear to have anything to do with grub
roberth has joined #nixos
jellowj has quit [(Ping timeout: 240 seconds)]
<clever>
ylwghst_: the grub on nixos will do 2 main things
<clever>
ylwghst_: first, it will inspect the list on nixos generations, and auto-generate grub.conf, while obeying boot.loader.grub.extraConfig
<clever>
ylwghst_: and second, it will check /boot/grub/state, and if the content is "wrong", it will re-run grub-install to update the binaries in /boot with the latest version