<ottidmes>
If I import e.g. a lib.nix in 40 files, will it actually reevaluate the contents of lib.nix 40 times, or is there some caching in place? I mean directly import, no via the imports of the module system
<hakujin>
I want to do something that I assume is incorrect: I want to run a derivation only for side effects, because it's convenient for my hydra setup. I want to call `nix-copy-closure` from within a shell derivation but it doesn't have access to the local nix store. is there a hacky way around this?
hakujin has quit [Read error: Connection reset by peer]
hakujin has joined #nixos
foo___ has joined #nixos
jrolfs has joined #nixos
orivej has quit [Ping timeout: 264 seconds]
<ottidmes>
Myrl-saki: You can just use pkgs.haskell.ghc802.ghcWithPackages (hsPkgs: with hsPkgs; [ mtl random ]) to get a GHC in which you have access to those libraries, should be plenty of examples to be found online
<hakujin>
lost connection; sorry for the repost:
<hakujin>
I want to do something that I assume is incorrect: I want to run a derivation only for side effects, because it's convenient for my hydra setup. I want to call `nix-copy-closure` from within a shell derivation but it doesn't have access to the local nix store. is there a hacky way around this?
<Myrl-saki>
ottidmes: Nooo, I'm trying to make a build system.
<Myrl-saki>
On my local machine, I tried using echo. Can anyone make a pull request? As much as I want to, I have a lot of packages in my nixpkgs that should have been an override, but I didn't know I could use an override. So, basically, my nixpkgs is too dirty.
<Myrl-saki>
Another is also to add \n
<Myrl-saki>
Oh wait, printLines also works?
<Myrl-saki>
Lmao. It actually used to be printLines, so it's a regression?
<thoughtpolice>
gchristensen: I misspoke, but figured out my question. I had an antiquote for an attrset name, e.g. x."${foo}", but I wanted 'foo' to be a nested attribute selector, e.g. 'foo.bar', yielding 'x.foo.bar'
<thoughtpolice>
But antiquotes are always coerced to strings, thus the string `"foo.bar"` and selector `foo.bar` are not the same
peterzky has joined #nixos
<thoughtpolice>
What I really wanted was something like: `f = path: obj: lib.getAttrFromPath (lib.splitString "." path) obj`
<gchristensen>
@disasm try triggering the eval again
<gchristensen>
Disasm someone broke master and master needs to be fixed. Ofborg is not at fault. Can you do it;
<{^_^}>
[nixpkgs] @disassembler pushed 2 commits to master: https://git.io/vxJuG
<{^_^}>
→ ba0e1882 by @disassembler: nixui: inherit libtool for derivation
<{^_^}>
→ 7b8fa347 by @disassembler: Merge pull request #36895 from disassembler/nixui
<disasm>
same issue as airfield ^^
<{^_^}>
[nixpkgs] @dtzWill opened pull request #36896 → metrics: fixup sed expr to work with Nix 2 → https://git.io/vxJuy
logzet has quit [Remote host closed the connection]
<dtz>
there's more "libtool" badness on master, dunno what it's about but please fix--or better yet avoid pushing broken things in the first place please :3
<dtz>
'zed' is broken apparently, idk, see the PR just above --I include the error there
<disasm>
dtz: do you have a better way of finding problems rather than doing one at a time and running borg eval?
<dtz>
haha "sorry", and ty for fixage. Wasn't sure if it was on your backlog and instead of pestering you further figured I'd fix since seemed straightforward (after looking how you fixed the others, especially) ^_^
<disasm>
dtz: you wanna take over? I really should go to bed
<dtz>
well, I was trying to fix the "metrics" job haha and it was triggering those , I think while doing a "nix-env -qa" so that'll get everything not unfree
<joepie91>
I think that's a mistake in the parser :)
hamishmack has joined #nixos
<dmj`>
clever: ping
<clever>
dmj`: pong
blonkhart has quit [Quit: WeeChat 1.9.1]
mrkgnao has quit [Ping timeout: 240 seconds]
<justanotheruser>
sudo nixos-rebuild switch --upgrade fails with "error: file ‘nixpkgs/nixos’ was not found in the Nix search path (add it using $NIX_PATH or -I), at (string):1:13"
<justanotheruser>
what should I do?
<justanotheruser>
I assume I screwed it up somewhere, and actually manually setting $NIX_PATH wouldn't be the right thing to do
<clever>
justanotheruser: what does `sudo nix-channel --list` say?
<clever>
justanotheruser: you must never run nixos from the nixpkgs channel
<clever>
justanotheruser: you need to switch to the nixos-unstable channel
robstr has joined #nixos
<justanotheruser>
ah right, that makes sense, I replaced nixos-foo with it
<justanotheruser>
ty
robstr has quit [Ping timeout: 246 seconds]
ryanartecona has joined #nixos
robstr has joined #nixos
ryanartecona has quit [Ping timeout: 256 seconds]
MercurialAlchemi has joined #nixos
Guanin has quit [Ping timeout: 260 seconds]
rogue_koder has quit [Quit: Konversation terminated!]
Guanin has joined #nixos
<kaychaks>
with nix 2.0 I am seeing the error `Peer certificate cannot be authenticated with given CA certificates` whenever am trying to install something. I've got a MITM certificate as being inside a corporate env. With Nix versions < 2.0 this never was a problem as I've got a `~/.curlrc` file with `insecure` flag in it. What changed and how can I get the same thing with nix 2.0.
<clever>
kaychaks: nix 2.0 uses libcurl, rather then the curl command, so it probably needs the certs in a different place
<clever>
[clever@amd-nixos:~]$ ls /etc/ssl/certs/
<clever>
ca-bundle.crt ca-certificates.crt
<clever>
kaychaks: what happens if you put it in this area?
<kaychaks>
but I want to just ignore the certificate check (hence the insecure flag), how can I do that
mounty_ has quit [Ping timeout: 260 seconds]
reinzelmann has joined #nixos
<clever>
kaychaks: i dont think nix2 has that option right now
<clever>
this generates a nix bundle that installs nix under ~/nix-installl/nix/store/
<clever>
and supplies a working namespace helper to chroot into it, without root
<sphalerite>
dtz: for "force-drop openssl ref, hopefully okay O:)" we have removeReferencesTo for that :)
<dtz>
so I don't have the data/figures easily linkable at the moment but what I'm doing is orthogonal to "packing" techniques like upx and kinda nix-bundle
<dtz>
but no only saw what went to nix-bundle itself! Looking now ^_^
<dtz>
it's .. 24MB? but it was smaller before including the other tools too xD
<dtz>
but nix-wise that path has no external dependencies
<dtz>
and full bash! and other things :)
<dtz>
also it's fun just to run all the commands in libexec/nix (generally "allmux" produces fun things haha) -- all of those programs from a single binary xD
<hyper_ch2>
clever: infinisil : with that proposed change to zfs force import - won't that be a problem when a pool was mounted elsewhere?
<dtz>
at a high-level it's super unimpressive, why would that be hard? haha. But it's impossible with conventional practices..... ALLVM or an ALLVM-like approach is what makes this possible for real-world software :)
<clever>
hyper_ch2: ?
<dtz>
well it takes as input a list of "allexe"s-- which are 100% LLVM IR programs
Intensity has quit [Remote host closed the connection]
<dtz>
currently they're self-exec'ing zip's of bitcode files --separate files are separate dynamic objects,
<clever>
dtz: ahh
<dtz>
I have... a ridiculous amount of software built this way xD, most of my daily stack is allexe-based
<clever>
dtz: ive just been packing up entire nix closures in nix-bundle
<clever>
my installer is 125mb
<clever>
but that includes a couple haskell binaries, and electron
<dtz>
yeah, and that's important/good work
<dtz>
(seriously)
<clever>
currently, i'm having some upgrade troubles
<clever>
dtz: namespacing works, but nesting namespaces doesnt
ma27 has joined #nixos
<clever>
hyper_ch2: ah, as long as the hostId baked into the initrd is the one to last import a pool, it will just work
<dtz>
anyway basically we all systematically convert our programs into obfuscated brittle things that we no longer understand... and so necessarily that limits what we can do with the result... this is particularly important/limiting as we progress through dev to dev-ops and beyond
<clever>
thats my understanding
<hyper_ch2>
clever: I don't follow
Lisanna has joined #nixos
<clever>
hyper_ch2: zfs uses the hostId to keep track of which machine is using a pool, incase its on a shared block device like iscsi/nbd
davidlt has quit [Ping timeout: 256 seconds]
<dtz>
said another way we all "know" that often what our computers execute is sub-optimal but lack the ability to easily take advantage of it or do something about it
<clever>
hyper_ch2: and force import disables the safeties for concurrent access detection
<dtz>
"take these libraries and programs and make a new one with all of them" is easy to describe but unthinkable in today's world :)
<dtz>
anyway
<dtz>
.....
<dtz>
<3
shikiryogy has joined #nixos
<hyper_ch2>
clever: so if you boot up a live iso, force import the pool... reboot... then you won't have an issue?
<clever>
dtz: yep, it was just a ram shortage, i gave it a gig of swap and it now runs
<clever>
hyper_ch2: because the wrong hostid was last using it
<dtz>
oh, yay
<dtz>
i mean kinda :( but yay easy fix
<dtz>
spent a few hours today reworking Nix's decompression logic
<hyper_ch2>
clever: so it seems contra indicated to apply this change
<clever>
hyper_ch2: i do agree that turning force off is better
<dtz>
our nar-fetching code currently downloads the thing... puts it into a std::string... copies it a few times (don't think more than 2 are ever live at once, but still) and then decompresses it in-memory
<dtz>
all of which is done via multiple worker threads
<etu>
rycee: I'm getting errors when installing home-manager on a new box. All my older boxes is fine. "/nix/store/g8hzn737vrcyir5jd851bk573m31mx1k-home-manager-generation/activate: line 70: oldGenPath: unbound variable"
<dtz>
so you end up with a silly amount of RAM required just to download the paths for your profile or w/e
<clever>
dtz: i was getting an OOM just executing an arx bundle
<hyper_ch2>
clever: so, if there's no force-import and you boot up live iso, force-import pool and then reboot.... how to fix this then?
<clever>
dtz: which is technically a 124mb "bash" script
reinzelmann has quit [Quit: Leaving]
<dtz>
lmao
<clever>
hyper_ch2: you have to tell the initrd to force-import once, to repair the state
<hyper_ch2>
clever: how to do that?
<etu>
rycee: Both of these are empty: /nix/var/nix/{profiles,gcroots}/per-user/$USER but existing directories owned by my user
<clever>
hyper_ch2: run the zpool import with the force flag, from the initrd shell
<hyper_ch2>
also, it seems every fresh zfs install will be affected
<clever>
hyper_ch2: and you get an initrd shell by booting with boot.allow_shell i think
<hyper_ch2>
clever: remote reboot of encrypted dataset?
<clever>
hyper_ch2: if your installation has the same hostid as the livecd, it wont have the problem
<clever>
and you can scramble the livecd hostid, before you create the pool, to mantain uniqueness
<clever>
or you can just choose to ignore the safety and turn force on for all of your machines
<hyper_ch2>
:)
<hyper_ch2>
btw, is there a good webirc client that can connect to znc?
<clever>
i just run irssi under screen over ssh
<hyper_ch2>
irssi is too complicated for my simple brain
<hyper_ch2>
and why screen and not tmux?
<clever>
i learned screen first
<hyper_ch2>
same here
generic has quit [Ping timeout: 240 seconds]
shikiryogy has quit [Remote host closed the connection]
shikiryogy has joined #nixos
<clever>
dtz: let me see where all my closure size is going to...
<dtz>
lol two versions of glibc are indeed hopefully unnecessary
vidbina has joined #nixos
<clever>
the issue, is that each github repo has its own nixpkgs pinned
<dtz>
yeah
reinzelmann has joined #nixos
<clever>
and i already broke everything once when trying to just merge everything into a single $out/lib/
<clever>
the glibc's didnt like being mixed up
<dtz>
>.<
<dtz>
* >.<
<dtz>
$#%! okay idk if that's rendered badly on normal IRC clients
<dtz>
...christ I need to revive my irssi+screen server lol
<clever>
exec 2>&1 | tee $DAEDALUS_DIR/Logs/pub/post-install.log
<clever>
dtz: i also need to look into seeing if this is possible
<dtz>
but so the allmux idea is .... if every program is fully specified, those problems go away. And indeed-- many aspects of conventional packaging are "incidental complexity", as opposed to "essential complexity", to borrow some vocabulary. That is to say.. why exactly do our programs require an unspecified and often unpredictable organization of files (dynamic libraries particularly, but other things too) in order to execute? Is
<dtz>
that a necessary part of what it means to be a program? Answer: no :)
<dtz>
that's just an implementation detail that's been accepted for decades for no particular reason other than momentum
<clever>
electron requires a surprising number of files to even start
<clever>
and often wont give clear errors
<clever>
or it fails after doing more things first
<dtz>
(one should note that nearly every high-level language basically replaces all of the traditional abstractions for libraries/modules/etc ... there is a reason they do this)
<clever>
for example, electron will print a dbus error, then some opengl errors, then crash
<dtz>
yeah I've never even tried electron this w ay lmao
<clever>
fixing opengl doesnt resolve the issue
<clever>
fixing dbus does
<dtz>
it just seems .... (sorry everyone that likes it) .... like a trashfire
Tucky has joined #nixos
FareTower has quit [Ping timeout: 264 seconds]
<dtz>
lmao re:error vs fix
<clever>
dtz: oh, but i have been thinking of something similar to your idea
<clever>
currently, there is a haskell binary, that has to run at the same time as the electron binary
<clever>
and all kinds of fun problems happen if you dont run both
<clever>
electron can load plugins that define custom javascript methods
<dtz>
btw my terminal, my shell, my editor.... all of these are allexe's that are JIT compiled (with compilation cache and optional AOT compilation) that I use every day and have for... 6-12 months xD :D. It's neat.
