<colemickens>
The problem is, some of the derivations that might be needed for building the toplevel might not actually be part of the output closure, right?
<colemickens>
if they're a build-time dep
ddellacosta has joined #nixos
<colemickens>
but I can't run `nix-store -q <some-unbuilt-outpath>` in order to collect the paths that already exist
<__monty__>
I'm not sure, but I assume that just makes sure the nix-shell produces a pure environment. direnv then simply adds that to the current environment.
<shreeya>
nix-shell --pure removes the direnv set environment vars
<shreeya>
maybe that's expected, but then what exactly is the use of nix-direnv, nix-shell by default is impure and inherits direnv set environment anyway
<dminuoso>
nix in general has way too many impurities by default.,
<shreeya>
with or without `use nix`
<dminuoso>
I wish we didnt have --pure but --impure rather.
<dminuoso>
purity should be opt-out not opt-in
<__monty__>
shreeya: The point is to avoid the nix-shell evaluation every time. nix-shell is pretty slow.
<shreeya>
hmm i think i just don't need to drop into nix-shell if i am okay with impurity, and all the project deps mentioned in the shell.nix file gets loaded into the path
<shreeya>
by direnv, so no need for nix-shell
<__monty__>
It's just a convenient way to get those deps into the envrc.
<shreeya>
shell.nix is pretty useless without --pure right, it is possible you missed some dep, but you have it in your local setup globally installed, i commit, and it will not work on a clean CI
vidbina has quit [Ping timeout: 268 seconds]
<mikroskeem>
i'd not say it's useless, but rather... error prone
<mikroskeem>
and one reason is what you just wrote
<__monty__>
Not to me, all shell.nix does for me is add dev tools to my environment.
<mikroskeem>
just have to be careful
<__monty__>
Actual build deps should be in default.nix.
<dminuoso>
shreeya: Consider that nix-shell without --pure leaves you completely naked. You dont have `grep` suddenly, or `awk`, or `sed`.. or `ps` or anything else.
<shreeya>
i see, maybe i am using it wrong them, i am just adding deps to buildInputs frmo the name
<dminuoso>
mmm, grep might be special cased interestingly
<dminuoso>
but you dont get `ps` or your editor
<shreeya>
even awk
<shreeya>
right ps aint there
<dminuoso>
ah hold on
<mikroskeem>
i think util-linux is special case yes
<dminuoso>
grep is a bash built-in
<mikroskeem>
no?
<dminuoso>
[nix-shell:~]$ which grep
<dminuoso>
The program ‘which’ is currently not installed. It is provided by
<dminuoso>
so what's going on here?
<mikroskeem>
command -v grep
<dminuoso>
oh
<__monty__>
You can use type instead of which btw.
<thibm>
no but it's in stdenv I suppose
<thibm>
(grep)
<dminuoso>
Anyway, the points isnt grep, its just that you lack a lot of tools you'd otherwise take for granted
<shreeya>
i dont think grep is bash builtin, if so is every shell rewriting grep lol
<shreeya>
right
<mikroskeem>
it's definitely not built in
<tertl3>
ok
<mikroskeem>
let's see pkgs/build-support/mkshell/default.nix
<atralheaven>
Hi, I'm looking for a way to have multiple instances of a program with different configs running, is there something for that in nix ecosystem?
<tertl3>
grep awk ggrep fib oimgh
<mikroskeem>
atralheaven: depends how you want to run your desired program
<mikroskeem>
as systemd service? create new service for every instance etc.
<atralheaven>
for example, multiple instances of vscode with different configs and extentions
<shreeya>
you can probably have multiple home-manager nix files that derive from something common and override the configs, but you will have to keep switching with -f
zertox has joined #nixos
<atralheaven>
shreeya: so it's possible with home-manager? does home manager do the same thing with config files that nix does with packages?
Qwerky has joined #nixos
<__monty__>
atralheaven: It's possible with makeWrapper if the program reads its config from an env var path or takes it as an argument.
<thibm>
atralheaven: I have an overlay which makes a binary $out/bin/prg-x that uses $x/bin/prg with an option specifying the config file. So I can use prg or prg-x in my $PATH.
<shreeya>
this seems better ^
<atralheaven>
what are overlays?
<thibm>
atralheaven: overlays are a way to add/modify packages in the nixpkgs package set
<__monty__>
,overlay
<{^_^}>
Overlays look like `self: super: { foo = ...; }`. Use the self argument to get dependencies, super for overriding things and library functions (including callPackage). More info: https://nixos.org/nixpkgs/manual/#sec-overlays-install
<atralheaven>
Ideally I want to have separate environment, each with access to specific packages and their own configurations
<__monty__>
If you want a completely separate environment I think you'd have to generate multiple profiles and switch to them. Not the greatest UX.
<atralheaven>
__monty__: do profiles have separated config files?
Qwerky has quit [Ping timeout: 265 seconds]
<__monty__>
atralheaven: You know what generations are? Profiles are what contain generations, so yes, they can be completely separate.
<atralheaven>
__monty__: like generations that each time I run "nixos-rebuild switch" would create? my understanding of profiles is that each profile can have specific packages (which are located on nix store) "activated" in shell, and profiles have generations too, just like how configuration.nix does
<atralheaven>
but how does it separate files that are generated from a package, like configs, cache, etc? does it store them in a separate directory?
<__monty__>
You can use nix to put those in place. Nix would then replace them whenever you switch. You can't separate state generated by programs unless you use filesystem trickery.
HenrikK has joined #nixos
henrikkarlsson has joined #nixos
<__monty__>
You'd have to be pretty motivated to get this all working nicely tbh.
