<simpson>
wolfman154: Oh, I thought that you were already using XMonad. I don't particularly care what you use, I'm saying that what you use likely determines what a default terminal is.
<wolfman154>
I am using xmonad, just wasn’t sure where to do the change, usually everything is done in configuration.nix
jasongrossman has quit [Quit: ERC (IRC client for Emacs 26.1)]
<judson__>
wolfman154, XMonad has a ~/.xmonad/xmonad.hs file it uses to configure itself.
aplainzetakind has left #nixos [#nixos]
<judson__>
Near the end of mine, I have something like "main = xmonad $ def { terminal = "alactritty" }"
<wolfman154>
judson__: thanks
<wolfman154>
judson__++
<{^_^}>
judson__'s karma got increased to 1
kvda has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<iclanzan_>
makefu++
<{^_^}>
makefu's karma got increased to 2
<pie_>
is there an easy way to get a derivation thats the contents of a fetchurl archive?
<pie_>
right now i just have the result of a fetchurl and i want to use a few files of it in sometihng else
jluttine has quit [Ping timeout: 250 seconds]
Andrey_ has quit [Quit: Page closed]
mbrgm_ has joined #nixos
mbrgm has quit [Ping timeout: 255 seconds]
mbrgm_ is now known as mbrgm
freeman42x has quit [Ping timeout: 258 seconds]
slack1256 has joined #nixos
orivej has joined #nixos
iclanzan_ has quit [Quit: leaving]
kvda has joined #nixos
realrokka has quit [Ping timeout: 250 seconds]
jluttine has joined #nixos
kleisli has joined #nixos
goibhniu1 has quit [Ping timeout: 244 seconds]
<kleisli>
running `nixos-rebuild switch --upgrade` typically takes very long for me, on the order of hours, is this normal?
<pie_>
ok i figured out that i should probably just use fetchzip
<pie_>
kleisli, are you recompiling the world?
<kleisli>
@pie_ not that i know of!
<kleisli>
i have a pretty simple configuration
<aanderse>
kleisli: definitely weird in that case
<pie_>
unless you have uber slow internet and end up downloading new packages, or you specifically caused something to need to rebuild a bunch of stuff all the time, i dont think that should happen
<pie_>
or oyu have some uber slow disk thats seeking a lot with symlinks? :P idk how that works
<pie_>
kleisli, how certain are you that you arent recomiling a bunch of things
<kleisli>
i believe last time i noticed the majority of the time was spent compiling some examples in the `sage` math package, now that i think of it
<pie_>
unrelated: how do I make one of those single file derivations that are just a file, like an archive, in the store, and not a directory?
<blackriversoftwa>
I have been able to build an image that starts on digital ocean, and I enabled cloud-init with the order of data sources recommended by digital ocean
<blackriversoftwa>
But my image still gets started with no hostname, no way to reset the root password, and without respecting the SSH keys given in the digital ocean console
<blackriversoftwa>
Right now I'm trying to upload another image with an ssh public key built in to the nixos config, so I can at least get in and look at the logs
fusion809 has joined #nixos
<blackriversoftwa>
but I'm wondering if anyone has been down this road before
<colemickens>
blackriversoftwa: have you peeked at the serial console output to see if cloud-init is logging about what datasources its trying, etc?
kleisli has quit [Read error: Connection reset by peer]
<blackriversoftwa>
colemickens: yeah the console just shows a login prompt
<blackriversoftwa>
it isn't even interfaced well enough to get me automatically logged in as root
<blackriversoftwa>
Unfortunately digital ocean seems to be having issues uploading custom images today. Once my progress bar finishes the image just vanishes
<colemickens>
blackriversoftwa: I don't know if I'd expect that, but if you're building your own images, you can always add a root/pw or another user/pw for debug purposes
<blackriversoftwa>
colemickens: yeah that is the next step if I can't ssh
<blackriversoftwa>
colemickens: my hope is that I go through this pain once and then there is a nice little module in nixpkgs/nixos/modules/virtualisation to build digital ocean images
<kleisli>
hm, it ran very quickly this time. I'm guessing it has been an issue with specific upgrades to packages in environment.systemPackages before having to do lots of compilation
<blackriversoftwa>
colemickens: the command I was using to build was (from the root of the nixpkgs repo) `nix-build nixos/lib/eval-config.nix -A config.system.build.digitalOceanImage --arg modules "[ nixos/modules/virtualisation/digital-ocean-image.nix ]" --argstr system x86_64-linux -o digitalocean -j 4`
<blackriversoftwa>
which magic should go into a script once I get this working
<blackriversoftwa>
colemickens: it takes quite a while even if just a config changes, because to generate the final image it actually boots it in QEMU and installs grub
<blackriversoftwa>
Oh dear it has my ssh public key in it
<blackriversoftwa>
oh well
<colemickens>
yeah, I did Azure images for a while so I'm sadly familiar.
<colemickens>
but iirc the nixos module uses the ancient version of the agent that mostly does everything itself in a single huge fragile python script (:
leotaku_ has quit [Ping timeout: 245 seconds]
<blackriversoftwa>
Ah lovely
<blackriversoftwa>
well ssh is a no go
<blackriversoftwa>
it's decided it doesn't even want to accept connections
<blackriversoftwa>
before it was just denying my public key
<blackriversoftwa>
colemickens: did you get a chance to look at that draft PR? See anything obviously dumb that I'm doing?
<colemickens>
yeah, it looked fine to be :(
<blackriversoftwa>
Well it builds the image fine. It just doesn't do cloud-init
<blackriversoftwa>
one of the other modules had a few things like wget in the system environment, presumably so the scripts that cloud-init grabs from the host and runs could use them?