<dtz>
it's also neat to be able to pull a lisp-machine-esque maneuver and say "show me the bitcode for the program I'm running, process it with this LLVM pass, and run the result"
<dtz>
only my love of Nix has basically done the same to me, but I've yet to bridge the gap to haskell wonderland
<clever>
seen it :D
<dtz>
*land of wonder
<dtz>
hehe
<dtz>
deep down? I think I'm super okay being an "unemployable" crazy. Although I question the "unemployable" part ;). Just employable in different circles lol
<clever>
[clever@system76:~/iohk/daedalus]$ exec 2>&1 | strace tee /tmp/test.txt
<clever>
read(0, "", 8192) = 0
<clever>
dang
Lears has quit [Remote host closed the connection]
Lears has joined #nixos
<dtz>
haha to "prove" that I build my paper/slides/poster/etc PDF's using only ALLVM code (I have to post-process it to pull in data files so there's a "risk" of native code execution) I have some code that --using a JIT'd strace, lol -- modifies mkDerivation to run itself (builder) under strace and verifies everything went through the allvm "interpreterer"
<dtz>
which is called "alley" because of ... reasons. ALLVM Execcution "Yengine", because "Allee" (Execution Engine) isn't a word :P
<infinisil>
Although, clever are you sure that it checks that the last hostid matches? Because I thought zpool import -f is only needed when the pool wasn't exported correctly by the previous system (this is also the error message when a normal import doesn't work)
<infinisil>
So I'd think -f is only needed when [It wasn't exported properly] and [The hostId doesn't match]
<adisbladis[m]>
This looks funky to me. Why are we copying golang src to $out :s
Itkovian has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
scribbler has quit [Ping timeout: 245 seconds]
zzamboni has quit [Quit: Leaving.]
<LnL>
because go
MP2E has quit [Remote host closed the connection]
<adisbladis[m]>
LnL: Yes but why is it in $out. Why not $dev
* dtz
nods
<hyper_ch2>
infinisil: does pool properly get exported when you properly shut down the system? If so, in a live usb it will still prompt you to force import
<adisbladis[m]>
With golang projects habits of vendoring every single source $out becomes huge
<LnL>
I agree but the go stuff uses bin/out insteas
<hyper_ch2>
LnL: btw, using now znc on nixos :)
<LnL>
and they reorder the outputs so bin is first
<infinisil>
hyper_ch2: I'm not sure myself, but will look into it, see comment on PR
<adisbladis[m]>
LnL: Yeah. I'm gonna have a look of that tonight. Potentially a good low hanging fruit for closure size reduction
<{^_^}>
[nixpkgs] @rnhmjoj opened pull request #36900 → noteshrink: init at 0.1.1 → https://git.io/vxJXo
zzamboni has joined #nixos
vidbina has joined #nixos
<infinisil>
Well damn, RIP this PR for at least a year hyper_ch2 clever
<garbas>
wilornel: sry, i was already out yesterday. not sure how much i can help with ib office stuff but i can always try
* hyper_ch2
gives infinisil a lot of cookies
* infinisil
munches on the cookies
* adisbladis[m]
munches on JWT
zzamboni has quit [Quit: Leaving.]
<infinisil>
Actually, all hope is not lost
<Lisanna>
woo, I might get an opportunity to use disnix for a project soon
periklis has joined #nixos
knupfer has joined #nixos
zzamboni has joined #nixos
<hyper_ch2>
infinisil: 'Abandon hope all ye who enter here' :)
<infinisil>
hyper_ch2: See PR comment, it's not that bad actually
zzamboni has quit [Ping timeout: 246 seconds]
rauno has joined #nixos
<LnL>
adisbladis[m]: out is basically dev for the go stuff and won't get into your closure unless you explicitly ask for foo.out
<hyper_ch2>
infinisil: you just wrote that to make yourself feel better :)
<elvishjerricco>
Does nix-shell not have `--add-root`?
<infinisil>
hyper_ch2: nah it's true! Just need to implement proper exporting now
<hyper_ch2>
but yes, if there's proper export on shutdown it should work well
<infinisil>
"Just"
<rauno>
Morning, what is the best way of making ethtool changes permanent in nixos?
<hyper_ch2>
well, I did try to set a mirror disk as "offline" before powering off... wasn't successfull
<hyper_ch2>
rauno: what's ethtool?
<clever>
hyper_ch2: offline isnt what you want
<clever>
hyper_ch2: you want export
<hyper_ch2>
clever: no, I do want offline for external usb at home and office
<rauno>
ethtool - query or control network driver and hardware settings
<infinisil>
But this has to be done at the very last point when it's the root fs
<hyper_ch2>
because then resilvering is faster... often they just resilver everything because they were not set to "offline"
<clever>
hyper_ch2: offline just stops writes, which can lead to the pieces of the mirror getting out of sync with eachother
<hyper_ch2>
clever: I know what offline does
<clever>
ahh
<clever>
thought you where talking about the pool not being closed properly at shutdown
zzamboni has quit [Ping timeout: 245 seconds]
<infinisil>
hyper_ch2: hmm, is `zpool offline -t` what you do? -t says: "Temporary. Upon reboot, the specified physical device reverts to its previous state."
<{^_^}>
→ a28464cb by @volth: frostwire: 6.4.5 -> 6.6.3-build-253
<{^_^}>
→ 0596a7c6 by @volth: frostwire: fixed meta.platforms
<{^_^}>
→ cedfe316 by @volth: frostwire: restore download-and-unpack derivation as frostwire-bin
<rauno>
hyper_ch2, but not the driver options?
zzamboni has joined #nixos
<hyper_ch2>
not sure what you mean by that
reinzelmann has joined #nixos
<hyper_ch2>
rauno: can't help you there
reinzelmann has quit [Client Quit]
<rauno>
ok then
zzamboni has quit [Ping timeout: 248 seconds]
reinzelmann has joined #nixos
nixer101 has joined #nixos
zzamboni has joined #nixos
asuryawanshi has quit [Remote host closed the connection]
<pjan>
I just tried installing NixOS with the live cd (non-graphical), but it seems broken. Throws a "unknown setting 'signed-binary-caches' while creating the nix derivation.
<nixer101>
hello! Can anyone who uses KDE desktop with sound card pastebin his configuration.nix? I have a problem, - no phonon backends. The cards are detected, but phonon backend selection in empty in System settings.
<{^_^}>
[nixpkgs] @svanderburg pushed to master « remarkjs: regenerate dependencies with node2nix 1.5.3 »: https://git.io/vxJ7N
davidlt has joined #nixos
davidlt_ has quit [Ping timeout: 240 seconds]
<{^_^}>
[nixpkgs] @svanderburg pushed to master « nixui: regenerate with node2nix 1.5.3 »: https://git.io/vxJ5J
<rauno>
anyone know do the system.activationScripts run on boot also? or only with nixos-rebuild ?
<{^_^}>
[nixpkgs] @adisbladis pushed to master « overmind: init at 1.1.1 »: https://git.io/vxJ5m
<rauno>
Didn't seem to work with reboot
<sphalerite>
rauno: yes, they should run at boot as well. What are you using them for though?
<{^_^}>
[nixpkgs] @svanderburg pushed to master « airfield: regenerate with node2nix 1.5.3 »: https://git.io/vxJ5u
<sphalerite>
They run very early on during the boot so thye may not behave the way you want them to. You probably want a oneshot systemd service instead.
<rauno>
sphalerite, currently trying to persist some network interface driver options
<rauno>
using ethtool
<rauno>
well i do need them early so the network would work normally :)
<sphalerite>
oh right they're necessary on power up
<Lisanna>
sphalerite I wonder if a plugin could be used to address this somehow...
<sphalerite>
in that case maybe make them wantedBy = [ "networking.target" ]; or something
<sphalerite>
Lisanna: I don't think so, I think plugins are only for adding builtins
<{^_^}>
[nixpkgs] @svanderburg pushed 4 commits to release-18.03: https://git.io/vxJ5H
<{^_^}>
→ f04f6132 by @svanderburg: remarkjs: regenerate dependencies with node2nix 1.5.3
<{^_^}>
→ 74f70f91 by @svanderburg: zed: regenerate packages with node2nix 1.5.3
<{^_^}>
→ 41f93c62 by @svanderburg: nixui: regenerate with node2nix 1.5.3
<Lisanna>
sphalerite could a builtin produce a function dynamically though that can capture --arg and --argstr?
knupfer has quit [Read error: Connection reset by peer]
<rauno>
ok, ill try the systemd also sphalerite
<sphalerite>
I don't think so… probably easier to just fix the issue. I agree that ... should capture --args
humanoyd has joined #nixos
<Lisanna>
like builtins.mkFn arguments fn
<Lisanna>
where arguments is something like [ "arg1" "arg2" ] and fn is (args: ...)
<Lisanna>
then the result is args @ { arg1, arg2 }: fn args
<Lisanna>
I don't know how primops work though
<sphalerite>
as a possible alternative, what about accepting `{ args }: args.foo or ""` and passing --arg args '{foo = 5;}'?
<Lisanna>
...oh
<Lisanna>
that's actually not a half-bad solution
<Lisanna>
yeah that's actually a really good idea!
<sphalerite>
it's not as pretty though
<Lisanna>
sure, but it'll work
<Lisanna>
as opposed to what I had before which is nothing
<sphalerite>
:)
<ij>
How much time does a rebuild take on your laptops? Mine was about five minutes just now. (I was a swappin', but just a little.)
<Lisanna>
ij that's not really possible to answer :p each rebuild is going to be different depending on what you changed since your last build
<ij>
I added a single package(xxd) and I hadn't/haven't updated the channels.
<Lisanna>
e.g. I deploy with nixops and it takes about a minute for trivial changes
<Lisanna>
but if I advanced nixpkgs then it takes a lot longer
<Lisanna>
and, even then it's not on my laptop, it's with remote building
<ij>
Could the large number of store paths slow it down considerably?
<Lisanna>
ij did it pull the package from a binary cache or did it compile from src?
<Lisanna>
not really, unless you're running out of disk space
roberth has joined #nixos
scribbler has joined #nixos
<sphalerite>
ij: do you have multiple instances of nixpkgs getting evaluated? Any containers?
scribbler1 has joined #nixos
<sphalerite>
ij: and how much RAM do you have? nix isn't exactly lightweight when it comes to resource consumption of evaluation, and nixos is a lot to evaluate :/
<ij>
No containers and it should be just one. I've 4 gigs, but didn't eat all of it up. But maybe five minutes isn't that much for a i3 laptop?
<sphalerite>
well you said it was swapping
<ij>
Okay, I'll kill the browser and try it again.
<sphalerite>
how long does `time nix-instantiate '<nixpkgs/nixos>' -A config.system.build.toplevel` take?
scribbler has quit [Ping timeout: 240 seconds]
scribbler1 is now known as scribbler
<{^_^}>
[nixpkgs] @mpickering opened pull request #36902 → Build QGIS and dependencies on darwin → https://git.io/vxJFI
<ij>
Lisanna, No, I found the path in channels/nixos-17.09/store-paths.xz, so not built.
<{^_^}>
[nixpkgs] @svanderburg pushed to release-17.09 « remarkjs: regenerate with node2nix 1.5.3 »: https://git.io/vxJFm
<{^_^}>
[nixpkgs] @zimbatm pushed commit from @volth to master « treewide: s/xlibs/xorg/g (#36889) »: https://git.io/vxJb8
tommyangelo has joined #nixos
<tommyangelo>
so how does one get rid of stuff left around by various nix-shell calls ? My nix folder is 122G, and I have run `nix-env --delete-generations old` and `nix-collect-garbage` and `nix-store --gc --print-dead`
<clever>
tommyangelo: `nix-collect-garbage` should have gotten rid of it
<clever>
tommyangelo: can you start with this command and tell me the output? `du -h --max=1 /nix/store | sort -h`
<tommyangelo>
sure, running it
<sphalerite>
ij: that is very slow…
<sphalerite>
ij: what are your specs? Can you share your full config?
<sphalerite>
and are you using nix 2.0? I think performance worsened significantly from 1.11 to 2.0
<ij>
Yeah, 2.0.
<mpickering>
sphalerite: I updated the pr, thanks for spotting that
<sphalerite>
Hm, seems to be a mass-rebuild on linux. Not sure that's appropriate for borg
<mpickering>
but building one attribute won't trigger a rebuild?
scribbler has quit [Remote host closed the connection]
<mpickering>
it will just rebuild the dependencies, ie the stuff I have changed
roberth has quit [Quit: Leaving.]
scribbler has joined #nixos
pkill9 has joined #nixos
<mpickering>
the mass rebuild is because I had to modify some qt packages
<mpickering>
not because of the qgis package
<sphalerite>
qt is a lot though, idk
<sphalerite>
gchristensen: ?
<tommyangelo>
clever: you want the whole output?
<mpickering>
right but what I'm saying is that if you trigger qgis to build. It will only build the dependencies of qgis rather than rebuild every package which depends on qt?