<atralheaven>
but if the program can get config file as an option passed to it, it would be as easy as just using profiles, right?
<__monty__>
You don't even need profiles in that case. You can simply use makeWrapper to create a script which'll invoke the program with that option passed.
xwvvvvwx has joined #nixos
<atralheaven>
so I will check makeWrapper out
<thibm>
atralheaven: in that case, a solution would look like this: http://ix.io/3n5h
<__monty__>
Take a look at wrapProgram too, it's a bit more convenient but I'm not sure it lets you pass options.
<atralheaven>
a question about profiles, when I create a profile, and specify the desired packages for it, when that profile is activated, what other packages (that I've not specified) would I have access to?
<CodeBitCookie[m]>
__monty__: When using nix with flakes we need a derivation to build that project right? But how can I use it without it interferring with others work and the git repo
henrikkarlsson has quit [Ping timeout: 252 seconds]
<CodeBitCookie[m]>
I mean we don't NEED a derivation
<CodeBitCookie[m]>
but It is very useful
<__monty__>
Well, you don't have to check in the expression.
<CodeBitCookie[m]>
meaning how do i reap the benifits without the files in the git repo
<CodeBitCookie[m]>
* meaning how do i reap the benifits without the files pushed in the git repo
<CodeBitCookie[m]>
* meaning how do i reap the benifits without the files pushed to the git repo
<CodeBitCookie[m]>
to*
<mikroskeem>
shreeya: not really
<__monty__>
Oof, is that a matrix typo fix feature? Please feel free to refrain from using that in bridged IRC rooms.
<shreeya>
what does it look like on the other side?
<dutchie>
it resends the whole message with a * at the front
<CodeBitCookie[m]>
oof
<dutchie>
my current way of working with a flake config that i don't want to push to the upstream repo is to use stgit
<CodeBitCookie[m]>
welp i guess no editing the message then
<CodeBitCookie[m]>
dutchie: by stgit you mean stacked git?
<CodeBitCookie[m]>
I guess you use matrix sorry meet
<dutchie>
presumably meet is shreeya on irc?
<CodeBitCookie[m]>
the other guy had this problem too
<CodeBitCookie[m]>
his name was Felix
<CodeBitCookie[m]>
It worked when we put some wierd stuff after his nickname and it worked
<shreeya>
yeah, it should be meet, doens't matrix bridge actually identify and login as user on the irc?
<CodeBitCookie[m]>
no
<CodeBitCookie[m]>
it puts something in brackets after too
<evils>
[m] because you're from the matrix bridge
<evils>
no idea why felix hads a 6 after that though
<CodeBitCookie[m]>
evils: yeaahhh
<CodeBitCookie[m]>
:D
<shreeya>
i see, what is actually the use of identifying with the nickserv user then, or have i just been sending out my passwd in plain text to some matrix user :O
<CodeBitCookie[m]>
meet[m]++
<CodeBitCookie[m]>
:(
<dutchie>
shreeya++
<{^_^}>
shreeya's karma got increased to 1
<dutchie>
there you go :p
vidbina has joined #nixos
<CodeBitCookie[m]>
what?
<dutchie>
presumably you did !nick shreeya to the bridge bot at some point to override the default user mapping
<CodeBitCookie[m]>
!nick meet
<CodeBitCookie[m]>
!nick
<CodeBitCookie[m]>
ohhh
<CodeBitCookie[m]>
i think I understand now
<shreeya>
oh weird, i should check the mapping, i indentified with the nickserv as meet though
<CodeBitCookie[m]>
dutchie: how did you find the name though?
<CodeBitCookie[m]>
since I am also thinking of switching to irc
<CodeBitCookie[m]>
I used irc
<CodeBitCookie[m]>
then started using element
<CodeBitCookie[m]>
and then forgot
<CodeBitCookie[m]>
lol
<dutchie>
CodeBitCookie[m]: just guesswork that the only other person you were helping was showing to me as that name
<atralheaven>
I installed vscode python extention "http://p.ip.fi/M0W2" but it's not working, it's disabled and can't be enabled, what am I doing wrong?
<atralheaven>
when a program gets updated, and the build process is not changed (can be done with same commands and dependencies but with different source code), does updating the nixpkgs .nix file on github only require changing the tarball file?
uznog has joined #nixos
<eyJhb>
atralheaven: Yeah, pretty much. Just the rev and sha
jimmiehansson has joined #nixos
<{^_^}>
[nixpkgs] @github-actions[bot] closed pull request #123350 → haskellPackages.cut-the-crap: use ffmpeg instead of ffmpeg_3 → https://github.com/NixOS/nixpkgs/pull/123350
<meet>
are nodePackages.* bound to getting outdated opposed to just getting straight from npm?
zaki has quit [Ping timeout: 250 seconds]
<eyJhb>
Ehmm... nodePackages are weird. But I think they are fairly often updated. As when you add/update a single package you actually have to update them all
civodul has joined #nixos
meh` has quit [Ping timeout: 245 seconds]
Qwerky has quit [Remote host closed the connection]
eoli3n_ has quit [Remote host closed the connection]
<meet>
i wonder why people decided to not write something like fetchFromNPM, fetchFromPyPI, etc. instead of creating packaging around another packaging
<meet>
my understanding with arch has been if you want to install some end-user app that uses python deps, you need to have those python deps in the aur / arch repo as well, hence, the python, etc. packages get created
<meet>
but in nix we could probably reuse functions the same way we do for fetchFromGit, etc. i have seen in places
supersandro2000 has quit [Quit: Ping timeout (120 seconds)]
<atralheaven>
how can I set the channel to unstable here? http://p.ip.fi/YeZX
Qwerky has joined #nixos
ddellacosta has joined #nixos
<evils>
meet: not sure what you mean, we have fetchPypi, but we provide the dependencies because most python packages don't guarantee reproducibility
<evils>
meet: in the case of cargo packages for example, they are reproducible so we let them fetch their dependencies and just set the hash of the output
eoli3n_ has quit [Ping timeout: 260 seconds]
<meet>
anything with lockfiles should be reproducible though, right?
