<ToxicFrog>
When in the boot process do the activation scripts run?
vmandela has joined #nixos
<ToxicFrog>
I have an activation script that doesn't seem to have any effect, and my best guess is that it's running too early, before /etc exists...but I thought /etc was created before those scripts run.
<ToxicFrog>
Actually, probably a better question: if I've defined environment.etc.foo, is there any way to refer to the generated file by path in an activationScript?
Rusty1 has joined #nixos
<gchristensen>
typically when people say they have written an activation script, I ask what it does and why -- because usually activation scripts are the wrong solution. /etc is set up as part of the activation script
<ToxicFrog>
Ok, this is probably the wrong place for it
<gchristensen>
what are you wanting to do?
<ToxicFrog>
It's an lmsensors initialization command (loads sensor thresholds into lmsensors, based on /etc/sensors3.conf, which is derived from the sensors3.conf that ships with lmsensors + a bunch of system-specific configuration)
<ToxicFrog>
It needs to run every time that configuration changes and every boot
<gchristensen>
gotcha
<gchristensen>
I would be inclined to make that a systemd service wanted-by something early
<ToxicFrog>
That said, if I can only pick one, running on boot is more important, because I'm more likely to remember to run it manually after a nixos-rebuild than after a reboot.
<gchristensen>
a Type=oneshot RemainAfterExit=true service should do the trick, and it'll automatically get restarted after it changes
<ToxicFrog>
The service definition wouldn't change, I think, only the contents of the file
<ToxicFrog>
So I'd want `restartTriggers = [environment.etc."sensors3.conf"];` or similar
<gchristensen>
if you're writing /etc/sensors3.conf with Nix, you can use the path of the sensors config in the nix store, as partof the definition
andreas303 has quit [Remote host closed the connection]
nst^ has quit []
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/fjH1p
srt^ has joined #nixos
andreas303 has joined #nixos
xcthulhu has joined #nixos
seanparsons has quit [Ping timeout: 246 seconds]
polman has joined #nixos
<vmandela>
I am trying to convert a binary installer into a nixpkg. The installer is linked against the system libc. How do I make it run during the nixpkg installation phases ?
seanparsons has joined #nixos
<jackdk>
can you patchelf it before you try to run it?
polman has quit [Ping timeout: 245 seconds]
<lambda-11235>
Can meson execute external files on build? Would there be something that prevents that?
joebobjoe has joined #nixos
Chiliparrot has quit [Ping timeout: 264 seconds]
___laika has joined #nixos
<vmandela>
jackdk, I get a open failed when I try to do a patchelf.
<vmandela>
looking into the autoPatchelfHook now.
<ashkitten>
does anyone use nut for their ups? i can't figure out how to get it to automatically shut down at critical
<ashkitten>
er, automatically shut down my computer
<DigitalKiwi>
between my printer and not being cut perfect and dealing with the adhesive and them being so close together in the design i'm glad a few of them turned out as ok as they did heh
ericsagnes has quit [Ping timeout: 264 seconds]
polman has joined #nixos
<gchristensen>
they look great :)
<{^_^}>
[nixpkgs] @MP2E pushed commit from @marius851000 to master « dolphinEmuMaster: update to the latest upstream version »: https://git.io/fjHMn
<DigitalKiwi>
thank you :D
<DigitalKiwi>
RIP my notifs
jb55 has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @MP2E closed pull request #65863 → dolphinEmuMaster: update to the latest upstream version → https://git.io/fjHTM
<danderson>
So, the NixOS manual has a warning about NixOS containers and their imperfect isolation from the host system. Are there details somewhere about what this means precisely?
<danderson>
specifically, how does this compare to docker containers and systemd containers
<gchristensen>
danderson: they're pretty much the same
<danderson>
okay, so the warning is "generic warning about containers not being VMs" ?
ericsagnes has joined #nixos
<gchristensen>
mstly
<danderson>
okay, got it. That I can live with :)
<danderson>
I was worried it was something more like "because Reasons, turns out /proc/kmem is world-readable in our containers"
mexisme_ has joined #nixos
<gchristensen>
do take a look, though, because sometimes nixos containercan access the whole store of the host, which you may not want
<gchristensen>
that would be thrilling :D
<danderson>
sometimes? Under what conditions? I'm not seeing that in the manual
<gchristensen>
sorry I don't actually fully remember the details
<gchristensen>
so that might have been wrong
<danderson>
not that it should matter, I _think_ ? Presumably the container can't then go and change the nix store...
<gchristensen>
definitely couldn't change the store
<danderson>
I'll create a container and poke around
<gchristensen>
but could add new paths to the store of course
mexisme_ has quit [Ping timeout: 245 seconds]
<danderson>
hmm. So, poking around a tiny bit, /nix/store is a read-only bind mount from the host nix store. So yes, seems like the container can see anything in the host's store.
<gchristensen>
it may not be, it might have a limited subset
<danderson>
the nix daemon socket is also bind-mounted in, which presumably is how the container and its users can get their stuff build
<gchristensen>
right
<danderson>
poking around the nix store, I see a directory for zfs-user, which is installed on the host but not in the container
<danderson>
so that suggests a shared view
<ashkitten>
i got nut working!
<gchristensen>
cool
<gchristensen>
danderson: I'm not surprised by that, fwiw, just didn't remember
ToxicFrog has joined #nixos
MmeQuignon has quit [Ping timeout: 245 seconds]
<drakonis>
ashkitten: that's nuts
<ashkitten>
thanks
<drakonis>
what exactly is nut tho
kp__ has quit [Quit: WeeChat 2.5]
<drakonis>
i just wanted to make a funny joke
<danderson>
gchristensen: yup. Just to confirm, I dug through the container startup scripts, and /nix/store is a straight read-only bind mount, no funny business
liberiga has joined #nixos
<danderson>
hm. That does mean it's not safe to use hashedPassword in my configuration though, because presumably that ends up burned into the store somewhere
<danderson>
... is that right? That might just be wrong.
<gchristensen>
it does :)
<danderson>
darn. Hm.
mexisme_ has joined #nixos
<gchristensen>
although you might decide your hashed password is "safe enough"
<gchristensen>
up to you
<danderson>
yup. I'm sitting with it and deciding what my threat model is
joebobjoe has quit [Ping timeout: 244 seconds]
<danderson>
mostly it's all about least privileges as far as I can push it, containers were just a handy way to get some of that
<danderson>
but maybe I'll tweak systemd units instead, and use all the mechanisms available there...
<CMCDragonkai>
I am using dockerBuildTools to build a container, this container has an executable that requires `/bin/sh`, what dependency do I need in nixpkgs to wrap the executable for this to work?
<CMCDragonkai>
I tried `bash` and `coreutils` but both don't provide `/bin/sh`
<CMCDragonkai>
Note that this is a python program that uses `subprocess.call` with `shell=True` which means its' not simply a string replacement
ddellacosta has joined #nixos
<CMCDragonkai>
maybe create a symbolic link from /bin/sh to the relevant shell?
<fasd>
Is nix's melpa mirror updated often? nix-env -f "<nixpkgs>" -qaP -A emacsPackagesNg.melpaPackages shows nothing released after 01/03/2019
Edes has quit [Ping timeout: 268 seconds]
<DigitalKiwi>
yay got it
<etu>
fasd: There's been changes in how they are updated very recently
<etu>
fasd: It's probably not in any channel yet, we're talking like last couple of days
Diagon has quit [Quit: Leaving]
rauno has joined #nixos
<fasd>
nix-haskell-mod isn't in any repositories atm. would this change fix it?