<gchristensen>
Go for it. If it doesn't work out it doesn't work out
<sphalerite>
mpickering: yeah it's not the other stuff, it's just qt I'm concerned about. But there, that's it triggered :)
<mpickering>
ah I see
<mpickering>
well now they won't be totally broken on darwin and it won't need to happen in future
<sphalerite>
gchristensen: whoa, only just saw that we have nix.ci for borg. I'm impressed! (at the domain name)
<clever>
tommyangelo: `nix-store --query --roots /nix/store/12725jwvrfjqkmpd58h9wnvpwl1zmscg-ghc-8.2.2` will tell you why a given ghc is still around, and `nix-store --delete /nix/store/12725jwvrfjqkmpd58h9wnvpwl1zmscg-ghc-8.2.2` will delete it (dont force it)
<clever>
tommyangelo: using query, delete, and du, you can figure out why the large paths are still around, and delete them
peterzky has left #nixos ["WeeChat 2.0"]
<Lisanna>
__Sander__ is there any advantage to using DisnixOS's NixOps integration if you are only using nixops to deploy to physical NixOS machines?
vidbina has quit [Ping timeout: 260 seconds]
<__Sander__>
Lisanna: not much, but it's probably slightly more convenient than a "ordinary" SSH connection because it allows you to reuse NixOps' settings
<Lisanna>
__Sander__ OK, and I guess NixOps has its own useful tools like nixops check etc.
<__Sander__>
yes
<__Sander__>
Disnix does not do any "machine management"
ixxie has joined #nixos
zzamboni has quit [Quit: Leaving.]
Itkovian has quit [Read error: Connection reset by peer]
Itkovian has joined #nixos
ottidmes has joined #nixos
Itkovian has quit [Remote host closed the connection]
<Lisanna>
__Sander__ and the services that disnixos deploys are the same kinds of services that regular disnix uses? i.e., it's just a program that disnix executes on the target host?
Itkovian has joined #nixos
zzamboni has joined #nixos
periklis has joined #nixos
<tommyangelo>
clever: cool, I'll write that down somewhere
<__Sander__>
Lisanna: yes
<Lisanna>
I'm thinking it might be cool to write disnixos services as imperative nixos containers
<__Sander__>
basically the main reason why DisnixOS is convenient is that NixOps/networked NixOS configurations are transparently converted into infrastructure expressions
<__Sander__>
but apart from that, it's the same stuff
<Lisanna>
yeah, I saw that part
<__Sander__>
just a thin wrapper around the core toolset
blankhart has quit [Ping timeout: 240 seconds]
xcmw has joined #nixos
ashgillman has joined #nixos
asuryawanshi has quit [Read error: Connection reset by peer]
asuryawa_ has joined #nixos
chisui has joined #nixos
scribbler has quit [Quit: scribbler]
spear2 has quit [Remote host closed the connection]
logzet has joined #nixos
<Izorkin>
how to change in binary file path from /nix/store/...-maxscale-2.1.14//nix/store/..-maxscale-2.1.14/lib/maxscale to /nix/store/..-maxscale-2.1.14/lib/
xcmw has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
xcmw has joined #nixos
<Lisanna>
Izorkin huh?
<Dezgeg>
you can't safely do a such change in binary files, but probably you can debug instead why the build system is generating such paths
ThatDocsLady has quit [Remote host closed the connection]
<yorick>
how do I tell nixos that it should restart my service when /etc files change?
<elvishjerricco>
yorick: I assume the etc files you want to have trigger the restart are defined in Nix?
<yorick>
elvishjerricco: yeah, they're in environment.etc."file/name.cfg".text = 'stuff'
<elvishjerricco>
There is `systemd.servicessystemd.services.myservices.restartTriggers = [config.environment.etc."file/name.cfg"];` would probably work.<name>.restartTriggers`. You can add the list of things you want to trigger a restart there. `
<yorick>
elvishjerricco: thanks!
<elvishjerricco>
Whoa my message got reordered somehow
<elvishjerricco>
There is `systemd.servicessystemd.services.myservice.restartTriggers = [config.environment.etc."file/name.cfg"];` would probably work..<name>.restartTriggers`. You can add the list of things you want to trigger a restart there. `
<elvishjerricco>
Huh. Same reordering... Something must be up with my client
<{^_^}>
→ 0fe236db by @FRidh: dtc: fetch over https
<{^_^}>
→ 7e3d1811 by @FRidh: netpbm: fetch over https
<{^_^}>
→ d9d21b67 by @FRidh: diffoscope: fetch over https
Neo-- has joined #nixos
<fadenb>
I am installing a new system and just got this message: warning: in configuration file '/etc/nix/nix.conf': unknown setting 'signed-binary-caches'
<fadenb>
Is this something one should worry about? Can not remember ever having seen that before
orivej has joined #nixos
jtojnar has quit [Quit: jtojnar]
jtojnar has joined #nixos
blankhart has joined #nixos
<barrucadu>
Is there a way to set an environment variable from the command-line for nix-build? I have a project which runs tests against an externally hosted API, so I'd like to be able to do something like 'API_TOKEN=... nix-build'
jtojnar has quit [Client Quit]
<barrucadu>
I guess I could just disable the tests when building and use nix-shell for running them
<{^_^}>
→ c476d2f5 by @teto: grub-install.pl: more details on blkid failure
<{^_^}>
→ 7b7860ff by @teto: build-vm: remove trailing dot from script path
John882 has joined #nixos
<tobiasBora>
Hello,
<tobiasBora>
I'd like to know, I don't remember why we need to recompile all the packages when the install dir change (~/nix instead of /nix for example)... For example when I download a firefox tar.gz, I can uncompress and run it in any directory, so why /nix has not the same property?
<gchristensen>
fadenb: looks like a nix 1 -> nix 2 glitch
<viric>
what program can I use to lock the X screen?
<gchristensen>
viric: I use i3lock
ixxie has quit [Ping timeout: 240 seconds]
<gchristensen>
tobiasBora: this is because the binary firefox gives you expects your libraries and libc to be in /lib and /usr/lib etc
<viric>
gchristensen: looks good! Thank you!
<gchristensen>
tobiasBora: whereas Nix expects your libc to be in /nix/store/d54amiggq6bw23jw6mdsgamvs6v1g3bh-glibc-2.25-123/lib/libc.so.6 (for example)
reinzelmann has quit [Quit: Leaving]
<viric>
gchristensen: do you have it automatic after some timeout?
<gchristensen>
tobiasBora: and your libz in /nix/store/dl66jvc19im9lqslzqszsp2gif2y0y8i-zlib-1.2.11/lib/libz.so.1 for example
ixxie has joined #nixos
<tobiasBora>
gchristensen: ok thanks. And we can't use LD_LIBRARY_PATH to avoid this hardcoding of libraries?
alex_ has quit [Read error: Connection reset by peer]
<gchristensen>
tobiasBora: not that we can't, we don't want to! the primary promise and guarantee of Nix is that its builds are pure and always use exactly what was provided, _not_ whatever the system happens to contain.
<tobiasBora>
gchristensen: well I mean, LD_LIBRARY_PATH could be set to point to nix library no?
<viric>
gchristensen: does the PR show that it works?
<gchristensen>
tobiasBora: sure, but then it will look at /nix/store/...
<LnL>
tobiasBora: libraries are only one example of relocatability problems
<gchristensen>
viric: well I'm just saying my setup doesn't work properly, and I need to look in to how to make it work properly, and that PR is probably helpful :)
<viric>
gchristensen: ah! :)
reinzelmann has joined #nixos
<tobiasBora>
gchristensen: well it would allow us to point to ~/nix/store if we move the base folder no, without recompiling everything no?
<viric>
gchristensen: but that's only a test
<gchristensen>
:)
reinzelmann has quit [Remote host closed the connection]
<tobiasBora>
LnL: do you have other examples of failure?
<LnL>
hardcoded paths like config files, etc. in scripts or even binaries
<LnL>
what you're saying is possible in theory, but in practice it's not really feasible
<LnL>
it's already hard enough to get build systems to ignore /usr/local :)
<tobiasBora>
LnL: ok thanks. It's too bad because it's a pain to recompile everything just because I don't have access to /nix... And proot is nice, but a bit slow
<Biappi>
gchristensen: congratulations!!!! 🎉 y'all deserve it =D
<gchristensen>
;)
<LnL>
tobiasBora: if your kernel is recent enough you can use mount namespaces
<tobiasBora>
LnL: don't you need to be root to do that ?
<LnL>
no, and nix 2.0 has native support for that with --store local?root=/tmp/foo and nix run
<{^_^}>
→ 34f12184 by @volth: treewide: s/saneBackends/sane-backends/g
<{^_^}>
→ 9f4aaf11 by @7c6f434c: Merge pull request #36891 from volth/aliases-2
fendor has quit [Ping timeout: 240 seconds]
asuryawanshi has joined #nixos
rauno has quit [Ping timeout: 246 seconds]
zzamboni has joined #nixos
asuryawanshi has quit [Ping timeout: 246 seconds]
<tobiasBora>
LnL: store cannot work with symlink ?
<sphalerit>
tobiasBora: during nix's installation? Or nixos?
blym has joined #nixos
<LnL>
tobiasBora: it's discouraged but you can set a magic environment variable
pie__ has joined #nixos
<tobiasBora>
sphalerit: during nix install
<clever>
bgamari-: you can limit what packages obey the modified glibc, but the entire closure of you program would still have to be rebuilt
<tobiasBora>
LnL: Hum... Let's try first without symlinks
<tobiasBora>
LnL: I tried this: ix-env -i isync --store /tmp/my-nix
<tobiasBora>
nix-env*
<tobiasBora>
but if failed after the downloading with "error: while setting up the build environment: cannot change root directory to '/tmp/my-nix/nix/store/riwgxba95sibkgpvv84mn278h5naa0i7-user-environment.drv.chroot': No such file or directory
<tobiasBora>
"
<bgamari->
clever, mm
John882 has quit [Ping timeout: 252 seconds]
<sphalerit>
LnL: isn't that for setting a different store path and results in rebuilding the world?
<sphalerit>
LnL: as opposed to a chroot store which still uses /nix/store but within something else
<bgamari->
pbogdan, indeed, I also stumbled upon that and it seems to do something useful
<bgamari->
it will likely be sufficient for my purposes
ericsagnes has quit [Ping timeout: 246 seconds]
rauno has joined #nixos
<instantepiphany>
Hey guys, I am having trouble setting up a pinned version of nixpkgs for a project I am starting. I can successfully pull in packages from a certain commit on nixpkgs on GitHub, but inside nix-shell, nix-env doesn't seem to be using this pin. What could I be doing wrong?
<instantepiphany>
Whats the preferred paste site here? I will paste my default.nix
<instantepiphany>
Line 15 & 18 are my current attempts to get it to work, with little success.
<clever>
instantepiphany: nix-env will only obey that if you do something like `nix-env -f nixpkgs.nix -iA hello`
<tilpner>
How would I let a module user configure the default value of options?
<clacke[m]>
nix-enf -f yourfile.nix -i asdf
<instantepiphany>
clacke[m] I will have a look
<tilpner>
(Other than making them do it manually as myDefaults // { /* changes */ }
<instantepiphany>
clever: should nix-env inside of nix-shell respect `-I /path/to/latestPkgs/output`?
<tilpner>
)
<clever>
instantepiphany: yes and no, if you `nix-env -f '<nixpkgs>' then it will search for a `nixpkgs` inside the `output` directory
<clever>
instantepiphany: but if you `nix-env -iA nixos.hello` it will check the nixos channel on the host system
<clacke[m]>
instantepiphany: ah, looked at your paste. With that file just install with nix-install -f yourfile.nix
<clacke[m]>
sorry nix-env
<tilpner>
clacke[m] - Oh, a lot happened to racket2nix since I last checked. Nice! :)
ericsagnes has joined #nixos
<instantepiphany>
anyway to use a path to set that to be the default inside the nix-shell? I thought thats what the `-I` flag or `NIX_PATH` envvar were for
loonquawl has joined #nixos
xcmw has joined #nixos
<instantepiphany>
clever: clacke[m] its not working :(
<instantepiphany>
What output will help debug this?
<clacke[m]>
tilpner: I'm pretty happy with how testing is set up now. calling nix-generated nix expressions is fun and can break your brain :-)
<clacke[m]>
instantepiphany: inside nix-shell or not is not the point
Sonarpulse has joined #nixos
<instantepiphany>
clacke[m] sorry I don't understand what you mean.
<clever>
instantepiphany: also, you shouldnt really be using nix-env inside nix-shell, its installing things into your user, that will persist even after you leave the shell
<instantepiphany>
clever: thanks, I didn't know nix-env inside nix-shell would effect user environment
<clacke[m]>
nix-env looks at NIX_PATH and can be overridden with -I. That's it. But the default.nix you showed doesn't even have attributes, so it's not useful for the nix path.
<clacke[m]>
inside nix-shell is a red herring
<clacke[m]>
nix-env always works the same
<{^_^}>
[nixpkgs] @fpletz pushed to master « samba: 4.7.5 -> 4.7.6 for multiple CVEs »: https://git.io/vxURw
<clacke[m]>
nix-shell just sets up environment variables for activating certain packages
<instantepiphany>
Ok - my end goal is simply to have a nix-shell open with a predetermined set of build inputs - I have done this many times. However, when discovering which dependencies I need, I would much rather search using nix-env instead of using GitHub search. For some reason I can't get nix-env to use the output from that .nix file - it has the same structur
<instantepiphany>
e as the other nixpkgs paths in my ENV vars and I am passing it as an argument to the -I flag for nix-env.