eoli3n_ has joined #nixos
<meet>
i forget if python setup.py's require version pinning or not
<evils>
atralheaven: you want a different pkgs in the module you're passing pkgs into?
ddellacosta has quit [Ping timeout: 268 seconds]
<evils>
meet: that's the general idea with lock files in package management, as i understand it
<meet>
then we could be skipping node packages with package-lock.json or yarn.lock?
<evils>
i'm quite unfamiliar with node
<meet>
and for some of the python projects that use poetry / pyproject.toml since poetry introduced lockfiles into python package management
<evils>
i know there's some facility for checking hashes in python, but that is rarely used
<lordcirth>
Github seems to have added namespaces / directories for branches... How do I specify them? "github:samueldr/wip-pinebook-pro/feature/2021-05-upgrades" says the URL is invalid
<ronthecookie>
hey so I'm trying to run nixpkgs 2007 firefox to beat the debian users into submission but `nix.cs.uu.nl` is dead so `nix-shell` fails, is there an archive somewhere?
<ronthecookie>
wow, I was two years old when this commit was made (7303789d889235566dbc109d62b0dd7bd27576f4)
uznog has quit [Quit: Connection closed]
Cyrinux33013863 has joined #nixos
gustavderdrache has joined #nixos
Cyrinux3301386 has quit [Ping timeout: 240 seconds]
<thibm>
Maybe you could write a page (or update one revelevant) on nixos.wiki with some of the things to know to do Nix speleology? Like the command to replace all the nix.cs.uu.nl references and maybe other things you found out.
<srhb>
I think you're all violently in agreement, but some people are reading past tense where others are writing hypotheticals :D
<simpson>
Even if software-makers were guilded or otherwise organized professionally, this sort of change in perspective would still be possible and good, just like how memory-safety was possible and good but not always a thing.
<ronthecookie>
srhb& I am 15 years old and I can only learn about the past so much, therefore the need for hypotheticals
<ronthecookie>
\nm
<srhb>
ronthecookie: Sure, it wasn't a criticism :) Just observing.
<ronthecookie>
srhb& i know, the \nm is for \not mad :P
<ronthecookie>
simpson& that article is crazy long so I'm not going to read it right now but, anyhow, I don't think what nix does is the only reason for why it is so good
<ronthecookie>
when I say "good" I really mean stable and reasonably-not-bloated and that is what shines in the implementation of nix for me
<simpson>
ronthecookie: Sure. But perhaps "stable" and "bloated" are subjective. I could argue that Nix is quite unstable in terms of flake support, and that it's quite bloated in terms of format support (XML support, etc.)
utente has joined #nixos
<simpson>
Also, the Guix folks maintain a Nix fork which has different ideas of "stable" and "bloated".
<ronthecookie>
thibm& it failed, logs are huge so it will take a bit to upload but the gist of it is: error: `cannot download curl-7.17.1.tar.bz2 from any mirror` which is quite odd to me
ddellac__ has joined #nixos
<simpson>
I earnestly think that Nix's trick is to treat packages as capabilities; if you know the path to something in the Nix store, then you're allowed to invoke/load it.
utente has quit [Read error: Connection reset by peer]
<thibm>
ronthecookie: it means that you can't find this curl version tarball from any mirror (website)
stree has joined #nixos
<ronthecookie>
thibm& I can read into it that much but why doesn't it build curl on it's own then? it was already using a different curl for all the other stuff
<thibm>
ronthecookie: that's something I wanted to tell (but I was lazy): I trust Nix and the only issue to do speleology is sources. They disappear and you can't do much about that (actually, you can, but that's another story).
johnny101 has joined #nixos
<thibm>
Or some sources hash change but nobody see it for years because almost everybody use cache.nixos.org. And anyway the Nix expression becomes "wrong".
* ronthecookie
waits for thibm to shout the magical acronym, I P F S
<thibm>
ronthecookie: IT tries to build curl but cannot find the source tarball.
<thibm>
ronthecookie: IFPS or/and other things.
<ronthecookie>
oh, I assumed that was a binary
<thibm>
.tar.bz2 here is the source coe
<ronthecookie>
simpson& thinking about it more I am not as sold on my opinion anymore
<thibm>
code*. You can either find this tarball somewhere (maybe some links just changed), or use a newer curl version and assume it will work (most likely)
growpotkin has joined #nixos
MmeQuignon has quit [Ping timeout: 240 seconds]
ddellac__ has quit [Ping timeout: 260 seconds]
exondation has quit [Ping timeout: 245 seconds]
exondation has joined #nixos
<thibm>
ronthecookie: https://curl.se/download/ there is no longer bz2'ed tarball for this release for some reason…
<ronthecookie>
heh I found a different mirror but it doesn't support HTTP :^)
<thibm>
ronthecookie: just change bz2 by gz in the curl src attribute
<thibm>
ronthecookie: you can just use https instead of http in the src attribute. It's probable that Nix fetcher didn't handle redirection correctly
<ronthecookie>
nope, curl wasn't built with ssl support
<thibm>
ok
<thibm>
2007 :)
<ronthecookie>
:^)
<thibm>
ronthecookie: are you using an old Nix or a recent Nix to build?