<etu>
fasd: There's also the newly started #nixos-emacs :)
<fasd>
nice. thanks
hyper_ch2 has joined #nixos
<etu>
fasd: short version: Future updates of melpa is much faster to run and will probably happen more frequently. Before it took hours to run the updater and you could get many failures. And the updater have recently been rewritten :)
ris has quit [Ping timeout: 245 seconds]
thc202 has joined #nixos
<Athas>
How do I make libclang available do a derivation? 'clang' seems to just be the CLI program.
<clever>
Athas: are you using clangStdenv.mkDerivation?
<Athas>
clever: no, I'm just listing clang in my buildInputs. How do I use what you are suggesting?
<clever>
Athas: instead of stdenv.mkDerivation, use clangStdenv.mkDerivation
lsix has joined #nixos
<Athas>
Thanks, that works. What does that do?
<clever>
provides clang by default, and makes sure clang actually works
<Athas>
It also worked to use stdenv.mkDerivation and llvmPackages.clang-unwrapped.
<Athas>
(Well, "worked" as in cmake finished without errors.)
wfranzini has joined #nixos
<clever>
there is also pkgs.clang.cc, which has some clang libraries, but no libclang.so
<clever>
clang.cc also has cmake files in its lib
Jackneill has joined #nixos
MinceR_ is now known as MinceR
hyper_ch2 has quit [Remote host closed the connection]
wfranzini has quit [Remote host closed the connection]
Jackneill has quit [Remote host closed the connection]
Jackneill has joined #nixos
ottidmes has joined #nixos
Jackneill has quit [Remote host closed the connection]
<Athas>
I'm trying to write a derivation for pocl. It's fundamentally a fairly well-behaved cmake user, but during building, I get a really weird error:
<Athas>
/nix/store/ih3v6znmawlxkic6r3rzj05dvxsaqf2h-clang-7.1.0/lib/clang/7.1.0/include/mm_malloc.h:27:10: fatal error: 'stdlib.h' file not found
mexisme_ has quit [Ping timeout: 246 seconds]
<Athas>
It manages to build 70% of the object files according to cmake, so it's not like the C compiler setup can be completely broken.
<Athas>
Well okay, 30%, but stdlib.h is so fundamental that I'm a bit surprised it can fail there, but not elsewhere.
<Mic92>
Izorkin: Two versions on the same node? The way I read this, is that you install the new galera version and then just restart it. But proof me wrong.
<Izorkin>
Mic92: Examle claster - 3 nodes on mariadb 10.3 + galera25. On the first node we update to mariadb 10.4 + galera26. Then update on the next node.
<sphalerite>
das_j: but yeah I don't think reflinking works on zfs
<das_j>
I still use it on nixos on my deskto
<das_j>
p
<das_j>
but I plan on switching to XFS as well
<das_j>
some day (TM)
<das_j>
sphalerite: Even better than the reflinking would be if the nix daemon used an overlayfs to combine the source derivations
<das_j>
but that's probably too linux-specific
jmiven has joined #nixos
<yorick>
sphalerite: yeah, sad. someone should make that
<hyper_ch2>
what's reflinking?
veske has joined #nixos
<yorick>
hyper_ch2: on copy-on-write filesystems, just making another reference to the current file. this costs zero time or space, and then you can modify the files separately
<Izorkin>
Mic92: no idea to corrcet check with nix-review?
<pingiun>
I get an error somewhere in nixpkgs when I import from it: error: value is a string while a set was expected, at /nix/store/glzx9gah7xgfwkm2bn4ia5p6c49kf1ry-nixpkgs-19.09pre187222.002b853782e/nixpkgs/pkgs/development/interpreters/python/cpython/2.7/default.nix:177:30
<pingiun>
although I'm not using the python package
<DariusTheMede>
Does anyone know why having a revision for a package would make nix decide to overwrite my explicit instructions to download the package from github?
<DariusTheMede>
clever: I believe I am on 19.03? (nix-shell -I nixpkgs=https://github.com/NixOS/nixpkgs-channels/archive/nixos-19.03.tar.gz --command "jupyter lab")
<DariusTheMede>
So you are right; it is ignoring 19.03
<DariusTheMede>
clever: I am not clear what you mean by hspkgs: do you mean replace pkgs.haskellPackages.inline-r.overrideAttrs byt pkgs.haskellPackages.hspkgs.inline-r.overrideAttrs?
<hyper_ch2>
hmmmm, still having issues with the new freenode webirc client.... need to check if another one is packed in nixo
<kaliumxyz>
Ah great, thanks symphorien
<symphorien>
kaliumxyz: also you have to disable the sandbox
nschoe has joined #nixos
ddellacosta has joined #nixos
<tilpner>
hyper_ch2: See <nixpkgs/pkgs/applications/networking/irc>
<{^_^}>
[nixpkgs] @domenkozar pushed commit from @mkaito to release-19.03 « haskellPackages.servant-client-core: apply patch only to 0.15 »: https://git.io/fjHbl
<hyper_ch2>
tilpner: webirc
<hyper_ch2>
I found shout but no idea how to get it to work
<tilpner>
hyper_ch2: If you some time, I hear glowing bear is good
<{^_^}>
[nixpkgs] @WilliButz opened pull request #66073 → nixos/unifi: create data directory with correct permissions → https://git.io/fjHb8
<tilpner>
*have some
<hyper_ch2>
tilpner: hmmm, only service I found is shout but can't get that to work
* tilpner
has not used shout
<hyper_ch2>
looking at glowing bear now
linuxdaemon has quit [Read error: Connection reset by peer]
<elux>
i tried to upgrade, and flatpak now errors/asserts: "To use Flatpak you must enable XDG Desktop Portals with xdg.portal.enable.". but what is xdg.portal anyways?
<simpson>
https://flatpak.github.io/xdg-desktop-portal/portal-docs.html#common-conventions "The general flow of the portal API is that the application makes a portal request, the portal replies to that method call with a handle (i.e. object path) to Request object that corresponds to the request. The object is exported on the bus and stays alive as long as the user interaction lasts. When the user interaction is over,
<simpson>
the portal sends a Response signal back to the application with the results from the interaction, if any."
<simpson>
Yet another reinvention of basic message-passing concepts.
<thoughtpolice>
elux: It's a mechanism to allow flatpak apps to talk to the outside system, by way of an exposed D-Bus service. xdg-desktop-portal is an application that creates those D-Bus services, so you need to enable it
halfbit has quit [Ping timeout: 252 seconds]
<thoughtpolice>
simpson: I mean, to be fair it's just using D-Bus, which is a workaround for poor IPC mechanisms in Linux, sure, but it's not like that's really new or anything, just working with what's there.
DariusTheMede has quit [Ping timeout: 268 seconds]
<thoughtpolice>
Guess it depends on where you draw the line.
<clever>
thoughtpolice: ive done something similar with nix-bundle, i called it the escape hatch
<simpson>
thoughtpolice: Oh, sure; D-Bus itself is a poor reinvention of concepts that would be native on other, higher-level systems. My gripe is mostly that people thought that writing C was an acceptable way to get an object system.
<thoughtpolice>
nix-bundle seems closer to Facebook's XAR but without the use of squashfs/fuse. I've wondered if something like that would be a good attempt for bundling Nix applications.
domogled has quit [Quit: domogled]
<clever>
thoughtpolice: main issue with nix-bundle, that squashfs would solve (but that needs root-ish), is the long startup time
<clever>
thoughtpolice: arx did recently add a cache thing, so it will reuse the previously unpacked stuff, but that introduces its own bugs
<clever>
if arx fails to unpack a thing, it wont clear the cache
<simpson>
thoughtpolice: nix-makeself looks pretty cool. I'll have to try it out.