<blackriversoftwa>
My suspicion is I'm going to see "command not found" or some such in the logs...once I can get to them
justan0theruser has joined #nixos
Erasmus has quit [Read error: Connection reset by peer]
<blackriversoftwa>
wishlist: for the cloud-init module to define an option for packages to go in a special env just for cloud-init scripts
justanotheruser has quit [Ping timeout: 255 seconds]
<blackriversoftwa>
colemickens: one frustrating thing is that there is almost 0 documentation of how to configure cloud-init on a client image
justan0theruser is now known as justanotheruser
<blackriversoftwa>
it is all about how you can customize it on the cloud platform/hypervisor side
<colemickens>
I have really never had anything but poor experiences with cloud-init.
<colemickens>
Long iteration cycles, easy mistakes (no thanks to YAML often), weird configuration bugs.
<colemickens>
When we were building images for various OSes inside Azure, we found numerous bugs in cloud-init's disk handling, including bugs that caused data loss. (and I don't say that lightly).
<blackriversoftwa>
yikes!
wolfman154 has joined #nixos
<wolfman154>
Should I install xmonad with nix or cabal?
<colemickens>
When I left that team they had someone that was starting to send patches somewhat regularly. Some of it had to do with certain interactions with Azure oddities (the way LUN devices are enumerated, udev rules changing over time, etc) but still.
<gchristensen>
I didn't know you were *at* azure, colemickens :)
<colemickens>
At one point in time, no longer.
<gchristensen>
aye
<wolfman154>
gchristensen: should I install xmonad with nix or cabal?
<wolfman154>
I am getting a recompile error
<gchristensen>
I don't really know about xmonad or cabal.
<wolfman154>
Oh
<wolfman154>
:(
<jasongro`>
wolfman154: FWIW, I've installed it with nix. I have no idea about the rest of the Haskell ecosystem though.
<blackriversoftwa>
I would have thought that at least QEMU as a native KVM would be a bit snappy. But no, it feels just like the armv7l qemu I use at work for Haskell builds for Raspberry Pi
<blackriversoftwa>
as in, molasses
<wolfman154>
jasongro`: I am getting a recompile error, I am going to remove everything and start from scratch again
<blackriversoftwa>
dammit
<wolfman154>
jasongro`++
<{^_^}>
jasongro`'s karma got increased to 1
<blackriversoftwa>
does anyone know how to make Ubuntu keep /dev/kvm permissions at 666
<blackriversoftwa>
I can build the image once and then it will fail if I don't chmod /dev/kvm again
<blackriversoftwa>
gchristensen: they still have them?
<blackriversoftwa>
I couldn't find them again last time I looked
<gchristensen>
they sell out very quickly :/ but if you get on their Slack, they can sometimes rustle one free
<gchristensen>
NixOS Foundation has a handful of them for the build infra
<blackriversoftwa>
Oh they are armv8 servers
shpx has joined #nixos
<acowley>
Checking in again about this PR <https://github.com/NixOS/nixpkgs/pull/57885>. Everyone who's looked seems to think its fine, but there is some uncertainty about how to test it.
<colemickens>
blackriversoftwa: that build command is building a bunch of X/i3 stuff surprisingly.
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/fjTqi
<blackriversoftwa>
colemickens: yeah I wondered about that
<blackriversoftwa>
some poking around with nix-store seemed to point to cloud-init (surprise surprise) as the culprit but I am not 100% sure of that
<simpson>
acowley: Yeah, that's about all I could say. I think it's fine! Do we know anybody with the HW that can test it meaningfully?
<colemickens>
somehow that sounds familiar? or just imminently possible? heh
<colemickens>
I got stumped elsewhere with Nix on Ubuntu: error: a 'x86_64-linux' with features {kvm} is required to build '/nix/store/p9zx4csv4yypw27bg0l5j2xw9ihjgb47-digital-ocean-image.drv', but I am a 'x86_64-linux' with features {benchmark, big-parallel, nixos-test}
obadz has quit [Ping timeout: 245 seconds]
<colemickens>
maybe that's the KVM permission bit though
reivilibre has joined #nixos
<blackriversoftwa>
colemickens: maybe? I didn't run into that
<blackriversoftwa>
one sec
<acowley>
simpson: Just myself and the other fellow on the linked PR
<blackriversoftwa>
colemickens: what nix version are you on?
<colemickens>
the latest. it was just the KVM bit, it's building now
<acowley>
simpson: The lingering question is, I think, if it is a problem for people without the relevant HW.
<blackriversoftwa>
ah ok
Erasmus has joined #nixos
nore has joined #nixos
<blackriversoftwa>
colemickens: I'm gonna be afk for the evening
<blackriversoftwa>
I built an image with a user nixos, password nixos
<colemickens>
blackriversoftwa: okay, I'll reply on the PR if I find anything of note
<blackriversoftwa>
I'll push that up to the PR
<colemickens>
I can build and push an image from this machine very quickly now :)
<blackriversoftwa>
colemickens: yeah I'm uploading from my home internet and it takes a while
<blackriversoftwa>
colemickens: thanks and tty again soon hopefully
<colemickens>
cheers
<gchristensen>
interesting. nixos' route options can't blackhole it seems
<simpson>
acowley: I have read some kernel code. Want my opinion here or in the issue? We should get a kernel maintainer to signoff but I don't think it's a big deal to enable by default.