<ronthecookie>
i am on a fairly recent nixos-unstable
endformationage has joined #nixos
<marnov>
Hello i am planning my homelab ... nixos server. i wanna some mirror for data (like databasses, webapps) but system can run from small ssd. WHere to mount that mirror ? /var ?
MmeQuignon has joined #nixos
civodul has quit [Ping timeout: 250 seconds]
<thibm>
ronthecookie: OK, it doesn't matter actually: fetchurl uses the nixpkgs checkout's curl, not an internal Nix's fetcher or anything like that
<thibm>
ronthecookie: you could also download the tarball yourself, add it to the store, and forget about it.
<simpson>
marnov: You could put it under /var, sure. I have a similar setup, and I use /mnt for the disk, plus a few bind-mounts to get stuff into /pub and /var. There's no hard rules on this; do what works.
<ronthecookie>
thibm& you do present a point :D
<marnov>
simpson: thx ... keyu question is, where applications usualy stores data ...
<marnov>
yea, $out is pointying to /nix/store/hash ... ?
<thibm>
yes
<thistledown>
Hey there. I'm fairly new to NixOS, and I'm trying to build a package for https://github.com/gilbertw1/drop which is a little screenshot tool in Rust. I am, however, running into a bit of a roadblock that no amount of Googling has helped me with.
<thistledown>
Namely: "Failed to run `"pkg-config" "--libs" "--cflags" "glib-2.0" "glib-2.0 >= 2.42"`: No such file or directory (os error 2)" I definitely have glib in the dependencies.
<ronthecookie>
thibm& I tracked down another mirror!
<thistledown>
I'm pretty certain I've done it entirely wrong, but the documentation is fairly sparse and I've pieced together what I can find.
<ronthecookie>
thibm& I thought this would be a bit more exciting but it is just resolving dependencies now
<thibm>
thistledown: yes, the file looks strange but that's not a problem :)
<Raito_Bezarius>
I'm getting an error of the sort: error: derivation '/nix/store/pgd5352spny4gy3sca9xxi6p1zak3vm6-OVMF-202102.drv' does not have wanted outputs 'fd,out'
<Raito_Bezarius>
though the drv has clearly out fd has outputs
<thistledown>
Yeah, need to fiddle with my emacs config to handle .nix files in a way that's a bit more sane, lol.
<Raito_Bezarius>
this looks like a regression in Nix because I didn't have it, but I cannot exclude nixpkgs unstable having done some changes that I don't understand as the outputs field has not changed
<thibm>
ronthecookie: you could also use the firefox 2 expression in the latest nixpkgs. That would need some work too but in some case it may be less work
<ronthecookie>
i'm not doing this for a legitmate reason :D
<ronthecookie>
also i've only had to change like 1 hash so far so i'd rather see this through
ddellac__ has joined #nixos
<thibm>
ronthecookie: for a legitimate reason, I had to use a very old LLVM. That's was difficult to do with Nix :(
<thibm>
thistledown: putting pkg-config into nativeBuildInputs instead of buildInputs fixes it.
<thistledown>
Aha, thank you!
<thibm>
(but you get another error :p)
<thibm>
thistledown: you should not use `with pkgs`. You can simply write all the dependencies in the function argument (the first line), and use them in buildInputs/nativeBuildInputs
<thistledown>
Gotcha, will keep that in mind.
<thistledown>
Giving it another go with the changed bits.
ddellac__ has quit [Ping timeout: 260 seconds]
bahamas has quit [Ping timeout: 252 seconds]
fendor_ has joined #nixos
<thibm>
thistledown: you will soon see that people not using Nix miss half of them dependencies in their README ;)
<thistledown>
No kidding.
<thibm>
(the README is one thing. The archlinux package is another…)
fendor has quit [Ping timeout: 252 seconds]
<ronthecookie>
thibm& it blew up again, this time because it depends on the kernel version...?
<thibm>
strange
<ronthecookie>
I am running 5.11.16-zen1 and it doesn't like that
<thibm>
can you paste the error?
<ronthecookie>
yup, just a moment
<thistledown>
Yeah, I just came over from Arch because I got super tired of AUR breaking everything with cyclic dependency hell.
mastarija has joined #nixos
<ronthecookie>
the log file is 12m
<thibm>
ronthecookie: paste just the last few lines with the error
<ronthecookie>
that sounds boring!
<thibm>
huh?
<ronthecookie>
rsync is already almost done anyway
<mastarija>
I'm using the minimal installer and when I idle for a bit, if I try to do some IO stuff, like a simple dir command, I get I/O error
<ronthecookie>
mastarija& maybe your usb stick is bad?
avaq has quit [Ping timeout: 240 seconds]
<simpson>
ronthecookie: Truncate your log next time, please. It looks like this software makes the poor choice of checking kernel version; perhaps the version check can be changed or disabled.
<mastarija>
ronthecookie, when I reboot everything goes back to normal, and it only happens when I idle for a while, if I continuously perform actions it works as it should
philr has quit [Ping timeout: 268 seconds]
<mastarija>
ronthecookie, could such behaviour be related to a faulty USB?
<ronthecookie>
maybe it's some power saving thing?
proofofkeags has quit [Ping timeout: 260 seconds]
<ronthecookie>
I guess you could try doing `echo -1 >/sys/module/usbcore/parameters/autosuspend` on boot
<ronthecookie>
simpson& the goal was to include detail but it seems that has the side effect of hanging the browser :D
<simpson>
ronthecookie: "Simply copying Linux2.6.mk to Linux3.0.mk fixes the problem," from that bug report. So perhaps you could change the patchPhase to perform a `cp`?