<clever>
and next time you run it, it uses the partially unpacked directory, causing various problems
<thoughtpolice>
simpson: Be warned that the internals are an absolute goddamn nightmare, but it works pretty OK
<clever>
my solution to the speed problem (and to support updating dynamically), is to use nix-bundle to package an "installer"
<thoughtpolice>
For some reason I thought "yes it would be a good idea to just put the C and Nix code inside the Perl program"
<thoughtpolice>
Presumably I was blacked out at the time I wrote it
<thoughtpolice>
Not sure
<clever>
and that installer copies a nix closure to ~/.daedalus/nix/store, and sets up scripts to run nix-user-chroot for you
<clever>
so its entirely transparent, and ceases to be nix-bundle once installed
<clever>
and the chroot has a fully working nix, so nix profiles can be used
<thoughtpolice>
clever: Yeah, all of the current solutions all have sort of various overlapping flaws/deficencies. I think an approach like the one XAR uses would be the best in the long run, but that's just a theory.
<clever>
main problem i have with nix-user-chroot is that debian made the silly choice to disable namespacing without root
<thoughtpolice>
One thing I really like about nix-makeself for example is that the packages are just described using Nix itself which is pretty convenient, most of the others just take some list of store objects. I'd like to see that in the other tools
<simpson>
thoughtpolice: I have strong expectations that aren't being met. I started writing a D-Bus client from scratch, but kept crashing other applications by accident, and on a modern Linux distro, crashing things over D-Bus can hang the system.
<simpson>
If I wanted that experience, I'd go back to GPU drivers~
linarcx has joined #nixos
<clever>
thoughtpolice: nix-bundle is also fully nix based, i can just `nix-build -A something` and it spits out the arx bundle
<clever>
simpson: i started with kernel code that delt with allocating chunks of memory, letting you mmap them into userland, letting you issue a render action, and waiting for an irq
linarcx has quit [Read error: Connection reset by peer]
<clever>
simpson: then i went to the userland side, with zero idea how mesa and opengl where "supposed" to work, and just implemented every function the opengl app was using, lol
<simpson>
clever: Have you looked into Mesa's structure yet? The C obfuscates, but the overall structure is not hard to grasp. It's a state machine and input validator, neither of which are well-expressed in C.
<dminuoso>
Mmm, what's the most nixos friendly language/environment to do embedded development? I was playing with the thought of using rust to target esp8266, but just looking at how the environment needs to be set up I fear this might throw me into the deepest and darkest holes of nixos and I may never crawl out of them again.
orbekk has joined #nixos
<dminuoso>
(And Arduino IDE is something Im trying to avoid desperately)
<simpson>
clever: You'd be surprised how close this is to the earliest GL drivers!
<thoughtpolice>
dminuoso: It really just depends on what toolchain you want. It was not very hard, for instance, for me to get a cross-compiler based on GCC to target RISC-V emulators/softcores. Rust/ESP8266 will be a bit harder; for one you'll certainly have to use forks of LLVM for Rust. (There is no Xtensa backend in LLVM upstream right now*)
<FRidh>
how can they, with XAR/squashfuse, ensure that references to native libraries that are bundled in the squashs, actually point to those native libraries and not to things elsewhere on /
<simpson>
dminuoso: I wish that there were a singularly-good answer. I think that it depends on which target you're working with. pkgs/development/arduino is a little barren, but maybe that's an opportunity.
<thoughtpolice>
* But there will be at some point, since an XTensa backend was proposed/accepted recently.
<clever>
FRidh: is any chroot or namespace magic going on?
<FRidh>
I mean, we typically need to patch python code to hardcode paths to .so's. I can't imagine they would not have to do that
<FRidh>
clever: as far as I can tell not
<clever>
FRidh: perhaps they patchelf'd things to behave more like nixos's initrd?
nhey_ has quit [Remote host closed the connection]
<Miyu-chan>
Well. Eep. gitlab-runner returns the wrong runner-name on startup, but returns the right runner name on shutdown.
<Miyu-chan>
This is big sad TBH
<clever>
FRidh: or the nix bootstrap tools
<FRidh>
clever: still won't help when you use ctypes.util.find_library. I have the impression that what can work are extension modules, because theyre in hte right folder so you can import them fine
<clever>
FRidh: yeah, that kind of stuff needs a chroot
<FRidh>
clever: exactly
<FRidh>
so it's not that great then. At least not as powerful as nix-bundle
<FRidh>
or snap
veske has joined #nixos
iyzsong has quit [Ping timeout: 268 seconds]
DariusTh` has quit [Ping timeout: 258 seconds]
<FRidh>
ah, looking in the code I do see now they do a chroot
<thoughtpolice>
Yes I would be surprised if they didn't, considering the README claims it is designed to bundle e.g. native objects, too. Though it's unclear from their python infrastructure how they determine that. (Realistically Facebook probably has internal components that smooth that out anyway so it's "irrelevant" for their XAR usage, I bet)
<{^_^}>
[nixpkgs] @zimbatm merged pull request #64242 → kops: restructure package to maintain multiple versions in the future → https://git.io/fj6JW
<clever>
FRidh: i see mention of namespaces, the same stuff nix-user-chroot uses
<FRidh>
indeed
johnny101m has quit [Ping timeout: 248 seconds]
<clever>
it will likely have the same problems on debian, where a custom kernel patch disables namespacing for non-root users
<thoughtpolice>
The real benefit of XAR even if that wasn't true though for whatever reason is that it has much better compression due to squashfs compressing across the entire filesystem image, and that the FUSE mounts can be shared between invocations. Those are big advantages none of our tools do.
<clever>
and they added a /proc flag to re-enable it
<thoughtpolice>
Though it adds a lot more moving parts, admittedly...
<clever>
i think they are forced to use fuse, because you cant mount actual filesystems when using user namespacing
<clever>
your only allowed to `mount --bind`
<thoughtpolice>
(Well, I guess nix-makeself sort of works around the second deficiency since it only has to unpack once, but I consider that more of an alternative solution)
<clever>
because a true mounted FS would open the kernel up to buffer overflow exploits in the FS driver, along with things like setuid binaries
<clever>
but now you rely on both fuse support (which itself has some security problems) and user namespacing (which debian claims has security problems)
_kwstas has quit [Remote host closed the connection]
<thoughtpolice>
I mean, a lot of people make the second claim, somewhat reasonably (I think at one point Arch's secure kernel just bans user namespaces entirely). The kernel just wasn't designed under the assumptions user namespaces imply ("uid == 0 does not really mean root")
hyper_ch2 has quit [Remote host closed the connection]
<thoughtpolice>
At least, as far as security considerations go. I have no idea what the consensus on userns is today but I'm willing to bet the bugs aren't all fixed.