<simpson>
When enabled, support for a couple extra chipsets is compiled in and the functionality you want is attempted. The changes seem local to the AMD drivers in the DRM subsystem.
obadz has joined #nixos
shpx has quit [Quit: shpx]
jmeredith has quit [Quit: Connection closed for inactivity]
<acowley>
simpson: Yeah, the desire for someone other than those of us using the HSA features is likely an excess of caution, but I also don't know how to push the PR forward.
drakonis_ has joined #nixos
<simpson>
acowley: "I talked to somebody on IRC who claims that they used to hack on `radeon` and they said it looks unlikely to interfere."
<simpson>
I dunno. These kinds of driver feature flags suck, because in a few kernel versions, the functionality might become stable and the flag would become a do-nothing.
kini has quit [Quit: No Ping reply in 180 seconds.]
buffet has quit [Ping timeout: 250 seconds]
<tdeo>
is the
buffet_ has joined #nixos
<tdeo>
is there a way to get an output path of a multiple-output derivation other than checking the environment? i want to make a (non-nixpkgs) package with an output with a dash in it, and you can't get the value of a variable with a dash in it in shell
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
kini has joined #nixos
hellrazor has joined #nixos
hellrazo1 has quit [Ping timeout: 250 seconds]
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
ddellacosta has joined #nixos
<tdeo>
turns out it's not possible anyway, some other bash code uses the variable
obadz has quit [Ping timeout: 246 seconds]
obadz has joined #nixos
<{^_^}>
[nixpkgs] @mnacamura opened pull request #58470 → rPackages: fix missing dependency to libiconv → https://git.io/fjTYY
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
MatrixBridge has joined #nixos
MatrixBridge has left #nixos ["User left"]
<infinisil>
tdeo: Bash has ${variable} you can use to access variables with special names
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
shpx has joined #nixos
npmccallum has joined #nixos
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
<colemickens>
as I'm reading through cloud-init stuff debugging this... it looks like the "vendor" script from DO for cloud-init includes workarounds for cloud-init bugs inside it
<colemickens>
heh
<colemickens>
I'm also starting to wonder if one really wants to use cloud-init with nixos...
<colemickens>
seems much easier to write a tiny script to pull ssh keys from DO's metadata
jasongro` has quit [Remote host closed the connection]
pie___ has joined #nixos
o1lo01ol1o has joined #nixos
Izorkin has quit [Ping timeout: 255 seconds]
jasongrossman has joined #nixos
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
pie__ has quit [Ping timeout: 246 seconds]
iqubic has quit [Quit: Page closed]
Izorkin has joined #nixos
o1lo01ol1o has quit [Ping timeout: 272 seconds]
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
kvda has joined #nixos
kvda has quit [Client Quit]
shpx has joined #nixos
<{^_^}>
[nixpkgs] @teto closed pull request #58341 → [WIP] ghc.withPackages: can add user arguments to wrapper → https://git.io/fjUcu
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
<leotaku>
Is it possible to build a nixops machine without connection to said machine?
<leotaku>
So like "nix build -f '<nixpkgs/nixos>' -A system" but for nixops instead of nixos.
d10n-work has quit [Quit: Connection closed for inactivity]
shpx has quit [Quit: shpx]
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
reinhardt has joined #nixos
endformationage has quit [Quit: WeeChat 2.4]
puffnfresh_ is now known as puffnfresh
domogled has joined #nixos
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
ninjin has quit [Remote host closed the connection]
mabel has joined #nixos
mabel has quit [Changing host]
mabel has joined #nixos
<mabel>
hey! I'd like to start tracking unstable rather than 18.09. do I need to do anything special other than updating my nix-channel and doing nixos-rebuild?
ninjin has joined #nixos
<mabel>
(or I guess sudo nix-channel --upgrade would accomplish the same)
vk3wtf has quit [Ping timeout: 240 seconds]
joachifm has quit [Remote host closed the connection]
<jasongrossman>
mabel: That's all.
<jasongrossman>
mabel: I assume you mean `nixos-rebuild switch`. And you might need to reboot.
<mabel>
this is excellent! I'm not even sweating bullets like I normally do :D
<mabel>
oh, another thing I was wondering about... is there an easy way for me to track my own nixpkgs fork, but still take advantage of hydra's cache?
<mabel>
I'm slowly starting to do a little contribution, so I was thinking that would be a good way of eating my own dogfood
<mabel>
but my understanding of channels is that they rely on tgz archives from the github
<jasongrossman>
It's good, isn't it? And you CAN make your own channel ... but I don't know the details.
<jasongrossman>
Or rather, some people use their own fork. I don't know whether they all use the channel mechanism. You don't HAVE to use a channel at all.
<mabel>
oh interesting! does nixos-rebuild not rely on channels?
<mabel>
I remember that its machinery is slightly different from nix-build but I don't remember the details
<jasongrossman>
I don't either. You might have to wait until someone who actually does this joins the chat room. They will eventually.
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
<mabel>
hm... actually, I guess if I don't mind doing git push, I can just add my forked repo's github url as a channel. I'd need to push to origin each time and download the archive, but that should work
phreedom has quit [Remote host closed the connection]
phreedom has joined #nixos
<jasongrossman>
mabel: Why not just use unstable + the specific packages you contribute to?
<jasongrossman>
mabel: Or you could use master.
<mabel>
hm, like an overlay or something?
shpx has joined #nixos
<jasongrossman>
mabel: Yes, or just run your own programs using nix-shell or whatever, until they're accepted into unstable.