<{^_^}>
[nixpkgs] @fpletz pushed 15 commits to release-18.03: https://git.io/vxURM
<{^_^}>
→ a14d18e3 by @primeos: androidStudioPackages.beta: 3.1.0.12 -> 3.1.0.14
<{^_^}>
→ 2271731c by @primeos: androidStudioPackages.{dev,canary}: init at 3.2.0.5
<{^_^}>
→ 75f8209e by @magnetophon: i3: 4.14.1 -> 4.15
<instantepiphany>
You are both right, I misunderstood nix-shell and thought it offered sandboxing as well. So I just need to get nix-env to let me search using a specific folder containing nixpkgs.
<clacke[m]>
if you are in nix-shell and want to activate more packages, just run nix-shell one level more
<symphorien>
instantepiphany: "would much rather search using nix-env instead of using GitHub search." << do you know this page ? https://nixos.org/nixos/packages.html
<symphorien>
the attribute name column is the one you want
<symphorien>
and btw adding -P to nix-env makes it output the atribute name ass well iirc
<instantepiphany>
Yeah I am aware of that page - but remember i am pinning a specific commit (to get a certain version of packages to keep dev environment stable). I assume that page lists packages from the latest successful Hydra build, is that right?
<symphorien>
no the list is for the last stable release of nixos
<symphorien>
currently 17.09
<clacke[m]>
if you want to make a nix file for nix-env, let it return a set of derivations, not a derivation
<clacke[m]>
the names don't change around all that much
ericsagnes has quit [Ping timeout: 264 seconds]
<instantepiphany>
Sure - but I might want to pin an older version, and use packages no longer present in 17.09. Thanks to this channels help, my question has changed and is now simpler - I am having trouble getting nix-env to respect the `-I` flag which should accept a path to particular version of nixpkgs - what am I doing wrong? Or how can I get nix-env to print w
<instantepiphany>
hat nixpkgs version it is currently accessing?
lopsided98 has quit [Ping timeout: 245 seconds]
<clacke[m]>
tilpner: what I'm struggling with in racket2nix currently is the circular dependencies and how to break them right
<ottidmes>
instantepiphany: If you want to know how nixpkgs is evaluated, wouldn't this do: nix-instantiate --eval --expr '<nixpkgs>'
s33se has quit [Ping timeout: 240 seconds]
jensens has joined #nixos
lopsided98 has joined #nixos
<instantepiphany>
ottidmes so the output of that command is the folder that nix commands will get <nixpkgs> from, right?
zzamboni has quit [Quit: Leaving.]
<clever>
ottidmes: there is also nix-instantiate --find-file nixpkgs
xcmw has quit [Ping timeout: 264 seconds]
raynold has joined #nixos
<instantepiphany>
If running that command twice, the second time prefixed with NIX_PATH=nixpkgs=/path/to/my/custom/nixpkgs gives a different output, then that means it is finding a valid nixpkgs, right?
<clever>
instantepiphany: yeah
<clever>
instantepiphany: you can also append -I nixpkgs=whatever to it
<instantepiphany>
So now my question is, why do I need to pass the -f '<nixpkgs>' flag to nix-env. If I don't pass that flag, where is it getting the nixpkgs value from?
<clever>
behind the scenes, <nixpkgs> gets translated into builtins.findFile builtins.nixPath "nixpkgs"
<clever>
instantepiphany: ~/.nix-defexpr/
<instantepiphany>
Oh ok, I thought it used to read from NIX_PATH. Thats all of my questions answered, thanks everyone for your help!
<clever>
yeah, nix-env is one of the weirder tools in the toolbox
<clever>
it can obey NIX_PATH, but it wont look in it by default
<instantepiphany>
If you pass -f '<nixpkgs>' that causes it to lookup using findFile (nixPath "nixpkgs") right?
zzamboni has joined #nixos
<instantepiphany>
So I could potentially pin a package set, add it to NIX_PATH as "customNixpkgs=/path/to/custom/nixpkgs" and then use nix-env -f '<customNixpkgs>'?
ericsagnes has joined #nixos
<clever>
instantepiphany: yep
<clever>
instantepiphany: but nix-env -i still acts on ~/.nix-profile/ and that will persist after you exit the shell
fendor has joined #nixos
<instantepiphany>
Yeah, thats a big thing i learned. nix-shell just sets up env vars to make building packages more convenient, it isn't a sandbox/vm at all. I will just use nix-env to find the packages I want from my pinned version of nixpkgs, and then add those packages to my default.nix for the project, so they are only in my profile when I run nix-shell in that
<instantepiphany>
folder. I guess I still have to be careful about programs such as editors that may store configs on the filesystem.
<clever>
instantepiphany: one thing you could do is mess with $HOME
<clever>
then ~/.nix-profile points somewhere new!
Synthetica has quit [Quit: Connection closed for inactivity]
Synthetica has joined #nixos
zzamboni has quit [Ping timeout: 264 seconds]
<instantepiphany>
Good point
<instantepiphany>
`export HOME=/path/to/project/.state` or something
<clever>
yep
zzamboni has joined #nixos
ma27 has joined #nixos
zzamboni has quit [Ping timeout: 240 seconds]
jensens has quit [Ping timeout: 246 seconds]
<fendor>
is the program KeePassXC packaged in nixos?
zzamboni has joined #nixos
<sphalerite>
fendor: looks like it's keepassx-reboot
<fendor>
sphalerite, how do you know?
<hyper_ch2>
hmmm, pass bash- path completion doesn't work
<hyper_ch2>
hmmmm
moriarty has joined #nixos
<hyper_ch2>
I mean tab-completion for pass
moriarty has left #nixos [#nixos]
<sphalerite>
fendor: there's a keepassx-reboot attr in nixpkgs, which points to github.com/keepassx-reboot/keepassxc which seems to be the repository for keepassxc :)
<fendor>
sphalerite, so i'd have to grep in the nixpkgs of nixos?
<Synthetica>
fendor: sphalerite: I use that, it works fine
<fendor>
to find it out by myself
<sphalerite>
fendor: well I used `nix search keepass` to find it in the first place
<Synthetica>
Or `nox keepass`
zzamboni has quit [Ping timeout: 260 seconds]
<fendor>
Synthetica, what is nox`
loonquawl has quit [Quit: loonquawl]
<fendor>
sphalerite, same, but didnt help me to identify what i am looking for
zzamboni has joined #nixos
<Synthetica>
I quote: "Nox is a small tool that makes the use of the Nix package manager easier."
<sphalerite>
fendor: feel free to submit a PR adding a `keepassxc` alias for it!
blankhart has quit [Read error: Connection reset by peer]
<fendor>
Synthetica, will try it out, thanks!
<sphalerite>
nox, at least for search, is mostly superseded by `nix search` though
<fendor>
sphalerite, there probably is, because you can install it via `nix-env -iA nixos.keepassxc`
<sphalerite>
the only thing it still has that nix search doesn't is support for multiple terms
<Synthetica>
PRs welcome?
blankhart has joined #nixos
<sphalerite>
fendor: oh yeah, indeed, in fact keepassx-reboot and keepassx-community are aliases of that!
<fendor>
then why didnt i find it via `nix search keepass`?
<sphalerite>
weird that keepassxc doesn't show up in the search resuts
<sphalerite>
yeah
<fendor>
nox shows it correctly ^^
<Synthetica>
fendor:
<Synthetica>
It does for me?
jtojnar has quit [Quit: jtojnar]
<fendor>
Synthetica, nix search keepass shows keepassxc with that name?
<sphalerite>
aaah, it's not in 17.09
<Synthetica>
... Oh
zzamboni has quit [Remote host closed the connection]
coot_ has joined #nixos
<Synthetica>
Yeah, I'm on unstable
<fendor>
me, too 0.o
jtojnar has joined #nixos
<fendor>
oh, nix only searches in stable?
<sphalerite>
maybe your nix search cache is stale, what about nix search -u keepassxc?
rauno has quit [Remote host closed the connection]
<Synthetica>
fendor:
<Synthetica>
Attribute name: nixpkgs.keepassxc
<Synthetica>
Package name: keepassxc
<sphalerite>
no, it searches in all the NIX_PATH entries I think. But it doesn't autmoatically update the cache
<fendor>
it is stale, but afterwards it still can't find it
zzamboni has joined #nixos
instantepiphany has quit [Remote host closed the connection]
zzamboni has quit [Ping timeout: 240 seconds]
zzamboni has joined #nixos
FareTower has joined #nixos
<hyper_ch2>
fpletz: hmmm, does pass bash-completion work for you?
<fendor>
another question, how can i remove all previous build generations and only keep the most recent one?
humanoyd has quit [Quit: WeeChat 2.0.1]
<fpletz>
hyper_ch2: yeah, works (18.03)
<fpletz>
hyper_ch2: do you have pass installed via nix-env?
zzamboni has quit [Ping timeout: 256 seconds]
spietz has joined #nixos
<hyper_ch2>
no, via nixos
<yorick>
elvishjerricco: ok, how do I make it update on a path? I have restartTriggers = [ ./thing.cfg ]; now, but that doesn't update it when thing.cfg changes
<hyper_ch2>
fpletz: pass show Web/[tab] doesn't display any entries
asuryawanshi has joined #nixos
zzamboni has joined #nixos
<sphalerite>
fendor: for nixos or your user profile? Or for both?
<sphalerite>
fendor: for your user profile: nix-env --delete-generations old; for your system: sudo nix-env -p /nix/var/nix/profiles/system --delete-generations old
<fendor>
sphalerite, both, just clean up some garbage, nixos currently takes up 70 gb while only firefox, gnome (also its stack), git and fish is installed
<sphalerite>
fendor: or you can use nix-collect-garbage -d, which also does a GC afterwards
hakujin has joined #nixos
<sphalerite>
what o.O that is a lot of space
<sphalerite>
how old is this installation?
<fendor>
a few months
<sphalerite>
ah fair enough, if you've never GC'd
<sphalerite>
you can also use --delete-older-than
<fendor>
sphalerite, i do regurarly, but only user space
<fendor>
now 7gb have been freed
<fendor>
*2gb
<sphalerite>
did you run it as root?
<sphalerite>
Unprivileged users can't touch the system profile
<sphalerite>
fendor: maybe you have some result symlinks lying around which are keeping large paths alive as well
<fendor>
weird, in my .local/trash, there are 13gb...
<fpletz>
hyper_ch2: yeah, with programs.bash.enableCompletion and pass in environment.systemPackages
<sphalerite>
fendor: try ls -l /nix/var/nix/gcroots/auto
<sphalerite>
oh, you mean it's not actually nixos using all that space
zzamboni has joined #nixos
<viric>
gchristensen: xautolock works great.
<fendor>
sphalerite, i think so
<sphalerite>
well then you need to delete files rather than just putting them in the trash in nautilus :) (or empty your trash)
<fendor>
i cant remember why there might be 13 gb in the trash.
<fendor>
oh, i forgot steam :D
<fendor>
hehehe, not nixos's fault
alexteves has joined #nixos
<fpletz>
viric: as for actual lockers, I can also recommend xtrlock-pam
<viric>
fpletz: oh, why?
<hyper_ch2>
fpletz: I^'ll try that
<fpletz>
xtrlock-pam also has a mode that won't hide your applications, that's great for playing movies :)
<hyper_ch2>
fpletz: what do you have in environment.systemPackages? programs can't be found
<viric>
fpletz: only keyboard/mouse?
<viric>
but movie players deactivate dpms and those things... won't them deactivate xauthlock?
<fpletz>
hyper_ch2: pass is in environment.systemPackages, and programs.bash.enableCompletion = true in your nixos configuration
<viric>
I usually run vlock as root, locking all terminals
<hyper_ch2>
ah, I read wrong
<fpletz>
viric: yeah, they will, I then lock via a shortcut… it displays a lock as your mouse cursor and blocks them from interacting with the x11 applications
<fpletz>
viric: you can only input your password and press enter :)
zzamboni has quit [Ping timeout: 246 seconds]
<sphalerite>
I use i3lock with a fancy wrapper script that pixelates my screen
<viric>
fpletz: interesting. Good to know that there are options
<sphalerite>
I don't think it actually looks that good but people seem to be impressed by it x)
<viric>
I don't care at all about screen contents but they should be hidden
<viric>
solid color is perfect
<fpletz>
yeah, that's also my default
zzamboni has joined #nixos
<hyper_ch2>
fpletz: it's recompiling stuff because sqlite patch :)
hyper_ch2 has quit [Quit: Page closed]
ryanartecona has joined #nixos
<fendor>
sphalerite, the commnads removed all my user space applications :(
<sphalerite>
fendor: then they weren't installed correctly
zzamboni has quit [Ping timeout: 248 seconds]
<coconnor>
good morning my dudes. What's new?
<fendor>
sphalerite, all instaleld via nix-env -iA
<sphalerite>
fendor: what does nix-env -q say?
zzamboni has joined #nixos
<fendor>
shows my installed applications
<viric>
fpletz: the i3lock && systemctl suspend is hard to beat though.
<fpletz>
viric: oh, do you also know xss-lock? that runs your locker of choice if systemd/logind wants to lock your session
<fpletz>
so systemctl suspend will automatically lock your session :)
<sphalerite>
fendor: then how were they removed?
zzamboni has quit [Ping timeout: 252 seconds]
shikiryogy has quit [Quit: Leaving]
<fendor>
sphalerite, well, i reinstalled them immediately
amir has quit [Read error: Connection reset by peer]
<fendor>
sphalerite, happened already once, when executing the command sudo nix-env -p /nix/var/nix/profiles/system --delete-generations old in combination with nix-collect-garabge -d. didnt check if it happens when i invoke only the first commnad
<sphalerite>
my<squish>fi</squish>rstpublish
<sphalerite>
fendor: neither of those commands do that.