_kwstas has joined #nixos
devalot has joined #nixos
_kwstas has quit [Remote host closed the connection]
hyper_ch2 has joined #nixos
<thoughtpolice>
I seem to remember at one point RedHat did the same but I'm not sure; last I remember I thought nix-makeself worked fine there, and it definitely does need unprivileged userns support
<clever>
thoughtpolice: from what i heard, its more about adding a large chunk of un-audited code to the kernel
<clever>
thoughtpolice: and they just didnt want to deal with auditing it for problems, so they just disable it
<colemickens>
What does it mean when I do `nix-store -q --roots /nix/store/path` and it returns: {memory:16}
npmccallum has joined #nixos
<worldofpeace>
,locate krb5-config
<{^_^}>
Found in packages: krb5.dev, heimdal.dev, kerberos.dev
<clever>
colemickens: a currently running process refers to that path
<thoughtpolice>
Mic92: Yes, I'm going to coordinate with gchristensen or niksnut to get a repository setup there. In the mean time I'll probably put the VCL somewhere else, or perhaps make my own repo and we can just migrate it
hyper_ch2 has quit [Remote host closed the connection]
acmilanfan has joined #nixos
<gchristensen>
thoughtpolice: ideally the config would live in nixos-org-configurations
<Miyu-chan>
arianvp: Hey. What happens if you have 2 machines gitlab-runner register'ing the same --name?
<thoughtpolice>
gchristensen: It might be possible to do that, and I see there is some Terraform stuff there (which we do have a Fastly provider for) that might be worth linking into, but I mostly figured it would be a separate repo since the cache configuration is more or less wholly separate from the actual Nix-based fleet deployment. It could go either way I suppose.
<thoughtpolice>
One thing I was going to explore was using GitHub Actions to deploy the configuration, so actual contributions do not require manual intervention for us to deploy.
<thoughtpolice>
(I've actually found GH actions works quite well to automate Nix-based setups. Truthfully I don't know if/what kind of automation tools we endorse, so rolling a simple one of our own might be beneficial, I'm not sure)
<alj[m]>
How do I set maven to use java 11 (both installed from nixpkgs). It doesnt seem to recognize `JAVA_HOME`
<alj[m]>
and I cant find the declaration in nixpkgs so i dont know how to override
<alj[m]>
please help :)
<gchristensen>
thoughtpolice: integrating with the existing terraform configuration would be preferable
<gchristensen>
if not that, putting the config in that repo with instructions
<srhb>
alj[m]: Try less `which mvn` -- that'll answer why it doesn't recognize JAVA_HOME
<alj[m]>
srhb: It uses the java version it installed into the nix store. 1.8
<srhb>
alj[m]: So the build process creates a wrapper script that points JAVA_HOME at the directory of the jdk maven was initially built with. You'll want to override that jdk.
<gchristensen>
thoughtpolice: it is :)
acmilanfan has quit [Remote host closed the connection]
<alj[m]>
rather its the one installed by nix ~/.nix-profile/bin/mvn
<alj[m]>
srhb: How?
<worldofpeace>
,locate kadmin.local
<{^_^}>
Found in packages: krb5
<thoughtpolice>
gchristensen: Ugh, tarballs.nixos.org is served as a bucket with S3's "Static website" option, apparently?
<srhb>
alj[m]: Have you created an overlay before?
<srhb>
alj[m]: Your goal is to use maven.override to feed it a different jdk than the default one
<gchristensen>
oh actually tarball.nixos.org is pointing to cloudfront still.
<gchristensen>
thoughtpolice: is that a problem? the static website option?
<alj[m]>
I look into it thanks srhb How do I know which Value to override?
<thoughtpolice>
gchristensen: Well, so one of the really important features my config uses is called "Streaming Miss". Right now, streaming miss does not support TLS for the Origin. There is a "Limited Availability" program I'd like to enroll us in to change that. But the S3 static website option *requires* non-TLS
<srhb>
alj[m]: try `nix edit nixpkgs.maven`
<srhb>
alj[m]: You'll see that one of its arguments is "jdk"
<thoughtpolice>
Streaming Miss is extremely important for semi-large objects, because it means you get the bytes streamed to you from S3. If that's not enabled, then if an object isn't in the cache, the POP has to download the object entirely, before it can serve anything.
<gchristensen>
thoughtpolice: tarballs does low-traffic serving, I wouldn't worry about tarballs for now.
<srhb>
alj[m]: You want an overlay that sets that argument to jdk11
<thoughtpolice>
That is why you sometimes see 'nix build' for instance, seemingly "waiting" on something to start downloading for multiple seconds.
<thoughtpolice>
It's waiting for the POP to fully download the object before it can even read the first response byte.
<alj[m]>
srhb: Okay I'll try! Thank you
proprobe has joined #nixos
<gchristensen>
well we should do streaming for cache.nixos.org for sure
<gchristensen>
but tarballs.nixos.org is less interesting for now
<thoughtpolice>
gchristensen: Another thing I'd like is information about what objects are in S3, i.e. the list of all file extensions, etc. I remember you said you had a copy of the metadata around; not sure if you can query it, but it'd be nice.
<proprobe>
hello everyone. I've just installed nix on a manjaro kde distro. Everything works good except that Ive gotten a problem with a package called `yarn2nix`
<thoughtpolice>
(Or, well, IDK if the metadata actually contains that info. I assume it does)
<proprobe>
running nix-shell in my project gives the following:
<proprobe>
$ nix-shell --show-trace
justanotheruser has quit [Quit: WeeChat 2.4]
<proprobe>
error: while evaluating the attribute 'buildInputs' of the derivation 'nix-shell' at /nix/store/gx5ds72p3kgdalyj1hnl73gp01dw2lfh-nixos-d98a165de78027202b8b12521bae472a6a47741d/pkgs/build-support/mkshell/default.nix:28:3:
<proprobe>
while evaluating 'getOutput' at /nix/store/gx5ds72p3kgdalyj1hnl73gp01dw2lfh-nixos-d98a165de78027202b8b12521bae472a6a47741d/lib/attrsets.nix:464:23, called from undefined position:
<proprobe>
while evaluating anonymous function at /nix/store/gx5ds72p3kgdalyj1hnl73gp01dw2lfh-nixos-d98a165de78027202b8b12521bae472a6a47741d/pkgs/stdenv/generic/make-derivation.nix:142:17, called from undefined position:
<proprobe>
while evaluating the attribute 'yarn2nix' at /home/proprobe/workspace/cd_concert_project/nix/yarn2nix.nix:21:10:
<proprobe>
linking '/nix/store/p4qhmjj57h3qx3i6hmapick63jis80wz-source.drv.chroot/nix/store/81ikflgpwzgjk8b5vmvg9gaw9mbkc86k-compress-man-pages.sh' to '/nix/store/81ikflgpwzgjk8b5vmvg9gaw9mbkc86k-compress-man-pages.sh': Operation not permitted
<gchristensen>
proprobe: please use a pastebin next time :)
<srhb>
proprobe: You'll want to use pastebins to avoid getting spam-kicked by bots in the future :)
<proprobe>
Ive tested other simpler projects and it works fine. this seems to be strictly related to the `yarn2nix` package above
<gchristensen>
thoughtpolice: we have .bz2, drv, ls, narinfo, txt, and xz files
<srhb>
proprobe: What filesystem is your store on?
zupo has joined #nixos
<alj[m]>
srhb: I created an overlay and it works within a nix-shell. How do I get it to work in standard shell?
laas has joined #nixos
<srhb>
alj[m]: The wiki page describes how to make the overlay be picked up by any nix tool :)
<proprobe>
srhb: its an encrypted ext4
<alj[m]>
ok
<thoughtpolice>
gchristensen: Ok, thanks. One more clarification: I assume .bz2 and .xz are only found under /nar/ paths for nars themselves? Is everything else under the root of the bucket?