<mabel>
ah, that could work. I wasn't sure if the overlay mechanism worked with nixos-rebuild, was all
<mabel>
ooh my rebuild is done. brb
mabel has quit [Quit: WeeChat 2.2]
Glider_IRC_ has joined #nixos
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
mabel has joined #nixos
mabel has joined #nixos
mabel has quit [Changing host]
<mabel>
it worked! unsurprisingly. fish changed its syntax, which seems very rude of it.
Glider_IRC__ has quit [Ping timeout: 250 seconds]
<jasongrossman>
mabel: \o/
jaeckel has quit [Read error: Connection reset by peer]
<jasongrossman>
And >---<>
<mabel>
hehehe. :p
ddellacosta has quit [Ping timeout: 272 seconds]
InFlames has joined #nixos
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
liori has quit [Remote host closed the connection]
<leotaku>
Sorry for repeating my question, but is it possible to build a nixops deployment like it is to build a nixos system? (so with nix-build/nix build, instead of the nixops command)
<immae>
leotaku: nixops creates on-the-fly derivations that you will usually need to build the whole system identical to what nixops does
<leotaku>
immae: interesting. So let me restate my question: I would like to build a nixops machine without connection to said machine, in order to pre-download and build its dependencies. Is that possible?
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
<immae>
Oh, then the building part happens without connecting to the machine ;)
<immae>
nixops deploy --build-only
<immae>
Though you cannot initialize it, since nixops needs information from the remote machine to build the first derivation
<immae>
(like ip address)
shpx has quit [Quit: shpx]
<mikeplus64>
does cabal2nix not have support for default-extensions? :-(
palo1 has joined #nixos
<immae>
so usually you’ll neeed to at least initialize nixops with a new server, but then subsequent derivations will only connect to the machine at the last moment
<immae>
is that what you want?
samrose has quit [Ping timeout: 245 seconds]
samrose has joined #nixos
palo has quit [Ping timeout: 246 seconds]
palo1 is now known as palo
<sb0>
is there no nix package for discourse?
<immae>
leotaku: If you also want to initialize, you’ll need to somehow build the "on-the-fly" one by hand (finding all the ip addresses and all), it’s doable but not very practical (and it’s much easier if you already have an identical machine in your nixops that you can "copy")
<immae>
sb0: not that I could find, discourse is only (officially) supported via docker anyway...
<sb0>
isn't nix a superior docker? =]
<immae>
that’s a point of view :p
<leotaku>
immae: --build-only seems to work fine, thank you.
<immae>
sb0: I tried to build a nix derivation for discourse some time ago, and I (temporarily) gave up when I saw the complexity of everything there is in their docker builds, so I ended up extracting only the "database" parts
<immae>
leotaku: ok good :)
<leotaku>
immae: also is it somehow possible to get the progress bar that the nix2 tools have with that command? Small issue, but maybe it is possible?
<sb0>
immae: can you elaborate?
<sb0>
their reasons for docker-only are the complexities of "ruby, redis and postgresql", all working fine on nixos already...
<immae>
leotaku: I have no idea what you’re talking about here, sorry, cannot help you for that...
<immae>
sb0: no, it’s ruby the complex part
<immae>
redis and postgres are simple
<leotaku>
immae: ok thank you again!
<sb0>
ok. just paraphrasing their rationale for docker...
<sb0>
so what is the issue with ruby?
<immae>
"system" dependencies (imagemagick for one thing)
<sb0>
ah. is there the same kind of issue that there is with python packages in nix?
<sb0>
I'm unfamiliar with ruby...
<immae>
it’s just that it will expect some system dependencies that may not be satisfied
<immae>
imagemagick is the one I remember, but I "saw" other ones and I didn’t take the time to build the whole tree
<immae>
(and I didn’t want to take the risk of forgetting one)
<sb0>
immae: do you have your derivation somewhere?
<immae>
nope, it was quite simple so I didn’t keep it
leotaku has quit [Ping timeout: 245 seconds]
reinhardt has quit [Quit: Leaving]
polman has quit [Ping timeout: 245 seconds]
<immae>
(simple in the sense too incomplete)
polman has joined #nixos
Anton-Latukha has joined #nixos
rauno has joined #nixos
<CMCDragonkai>
What to do when you encounter an error like this: error: packages '/nix/store/r22m1s5i8xpla881xnmcj86h6g8sd9pm-env-development/share/man/man3/err.3.gz' and '/nix/store/m3pcpc78an04wzvbg5rvzc4zky13cc1d-env-network/share/man/man3/err.3.gz' have the same priority 5; use 'nix-env --set-flag priority NUMBER INSTALLED_PKGNAME' to change the priority of one of the conflicting packages (0 being the highest priority)
<CMCDragonkai>
My config.nix has separated bunches of packages from each other into specific environments. By doing this, I believe some things are conflicting between them, such as in the case for these man page files.
<immae>
CMCDragonkai: the error line tells you what to do ;)
<immae>
in you case, INSTALLED_PKGNAME should be env-development or env-network I guess
<CMCDragonkai>
Is there a way to do that inside config.nix?
<CMCDragonkai>
I'd want to make sure I don't need to do this in the future?
<immae>
I’m sure you can, but I don’t remember how
<CMCDragonkai>
what about ignoreCollisions? What does that do?
<CMCDragonkai>
immae: I found that you just have to do `meta.priority = 50;`
<CMCDragonkai>
In the buildEnv and it works
polman has quit [Ping timeout: 245 seconds]
<immae>
ok cool
polman has joined #nixos
<immae>
I may remember it for next time :D
<CMCDragonkai>
I found there's also an option for ignoreCollisions for buildEnv
<CMCDragonkai>
but what does that actually do?