<gchristensen>
hah!
amir has quit [Read error: Connection reset by peer]
zzamboni has quit [Client Quit]
amir has joined #nixos
<fendor>
but it does
<fendor>
no i checked, both in combination delete all application, the former command alone does not
<sphalerite>
are you sure it didn't just delete the store paths for old versions of your applications?
<shlevy>
Switching to 18.03... Let's see if I can make it to May without switching back to master :D
<sphalerite>
it shouldn't have affected the contents of your profile at all.
<fendor>
sphalerite, idk how would i notice? i used unstable. whatever application i installed
<sphalerite>
what led you to believe that all your applications were gone?
<fendor>
because i cant invoke them anymore via command line
<fendor>
and nix-env -q does not output anything
<sphalerite>
/:|
<infinisil>
Alright I'm free now, need to look into this ZFS thing..
<rawtaz>
thanks sphalerite. so aside from letting you use later defined variables, it also introduces a local scope check before outside scope
<sphalerite>
yep pretty much
<tobiasBora>
By the way, I think to remember talked about ways to install/configure stuf from each user directory, more or less like in nixos. Do you know if it's now usable and/or official ? I can't find the seferences I use to have.
zzamboni has quit [Quit: Leaving.]
<elvishjerricco>
Oh wow. Libinput is much much better than synaptics
<sphalerite>
LnL: yeah but it's your overlay that's linked from the wiki page so you get all the credit :p
<infinisil>
elvishjerricco: One point against this is that a lot of modules are just duplicated in home-manager..
boomshroom has joined #nixos
jensens has joined #nixos
<elvishjerricco>
infinisil: Yea. I don't think we'd actually want actual home-manager rolled in. A minimal version of something much like it though
<sphalerite>
infinisil: because it's not really that great. No way to remove something from systemPackages for instance. Confusing/hard-to-debug cases of infinite recursion. Lack of actual modularity
<boomshroom>
Hello!
<FareTower>
from I use nix-copy-closure --from foo, why does the user on foo need to be privileged?
<FareTower>
I understood for --to foo, but not --from foo
<{^_^}>
[nixpkgs] @efx opened pull request #36913 → timewarrior: enable for darwin → https://git.io/vxU10
<sphalerite>
infinisil: and all the standard modules get loaded when evaluating systems which typically only use 5% of them. It's rather expensive to evaluate
<infinisil>
sphalerite: That can be solved, I'm pretty sure
<elvishjerricco>
sphalerite: I actually find it to be exceptionally modular. The inability to remove stuff is a little unfortunate but I think it's probably for the best. Would be very hard to get predictable behavior if that were the norm.
<sphalerite>
ref ij's issues earlier today
<infinisil>
Yeah I think it's also really modular, not sure what you mean by its not
<elvishjerricco>
FareTower: The user just needs to be trusted by nix. Escalating to root is an easy way to do that :P
zzamboni has quit [Client Quit]
<{^_^}>
[nixpkgs] @limeytexan opened pull request #36914 → perlPackages.FileTail: init at 1.3 → https://git.io/vxU1V
zzamboni has joined #nixos
Mateon3 has joined #nixos
* sphalerite
calls for expert on shortcomings of the module systems: shlevy shlevy shlevy !
<elvishjerricco>
FareTower: You can add the user you want to trusted-users in /etc/nix/nix.conf
Mateon1 has quit [Ping timeout: 240 seconds]
Mateon3 is now known as Mateon1
<infinisil>
sphalerite: Well ya can't be throwing arguments if you don't know the reasoning behind them!
<sphalerite>
infinisil: it's mostly that I'm leaving just now so I can't go into detail :p
<sphalerite>
ttyl
<infinisil>
I have my system config designed around modules, super modular
<infinisil>
See ya!
<boomshroom>
I'm trying to install a rust package that depends on the crate openssl-sys which has a whole bunch of dependencies. Is there a way to specify the dependencies for that one crate and have them get added to any crate that depends on it?
winem_ has quit [Ping timeout: 256 seconds]
zzamboni has quit [Ping timeout: 246 seconds]
<infinisil>
boomshroom: Isn't that how dependencies work?
fendor has quit [Ping timeout: 246 seconds]
xcmw has joined #nixos
knupfer has quit [Remote host closed the connection]
knupfer has joined #nixos
<boomshroom>
infinisil: Currently, to build using rust, you specify the source directory and the hash of the crate dependencies. The problem is every time you try to use a given crate, it's just as one of the vendored crates. openssl-sys doesn't have a lot of rust dependencies, it has a lot of native dependencies, all of which need to be in the path when building the final binary.
<sphalerit>
infinisil: the biggest drawback is that 90% of these "modules" depend on options declared and implemented by other modules
<sphalerit>
But these dependencies are completely implicit and undeclared
<{^_^}>
[nixpkgs] @limeytexan opened pull request #36915 → perlPackages.FilePid: init at 1.01 → https://git.io/vxUDO
dan_b has quit [Ping timeout: 264 seconds]
<roconnor>
for nix run, how do I specify an attribute of the local default.nix as my installable?
<elvishjerricco>
sphalerit: Although that seems inelegant, it's never been a problem for me in practice
<elvishjerricco>
roconnor: Can't you do `-f .`?
<roconnor>
that does work.
<roconnor>
Seems nix run doesn't set IN_NIX_SHELL
<roconnor>
and haskell's .env doesn't seem to be compatible with nix run :D
blankhart has quit [Ping timeout: 264 seconds]
<roconnor>
back to nix-shell for the time being.
<elvishjerricco>
roconnor: nix run is different than nix-shell
<roconnor>
I understand that nix run is slated to be a replacement for nix-shell.
<elvishjerricco>
nix run foo adds foo to the shell. nix-shell -A foo gives you a shell with all of foo's dependencies
<elvishjerricco>
so nix-shell -A haskellPackages.foo.env gives you an environment where you could build foo yourself. But nix run haskellPackages.foo.env would try to actually build the env derivation and put you in a shell with the build result
<roconnor>
I see your point.
srdqty has joined #nixos
<elvishjerricco>
I won't actually consider it a replacement for nix-shell until it has this use case covered, as it is the primary use case to me.
<elvishjerricco>
It is a replacement for nix-shell -p though
sigmundv has quit [Ping timeout: 256 seconds]
spietz has quit [Quit: WeeChat 1.9.1]
blankhart has joined #nixos
<tobiasBora>
sphalerit LnL : Interesting, I'll give it a try!
<{^_^}>
[nix] @edolstra pushed to master « build-remote: Don't substitute the build result »: https://git.io/vxUSg
asuryawanshi has quit [Remote host closed the connection]
fendor has joined #nixos
<FareTower>
elvishjerricco, thanks for the pointer. I still don't understand why I need to be in either trustedUsers or allowedUsers on the machine I'm copying packages *from*.
asuryawanshi has joined #nixos
<elvishjerricco>
FareTower: Oh. That is odd. Not sure
sigmundv has joined #nixos
boomshroom is now known as boomshroom-away
<{^_^}>
[nixpkgs] @jtojnar pushed to master « maintainers/scripts/update.nix: use saner nubbing »: https://git.io/vxU9o
<{^_^}>
[nixpkgs] @bjornfor pushed commit from @aszlig to release-17.09 « setup-hooks: Add autoPatchelfHook »: https://git.io/vxU9b
civodul has quit [Quit: ERC (IRC client for Emacs 25.3.1)]
magnetophon has quit [Read error: Connection reset by peer]
blym has quit [Read error: Connection reset by peer]
Neo--- has joined #nixos
JosW has quit [Quit: Konversation terminated!]
tommyangelo has quit [Quit: Leaving]
Neo-- has quit [Ping timeout: 240 seconds]
Neo--- has quit [Remote host closed the connection]
Neo--- has joined #nixos
<FareTower>
I tried to play with nix.conf (via settings in the configuration.nix), and now I get this error trying to nixos-rebuild --show-trace switch : while evaluating the attribute ‘args’ of the derivation ‘coreutils-8.28’ at /nix/store/4m8xpaigxvpdq88lrgzaxd5v7ydypn4j-nixos-17.09.3037.883a4ebe7ee/nixos/pkgs/stdenv/generic/make-derivation.nix:98:11: cannot start daemon worker: writing to file: Broken pipe
<{^_^}>
[nixpkgs] @yegortimoshenko opened pull request #36917 → Drop Pale Moon → https://git.io/vxU5Q
<FareTower>
or this: cannot start daemon worker: writing to file: Broken pipe
<FareTower>
and rebooting doesn't fix it, so it's not just a daemon being stuck :-(
<dtz>
o_O just curious what nix.conf changes did you make?
<clever>
FareTower: simple answer: pick an older generation from the bootloader when rebooting, that will undo all changes to nix.conf
<dtz>
(did you successfully rollback or otherwise get into a working state? ask if you need help)
<dtz>
(although I'm very interested in understanding what broke xD)
<rawtaz>
im trying to run Falkon on NixOS stable. it runs, but there's icons missing in the theme. according to someone i have to set the environment as explained in https://community.kde.org/Get_Involved/development#Test_your_patch , but those paths will of course match NixOS. how can i identify the proper paths for those environment variables?
<{^_^}>
→ 134af4c5 by @fpletz: Merge pull request #36659 from Ma27/fix-freeorion
<ottidmes>
rawtaz: This is the module I am using for my theme/icons for GTK and Qt, not sure if that is actually your question, but here you go: https://pastebin.com/5P2a66d6
<{^_^}>
[nixpkgs] @fpletz pushed commit from @Ma27 to release-18.03 « freeorion: fix build »: https://git.io/vxUbE
xcmw has joined #nixos
Itkovian has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<mjacob>
clever: thanks!
<mjacob>
different question: imagine i'm working on, say, 3 projects with their own nix environment defined in a default.nix. how can i ensure that a garbage collection won't delete packages i'll need next time i work on one of these projects?
<rawtaz>
ottidmes: looks relevant, ill give it a shot. right now trying stable falkon instead of unstable.
<clever>
mjacob: simplest answer, just dont run garbage collection
<infinisil>
sphalerite: I'm actually pretty certain these dependencies are inspectable
<infinisil>
(Talking about module system dependencies)
jensens has quit [Remote host closed the connection]
<ottidmes>
infinisil: You can at least inspect where a value was defined, I am using that to define a default name based on the file location
<mjacob>
clever: the problem is that because of the way nix works, an openssl security update will force a rebuild of half of the packages and that can add up quickly on a small-ish ssd
jensens has joined #nixos
<goibhniu>
mjacob: the less simple answer is to add garbage-collection roots for them
<clever>
though you will also need to update those roots any time something like openssl changes
<clever>
or it will keep the old version around, not the new one
<ottidmes>
clever: But couldn't you add that to an activationScript if you want to keep it up to date with the rest of your system (never tried, I do not really use nix-env to begin with)
<clever>
ottidmes: an activationScript is the worst place to do that
<clever>
ottidmes: first, the machine will refuse to finish booting until it has fully build everything
<elvishjerricco>
mjacob GC roots are pretty simple. Unfortunately nix-shell seems to have lost the ability to add roots in nix 2.0. But `nix-build` or `nix build` will output symlinks that will act as GC roots by default
<clever>
ottidmes: second, this happens so early in bootup, that systemd isnt even running, so you have zero network access
<rawtaz>
question: when installing open-vm-tools-headless using environment.systemPackages, how does the "-headless" addition in the package alias affect the withX flag in the .nix expression for this package? like, what makes withX defined in there, so to speak?
<John882>
hello guys what is the name of the python qt4 core package in nix packages?
<mjacob>
goibhniu, elvishjerricco: thank you for the pointers, i found some resources on the website
<ottidmes>
clever: I find it a pity that there is no, "activate on build" equivalent that only runs on rebuild, I was thinking about that when I mentioned activationScript, I suppose you could add something to activationScripts to check for this
<clever>
ottidmes: a one-shot systemd service is the closest thing to what you want
<{^_^}>
[nixpkgs] @dtzWill opened pull request #36919 → busybox: backport updates and proper sandbox shell configuration → https://git.io/vxUNF
<ottidmes>
clever: But how to configure that to only run on build? Or does it work simply by not having a wantedBy and defining those restartTrigger attribures?
boomshroom-away is now known as boomshroom
<clever>
ottidmes: restart triggers only work when you do a rebuild-switch, and it still needs a wantedBy
<clever>
ottidmes: a systemd timer might be better then, but now your starting to re-invent hydra
coconnor has quit [Quit: Lost terminal]
xcmw has quit [Ping timeout: 268 seconds]
<ottidmes>
clever: How does wanting to run something on (re)build leads to re-inventing Hydra?
<clever>
ottidmes: on (re)build of what?, the project, or nixos?
<boomshroom>
My laptop (running NixOS) has my desktop set as a remote builder with a deticated builder user on my desktop. Attempting to use the builder gives the cryptic message of `serialised integer 7161674624452356180 is too large for type 'j'`. Where is that number coming from?
<ottidmes>
clever: nixos
<clever>
ottidmes: ah, thats less re-inventing hydra, but you still need to build the closures of every project on every system update
<clever>
ottidmes: and then broken projects prevent you from updating the OS
<clever>
> new Buffer('6363612073696800','hex').toString('ascii')
<clever>
'cca sih\u0000'
<boomshroom>
The good news is that the large number apears to be consistent across invokations.
<clever>
boomshroom: what is the hostname of the build slave?
<ottidmes>
clever: just forget about what I want to run on rebuild of nixos, theoretically, how does one run a script outside of the nixpkgs environment after each rebuild, you mentioned one-shot services and timers, but I fail to see how to configure them to make it happen, what is the hook/trigger
<boomshroom>
clever: It's a raw ip address that I set an alias of "desktop" for in programs.ssh.extraConfig.