<thoughtpolice>
(I know for a fact .ls and .narinfo's are in the root, for instance, but not .drv or .txt)
<gchristensen>
those questions are getting trickier to ask :P
* gchristensen
writes some awk
<thoughtpolice>
:P
<thoughtpolice>
gchristensen: Basically, I just need to know: what file extensions can occur under the top root of the bucket, and what can occur under /nar/
<gchristensen>
right
<thoughtpolice>
That way I can just outright deny anything else
<gchristensen>
also, I'll go ahead and update my metadata sync
<thoughtpolice>
gchristensen: Cool! And if you can dump some example paths so I can test them that would of course be excellent :)
<thoughtpolice>
But I can probably find some of them on my own
<gchristensen>
sounds good (btw, I'm running out of time to work on this this morning, but I'll get the sync running and come back to it this evening)
<proprobe>
srhb: it appears to be related to user_namespaces https://github.com/NixOS/nix/issues/2633. Not sure what the problem actually is but after enabling this, it seems to properly build now
<{^_^}>
nix#2633 (by x80486, 28 weeks ago, closed): Error: cloning builder process: Operation not permitted installing Nix 2.2.1 in (Arch) Linux
<alj[m]>
srhb: I really dont get how to make my overlay work... It worked when it was a packageOverride but it seems i dont want that, right? so now the overlay is written as in the wiki but i cant figure out how to get to `pkgs` to set `jdk = pkgs.jdk11`
<fresheyeball>
do I need to reboot to get it to start for real?
<fresheyeball>
or did switch do it already?
<srhb>
alj[m]: You'll want self.jdk11 :)
<srhb>
alj[m]: self is "pkgs" after the overlays apply.
<gchristensen>
thoughtpolice: the joy of syncing these cache stats files is it perpetually says 430 files remaining.
<srhb>
alj[m]: (super is pkgs before your current overlay)
<alj[m]>
srhb: oh.. okay... ill try
<srhb>
fresheyeball: I think you need to reread the instructions on the limits, but I don't think a reboot is necessary :)
<srhb>
alj[m]: In the end it'll look like this: self: super: { maven = super.maven.override { jdk = self.jdk11; }; }
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<alj[m]>
srhb: now it says unexpected end 😢
<srhb>
alj[m]: Count your brackets :)
<srhb>
Or mine. Whomever made the mistake ;-P
<alj[m]>
i have exactly 4
<srhb>
alj[m]: Pastebin please. :)
<alj[m]>
mine is like yours
knaun has joined #nixos
proprobe has left #nixos ["Konversation terminated!"]
<srhb>
alj[m]: Mine is correct, so your error must be elsewhere.
<fresheyeball>
srhb: what did I miss?
<knaun>
is bash installed by default? i can't find where i specified it's installation?
<srhb>
fresheyeball: The note on security.pam.loginLimits and systemd.extraConfig (not systemd.coredump.extraConfig)
<{^_^}>
[nixpkgs] @angristan opened pull request #66083 → coinomi: init at 1.0.11 → https://git.io/fjHx9
<colemickens>
The weird thing is, the path that nix-optimize complains, and is supposedly in use in memory was deleted (oops) and shouldn't be in use anywhere...
<colemickens>
I only have one blank user profile and one current system profile that shouldn't refer to that outdated nixpkgs path. I might need a reboot though.
<alj[m]>
every f**king day i learn something new about nix. and it works so well after i understand. i just love nix
<srhb>
alj[m]: You're welcome. Sorry for the roundabout approach, teaching to fish etc. :-P
<alj[m]>
srhb++
<{^_^}>
srhb's karma got increased to 60
<srhb>
And yes, it's great! :)
<alj[m]>
no problem at all!
<srhb>
fresheyeball: The core limit ... "limits the core file size"
<srhb>
fresheyeball: Like the systemd one you want to be permitted to grab ones big enough for your purpose
<srhb>
fresheyeball: ie. on a single user system, set it to something Large Enough™ with little other concerns
<fresheyeball>
it takes a set with domain type item and value
<fresheyeball>
are we saying domain is my user name
<fresheyeball>
type is ? hard soft?
<fresheyeball>
item is core
<fresheyeball>
and value is 1000000
<fresheyeball>
just to get it working for user started processes?
<srhb>
I mean, the man page describes how you'd decide on it, I can't really decide for you, but here's one that applies for anything and gives you 100MiB for either limit: security.pam.loginLimits = [ { domain = "*"; type = "-"; item = "core"; value = toString (1024*1024); } ];
<fresheyeball>
ok that looks reasonable
<fresheyeball>
srhb: thank you, that saved me potentially a few hours trying to understand these docs
<fresheyeball>
srhb:++
<fresheyeball>
srhb++
<{^_^}>
srhb's karma got increased to 61
<srhb>
You need to relog for that to take. :)
<srhb>
Sure thing.
<{^_^}>
[nixpkgs] @worldofpeace pushed 7 commits to master: https://git.io/fjHpf
<fresheyeball>
srhb: I want to debug a specific crash of the OS
<fresheyeball>
not just crash a process
<srhb>
fresheyeball: OS?
<ivegotasthma>
I want to ass the plugdev group to my user, but the user already exists and nothing happens when I add it to extraGroups in /etc/nixos/configuration.nix
<ivegotasthma>
s/ass/add*
<srhb>
fresheyeball: I thought you were talking about an application :)
<fresheyeball>
srhb: no, the system is crashing
<srhb>
ivegotasthma: Did you log out and in again?
<ivegotasthma>
yes
<ivegotasthma>
I rebooted
<fresheyeball>
as in I am drawing in krita, and all of the sudden the tablet goes dark and has to be turned back on again
<srhb>
fresheyeball: Here I thought wacom studio was some sort of design application
<fresheyeball>
it crashes roughly every 20 minutes or so durring use
<clever>
,tofu
<{^_^}>
To get a sha256 hash of a new source, you can use the Trust On First Use model: use probably-wrong hash (for example: 0000000000000000000000000000000000000000000000000000) then replace it with the correct hash Nix expected.
<{^_^}>
[nixpkgs] @worldofpeace pushed 4 commits to master: https://git.io/fjHpE
<srhb>
fresheyeball: OK, well, your best bet is to start by analyzing the logs from your last boot (the one that crashed.) Hopefully something is there. If not, you'll probably want boot.crashDump.enable and things get somewhat hairier.
nschoe has quit [Read error: Connection reset by peer]
ZD[m] has joined #nixos
<nixbitcoin>
Hi, I'm getting [this issue](https://github.com/ofek/coincurve/issues/47) while packaging coincurve for NixOS. I wanted to make sure this issue isn't in my nix expression. Could anyone double-check my expression pasted in the issue?
<{^_^}>
ofek/coincurve#47 (by nixbitcoin, 3 minutes ago, open): libsecp256k1.dll on a non-Windows system
akaWolf has quit [Read error: Connection reset by peer]
<{^_^}>
[nixpkgs] @veprbl opened pull request #66091 → bluejeans-gui: use patchShebangs, add me to maintainers → https://git.io/fjHho
MmeQuignon has joined #nixos
ottidmes has quit [Remote host closed the connection]
<colemickens>
so I manually deleted a store path that can't be in use anymore, and it seems to block me from running `nix-store --optimize` and I can't --delete it either.
<colemickens>
It says its in use in memory, but I don't actually believe it :/
<azazel>
hey guys, what can it be that having enabled memtest in grub configuration, choosing its entry on book freezes the system without printing anything on the screen?
<azazel>
i.e. no traditional blu screen of memtest is shown and the system is unresponsive
<colemickens>
is there a shorthand for nix-collect-garbage that takes a number of profiles to keep, rather than a number of days to keep?