<pta2002>
uh
<ronthecookie>
simpson& yep, I just opened it in emacs to do that :P
<pta2002>
second issue that's blocking my update
<pta2002>
nvm i can just turn this off
<thistledown>
Well, I'm getting closer, but now it's failing on the very last bit of linking because it really, *really* wants clang for some unknowable reason. I've tried supplying it with llvm + clang, but this does not please it.
<ronthecookie>
thistledown& a wild guess but LIBCLANG_PATH?
<mastarija>
ronthecookie, thanks, disabling autosuspend seems to be working
<ronthecookie>
mastarija& no problem~!
<thistledown>
ronthecookie: yes.
<pennae>
oh elements. bumped glib locally in accordance to #122920, mass rebuild incoming
<thibm>
rotaerk: FYI I'm able to launch a firefox build from the latest nixpkgs with `callPackage ./firefox { gtk = gtk2; inherit (gnome2) libIDL; }` (that was easier that what I thought) (./firefox comes from the old nixpkgs)
<bbarker>
Hey there - I recently migrated a nix expression that worked fine in WSL and Linux to OS X, and am running into this issue: "collision between `/nix/store/lkkmjr02ybg6x9005ljpl4y8j0pxfziq-zulu16.28.11-ca-jdk-16.0.0/share/man' and `'"
<bbarker>
Not entirely sure how to read that error, does the second package just mean the current derivation?
<thistledown>
thibm: After much wailing and gnashing of teeth, I finally got it to build. I am not, however, sure how to actually *use* my local package.
<KarlJoad>
Is there a way to selectively disable tests run using pytestCheckHook? I am using `disabledTests = [ "test_quadrature" ]`, but `test/unit/test_quadrature.py` is still being run.
<hyperfekt>
bbarker: you're looking at an error thrown by buildEnv, the second output means you've got an empty symlink in the trees you're trying to merge, i think
jjakob has quit [Read error: Connection reset by peer]
gvmnm has quit [Client Quit]
respawn has joined #nixos
<bbarker>
hyperfekt, ah, ok, so I do see that `/nix/store/lkkmjr02ybg6x9005ljpl4y8j0pxfziq-zulu16.28.11-ca-jdk-16.0.0/share/man -> zulu-16.jdk/Contents/Home/man`, which I suppose may be an issue in OS X. I'm not directly depending on zulu, so I need to find out what is causing it to be installed into the /nix/store
<mjrosenb>
(reposting from last night, since it was a low-traffic time) I'm running on macos. I recently updated it, and am now getting: dyld: Library not loaded: /usr/lib/system/libsystem_network.dylib
johnny101 has quit [Ping timeout: 240 seconds]
<mjrosenb>
I assume there's something simple that I need to do to get the libraries set up correctly again.
hiro98 has quit [Read error: Connection reset by peer]
hiro99 is now known as hiro98
hiro98 has quit [Client Quit]
zupo has quit [Ping timeout: 260 seconds]
<SumnerEvans[m]>
PSA: any users from the Matrix side who are using the Spaces beta can now join #nix-space:matrix.org which has all of the Nix-related rooms.
johnny101 has quit [Ping timeout: 250 seconds]
bvdw has quit [Remote host closed the connection]
johnny101 has joined #nixos
bvdw has joined #nixos
ddellaco_ has joined #nixos
<hexa->
SumnerEvans[m]: what are the specific version requirements when selfhosting?
bbarker has quit [Remote host closed the connection]
<hexa->
i have element-web 1.7.27 and matrix-synapse 1.33.2
ddellacosta has joined #nixos
<hexa->
and what you mentioned just appears as a room to me, even after enabling spaces in labs
<pennae>
apparently you need to enable it on the server side too, and that's new in 1.34
<SumnerEvans[m]>
hexa-: you need Synapse 1.34 and Element 1.7.28. The reason why it shows up as a room is because the Element version you are using expects the unstable prefixes.
<thibm>
thistledown: congrats! With home-manager, you need to put your package in home.packages. You can directly use `(callPackage ./path/to/drop.nix {})` or add your package to the package set using an overlay (and the same syntax) and just use `drop` in home.pacakges
<maralorn>
hexa-: I don’t know. I am pretty certain it’s the next thing after spaces. Last time I heard about threads they said spaces will land end of 2020 …
<maralorn>
hexa-: So I would say probably in about 9 to 12 months?
<maralorn>
andi-: I think that’s a bit orthogonal to the main features developed by Element. I think conduit and dendrite are both making good progress. Maybe within one or two years?
<jasom>
What's the best way to get an ubuntu-like setup for testing a python package? A package I use (Nikola) fails tests in nixpkgs trunk, and I want to see if the behavior in something upstream will look at is the same or different; my intuition is that it's just a bad test...
<mjrosenb>
ok, I want to reinstall nix to fix these errors, how do I get a list of what I have installed so I can easily reinstall it with the new nix?
<thistledown>
thibm: Weird, I did that and it's giving me "error: stack overflow (possible infinite recursion)"
<{^_^}>
[nixpkgs] @SuperSandro2000 merged pull request #123086 → hplipWithPlugin: correct arch name for the plugin file on aarch64 → https://github.com/NixOS/nixpkgs/pull/123086
dev_mohe has joined #nixos
thistledown has joined #nixos
johann9 has joined #nixos
thistledown31 has joined #nixos
thistledown31 has quit [Client Quit]
johann9 has quit [Quit: Connection closed]
johann9 has joined #nixos
<mjrosenb>
... really?