<ottidmes>
boomshroom: Did you configure a proper login shell for the SSH users used by the builder? (assuming you use SSH), I had the exact same error, which was caused by using nologin for the SSH user that connected on the remote machine
<clever>
ottidmes: a one-shot systemd service would run on bootup, and with the right restart triggers, on every rebuild-switch
<boomshroom>
ottidmes: That sounds like it's probably the issue.
<clever>
boomshroom: what happens if you try to manually run something like `ssh user@host nix-store --version`
xcmw has joined #nixos
notdaniel has joined #nixos
<boomshroom>
(After adding sudo and the id file) This account is currently not available.
saschaveres[m] has joined #nixos
<FareTower>
great. I broke nix, and I don't even know how.
<clever>
> new Buffer('6363612073696854','hex').toString('ascii')
<clever>
'cca sihT'
<clever>
boomshroom: floating point in nodejs ate the first few bytes
<boomshroom>
I have the builder set as systemUser = true.
<clever>
boomshroom: that number, is the string "This acc", backwards
<clever>
c2d ~ # nologin
<clever>
This account is currently not available.
<FareTower>
I tried manually changing the nix.conf, and I still have error: cannot start daemon worker: writing to file: Broken pipe
<clever>
boomshroom: nix is parsing the output of `nologin` as an int
coconnor has joined #nixos
<ottidmes>
boomshroom: Just set the shell of the user you SSH with to something other than nologin and the error should be fixed: useDefaultShell = true; for example
<clever>
boomshroom: ah, and that slao what you get when you manually ssh
<boomshroom>
It has a value of bash, but not a default. That should be good enough.
<clever>
messy!, bash sets it with a default priority, but its not the .default of the option type
<boomshroom>
rebuilding...
<ottidmes>
clever: About the rebuild trigger, I meant call something _only_ on rebuild, so not at start as well, for which I can think a few uses cases
FareTower has quit [Ping timeout: 246 seconds]
<{^_^}>
[nixpkgs] @LnL7 pushed to master « v8: only use -Wno-error=unused-lambda-capture with clang »: https://git.io/vxUxA
<boomshroom>
Now ssh works.
<{^_^}>
[nixpkgs] @LnL7 pushed to release-18.03 « v8: only use -Wno-error=unused-lambda-capture with clang »: https://git.io/vxUpI
<John882>
guys is there a pyqt4core in on nix?
<boomshroom>
Now I'm back to an old error where xargo can't find rust-src.
<John882>
no idea how is it called on nix search
ryanartecona has quit [Quit: ryanartecona]
<gchristensen>
John882: how about just pyqt4
<ottidmes>
boomshroom: That at least is more sensible than some weird number :P (Initially at least)
<John882>
gchristensen: yeah found it but still the app it's not working
<boomshroom>
I did sign all packages earlier, but I didn't set the key file in the configuration.
xcmw has joined #nixos
<ottidmes>
boomshroom: I did it the other way around, being surprised I still got the error even though I had specified the key, thinking it would do the sigining on-demand
<{^_^}>
[nixpkgs] @dtzWill pushed to master « metrics: fix "values" to be count not bytes, to match prev behavior »: https://git.io/vxUhk
<John882>
gchristensen: actually I am testing bitcoinarmory since I have found the package with nix and since it's broken and missing some dependencies i am trying to find them and update the pkg
<boomshroom>
I remember not knowing where to put to signature files.
<boomshroom>
I guess /etc/nix/signing.{sec,pub} is the standard?
spietz has joined #nixos
<{^_^}>
[nixpkgs] @dtzWill merged pull request #36913 → timewarrior: enable for darwin → https://git.io/vxU10
<rawtaz>
ottidmes: where do i get that fancy mkMerge you have in your config? i tried to use it but it tells me its an invalid var
<rawtaz>
undefined*
<ottidmes>
rawtaz: It is in lib
<rawtaz>
ah, so add lib at the top (which i guess is some kind of import)
fendor has quit [Ping timeout: 260 seconds]
<ottidmes>
rawtaz: so if you include lib in your argument list of your module: { config, lib, pkgs, ... }: you can then use lib.mkMerge, of if you use with lib; like I often do, you can just use mkMerge directly
<{^_^}>
→ 4b00e5c1 by Averell Dalton: teamviewer: 12.0.85001 -> 12.0.90041
<{^_^}>
→ 6e55d835 by @fpletz: Merge pull request #36673 from averelld/upgrade-teamviewer
vidbina has joined #nixos
<boomshroom>
I probably should have seen this coming, but my laptop tried to build its own /etc on my desktop.
<rawtaz>
well that's cute. "Module `/etc/nixos/configuration.nix' has an unsupported attribute `boot'. This is caused by assignments to the top-level attributes `config' or `options'.". why should assignments to config or options affect boot?
<{^_^}>
[nixpkgs] @fpletz merged pull request #36888 → xfce: do not use old names → https://git.io/vxJYk
<infinisil>
sphalerite: You were probably right, I can't find a way to get dependencies of modules out of the module system
<clever>
rawtaz: but if you lack both config and options, nix will automatically wrap the entire thing with config = { ... };
<shlevy>
sphalerite: What'd I do? :D
<clever>
rawtaz: that is a nixos module, which is what configuration.nix also is
<rawtaz>
clever: ah. so i should just move my stuff from within the config i added to outside it, and remove the config
<clever>
rawtaz: the simplest way to use it, is to add its path to the imports list
ryanartecona has joined #nixos
<clever>
rawtaz: then nixos will merge the 2 files for you
xcmw has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<{^_^}>
[nixpkgs] @fpletz pushed 5 commits to release-18.03: https://git.io/vxTfU
<{^_^}>
→ 038560d6 by @dtzWill: metrics: fix "values" to be count not bytes, to match prev behavior
<{^_^}>
→ 2a5231c4 by @efx: timewarrior: enable for darwin
<{^_^}>
→ 2d3e899b by @magnetophon: qutebrowser: 1.1.1 -> 1.2.0
<rawtaz>
clever: right. if i were to do that, where should i put such a file as in the paste, and how would i import it?
<clever>
rawtaz: imports = [ ./thatfile.nix ];
<boomshroom>
clever: All I did was generate a key pair for my laptop and added nix.extraOptions with the key to my laptop. It was mostly copy-paste from the commands on my desktop.
<clever>
rawtaz: and it would be in the same directory as configuration.nix
<rawtaz>
boomshroom: my sincerest apologies for clever's behavior ;)
xcmw has joined #nixos
<boomshroom>
allowedUsers are allowed to build derivations right? And systemUsers are by-default not?
Judson has joined #nixos
<Judson>
If I want to pin a revision/tag of nixpkgs for a nix-shell, the way to do that is to do a separate checkout, right? There's not a clever NIXPATH thing or something?
<Judson>
Or... wait, could I use a fetchFromGithub and use the result as pkgs?
<tilpner>
That'll get you hello from pinned recent master
<Judson>
tilpner, I'm looking to put this in a shell.nix file...
<clever>
Judson: check the shell.nix i linked
xcmw has quit [Client Quit]
<Orbstheorem[m]>
Hello, how can I install a package without adding it to top-level/all-packages.nix ? ie. I have a .nix file that should go in applications/.../mypackage/default.nix, but I want to install it without clonning all the tree, adding it and adding it in all-packages
<Judson>
clever, that looks kinda rad.
<boomshroom>
Orbstheorem[m]: What I've been doing is creating overlays.
<rawtaz>
ottidmes: thanks to your config paste there i found that just adding environment.pathsToLink = [ "/share" ]; to my config is all that was needed to make themes work (duh)
<Orbstheorem[m]>
Trying with build -E
Intensity has joined #nixos
<rawtaz>
i still get a segfault if i run the stable version of falkon though, but if i install the one from unstable/master it works
spietz has quit [Quit: WeeChat 1.9.1]
<hyper_ch>
fpletz: now it works :) awesome :)
Neo--- has quit [Ping timeout: 264 seconds]
xcmw has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<boomshroom>
It looks like the remote build user needs to be a trustedUser.
<boomshroom>
Distributed builds FTW! Now I just need to work on a fully distributed OS. Ideally one that can use nix or a nix compatible build system.
coot_ has joined #nixos
alanz has quit [Client Quit]
alanz has joined #nixos
jtojnar has quit [Read error: Connection reset by peer]
jtojnar has joined #nixos
coot_ has quit [Client Quit]
<boomshroom>
Why bother building on my tiny Macbook Air when I can have my gaming rig handle the dirty work? :D
<Orbstheorem[m]>
clever: nix-build -E ... worked \o/, now how can I install it 😛 (I'm trying `nix-env -E ... -i vsakm` where ... is the same line you pasted and vsakm is the name in mkDerivation
rardiol1 has joined #nixos
<boomshroom>
Orbstheorem[m]: nix-build should have made a result link. run nix-env -i result (or nix-env -i $(readlink result)) and it should install.
spietz has joined #nixos
<Orbstheorem[m]>
thanks
arximboldi has joined #nixos
<arximboldi>
hmmm, is there a way to specify in a configuration.nix some files that need to be copied to users /home when creating the account?
<arximboldi>
I guess I could directly specify the user home to be the result of some derivation but I am not sure an immutable home is a good idea for my use-case...
<symphorien>
on another distro that would be the content of /etc/skel, no idea if NixOS respects that
<Orbstheorem[m]>
:D
<arximboldi>
hmmm
<arximboldi>
so there is: security.pam.makeHomeDir.skelDirectory
<arximboldi>
but I want this for just one particular user...
MichaelRaskin has joined #nixos
<Myrl-saki>
*** abort because of serious configure-time warning from Cabal
<symphorien>
(or whatever package provides the tool you want to put in the daemon PATH)
<ghostyy>
that will append zfs to the path?
<ghostyy>
or replace the path with only the things in that list?
<clever>
ghostyy: nixos will append that list to the existing one, and add it to the PATH env for the service
<ghostyy>
ahh great thank you
<hyper_ch>
infinisil: *thumbs up*
* hyper_ch
gives infinisil chocolated coated super cripsy cookies
zzamboni has quit [Quit: Leaving.]
<gchristensen>
oddly specific :)
<infinisil>
hyper_ch: You mean the comment?
<{^_^}>
[nixpkgs] @peti pushed to release-17.09 « ghc-mod: jailbreak to make the build accept hlint version 2.1 »: https://git.io/vxTkx
davidlt_ has joined #nixos
zzamboni has joined #nixos
<infinisil>
For anybody using ZFS and interested in the not force importing pools by default discussion should read this: (tl;dr: force import should stay the default, and the recommendation to turn it off should be removed): https://github.com/NixOS/nixpkgs/pull/36882#issuecomment-372787999
ashgillman has quit [Ping timeout: 256 seconds]
<gchristensen>
how could two computers possibly use the same pool?
<clever>
gchristensen: in the past, i have used nbd (a relative of iscsi) to jurry-rig an lvm array together across 2 machines, so i could replace one of the drives with a bigger one
<clever>
infinisil: iscsi is basically sata over tcp
<infinisil>
i see
<clever>
infinisil: i was once booting my laptop over that
<gchristensen>
yeah but ... can you really hook up the same block device to two hosts and have it come out ok?
<clever>
gchristensen: thats why zfs wont import the pool if the hostId is wrong
<infinisil>
gchristensen: i guess read-only for sure
<hyper_ch>
gchristensen: what's oddly specific?
<clever>
infinisil: even reads are unsafe, because the other machine may change things your not expecting
<clever>
infinisil: which may confuse and crash the reading kernel
<infinisil>
clever: read-many, write-one then
<infinisil>
write-exclusive or how it's called
<clever>
infinisil: your read cache may wind up claiming something that is false, because your un-aware of writes another machine did
<infinisil>
Oh, caches.. right
<{^_^}>
[nixpkgs] @roosemberth opened pull request #36928 → pcm: init at 201710 → https://git.io/vxTIN
<clever>
so you need to export the pool and flush the write caches, then flush the read caches and import the pool
xcmw has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<clever>
and properly hand-off full control of it
<infinisil>
Cache invalidation, one of the 4 hardest problems of computer science, along with naming things and one-off errors
<clever>
that list appears to suffer from the 3rd problem :P
<infinisil>
;P
<hyper_ch>
infinisil: I know you worked very hard on that PR and I just trotted along and made you reconsider
fendor has joined #nixos
<infinisil>
hyper_ch: Well #zfsonlinux made me consider really, I just asked until I understood it
<infinisil>
and looked up some source code
<hyper_ch>
ok, let's blame them :)
<infinisil>
Yarr :P
<hyper_ch>
als, what's WIP? Wery Important Person?
<infinisil>
hyper_ch: Work In Progress, a very common abbreviation for PR's actually :)
<MichaelRaskin>
Work-in-progress
<hyper_ch>
how... intuitive :)
<MichaelRaskin>
Intuition is earned, not inherited
ixxie has quit [Ping timeout: 256 seconds]
aarvar has joined #nixos
<MichaelRaskin>
Although «Wildly insane project» would also apply in some cases…
<coconnor>
haha
muzzy has quit [Read error: Connection reset by peer]
<hyper_ch>
well, I'm off to bed... gotta get up really early tomorrow... like at 4:30 :(
<MichaelRaskin>
Ouch
muzzy has joined #nixos
<infinisil>
hyper_ch: Good luck and good night
<hyper_ch>
cya$
qewzee has quit [Ping timeout: 264 seconds]
pkill9 has quit [Ping timeout: 240 seconds]
humanoyd has quit [Quit: WeeChat 2.0.1]
<clever>
infinisil: testing some things out in my kexec env to see what happens...
obadz has quit [Ping timeout: 264 seconds]
<infinisil>
clever: With no forced import?