Lukas4452 has quit [Ping timeout: 268 seconds]
shibboleth has quit [Quit: shibboleth]
erba has joined #nixos
<srhb>
colemickens: I think you need to go through nix-env for that
<aveltras>
how to find where the settings for a gnome extension are located ? i can find settings from default extensions with "gsettings list-schemas" but not the ones which i have installed via environment.packages
<srhb>
colemickens: As in nix-env --profile someprofile --delete-generations +10
<colemickens>
srhb, okay, I'm already setting the profile manually, but oh wow, I didn't know nix-env had an option like that :) cool
<srhb>
erba: As in, emacsPackagesNg.mood-line ? If so, are you on a commit that actually includes that change?
<srhb>
Or, to answer it differently: Stable users (ie. 19.03, currently) never see changes on master unless they are explicitly backported. unstable users see it when the channel catches up to master.
<dmj`>
hello, I have 4 python packages, 1 executable that depends on 3 libraries. All are built with nix using buildPythonPackage. When I attempt to run the executable it says it is missing a module from one of the libraries. Anyone know why this would be? I have all deps in both buildInputs and proprogatedNativeBuildInputs
<adisbladis>
Right, that's it then (also please use a pastebin next time)
jlv has joined #nixos
<erba>
adisbladis: I was just gonna excuse the spam :) I'll remember that, thanks.
<etu>
adisbladis: remove -m and it's single line
<recj>
after i installed django using nix-env -iA nixos.python37Packages.django_2_2
<adisbladis>
You're on 19.03, mood-line is not there
<recj>
i can't import it using python3. i was wondering why this might be
ozychhi has joined #nixos
Jackneill has quit [Remote host closed the connection]
<adisbladis>
And emacs packages are not backported (especially not since we just merged a rewrite of the melpa infra last week)
<srhb>
,libraries recj
<{^_^}>
recj: Don't install libraries through nix-env or systemPackages, use nix-shell instead. See https://nixos.wiki/wiki/FAQ/Libraries for details.
<adisbladis>
I'd suggest grabbing emacs from unstable :)
<erba>
adisbladis: Ah ok, thanks! :)
<srhb>
recj: Or see python.withPackages :)
chloekek has joined #nixos
<erba>
adisbladis: I'm very happy with NixOS so far btw (used for about a week)
<adisbladis>
\o/
<etu>
erba: Welcome! :)
<erba>
etu: Thanks! :)
<recj>
ok thank you {^_^} and srhb
<adisbladis>
recj: {^_^} is a bot
<srhb>
,botsnack
<{^_^}>
Oh thanks, have a cookie yourself
<srhb>
Bots deserve thanks too!
<adisbladis>
If corporations can be people so can bots
<iqubic>
Hehe.
<azazel>
:-)
nurelin[m] has left #nixos ["User left"]
<recj>
ok so in general only install applications into my user profile
<iqubic>
adisbladis: Does your EXWM overlay fetch emacs from unstable?
<iqubic>
Or would that be done by the use-package hack?
Chiliparrot has quit [Quit: My iMac has gone to sleep. ZZZzzz…]
<azazel>
someone should add the link to a {^_^} help page on the topic, I still can't rember how to ask things to it
<adisbladis>
iqubic: It has emacs from git as an opt-in
<iqubic>
I see.
<srhb>
recj: yep!
<iqubic>
adisbladis: Should I opt-in to that?
<recj>
noted :-)
xkapastel has quit [Quit: Connection closed for inactivity]
<adisbladis>
iqubic: If you want to?
<adisbladis>
I'm using it and only had to patch one package
<srhb>
recj: (Development environments get very impure if libraries are just magically picked up from the environment, we try to avoid that)
<iqubic>
I think I'll stick with emacs from unstable and not directly from git.
* etu
is running emacs from git through that overlay
<iqubic>
Also, what's the status of the new overlay?
<jlv>
I'm working on adding a package and hit a bit of a snag. I am getting "error while loading shared libraries: libevent-2.0.so.5: cannot open shared object file: No such file or directory". I am including libevent. It is a debian package installed using patchelf and rpath set, in case that matters.
<jlv>
I added it to a fork of the nixpkgs master branch, if that matters
_kwstas has quit [Quit: _kwstas]
<srhb>
jlv: Huh, looks like that pastebin snuck in some weird whitespace characters. My bad.
<hyperfekt>
jiv: If you're using a reasonably recent version of Nix sandboxes should be enabled by default. Check `nix run -f '<nixpkgs>' nix-info -c nix-info -m`
<srhb>
jlv: So what I'd do from here is start by adding autoPatchelfHook as an argument and as a nativeBuildInput
<srhb>
jlv: And remove the custom postFixup
<srhb>
jlv: It will emit a nice list of things it didn't found (and there's more than one)
detran has quit [Ping timeout: 268 seconds]
<srhb>
s/found/find
<jlv>
srhb: I'll try that and get back to you :)
<srhb>
jlv: Regarding libevent specifically, I don't think we actually have that version.
ris has joined #nixos
_kwstas has joined #nixos
<srhb>
jlv: (It's specifically asking for 2.0, we're on 2.1)
<jlv>
srhb: Ok, that was easier than I expected. Looks like autoPatchelfHook just works? Not exactly sure how. Regardless, looks like libevent-2.0.so.5 is the only missing library (after adding the rpath packages to buildInputs)
<srhb>
jlv: ~*magic*~
<srhb>
jlv: It reads the missing libraries from each file and tries to find them in the inputs, then patches the files accordingly :)
<jlv>
srhb: So, I guess I need to specify the specific libevent 2.0 version as a dependency, and add it if it doesn't exist?
rouma has quit [Remote host closed the connection]
<jlv>
That's pretty cool :)
<Miyu-chan>
Is there any code for nginx+uwsgi on nixos-unstable? I don't see one for nixos-19.03.
__monty__ has quit [Ping timeout: 245 seconds]
<srhb>
jlv: You can _try_ just copying the libevent.so to the build directory under the old name, but this might break things spectacularly at runtime of course
rouma has joined #nixos
<scoates>
gchristensen: is it cool to ask Lorri questions in here or is there a better place?
rouma has quit [Read error: Connection reset by peer]
<srhb>
jlv: The right solution is as you say, provide an older libevent :)
<jlv>
srhb: I think I'll try the quick fix for now and see how it does :)
<srhb>
jlv: It might eat your data. Fair warning. :P
<jlv>
srhb: Hard choices...
<srhb>
jlv: no guts no glory: cp ${libevent}/lib/libevent.so libevent-2.0.so.5
<srhb>
:-P
detran has joined #nixos
<Miyu-chan>
Okay. So, if I make an nginx + uwsgi code, do I put it in what file, uwsgi, nginx, or a new file?
<srhb>
jlv: libevent as it were before the 2.1 bump: git show 3be19de8e4eba036d1c6ae81295beee1871a7fef~1:pkgs/development/libraries/libevent/default.nix in a nixpkgs checkout
rouma7 has joined #nixos
Lukas4452 has quit [Ping timeout: 244 seconds]
eraserhd has quit [Ping timeout: 245 seconds]
hyperfekt has quit [Ping timeout: 260 seconds]
Lukas4452 has joined #nixos
<jlv>
srhb: Ok, so looks like nixpkgs does have the 2.0 version around. Is it possible to reference it directly, or do I need to add a libevent2 (or something like that) package?