<mjrosenb>
error: could not set permissions on '/nix/var/nix/profiles/per-user' to 755: Operation not permitted
<mjrosenb>
(on nix-env -q)
<__monty__>
Something got messed up.
respawn has quit [Quit: Leaving]
bennofs__ has quit [Read error: Connection reset by peer]
bennofs_ has joined #nixos
ddellaco_ has quit [Remote host closed the connection]
ddellaco_ has joined #nixos
ddellaco_ has quit [Remote host closed the connection]
<mjrosenb>
drwxr-xr-x 4 root wheel 128 May 31 2018 /nix/var/nix/profiles/per-user
<mjrosenb>
so, it is already 755
<mjrosenb>
I've heard this can be caused by nix-daemon not running on macos.
<__monty__>
That sounds like a possibility.
<Kinnison>
gchristensen: Just because I love to sound like a broken record - https://hydra.nixos.org/build/143010346 <-- read errors on usb boot VM thingy again :D
<__monty__>
Note that if you didn't install things using nix-env you should have a configuration.nix that already lists everything you installed.
<johann9>
I noticed this morning that the latest versions of nixpkgs and home-manager are causing a segfault with `home-manager switch`. I was able to work around it by `nix-channel --rollback` a few times. Is there an easy way to check what version of those channels this generation is using and pin them?
vk1 has quit [Ping timeout: 240 seconds]
<gchristensen>
johann9: nix-info -m should tell you the channel info
<gchristensen>
Kinnison: interesting ... I'm suspicious of something else causing this at this point
<Kinnison>
gchristensen: the failure on the 20.09 branch looks suspicious too
<pennae>
gchristensen: and fwiw the test runs totally fine on our (amd) boxes
<johann9>
gchristensen: Just tried installing that tool and ran into the issue again. Maybe I should do a clean install of nix on my mac and just pin it immediately to a stable version...
<__monty__>
mjrosenb: If you don't know where configuration.nix is you probably don't have one. How have you installed things using nix before? I assumed you used nix-env -i because of your question. If you haven't been using that then I don't see how you could not have a list of the packages you installed yet.
<thistledown>
Everything else works, but home-manager somehow got bricked.
<thistledown>
It's doing it without the new package line now, too.
<thibm>
ah. That was my next question :)
<thibm>
did you put the package in an overlay?
<thistledown>
No, I just did a local package because I didn't really want to spend another hour learning how to do overlays.
ManiacOfMadness has joined #nixos
<thistledown>
Maybe eventually, but I just don't have enough mental energy, lol
<matthewcroughan>
What is the word Nix gives to the fact that you can't abstract packages and put them into groups?
qbit has joined #nixos
<matthewcroughan>
Or can you in fact do that, is it just a stylistic thing in Nixpkgs?
<matthewcroughan>
The fact that I can't use python3-dateutil in place of being explicit and saying that I want python3-six
<hyperfekt>
ronthecookie: current problem is that the xlibs.patch isn't applied for some reason
<thibm>
thistledown: yeah, sure. I was just looking for reason for the infinite recursion. So I think it has nothing to do with your package. (I have to go, good luck!)
<matthewcroughan>
python3-datetutil depends on python3-six. So if I included python3-dateutil in something, I'd basically be being very messy and including unnecessary stuff.
<thistledown>
thibm: Thanks for the help, see you later!
<matthewcroughan>
And also obscuring the fact that whatever I'm building really needed python3-six. I do not see that happening in Nixpkgs. Why? Is it style?
<qyliss>
how would putting them in groups help with that?
<matthewcroughan>
Or is there some principle of nix that prevents/halts that behavior?
<matthewcroughan>
qyliss: It wouldn't. I am saying it's bad. I just found a bad case of it in Yocto/OE.
<matthewcroughan>
And I'm curious what the lingo is to describe this behavior.
<mjrosenb>
so, I think python is being pulled in as a dependence of emacs
<phalrax>
just updated to the latest nix-unstable-small channel and now lorri/direnv isn't working anymore: _direnv_hook:2: no such file or directory: /nix/store/m8jlpa7am9jkwpcmms2anb2q6bsfry5n-direnv-2.28.0/bin/direnv. any ideas what might be causing this?
<colemickens>
nix-build -A packages.x86_64-linux.kata-containers should build what I'm building
<colemickens>
I'm currently just trying to even build src/agent of kata-containers, which is their rust agent. Preferably with `make` too, unfortunately.
Guest68412 is now known as Miyu-saki
Miyu-saki has quit [Changing host]
Miyu-saki has joined #nixos
noudle has quit []
<qyliss>
colemickens: what am I nix-build ing here?
<qyliss>
oh sorry, you said already
<colemickens>
src/kata/default.nix is the actual source :)
<cole-h>
colemickens: What if you try `source/src/agent`?
<colemickens>
er $checkout/pkgs/kata/default.nix, rather
<cole-h>
I just know in the past I've had to specify `source/` for sourceRoot on some things
<qyliss>
that was my instinct too but wanted to test it first
<cole-h>
it's really weird. some phase probably does `pushd source` / `cd source`, but it happens really early or something
<qyliss>
cole-h: unpackPhase
<qyliss>
if you don't have a sourceRoot set
<cole-h>
ah, ok
<qyliss>
it'll check that only one directory exists, and then cd to it
<marnov>
hello, i am planning my nixos test server, with / on ssh, and /var on 2 500G hdd in raid 1. do i have to use zfs for /var ?