<CMCDragonkai>
Are conflicting packages solved by priority a different thing from ignoring collisions?
<immae>
I don’t know, the best way is to test. I suppose he will pick one using an arbitrary rule
<immae>
so if I understand correctly, it checks content/rights, and if they match he’s happy. otherwise it’s checkCollisionContents that will be taken into account
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
<CMCDragonkai>
also is there a way to find out which package is causing the collision
<CMCDragonkai>
it says that env-base, but that's a buildEnv
<CMCDragonkai>
consisting of lots of other packages
<immae>
well, in your case he’s trying to merge r22m1s5i8xpla881xnmcj86h6g8sd9pm-env-development and m3pcpc78an04wzvbg5rvzc4zky13cc1d-env-network, and the error happens here, so he has no clue where they come from and he cannot know. If you want to investigate, do `ls -l /nix/store/r22m1s5i8xpla881xnmcj86h6g8sd9pm-env-development/share/man/man3/err.3.gz` and it will probably be a symlink to the origin package
taktoa has joined #nixos
<immae>
My guess is that you have a package identical in env-development and env-network, so when you try to merge you have duplicates
<CMCDragonkai>
well right now priority solves it
<CMCDragonkai>
but its really annoying
<CMCDragonkai>
qt is always conflicting everywhere
<CMCDragonkai>
too many global libraries...
<immae>
yes
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
<immae>
May I ask why you’re trying to do that? Last time a colleague asked me about a similar issue, and the response was that he did unnecessary things because of a misunderstanding
<CMCDragonkai>
just to separate my package list into more consumable chunks
<immae>
ok
<immae>
then you’ll likely have more of them sorry
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
<immae>
If you really want to avoid it, you’ll have to rethink the way you’re writing your buildEnv’s
<CMCDragonkai>
like put all the qt apps together?
<CMCDragonkai>
or maybe compose the buildenvs together dynamically according to some input expression
<immae>
the second one yes
<immae>
First one I have no opinion on it, it’s your way of seeing your stuff :)
<{^_^}>
[nixpkgs] @thefloweringash opened pull request #58472 → emacs: only use patchelf on linux → https://git.io/fjTGn
<immae>
rather than building two buildEnv and merging them, you could build one buildEnv using combined lists, that would solve your conflict issue since you would only have one env.
<immae>
(it might not be as easy as it seem though, it depends on the complexity of your env’s)
<immae>
so the "composing atoms" would be list of packages rather than list of env’s
<CMCDragonkai>
yes, but the reason to separate them was so that sometimes i only want to install 1 or 2 package lists, instead of all of them
<CMCDragonkai>
like i have package lists only relevant for certain kinds of work
<immae>
yes, that’s what I understood, and you should be able to do it the way I suggested
<immae>
say you have a list "base", a list "work", a list "leasure". Then you build one env using base+work on your work laptop, and one using base+leasure on your gaming computer
<immae>
that’s the kind of thing you want, right?
mabel has quit [Quit: WeeChat 2.4]
polman has quit [Ping timeout: 245 seconds]
polman has joined #nixos
shpx has joined #nixos
root has joined #nixos
root is now known as Guest29214
<Guest29214>
when i adding my user to wheel and network manager group by extraGroups command in configuartion file and build it, i cant login to my user; the output of "su username" command is "this account currently not available"! what should i do?
<adamt>
Or well, maybe, but both got layout defined as symphorien says
<symphorien>
> { test.foo = 1; test = { baz = 2; bar = 3; }; }
<{^_^}>
error: attribute 'test' at (string):225:17 already defined at (string):225:3
shpx has quit [Ping timeout: 250 seconds]
cfricke has joined #nixos
<mehran>
adamt: thanks ! it built sucsessfully but stil i can't change my keyboard layout with alt + shift with this line on my i3wm configuration: "exec --no-startup-id setxkbmap -layout us,ir -option 'grp:alt_shift_toggle'"
realrokka has quit [Read error: Connection reset by peer]
<mehran>
oh my bad i forgot to restart my i3wm. solved. thanks
<jasongrossman>
mehran: \o/
<symphorien>
when you change the xserver configuration, you should restart the xserver for changes to have an effect
elgoosy has quit [Remote host closed the connection]
pmiddend has joined #nixos
Serus has joined #nixos
<pmiddend>
I juts tried to use "steamcmd" with a Nix-built Docker image. This failed because "mkdir" wasn't found. Does steamcmd need an explizit dependency on coreutils?
<pmiddend>
Seems to work fine on its own, without Docker.
fendor has quit [Read error: Connection reset by peer]
<Qubasa>
I have setup gitlab on stable with gitlab version 1.7.5 which was too old for me so I switched nixos to the unstable channel but the gitlab service runs still at version 1.7.5 even though in nixpkgs the gitlab version is already 1.9.1 changed 1 day ago and 1.8.x changed 11 days ago. I looked at the hydra page and the unstable channel got merged 1 day ago. Does someone have an idea what is going on?
<musicmatze>
maybe it is a new attribute? Like gitlab19 or something like that?
<Qubasa>
Hmm its a service so that shouldn't be the problem?
<musicmatze>
in `nix repl` I get "v11.7.5" for `pkgs.gitlab.src.rev` on nixpkgs-unstable channel
<Qubasa>
yeah 11.7.5 thats what I meant, my bad
<musicmatze>
I am confused now... 1 and 11 are a bit of a diference
<gchristensen>
do less stuff inside the nix-shell call
<sphalerite>
does ./bar 75 work?