<asymmetric>
yeah.. i've tried that but somehow the nixos module i've developed isn't found, even though i've been testing it with nixos-rebuild and it works
<clever>
and i just leave a comment to remind myself if i ever forget
<clever>
the comment also keeps track of what rev i should be using
<jlv>
srhb: Looks better than renaming the 2.1 library at least XD I must just add a libevent-2_0 package to nixpkgs if that is considered ok. That seems like the "right" was to do it to me.
<jlv>
*might
<jlv>
*way
<jlv>
Typing is hard
<Miyu-chan>
...
<Miyu-chan>
I just realized something.
<Miyu-chan>
Oh whoops. Wrong chat. Lmao
<asymmetric>
gchristensen: how do you set NIX_PATH in the shell.nix? i just realized that the way i do it doesn't work, whereas -I does
<clever>
asymmetric: can you pastebin your shell.nix?
<jlv>
srhb: Ok. Your fetchurl trick is working :) I think I will leave it at this then. Next up is adding seadrive-gui (which has seadrive-daemon) as a requirement. That one is open source and should be easier at least. Then I'll look at doing a pull request to get seadrive in nixpkgs :)
Shoubit has quit [Ping timeout: 268 seconds]
kbZ has joined #nixos
<srhb>
jlv: Cool. :)
<Miyu-chan>
> /foo + /bar
<{^_^}>
/foo/bar
<srhb>
jlv: Hmm, maybe I've changed my mind. looks like libevent recently got a 2.0.23 beta.
<Miyu-chan>
Okay... I think I'm stupid?
<Miyu-chan>
> "/foo" _+ /bar
<{^_^}>
attempt to call something which is not a function but a string, at (string):256:1
<clever>
asymmetric: are you on nixos or another distro?
<asymmetric>
clever: nixos 19.03
<Miyu-chan>
> "/foo" + /bar
<{^_^}>
access to path '/bar' is forbidden in restricted mode
<Miyu-chan>
Oh okay.
rouma7 has quit [Remote host closed the connection]
<Miyu-chan>
So copyPathToStore gives you a strsing, huh.
rouma7 has joined #nixos
_kwstas has quit [Quit: _kwstas]
rouma has joined #nixos
<asymmetric>
seems like `NIX_PATH=nixpkgs/foo/bar nixops deploy` is not equivalent to `nixops deploy -I nixpkgs=/foo/bar`
<asymmetric>
the former doesn't see my new module, the latter does
<Miyu-chan>
Interesting.
<clever>
asymmetric: what does `nixops info` say?
<Miyu-chan>
Also, are you copying verbatim, since those are 2 different code.
<clever>
asymmetric: does it show a `Nix path:` ?
<Miyu-chan>
The left one sets the default search path to start on nixpkgs/foo/bar, I believe.
<Miyu-chan>
(Asking in case of typo.)
<asymmetric>
Miyu-chan: no, i typed it in, but i typed it correctly in my shell
<andi->
Anyone else having issues with slack on both unstable and 19.03? I can start it but the login button and basically anything else doesn't work... Not sure if that's waylands fault but I fail to see why.
<gchristensen>
yes
<gchristensen>
I can't log in to new workspaces either
<Guest88>
hi, i'm kinda new with nixos. I'm getting some throuble when connecting to a NFS share on my network. When I try to mount I receive the following message: bad option; for several filesystems (e.g. nfs, cifs) you might need a /sbin/mount.<type> helper program.
byteflam1 has joined #nixos
___laika has joined #nixos
<Guest88>
I installed libnfs and nfs-utils
<Guest88>
i've installed thought nix-env -iA. I dont want to install this packages permanetnlty right now
<{^_^}>
[nixpkgs] @worldofpeace pushed 3 commits to master: https://git.io/fjQfn
Blessjah has quit [Remote host closed the connection]
<gchristensen>
andi-: ??? ...
throwaway7777777 has joined #nixos
<andi->
gchristensen: step 1) add xdg-open (or `echo` aliased to xdg-open) to your path 2) open the website it tries to open and type in the slack you want to access. 3) On the page you are being redirected to open the dev console and set a breakpoint at line 237 of the HTML document. When it breaks there extract the value of the `link` variable and call slack with it (slack ${link}).
ericsagnes has quit [Ping timeout: 276 seconds]
<gchristensen>
:o
hmpffff has quit [Quit: Bye…]
<throwaway7777777>
Hello, i am currently attempting to configure networkmanager to pass all DNS requets to Stubby (DNS-over-TLS) with dnsmasq as a caching proxy.)
<throwaway7777777>
"It is possible to pass custom options to the dnsmasq instance by adding them to files in the /etc/NetworkManager/dnsmasq.d/ directory."
<{^_^}>
[nixpkgs] @vcunat pushed to master « lua*Packages.cqueues: fixup darwin build »: https://git.io/fjQJn
DariusTheMede has quit [Ping timeout: 244 seconds]
<jw358>
hey folks
<jw358>
what should i do when i get `error: getting status of \'/nix/store/1manqjxzqpriw9hq5lfqlc6pgdwfmgbh-runtime-closure.nix\': No such file or directory` when building things
<jw358>
what do i need to look at?
<hyperfekt>
How would I pin NixOS using niv without needing a double rebuild? Putting something in the NIX_PATH that pulls the pinned nixpkgs from sources.nix gives me infinite recursion because Niv thinks t can use nixpkgs for its fetchers instead of having to use the builtins.
<recj>
i'm trying to look at an example of a config with a python.withPackages in it because I tried to put on in my system packages but still can't import django in the interpreter or run my project
<infinisil>
jw358: Hard to say without knowing more about the problem
___laika has quit [Ping timeout: 246 seconds]
<jw358>
what builds runtime-closures
<jw358>
and how do i know is it fails silently or whatever
<jw358>
i assume something was supposed to build that but it didn't, thus the generic "no such file"
<ajs124>
recj: the interpreter that results from nix-build -E '(import <nixpkgs> {}).python3.withPackages (ps: with ps; [ django ])' can successfully import django (at least on my system)
npmccallum has quit [Quit: npmccallum]
<recj>
yes, that worked when i tried it on my system as well, maybe i should just put that in a script for easy access
dansho has quit [Ping timeout: 248 seconds]
<Gopal[m]>
hey guys I've been trying to get borg working for 2 hrs
<Gopal[m]>
is it possible to replace `docker-compose` with NixOS' `docker-containers.<nameofservice>` ?
<jsgrant_>
Saw podman is packaged, kinda want to look into how hard it'd be to get 'fedora-toolbox' working for it -- having an optional mutable state seems like a sweat-spot to ease into things from within NixOS 'on real hardware'; Compared to the path I see a lot of people going where they install it on an existing distro (nix) or a VM beforehand. :^P
<jsgrant_>
sweet-spot* lol
<simpson>
jsgrant_: Depends on what you're building. NixOS on bare-metal is pretty straightforward.
<jsgrant_>
simpson: Well, yeah; I was able to re-piece together a config I had about 2 years ago when I messed with NixOS 'more seriously' the first-time -- I'm saying for new-users to go "all in" and still have a trivial development environment in the meantime without encuring the cost of transfering everything to a nix-ified way immediately
lambda-11235 has quit [Ping timeout: 272 seconds]
<simpson>
jsgrant_: Ah, sure. The solution might be to stop teaching non-NixOS as a first Linux distro; who knows~
<jsgrant_>
simpson: Ease first step, time to steal all the market-share of other distros. :^)
abathur has joined #nixos
<simpson>
jsgrant_: Sure, but "avoid (success at all costs)", right? We should be careful about emulating the missteps of other distros; we ought not to choose an also-ran path of cleanly and perfectly supporting tools tailored for other distros.