<clever>
infinisil: yeah
obadz has joined #nixos
<rycee>
etu: That is a very odd error that I've never seen before! I pushed an update the HM master that hopefully will print a more useful error message. Please try it and let me know what it says.
<clever>
Myrl-saki: nix-copy-closure already gets the closure, its in the name
<infinisil>
Synthetica: no idea, maybe ask #idris? Nothing nix(os) specific
<infinisil>
Synthetica: I personally use an idris alias to supply all available packages
<Synthetica>
... Hmm, that might work
<Myrl-saki>
clever: I think that's the problem.
<Synthetica>
But I mostly want it to work in conjuntion with atom-language-idris
<Myrl-saki>
clever: I need build-time dependencies.
<Myrl-saki>
clever: And I don
<Myrl-saki>
clever: i don't think there is a One True Way to get build-time dependencies?
<Myrl-saki>
It requires checking the default.nix file itself?
<MichaelRaskin>
Myrl-saki: you can use nix-instantiate then nix-store -q
<clever>
Myrl-saki: the issue, is that the deps in the default.nix may themselves not be built yet
<clever>
and there is no one magic tool to list off everything you currently have that can be copied to a remote machine to speed up its building
<clever>
nix-store -qR on the .drv would give you ALL the build-time things, including stuff it no longer needs (like the gcc source!)
<Myrl-saki>
clever: While we're at this topic, what are the .drv files?
<clever>
the raw directions telling nix-daemon how to build something
<clever>
the nix expressions are used to generate the .drv files
<Myrl-saki>
clever: Thanks. I think I get it now.
<Myrl-saki>
MichaelRaskin: Thanks.
swolffs has joined #nixos
<Myrl-saki>
MichaelRaskin: clever: If I understand this correctly, if I want to prevent GC on my build tools, I could also use nix-instantiate --add-root?
<{^_^}>
[nixpkgs] @yegortimoshenko opened pull request #36937 → ssh: drop DSA support, fixes #33381 → https://git.io/vxT85
hakujin has joined #nixos
prismate has joined #nixos
<Myrl-saki>
Oh okay. I get what's happening now.
<Myrl-saki>
This is a stupid question.
prismate has left #nixos ["WeeChat 0.3.8"]
maximiliantagher has joined #nixos
Itkovian has quit [Ping timeout: 260 seconds]
<Synthetica>
infinisil: Turns out adding `idris = with pkgs.idrisPackages; with-packages [prelude base contrib lightyear];` to my `packageOverrides` works.
<infinisil>
Synthetica: Ah you mean that, right
<infinisil>
But then you also need to pass -p for each package, which is what I meant initially
<Synthetica>
Oh, no, that's fine
Fare has quit [Ping timeout: 256 seconds]
Itkovian has joined #nixos
hiratara has quit [Ping timeout: 240 seconds]
<Myrl-saki>
clever: All these times, I never knwe that I could have prevented all my woes by using --include-outputs
<Myrl-saki>
clever: lmao
dtulig has joined #nixos
* clever
heads off to bed
hiratara has joined #nixos
spietz has joined #nixos
Itkovian has quit [Ping timeout: 246 seconds]
alex`` has quit [Quit: WeeChat 2.0.1]
qewzee has joined #nixos
<Myrl-saki>
Is there a patchelf script ot automatically convert the result to FHS?
<infinisil>
clever: good night :)
<infinisil>
oh that was late huh
<infinisil>
Myrl-saki: I'm pretty sure there isn't
<Myrl-saki>
Awww.
civodul has quit [Read error: Connection reset by peer]
srdqty has quit [Quit: WeeChat 1.9.1]
civodul has joined #nixos
hamishmack has quit [Quit: hamishmack]
Meizikyn has joined #nixos
<infinisil>
Lol, it's 6pm for clever btw, what a guy
* infinisil
is jealous
<disasm>
is anyone using libinput with a macbook pro touchpad? It's overly sensitive ( It somehow detects my palm hovering slightly above the touchpad when typing), I switched back to synaptics for now.
<Myrl-saki>
It's 6 AM for me.
<infinisil>
disasm: I have tried it with disappointing results, using the multitouch driver
dan_b has joined #nixos
<Meizikyn>
Hi all, looking for help regarding systemd-nspawn DeviceAllow option. I peeked in configuration.nix(5) and found DeviceAllow options for NixOS specific containers, but no similar option for systemd-nspawn (say for a Ubuntu container)
<disasm>
infinisil: you have your config your using on github somewhere?
<{^_^}>
→ c35095a1 by @jtojnar: libsoup: fix build & clean up
<{^_^}>
→ df7ae304 by @jtojnar: modemmanager: 1.6.8 → 1.7.990
<{^_^}>
→ 4ef429b3 by @jtojnar: gnome3.eog: update dependencies
<infinisil>
Meizikyn: You can set service configs with `systemd.services.<name>.serviceConfig.User = "foo";`, even though the User option isn't in configuration.nix
lord| has joined #nixos
<disasm>
infinisil: love `config.mine.touchpad` :)
<infinisil>
Which is because serviceConfig is an option that holds an attribute set
<infinisil>
disasm: I redid my whole config recently with a whole lot of config.mine.* settings :) Everything in the new-modules folder
<disasm>
infinisil: yeah, I have something similar except use `profile.something`
<infinisil>
You may ignore the "MONSTER COMMIT" going over pretty much all my files in the repo
<Meizikyn>
infinisil: So for example I could add systemd.services.systemd-nspawn@.serviceConfig.User= "DeviceAllow=/dev/foobar" and it will "populate" under [Service] in the unit file?
<infinisil>
Meizikyn: Nah, the User one was an example for the User setting, so it would do `User=foo` in my example
<disasm>
although I intend to modularize my x config stuff here shortly...
<infinisil>
Meizikyn: I'm not sure in which section DeviceAllow is supposed to be though
<infinisil>
disasm: It's kinda hard, I haven't done it really well
<Meizikyn>
I believe in [Service]
<Meizikyn>
The configuration.nix(5) seems to indicate most settings are placed under [Exec], but giving it a try to confirm
<disasm>
Meizikyn: it's not abstracted enough in nixos/modules/system/boot/systemd-nspawn.nix -- you'd need to pull request an attribute for extraServiceOptions or something.
<infinisil>
Meizikyn: It does? Can't confirm
<Meizikyn>
specifically for systemd-nspawn options it does anyway
<Meizikyn>
hold on, citing
<hakujin>
is there a way to view hydra's evaluator status? I have a process that has been running for an hour.
<Meizikyn>
ah, I read this under systemd.nspawn.<name>.execConfig, only option that seems to modify the unit file
<hakujin>
s/process/jobset
<infinisil>
Ah
newhoggy has joined #nixos
newhoggy has quit [Client Quit]
<Meizikyn>
actually, paying more attention, their are options to modify [Files] and [Network] too, just not [Service]
newhoggy has joined #nixos
Neo-- has joined #nixos
<disasm>
Meizikyn: yeah, that's what I was saying. It would need to be abstracted out a bit more to manage custom service configs.
xcmw has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<{^_^}>
[nixpkgs] @aszlig pushed commit from @dtzWill to master « virtualbox: 5.2.6 -> 5.2.8 »: https://git.io/vxTu1
jensens has quit [Ping timeout: 260 seconds]
<sphalerite>
shlevy: you're not a big fan of the module system either iirc, was hoping you would be able to provide some input to the discussion too :p
<shlevy>
Aaah :D
<infinisil>
shlevy: Yes please
<ottidmes>
infinisil: I've got some questions about your config, why include all the files like you do? Is it just so you do not have to import them yourself, or is there some bigger idea behind it? Why use 'mine' everywhere in the modules, is it meant as a namespacing thing to prevent naming conflicts?
<{^_^}>
[nixpkgs] @aszlig pushed commit from @dtzWill to release-18.03 « virtualbox: 5.2.6 -> 5.2.8 »: https://git.io/vxTzL
<infinisil>
ottidmes: to the second question: indeed that's why
<shlevy>
Can you summarize what the question was? :)
<infinisil>
ottidmes: To the first: by importing everything I can just add/move/delete modules in new-modules any way i want and don't have to care about file names changing
<infinisil>
ottidmes: that's also the reason i use options for enabling all those modules, to be able to choose the ones i need
<infinisil>
ottidmes: No need to import the files on their own like this
<ottidmes>
infinisil: Thanks! 1) I guess that can indeed be convenient, if you make them all modules with an enable condition. I make the distinction between config and modules, where config are just config snippets and modules contain options (although I know technically they are both modules). And I do not always want to include the config files, it will depend on the machine, but I guess you could create an extra
<ottidmes>
directory for that
<infinisil>
shlevy: I think just why you don't like the module system, what's bad about it
ryanartecona has quit [Quit: ryanartecona]
spietz has quit [Ping timeout: 260 seconds]
<sphalerite>
shlevy: it was pretty much "what's wrong with the module system". I briefly said expensive evaluation, especially since all the nixos-provided modules are included and there's no simple way to get only the ones you want; the restrictive merging model which essentially only allows addition and wholesale replacement; and the completely implicit dependencies between modules
nallar has joined #nixos
<ottidmes>
infinisil: And if I rename a file, the tooling will yell at me, so I am not afraid for refactoring issues like that, but I like I said, I can see the convenience of your approach, I do it as well, but on a smaller scale
Ross has quit [Ping timeout: 260 seconds]
nallar is now known as Ross
<infinisil>
ottidmes: Well I think of the whole module system and every option set as config, and I just aggregate more and more config behind more higher level options, resulting in e.g. the option mine.profiles.desktop.enable
<disasm>
infinisil: just tried that, what's your middle click gesture for that config?
civodul has quit [Quit: ERC (IRC client for Emacs 25.3.1)]
<infinisil>
disasm: Haha, don't have one, never used middle click at all
<shlevy>
Ah. I mean ultimately my problems stem from lack of real modularity. Modules aren't self-contained in any meaningful way, they don't have a well-specified interface they use for affecting the system (I mean, the module system itself is well-specified, but not how it's composed into a config), they can't be reused or composed or anything.
<disasm>
infinisil: how do you paste in a terminal? :)
<infinisil>
disasm: Shift-Ctrl-V
<shlevy>
That all modules are included by default is an additional pain.
<disasm>
ouch, lol
<shlevy>
I don't think this is a project I'm likely to undertake myself any time soon though... If someone is interested I'd be happy to advise/review/etc.
<infinisil>
disasm: Well you can get used to pretty much everything
<ottidmes>
infinisil: 2) Just like I am not afraid for renaming, it will notice soon enough if I am conflicting with some pre-existing name
<infinisil>
shlevy: I've been thinking about explicitly layering the module system, so that every layer can only affect the one below it, more higher-level ones at the top
<infinisil>
Would be a step in that direction
<LnL>
everything beeing included by default is kind of a result of the unclear boundaries
<infinisil>
ottidmes: Yeah, I'm just gonna avoid all that, then I know when I see a mine.* that it's some option of my own modules, declared in new-modules
<shlevy>
A secondary, though definitely real, concern is that much of NixOS and to a lesser extent nixpkgs is aimed at the end-consumer personal computer case. Not all, of course, and to some extent this is a natural consequence of limited resources and the interests of people in the community, but it does at least seem difficult to get a system tuned for production deployment usecases, and IME the configuration model doesn't quite fit what you
<shlevy>
want from a maintenance/team perspective in that environment.
<infinisil>
LnL: shlevy: Oh, so how about we declare dependencies of modules like we do with packages, needs some maybe mindbreaking changes to the fixed point evaluation though
<shlevy>
infinisil: Hmm... Not sure I'm seeing it?
<ottidmes>
infinisil: But if you were to share your module, it will be clearly one of yours, but I guess it is easy enough to modify if you plan to share something
blankhart has quit [Ping timeout: 246 seconds]
<MichaelRaskin>
NixOS module system allows to fully express spaghetti code inside a formerly referentially transparent substrate
<shlevy>
I'm just not sure we even need anything as complex and powerful as the module system at all. Why not just functions?
lord| has quit [Read error: Connection reset by peer]
lord| has joined #nixos
<MichaelRaskin>
Well, for my own small system that runs on my notebook I try to use just the makeExtensible
<infinisil>
shlevy: Well I certainly use the module system for a lot of things that wouldn't be possible with ordinary functions..
<gchristensen>
it is all functions =)
<gchristensen>
s/=)/:trollface:/
<shlevy>
infinisil: I'd like to see some examples and how they'd look in your layered/explicit dependencies approach
<MichaelRaskin>
coordinated changes to arguments to multiple parts can be somewhat annoyinh
<boomshroom>
Hi! As a rust developper, I love simply being able to use buildRustPackage, but at the same time, much rust code depends on nightly-only features (for better or worse) and it can be very anoying trying to build a crate on nix using the nightly compiler.
<infinisil>
shlevy: Example: (it's not done yet, but starting): Have an option for a dev set up for {rust,haskell,etc.}, and an option for enabling {vim,emacs,etc.}, then it can be made so that each of those dev setups do their thing to the emacs/vim config
abathur has joined #nixos
rardiol1 has quit [Remote host closed the connection]
<infinisil>
shlevy: I haven't thought about layered modules and explicit dependencies a lot, but I'll remember to ping you when I do
<manveru>
like spacemacs layers?