<gchristensen>
doing more things in there makes quoting more complicated
<jomik>
in cli, yes. let me try in the config..
<jomik>
It does not... :D
<sphalerite>
there's also the option bat=/sys/class/power_supply/FILL_THIS_IN ; echo $(( $(<$bat/energy_now) * 100 / $(<$bat/energy_full_design) )) if you want a percentage of the design capacity rather than the current capacity of the battery
<sphalerite>
__monty__: well, no, not all of nixpkgs. But it does go through all the overlays and stuff to evaluate the derivation.
<sphalerite>
__monty__: not much. Compare `nix run -vv nixpkgs.hello` and `nix-shell -vp hello`. The main difference is that nix run won't instantiate a new drv per shell environment, since all it does is add one derivation's bin dir to the PATH
<sphalerite>
__monty__: so for stuff that gets run periodically/frequently/automatically, prefer installing it into a profile or wiring the path in using a wrapper or similar
cfricke has quit [Ping timeout: 272 seconds]
cfricke has joined #nixos
<jomik>
sphalerite, gchristensen: soo... My issue was that my pango markup had an error ! :D
<gchristensen>
ah!
<gchristensen>
please do less inside the --run argument, though!
<Qubasa>
lassulus: thanks I will try it out
<sphalerite>
or, y'know, don't use nix-shell for this :p
<jomik>
Yeah, that makes sense gchristensen :P
<Qubasa>
!
<jomik>
I went with what sphalerite suggested in the end tho, just catting that file :P
reinhardt has quit [Ping timeout: 246 seconds]
<sphalerite>
or if it's running in bash, use $(< $file) to avoid a fork+exec :D
reinhardt has joined #nixos
<{^_^}>
[nixpkgs] @dotlambda pushed 2 commits to release-19.03: https://git.io/fjT82
<jomik>
I think it is passed to sh -c, but on NixOS that is bash? :P
<jomik>
I have `./bar $(cat /sys/class/power_supply/BAT0/capacity)` now
<jomik>
sphalerite: Any nice way to read % of screen brightness? :P
<sphalerite>
jomik: same style as the slightly more complex design-capacity-percentage thing, but with /sys/class/backlight/foo/{max_,brightness}
zupo_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
iyzsong has quit [Ping timeout: 250 seconds]
<jomik>
Thanks :D
<adamt>
I'm trying to set RequiredBy on a system unit using the systemd module, but the generated unit doesn't have a RequiredBy= line. Wants/Before/After all works. Suggestions?
<musicmatze>
How can I combine a list of outputs into a single output? E.G.: I have a "buildme.nix" file which has a list where I can build each element using `nix-build`... but how to combine all of them to build them all using `nix-build ./buildme.nix`?
cfricke has quit [Quit: WeeChat 2.4]
<buckley310>
musicmatze, so you want to have a literal copy of two other outputs inside of your output?
Rusty1 has joined #nixos
<musicmatze>
I don't care whether it is a copy of everything or just symlinks to other stuff, I basically just want to build a list of things and get all outputs ... whether this results in several ./result symlinks or one ./result with everything inside ... I don't mind eitehr
<ar>
er/36
<musicmatze>
(context: I'm trying to use `pkgs.releaseTools.debBuild` recursively to build a whole dependency-chain as deb packages).
iclanzan has joined #nixos
npmccallum has joined #nixos
<buckley310>
ok, shouldnt be too terribly difficult
<buckley310>
went the symlink route, although im not sure what process you're using to pull all the outputs together indo a deb...
<musicmatze>
well, I cannot hardcode the list, because for that I would need to know all transitive dependencies of the package I am about to build. And I don't want that... because I want to write that later as an nixpkgs overlay so that I can build any nixpkgs package into a .deb and it automatically builds all dependency (down to libc and everything) for the package and I can use the resulting .deb files and put them into a real
<blackriversoftwa>
srhb: I am considering making a little module generator (sort of like what is done with hardware-configuration.nix) that would poll the metadata service for the cloud provider I'm working on
jbgi has quit [Ping timeout: 250 seconds]
<{^_^}>
[nixpkgs] @flokli pushed 2 commits to release-19.03: https://git.io/fjTBx
jmeredith has joined #nixos
<{^_^}>
[nixpkgs] @vcunat pushed 318 commits to staging-19.03: https://git.io/fjTRv
o1lo01ol1o has joined #nixos
o1lo01ol1o has quit [Remote host closed the connection]
o1lo01ol1o has joined #nixos
<d10n-work>
I installed sublime3 with Nix on CentOS6. When starting subl, I get a message from Package Control saying my system's locale doesn't support non-ASCII characters. My system locale is en_US.UTF-8, but when I add the locale command to the sublime_text wrapper, it outputs "locale: /lib64/libc.so.6: version `GLIBC_2.14' not found (required by /nix/store/avw6pay7yq18pg42i36i3vnldp97idxq-libredirect-0/lib/libredirect.so)"
<mog>
gchristensen, what happened that caused r13y.com to crash so hard?
<{^_^}>
[nixpkgs] @dotlambda opened pull request #58485 → [18.09] dovecot: add patch for CVE-2019-7524 → https://git.io/fjTR8
<gchristensen>
ouch
<gchristensen>
mog: it looks like the builder it is on is just not really able to do it
<gchristensen>
and instead of failing the build of r13y, it just powers on
* mog
nods
<gchristensen>
it is a bug in r13y's build tools I guess. I'll fix that up when I can
<d10n-work>
To put my statement in question form: how can the sublime3 nix package be made to support my locale?