<qyliss>
so it doesn't have to be called "source"
zupo_ has quit [Ping timeout: 265 seconds]
<qyliss>
marnov: you don't have to use zfs at all
TK__ has quit [Remote host closed the connection]
<hyperfekt>
the 'oops i wanted to pass an argument to the builder but it didn't actually get picked up' seems like something that needs throwing away the stdenv magic build scripts to fix
<marnov>
qyliss: yeah, thats what i think about :-)
<marnov>
does it worth ..
<hyperfekt>
they were a good way to get started with nix but are arguably holding us back
<Herdinger>
Hi, I'm a recent (2 days ago :)) user of NixOS and currently trying to use flakes as my NixOS config. How can I point the classic nix tools (nix-shell etc.) to use my /etc/flake.nix modules and overlays instead of nixpath?
<immae>
Herdinger: I think you’re mixing some things, but it should be sufficient to put it as /etc/nixos/flake.nix (note the /nixos/ in the middle) to have it picked automatically by nixos-rebuild
ddellac__ has joined #nixos
<Herdinger>
immae: Yes, that part is working for me without a hitch.
<Herdinger>
immae: The issue is that nix-shell does not use the overlays defined in there, and also not the additional modules.
<immae>
ok then you’re not mixing things :)
<immae>
To partially answer your question, you can set nixpkgs-overlays in the NIX_PATH, but I don’t know how to specify a flake overlays output there
ddellac__ has quit [Ping timeout: 246 seconds]
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
jimmiehansson has quit [Remote host closed the connection]
<Herdinger>
Thats how far I got :P
<__monty__>
Flakes is still experimental so you have to find the right people.
<Baughn>
Is there a way to specify a timeout on a package build?
<Baughn>
Something more reliable than doing it inside the build script, with `bin/timeout` and the semi-broken shell code that seems to be all I ever manage?
pushqrdx has quit [Remote host closed the connection]
BenjiProd has joined #nixos
pushqrdx has joined #nixos
<colemickens>
okay, I got kata-agent built by just invoking `make` manually to generate a file. the runtime is go and more involved, trying buildGoModule and make, but not a lot of hits for that in nixpkgs either...
bbarker has quit [Remote host closed the connection]
<mjrosenb>
ok, does error: build of '/nix/store/rkmipzrzahk6xmhxa4b3lhln4b2wayh2-adoptopenjdk-hotspot-bin-16.0.0.drv', '/nix/store/rrl6r0r4yrxmpk5w7g8r4jpad2l9g7bs-python3-3.10.0a5.drv', '/nix/store/x000yyg6kk06nv8rf9fk9w6kll40g4yx-emacs-mac-with-packages-27.2-8.2.drv' failed give some sort of dependency information, or is it just "these three expressions failed to build"?
iMatejC has joined #nixos
fuiltilt has joined #nixos
iMatejC has quit [Client Quit]
bbarker has joined #nixos
bbarker has quit [Remote host closed the connection]
<Baughn>
fuzzy_id: Not reliably. The point of nixos-stable is that software versions don't change; if you're ok with running newer software, you should be using nixos-unstable.
<Baughn>
fuzzy_id: Most services provide a 'package' setting you can override (to a docker imported from nixos-unstable, say), but there's no guarantee that the rest of the module will work with that.
zupo has joined #nixos
bbarker has quit [Remote host closed the connection]
<fuzzy_id>
it'll also compile all dependencies of docker from scratch, right?
<Baughn>
No, it'll pull those from the nixos-unstable binary cache.
<fuzzy_id>
ah, great.
<Baughn>
Really though, you're less likely to run into issues if you switch the entire system to -unstable.
<jonringer>
it will only build what's not available in the cache. If you're not using overrides or an overlay. You should always get cache hits (Assuming the package is working to begin with)
<Baughn>
"Unstable" here doesn't mean "Will have bugs and crash". It means "Will have rapidly changing software versions".
<fuzzy_id>
Ah ok.
<Baughn>
(*Sometimes* it means the former, but IME nixos-unstable is significantly stabler than e.g. Arch, mostly thanks to the test infrastructure.)
<fuzzy_id>
Yeah, I might give it a try. I can always roll back. :D
lsix has quit [Ping timeout: 260 seconds]
<jonringer>
yea, stable/unstable refers to versioning. Not usability
<Baughn>
fuzzy_id: You may have to blow away /var if you want to roll back.
<Baughn>
Lots of software gets Very Unhappy about that sort of thing.
<jonringer>
oh, I'm well aware *glares at python ecosystem*
<hyperfekt>
another alternative would be to use a newer module in combination with the package from unstable. but since there aren't really overlays for nixos you'll have to create your own tree for that
<hyperfekt>
if you wanted to stay on stable nixos, that is
<Baughn>
Yeah. But someone who explicitly wants a newer Docker, probably isn't the target audience for -stable.
orivej has joined #nixos
<hyperfekt>
hm actually it might be doable without a changed tree by using import and disabledModules 🤔
<cole-h>
I think we're having a near-miss in communication x)
<CodeBitCookie[m]>
I put it in quotations, you said that
<CodeBitCookie[m]>
<cole-h "I think we're having a near-miss"> yeah
<Herdinger>
Hi, I had a disconnect earlier but that gave me some time to ponder on my question :). Is there any way to let the nix (flake) commans resolve the same as nixos-rebuild? I basically do not want to resolve against nixpkgs but this custom package set.