<infinisil>
manveru: Yeah, just for all editors with a single option
<shlevy>
infinisil: So something like "global editor support" makes sense to be available in a global namespace somehow
<manveru>
what i find the biggest pain atm is getting my emacs working with nix-shell... having to redefine a lot of functions so they use the nix-sandbox functions...
<manveru>
and i don't think that's solved by your approach :|
ma27 has joined #nixos
<shlevy>
But I don't think that means *all* knobs should be, or that we need fully general merge logic
<{^_^}>
→ 4f225f96 by @Ma27: xautolock: implement basic test to confirm functionality
<{^_^}>
→ 1a1abb38 by @grahamc: Merge pull request #36706 from Ma27/xautolock-testcase
<ottidmes>
I tried to do things like generate imports depending on the users available in the config, but you cannot do that, since the users will depend on the imports, and thus an infinite loop occurs. Would be nice if the language would allow for this. For example, then mark the users as evaluated, so the imports being generated based on it are not allowed to impact the value anymore, only then you would get an error
<infinisil>
manveru: I share your pain, nix-shell can be a burden sometimes
srdqty has joined #nixos
<manveru>
so that dev env and desktop approach i think is best handled by home-manager, while nixpkgs could focus on more serious stuff?
<shlevy>
Though a separate concern I have, unrelated to the module system per se, is that "how my system is configured" seems orthogonal to "how my user env is configured" and editor support belongs in the latter :P
<infinisil>
manveru: An approach that would almost certainly work is: Wrap all the binaries you need with nix-shell with a simple script that just passes along arguments
<manveru>
shlevy: exactly
<infinisil>
shlevy: Yeah, that's something I've been thinking about for a bit longer now
zzamboni has quit [Quit: Leaving.]
<MichaelRaskin>
shlevy: well, if you want an example of an annoying cross-cutting concern surfacing in a ton of places, there is always hostname.
<infinisil>
shlevy: A lot of modules are really just package configurations, all of those shouldn't really be modules, but rather packages you configure with a nix function, then install it into the desired env
<shlevy>
infinisil: +100
<infinisil>
MichaelRaskin: What's the thing with hostname?
<infinisil>
shlevy: The same with systemd services
<shlevy>
MichaelRaskin: hostname should be late-bound IMO :P
<MichaelRaskin>
infinisil: you need to pass it to a lot of places.
<MichaelRaskin>
shlevy: erm, what about /etc/hosts ?
<infinisil>
When those modules are separated from the nixos folder of nixpkgs, it allows for other distros, home-manager, and more to use them
<infinisil>
i mean moving them to pkgs where they belong
<MichaelRaskin>
I am annoyed enough by what I do with resolv.conf and passwd
<shlevy>
MichaelRaskin: Use DNS if your'e trying to compose systems :P
<MichaelRaskin>
I want to compose a single small notebook that can happen to be in multiple networks…
<infinisil>
MichaelRaskin: Yeah same
<shlevy>
What do you want to do with /etc/hosts exactly that requires threading around your hostname?
<MichaelRaskin>
Ideally, the hostname would be there as an alias for 127.0.0.1
<infinisil>
MichaelRaskin: Would be nice to configure this by specifying the network topology in a nix expression!
<shlevy>
manveru: Essentially yeah. Basically not everything boils down to env vars and I don't think it's tenable long term to try to shove everything into them
<MichaelRaskin>
Also, there are some things that have to be added to pam.d in addition to whatever you do with them.
<shlevy>
MichaelRaskin: Fine, I think there are other ways to do that (e.g. nss plugins somehow) but I'm not opposed to *some* things knowing about hostname in advance
<shlevy>
I just think in general we'd benefit from trying to bind more things later in the Nix world
<shlevy>
I mean, my ideal world would have building only depend on headers/interface definitions and all runtime dependencies resolved by capabilities :P
<MichaelRaskin>
In the meantime, I just run every browser instance (often with a single page) in a separate namespace sandbox…
Neo-- has quit [Ping timeout: 246 seconds]
<infinisil>
whoa
lord| has quit [Ping timeout: 264 seconds]
<infinisil>
Alright I have a crazy idea..
<infinisil>
Hold on, needs to brew for some minutes
lord| has joined #nixos
<abathur>
related, but not quite on-topic--I've been picking at getting myself set up to work on a long-term project to reproducibly configure a server/host for an old MUD I admin for
<{^_^}>
[nixpkgs] @jtojnar pushed 21 commits to gnome-3.28: https://git.io/vxT2D
<{^_^}>
→ 1730eca7 by @jtojnar: meson: fix rpath fixup
<{^_^}>
→ 1f93fc57 by @jtojnar: deja-dup: remove unneeded rpath fixup
<{^_^}>
→ 3e4ce466 by @jtojnar: epiphany: remove unneeded rpath fixup
<MichaelRaskin>
At least this way I get to run Marionette without managing the port space.
<abathur>
and I've figured out how to set up shells to hack on some of the individual software pieces we're using
<abathur>
but I'm a little less sure about configuring database/authentication/web servers/services and such
jtojnar has quit [Read error: Connection reset by peer]
<manveru>
abathur: usually there's nixops for this kind of thing
<abathur>
that may be what I'm missing
jtojnar has joined #nixos
<abathur>
before I just had a VM and chef
<abathur>
but chef has been wearing me down over the years
<abathur>
sooooo tired of my builds blowing all apart and needing serious work because I need to bump a cookbook for some specific feature in the underlying library that just happened to be tied to epic API changes at the chef layer
<abathur>
all I really need/want, for _now_ is to figure out what the lower-friction way of setting up an environment to hack on these components is
<{^_^}>
→ 833851cd by @badi: qtlocation: fix darwin build
<{^_^}>
→ ecbb9027 by @LnL7: Merge pull request #36847 from badi/qtlocation
<abathur>
and it's seeming like I either need to circle back to making a VM, or just let my desktop be the MUD server while I work on this stuff, and then later pull out those pieces into a separate config
<{^_^}>
[nixpkgs] @LnL7 pushed commit from @badi to release-18.03 « qtlocation: fix darwin build »: https://git.io/vxTaZ
<manveru>
i've done either
<manveru>
s/either/both :)
<abathur>
yeah
<infinisil>
Alright I think my crazy idea would work
<manveru>
so first writing your modules to work on your desktop first, and then using them in your VM works with nix too
erasmas has quit [Quit: leaving]
<manveru>
once the VM works, deploy the thing
<infinisil>
An overhaul of the module system that should make it faster and easier to see dependencies, maybe more benefits even
<infinisil>
(I'll wait a bit to present it)
<ottidmes>
infinisil: Would be cool if it also allowed you to define the same property twice (implicit merge), it is quite annoying having to use mkMerge blocks to allow mixing setting a set at once and a single attribute in the same file
<abathur>
manveru: good to know you've had success with both; I'm a little gun-shy about leaping down the wrong path, since I imagine I'm looking at a few weeks of spare-time work to translate all of my existing chef cookbooks
* infinisil
thinks about that
elasticdog has quit [Ping timeout: 268 seconds]
<manveru>
abathur: well, always make sure you look into nixpkgs modules first before :)
<manveru>
almost everything you'd wanna do has been done there before, except for proper deployments of erlang servers...
<abathur>
manveru: a few of the pieces are there, so those I'm just customizing; it's a bit of a mess/next :)
knupfer has quit [Ping timeout: 246 seconds]
<infinisil>
Wait my idea has some holes, I'll be back..
disasm has quit [Remote host closed the connection]
<shlevy>
infinisil: Can you ping me when you present soemthing sharable? I can't keep up with the backlog these days :D
<infinisil>
shlevy: Yeah I'll try to remember (if i ever get there :))
noam has quit [Read error: Connection reset by peer]
<abathur>
manveru: most of it's just install and config, but a few of the applications are custom forks for now as I work on a set of inter-related contributions to them all
noam has joined #nixos
Ross has quit [Ping timeout: 246 seconds]
Ross has joined #nixos
<manveru>
abathur: well, the only difference is that you'll set `src = /path/to/my/fancy/fork` instead of `src = fetchGit { url = https://github.com/the/original; }` more or less?
pkill9 has joined #nixos
<MichaelRaskin>
shlevy: I guess with careful structuring, just .extend and lib.recursiveMerge could be enough.
<shlevy>
For hostname? Or nixos config in general?
<infinisil>
MichaelRaskin: enough for what?
<MichaelRaskin>
For system config in general
<infinisil>
am not convinced
<tobiasBora>
Hello again!
logzet has quit [Read error: Connection reset by peer]
<abathur>
manveru: I think mostly, I've only set up 2 or 3 components so far; the main asterisk so far is adding in custom support to enable and build optional features
<infinisil>
MichaelRaskin: shlevy: So, I just came to realize that the dependencies in the module system are pretty much just a directed graph
<infinisil>
And this can be cyclic of course which is where infinite recursion can come from
<tobiasBora>
Just a quick question, do you know if it's possible to compile nix so that the local user folder are changed? For example .nix-channel-2, .nix-defexpr-2 and .nix-profile-2? I would like to install several parallel "version" of nix
<MichaelRaskin>
I like how there is no «acyclic» in your characterisation
<infinisil>
MichaelRaskin: Indeed..
vidbina has quit [Ping timeout: 240 seconds]
<infinisil>
So now the question is: Do we need cyclic dependencies?
<infinisil>
Because otherwise we could just use a DAG (directed acyclic graph) and eliminate infinite recursion at last
<infinisil>
"just"
disasm has joined #nixos
<infinisil>
I mean cyclic dependencies are necessary to allow for fixed points, but we might not need this
<abathur>
manveru: those are less the problem than figuring out what the ideal path to defining and configuring supporting services was going to be without reinventing the wheel, but I think that'll be workable if I just set them up at the system level as discussed above
aborsu has quit [Quit: aborsu]
<MichaelRaskin>
infinisil: note that once you have a DAG, you can structure it closer to Nixpkgs and .extend becomes enough to express what is going on…
<infinisil>
MichaelRaskin: i assume you mean .extends?
<infinisil>
oh wait no
<MichaelRaskin>
No, .extend
<infinisil>
almost the same though
<MichaelRaskin>
lib.makeExtensible etc.
<infinisil>
Ah so that's the magic behind overlays
blankhart has joined #nixos
<infinisil>
but it still uses a fixed point
Meizikyn has joined #nixos
<MichaelRaskin>
Yes, but if you add the condition of the construction actually evaluating…
<electrocat>
is there a declarative way to configure user shell, e.g. set it zo zsh?
<MichaelRaskin>
I mean, it needs a fixpoint because it cannot know in advance what is the depth of a DAG
<infinisil>
Ah
<Meizikyn>
is there a way to have 'symbol' characters in a nixos definition (configuration.nix) such as "@" i.e. If a systemd unit file has "@" in the name (systemd-nspawn@ubuntu1.service)
<manveru>
electrocat: yeah, users.extraUsers.electrocat.shell = "${pkgs.zsh|/bin/zsh" for example
<manveru>
err, ${}, not ${|
<electrocat>
manveru: ah tnx
<infinisil>
MichaelRaskin: So self would be used to refer to options and super would be used to ?
<manveru>
Meizikyn: put it in quotes?
<MichaelRaskin>
Well, the initial system definition defines all the parts, and all the settings.
<MichaelRaskin>
And it uses self for resolving the references
<MichaelRaskin>
Configuration mostly uses super, unless it does want to define something and reuse it.
<infinisil>
MichaelRaskin: Why not self though? self would be like the final config
<infinisil>
MichaelRaskin: And wouldn't super be dependent on the order of evaluation of the "modules"?
<Meizikyn>
I've tried as "systemd.services.systemd-nspawn@ubuntu1.serviceConfig.DeviceAllow"="/dev/snd/seq" and systemd.services.systemd-nspawn"@"ubuntu1.serviceConfig.DeviceAllow="/dev/snd/seq"
<Meizikyn>
still fails build
<manveru>
Meizikyn: like systemd.services."systemd-nspawn@ubuntu1".serviceConfig.DeviceAllow = "/dev/snd/seq"
<MichaelRaskin>
Well, the config would need both self and super.
<MichaelRaskin>
Modules? What modules.
<infinisil>
MichaelRaskin: maybe i haven't understood extend fully, and how overlays internally work for that part
<tilpner>
electrocat, manveru - users.users.electrocat.shell = pkgs.zsh; is sufficient
<manveru>
oh :)
<tilpner>
(But make sure to programs.zsh.enable = true;, or some things may go horribly wrong)
<MichaelRaskin>
infinisil: One large nixpkgs-syle Thing. There are no modules.
<electrocat>
tilpner: indeed it is
<electrocat>
it works without "programs.zsh.enable = true;" though
<manveru>
MichaelRaskin: would you put both the package and service into one layer?
<tilpner>
electrocat - It can work, but I almost locked myself out of a server by forgetting programs.zsh.enable
<infinisil>
MichaelRaskin: Hmm.. But then everything needs to be evaluated right? Every such overlay needs to go through the fixed point. And I think infinite recursion can still happen
<MichaelRaskin>
manveru: probably not always.
<MichaelRaskin>
We already put some service-like things in Nixpkgs, like makeFontsConf
<MichaelRaskin>
(I would like more)
<infinisil>
same
<jtojnar>
cool, gnome3 attribute set contains gnome3 attribute so you can install packages even if you accidentally paste more gnome3. prefixes than necessary
<gchristensen>
is that a feature? :P
<MichaelRaskin>
Then, just like Nixpkgs imports python-packages separately then inherits some stuff from there, system config could import nixpkgs and inherit something
<hakujin>
what are the user/group/permissions required for nix distributed build ssh keys?