<mog>
that is much better than lots of random changes
<gchristensen>
mog: see that "714 unchecked" line
<symphorien>
d10n-work: I am surprised it uses /lib64/libc.so.6 instead of /nix/store/something
<mog>
ahh i assumed that meant that number wasnt part of the 1276
<gchristensen>
ohh
<gchristensen>
maybe should change that wording
<symphorien>
d10n-work: maybe put nix's libc in LD_PRELOAD ?
rauno has quit [Remote host closed the connection]
<d10n-work>
symphorien: I added one of the libc.so.6 in /nix/store to the sublime_text wrapper like: export LD_PRELOAD='/nix/store/avw6pay7yq18pg42i36i3vnldp97idxq-libredirect-0/lib/libredirect.so /nix/store/sjsvwd9sqxjflmslnf8xs48aqpkb2sjy-glibc-2.27/lib/libc.so.6' --- and I get: ERROR: ld.so: object '/nix/store/sjsvwd9sqxjflmslnf8xs48aqpkb2sjy-glibc-2.27/lib/libc.so' from LD_PRELOAD cannot be preloaded: ignored.
mmlb977450 has joined #nixos
boogiewoogie has joined #nixos
<d10n-work>
when adding that libc.so.6 to LD_PRELOAD and then starting locale as `/nix/store/sjsvwd9sqxjflmslnf8xs48aqpkb2sjy-glibc-2.27/lib/ld-2.27.so locale` I get "locale: error while loading shared libraries: locale: cannot open shared object file: No such file or directory"
<sphalerit>
musicmatze: I suppose I don't understand what it is that you need, buildEnv *sounds* like what you want but I'm not sure :)
<musicmatze>
Linus: I have to investigate that. It might help, yes. But first I have to get the recursive build working
<musicmatze>
xD
<sphalerit>
oh that sounds like Fun in and of itself.
<musicmatze>
it does.
<{^_^}>
[nixpkgs] @bgamari opened pull request #58487 → gitaly: Run gitaly with procps in scope → https://git.io/fjTRN
domogled has quit [Remote host closed the connection]
boogiewoogie has quit [Quit: Leaving]
<d10n-work>
Adding 'export LC_CTYPE=en_US.UTF-8' to the sublime wrapper shows this warning: openat(AT_FDCWD, "/nix/store/681354n3k44r8z90m35hm8945vsp95h1-glibc-2.27/lib/locale/en_US.UTF-8/LC_CTYPE", O_RDONLY|O_CLOEXEC) = -1 ENOENT (No such file or directory) [...] write(2, "./sublime_text: line 9: warning:"..., 116./sublime_text: line 9: warning: setlocale: LC_CTYPE: cannot change locale (en_US.UTF-8): No such file or directory
<{^_^}>
Found in packages: ncurses.dev, ncurses5.dev
ma27 has quit [Remote host closed the connection]
<fascinated_>
Bored? Call the official freenode IRC partyline at +4521137886
fascinated_ has joined #nixos
stateless23 has joined #nixos
<stateless23>
Bored? Call the official freenode IRC partyline at +4521137886
stateless23 was kicked from #nixos by gchristensen [stateless23]
fascinated_ was kicked from #nixos by gchristensen [fascinated_]
ma27 has joined #nixos
<infinisil>
gchristensen++
<sphalerite>
oh boy, is this starting again?
<{^_^}>
gchristensen's karma got increased to 88
<edlin>
I'm having difficulties with virt-manager and QEMU on stable nixos, I can't seem to get it to recognize that I have KVM and libvirt available.
<edlin>
libvirt appears to be running
<edlin>
and I'm loading kvm and kvm-intel in my modules
<sphalerite>
EdLin: your user probably doesn't have access. Try adding yoruself to the libvirtd group and logging out and back in
<edlin>
sphalerite: ah, thanks.
<sphalerite>
at least, that would be my best guess.
<sphalerite>
An actual error message could confirm/correct my suspicion ;)
Mr_Keyser_Soze has joined #nixos
edlin has quit [Quit: Konversation terminated!]
hoijui has quit [Ping timeout: 245 seconds]
edlin has joined #nixos
<edlin>
sphalerit: it didn't help
<edlin>
sphalerite ^
<edlin>
I confirmed with "groups" also that I was part of the group, even restarted libvirtd
<adamt>
What's the preferred way of packaging a java application these days? Right now I'm using maven for dependencies, but I can change it to whatever is easist with nix.
<sphalerite>
EdLin: hm ok, is it giving you an error of some sort?
<srhb>
adamt: I can give a recommendation for what to use, but avoid sbt like your life depended on it.
<edlin>
sphalerite: it says that no hypervisor is detected on startup
<srhb>
adamt: Anything that will allow you something like a lockfile and where you can generate a list of dependencies/urls for use with fetchurl and friends should suffice.
<sphalerite>
is there a details thing or something?
<edlin>
in virt-manager?
<edlin>
no
<edlin>
"could not detect a default hypervisor"
<edlin>
with the advise to install one with various packages
<srhb>
adamt: Er, can't give you a recommendation on what to use* ;)
<edlin>
all of which are installed
<d10n-work>
Installing nixpkgs.glibcLocales and adding 'export LOCALE_ARCHIVE=$(nix-env --installed --no-name --out- path --query glibc-locales)/lib/locale/locale-archive' to the sublime_text wrapper fixed the locale problem. Is there a good way to make this a non-manual step?
<adamt>
srhb: Damnit, I was waiting for the recommandation :P
<srhb>
adamt: Sorry. Migraine, not very useful today. :P But honestly, i think any of the lockfile-things will work.