<cole-h>
Ah. I meant: if you have a project that has multiple packages (say, one package provides a service for something, and the other provides a CLI to interface with that service), you'd want them both under `packages.<platform>`
<hyperfekt>
ronthecookie: probably an issue with fontconfig. next step would be to build with debug info, but i think i'll leave it here. no idea how hard it actually is to get it to run in a modern environment. but it does build which is already a pretty nice achievement
<cole-h>
And then you would set `defaultPackage.<platform>` to whatever you want (most likely the CLI, since that will probably be run more often)
<{^_^}>
[nixpkgs] @Ma27 pushed commit from @TredwellGit to release-20.09 « element: 1.7.27 -> 1.7.28 »: https://git.io/JsElg
<ronthecookie>
hyperfekt& I'm writing the wiki article now
<ronthecookie>
fontconfig is evil. i do not want to see another fontconfig issue ever again.
<hyperfekt>
about getting old stuff to run?
<ronthecookie>
"Building anicent nixpkgs", I promised thibm that much :P
<CodeBitCookie[m]>
<cole-h "And then you would set `defaultP"> Ah I think I understand now
<CodeBitCookie[m]>
so something like this:
lunik1 has joined #nixos
<hyperfekt>
one helpful hint to include may be that since nixpkgs didn't have post and pre hooks back then, you can use "unset somePhase; somePhase" to run the original one from your changed one
<cole-h>
CodeBitCookie[m]: That's not quite right. `packages.<platform>` is an attrset, so it should be `packages.<platform> = { hello = legPkgs.hello; cowsay = legPkgs.cowsay; };`
<CodeBitCookie[m]>
ohhh yeah yeah mixing two languages
<cole-h>
And also `defaultPackage` requires you to pick one of them -- so you'd do `defaultPackage.<platform> = self.packages.<platform>.hello;` (or `.cowsay`)
<atralheaven>
Hi, I'm having a problem enabling vscode extentions, all specified extentions are installed, but only some can be enabled. here is the code: http://p.ip.fi/wxzx
<CodeBitCookie[m]>
<cole-h "And also `defaultPackage` requir"> why can't nix-build just build everything
<CodeBitCookie[m]>
shouldn't it do that
<atralheaven>
trying extensionsFromVscodeMarketplace didn't help
<cole-h>
`nix build` != `nix-build`
<CodeBitCookie[m]>
still though
<CodeBitCookie[m]>
I know that one is intrduced with flakes (space) and one is always there
<CodeBitCookie[m]>
but I guess some other difference that I am missing out on
<cole-h>
Unfortunately, that's not how `nix build` works. AFAICT, it can only build one at a time.
<mikroskeem>
how would i get wasm32-unknown-unknown rust toolchain up and running next to my default one?
<CodeBitCookie[m]>
Im so lost
<CodeBitCookie[m]>
So you are saying it doesn't even build your poject
<mikroskeem>
i need to cross compile few wasm blobs, but i wasn't able to figure out how to do so
<cole-h>
CodeBitCookie[m]: No?
<{^_^}>
[nixos-homepage] @garbas pushed to master « disable (temporarily) updating of nix-dev flake input »: https://git.io/JsEBC
<cole-h>
I just said you can't use `nix build` to build every single attribute *at the same time*
<CodeBitCookie[m]>
and it just builds one dependency at a time
<cole-h>
Nonono
<cole-h>
You can only specify it one *package* to build
<CodeBitCookie[m]>
then isn't that just one package
<CodeBitCookie[m]>
wait what
<cole-h>
`nix build .#package` and `nix build .#package2`, but not `nix build .#packages.<platform>` to build all of `packages.<platform>`s packages
<CodeBitCookie[m]>
what about
<mjrosenb>
ok, nix-env -u didn't fix the dyld problems on MacOS. Any idea what needs to be done to get binaries that link against the more modern dylds?
<Herdinger>
Not meaning to be rude but my question got a little drowned out so I'm going to ask again. I'm configuring by NixOs through flakes. There are overlays for the nix command it self and a few custom packages. Is there a way for me to rexport the overlayed nixpkgs that get used by nixos-rebuild as something that can be used by other nix commands?
<{^_^}>
[nixpkgs] @sternenseemann pushed to haskell-updates « haskellPackages.taglib: add zlib to librarySystemDepends »: https://git.io/JsEED
<noobstar>
hi! I'm again having trouble getting a project to build with glib, this time using meson+ninja. meson's setup executes fine, but when building, it can't find `gio/gdesktopappinfo.h`. I believe this is part of glib, which is in buildInputs, and I do have pkg-config this time. does anyone have any suggestions?
<judson_>
I'm trying to set up a custom boot image. I've been able to follow the manual accurately, but I'd like to have a parallel repo for my custom stuff. It'd be possible to fork nixpkgs, I suppose, but I'd rather reference them. My first attempt, setting Nix path entries didn't work as I'd hoped.
<{^_^}>
[nixpkgs] @sternenseemann opened pull request #123419 → gcc*: force building a cross compiler if host and target platform differ → https://github.com/NixOS/nixpkgs/pull/123419
<{^_^}>
[nixpkgs] @Ericson2314 merged pull request #123419 → gcc*: force building a cross compiler if host and target platform differ → https://github.com/NixOS/nixpkgs/pull/123419
ddellacosta has quit [Ping timeout: 265 seconds]
Reiko2 has joined #nixos
supersandro2000 has quit [Killed (orwell.freenode.net (Nickname regained by services))]
supersandro2000 has joined #nixos
ddellacosta has joined #nixos
stree has quit [Ping timeout: 260 seconds]
Reiko2 has quit [Read error: Connection reset by peer]
<{^_^}>
[nixpkgs] @Gabriel439 pushed 26 commits to gabriel/specialArgs_patch: https://git.io/JsEil
<ToxicFrog>
Note that old versions do eventually stop getting updates, so if you're on like nixos 19.x or something you'll need to switch to a channel that still gets updated