* jsgrant_
really wants to avoid success and wants that full-haskell rewrite, and a functional kernel while we're at it. :^)
<jsgrant_>
But yeah, I agree it's a difficult line to tow/ walk the border of
<jsgrant_>
I guess having a mutable state like that, so easy to install -- could in-theory teach people not to be as ideological (I think rightly) about a fixed/deterministic state
<jsgrant_>
simpson: A fully-mutable priveleged container; But at that point it'd certainly not be NixOS -- but I wouldn't want/suggest that be a feature of nix/nixos; Just a niceity to get people to install NixOS on their system with 'less initial investment'.
mexisme_ has joined #nixos
<jsgrant_>
That's why I'm interested in Fedora-toolbox (which is what silverblue is using) and doesn't seem especially tied to Fedora in anyway besides container defaulting to a Fedora system
<simpson>
jsgrant_: The filesystem's quite mutable, aside from /nix. What kinds of habits, specifically, are you thinking of encouraging? I'm not sure why mutability of, say, /usr is a good thing.
<jlv>
Anyone know what package might have mount.pc? cmake says the package is 'mount'. Doesn't look like the 'mount' nix package is what I'm looking for.
<jsgrant_>
simpson: It's VERY VERY possible it's mostly ignorance on my end at this point; But I guess it's more-so an 'issue' of not being accustomed to the workflow and ecosystem and slowly introducing someone to that -- instead of everything being thrown at you at the get-go , more than any actual limitations of the system as a whole
<ashkitten>
,locate mount.pc
<{^_^}>
Found in packages: libuuid.dev, utillinux.dev
<niso>
a path has to be explicitly casted to a string right? (via toString)
gentauro has quit [Read error: Connection reset by peer]
lambda-11235 has joined #nixos
<simpson>
jsgrant_: Yeah. I'm not sure if there's any good way to deal with that; ultimately, the kind of safety that Nix introduces only comes with a rewrite; in this case, a rewrite of build and integration systems.
gentauro has joined #nixos
<jlv>
ashkitten: Is there an easy way to use that locate command outside of irc? Seems very hands, and I wouldn't want to spam irc.
<ashkitten>
jlv: nix-index and nix-locate
<jsgrant_>
simpson: Oh yeah, absolutely -- not suggesting this should even be a goal of nix in the first place; I just guess wondering out loud if attempting to package something like Fedora-toolbox would be worth it, as an optional path to get people trying out NixOS on bare-metal more. I mean, I now have NixOS on 3 (soon 4, with an offsite machine I'm staging in the next few days) -- but I have this Ultrabook I'm
<jsgrant_>
keeping on Fedora for the next year "just in case" I can't figure something out real quick devel wise. And as I learn the ecosystem more ... think this is probably a non-issue after first few weeks, but still; The idea of being able to go in NixOS and just call up 'toolbox' to enter a mutable Fedora or something would probably let me free up that Ultrabook (probably).
<jsgrant_>
If I didn't have this Ultrabook I could dedicate to "fullt-ime emergency/backup machine" I'd be on that what I define as the 'more traditional path' of either installing Nix on my host-system or trying to run in a VM; And suspect (because this happened from personal experience) the retention rate is probably a bit lower. :^P
<jsgrant_>
Took me near 2 years (with a few tries in-between and quickly bailing each time) to give NixOS a real shot; And probably this is somewhat due to a personal failing of mine -- but I have a pension to all or nothing; So there's that. lol
<simpson>
I guess. After tasting Nix, I rapidly switched to NixOS on my personal daily-driver machines. I'm not sure why one would *want* Fedora, TBH, other than nebulous proprietary vendor requirements.
<azazel>
jsgrant_: If it took you 2 years just to try out NixOS, that maybe means you wil ditch Fedora in 10 years
byteflam1 has quit [Ping timeout: 248 seconds]
<azazel>
that is something
linarcx has quit [Quit: WeeChat 2.5]
<{^_^}>
[nixpkgs] @jdanekrh opened pull request #66134 → [WIP]msbuild: Init at 16.3 → https://git.io/fjQUN
<immae>
Hello there! Is there a way to add python packages as a dependency of a mkShell but not have it modify the PYTHONPATH? (the reason is that I have a mix of python2 and python3 applications, and I mixing of the PYTHONPATH doesn’t please them)
<jsgrant_>
azazel: Well, I guess to define "try" a bit; I have used NixOS probably about a dozen diff times on real hardware for a few weeks here and there over the past 5 years -- about to years ago, it was a few months (2 to 3). But (and it is very possible this was due to being in a schedule which was the eq of near double full-time work during the latter period) I never found I had enough time to work though any
<jsgrant_>
snags I ran into -- and again, this may be due to personal failings, it'd be a non-problem if each time I tried to go all-in instaed of going this traditional path I mentioned of either nix on host-system or in a vm before doing that
<jsgrant_>
about two years ago*
<jsgrant_>
I do think "having one emergency box" is probably the way forward for me, though; Shocked I didn't think of it before ...
ambro718 has joined #nixos
<jsgrant_>
Well and it was 2 or so times of "several months" too. :^P
fusion809 has quit [Remote host closed the connection]
<jlv>
Adding a package. Getting error "Can not use "WebEngine" module which has not yet been found.". I added 'qtwebengine' to buildInputs. Any ideas?
<jtojnar>
yeah, the apps would need to tell registry
<jtojnar>
pie_: I think there might be some X protocol calls for obtaining window icon but again, it will not work on Wayland
<pie_>
yeah i dont know much about most of this stuff
<pie_>
dbus has just been this word i didnt really understand for a long time that does some kind of weird application interaction magic and i just didnt get into figuring out whats going on there
<pie_>
(would love to work on this some more but my mind is currently preoccupied, you have my attention, just not undivided :D)
<pie_>
jtojnar: wow that seems...messy :I
<pie_>
jtojnar: so what if different windows for the same application have different icons? :P
<pie_>
or idk
<{^_^}>
[nixpkgs] @orivej-nixos pushed commit from @orivej to master « llvmPackages_8: 8.0.0 -> 8.0.1 »: https://git.io/fjQTM
<pie_>
getting stuff out of desktop files just seems horribly disconnected from whats actually going on :(
<jtojnar>
pie_: maybe we could create a program that would watch ps and when you start a process from /nix/store and it is not installed and has a desktop file, it would link the file to ~/.local/share/applications
<jtojnar>
and remove it when the process exits
<pie_>
well ok but i guess thats a bit different than the _application icon_ paradigm, - and getting the higher res icons from the icon directory
lambda-11235 has joined #nixos
<pie_>
jtojnar: or tmp
<pie_>
idk how it works but bamf does seem to drop some icons into tmp under some circumstances that i dont know
<jtojnar>
the desktop file needs to be somewhere under XDG_DATA_DIRS
<pie_>
rght you actually have to get the thing to find it
mckeankylej has quit [Remote host closed the connection]
<jtojnar>
so do you have quassel installed? does gnome-pie get at least some icons?
<pie_>
gnome pie gets konsole somehow
<pie_>
but i thik misses everything else i have open right now
<pie_>
i think it might just be using some default icon for konsole
ambro718 has quit [Quit: Konversation terminated!]
<alienpirate5>
is there an effective way to run something like this: `nix-shell -E 'with import <nixpkgs> {}; pkgsCross.armv7l-hf-multiplatform.linuxPackages_rpi.overrideAttrs (o: {nativeBuildInputs=o.nativeBuildInputs ++ [ pkgconfig ncurses ];})'`