<srhb>
Might just not have nice tooling yet.
fendor has quit [Ping timeout: 246 seconds]
<adamt>
I have no idea which tools support lock files though. I wouldn't have written it in java if I had a choice :P
illegalprime has joined #nixos
endformationage has joined #nixos
<edlin>
OK, in Nixos, how do I edit qemu.conf?
<edlin>
or better yet, change parameters using configure.nix
<edlin>
sigh, still no detection, and I did restart libvirtd....
<{^_^}>
[nixpkgs] @LinArcX opened pull request #58488 → nahid-fonts:init at v0.3.0 → https://git.io/fjT0a
<edlin>
tried changing user for qemu.conf to my user.
Mr_Keyser_Soze has quit [Ping timeout: 245 seconds]
<{^_^}>
[nixpkgs] @aanderse opened pull request #58489 → nixos/mailcatcher: fix test to be compatible with mailcatcher 7.x series → https://git.io/fjT0V
buffet_ is now known as buffet
<illegalprime>
I want to add a defaultBuildInput (bash) to stdenv, but that input _depends_ on stdenv! what should i do?
civodul has quit [Quit: ERC (IRC client for Emacs 26.1)]
fusion809 has quit [Remote host closed the connection]
<sphalerite>
aaargh, this is dirving me mad. How am I supposed to set a kernel config option to the empty string in nixpkgs? Using freeform "" results in it getting set to an empty string
<matthewbauer>
illegalprime: I looked into that but decided it wasn't easy to do
<sphalerite>
samueldr: maybe you have some enlightenment to provide (once again)?
<illegalprime>
matthewbauer: yea! i was trying to hack on your patchshebangs work
<matthewbauer>
there may be some stdenv tricks to get it to work though. for now the patchshebangs has a fallback to use the native bash
<matthewbauer>
honestly i wonder if we shouldn't patch and just do ```#!/usr/bin/env sh```. that should work *everywhere*
zupo_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
andi- has quit [Ping timeout: 258 seconds]
<illegalprime>
matthewbauer: lol that's true
johanot has quit [Quit: WeeChat 2.4]
<samueldr>
sphalerite: no idea about that one, that seems... different?
<illegalprime>
the other thing i was wondering about was that i get the cross-gcc in my cross-compiled closures. so in an aarch64 build of tree, my cross-compiling gcc is a _runtime_ dependency
adamt has quit [Quit: WeeChat 2.4]
<illegalprime>
this is just because the cross comp gcc has some aarch64 libs in it that packages need, but it also has x86 binaries that it definitely shouldn't need, so maybe it should be spliced?
rcshm has joined #nixos
<samueldr>
sphalerite: and I'm not sure how "set to an empty string" is different than the goal of setting it "to the empty string"
<sphalerite>
samueldr: I want: CONFIG_CROSS_COMPILE="". I get: CONFIG_CROSS_COMPILE="\"\""
<samueldr>
oh!
<illegalprime>
I'm not sure where the cross gcc gets built, it has a weird name like armv6l-unknown-linux-gnueabihf-stage-static-gcc-debug-7.4.0
<sphalerite>
oops, only just realised I mistyped my problem previously.
rprije has quit [Remote host closed the connection]
rprije has joined #nixos
<matthewbauer>
illegalprime: yeah gcc has some runtime libraries like libgcc that need to be included. they should just be in the "lib" output though so it shouldn't inflate closure size too much. This is partly why we want to move to using LLVM which has these things separated out better
<{^_^}>
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.
<Taneb>
infinee: the 8ac39db... is a reference to the Git commit
<Taneb>
And is 40 characters
andi- has joined #nixos
<srhb>
But if you're using fetchFromGitHub, just use v3.0.6-stable for the rev, if that's what you're asking for.
rcshm has quit [Read error: Connection reset by peer]
rcshm has joined #nixos
<illegalprime>
matthewbauer: there are some libraries in the non-lib output (the one with prefixed gcc like armv6l-unknown-linux-gnueabihf-gcc)
rprije has quit [Ping timeout: 246 seconds]
<infinee>
srhb: I'm trying to fetch 3.1
<{^_^}>
[nixpkgs] @srhb pushed commit from @johanot to release-19.03 « kubernetes: 1.13.4 -> 1.13.5 »: https://git.io/fjTEm
Neo-- has quit [Ping timeout: 268 seconds]
<srhb>
infinee: 3.1-stable then.
rprije has joined #nixos
jomik has quit [Quit: WeeChat 2.4]
zupo has joined #nixos
<illegalprime>
like this one: /nix/store/HASH-armv6l-unknown-linux-gnueabihf-stage-final-gcc-debug-7.4.0/armv6l-unknown-linux-gnueabihf/lib/libatomic.so
<matthewbauer>
illegalprime: that sounds like a bug. what gcc version are you using?
<illegalprime>
matthewbauer: thanks for checking that, i'm still rebasing! do you think its just a matter of moving the right libraries to the lib output? i could try to take a look at this
edlin has quit [Quit: Konversation terminated!]
<matthewbauer>
yeah i'm pretty sure that's all it is. i'm surprised no one has noticed that before!
<sphalerite>
idea: patchSource or something like that, in pkgs/build-support, which takes a source and applies a patch/list of patches, copying the result to the output.
<srhb>
eyJhb: In exactly the same way? Not sure what you're asking.
<illegalprime>
matthewbauer: i have a script that tries to find build system contamination, this is what i found so far: https://pastebin.com/raw/9QGLWMx1
<srhb>
eyJhb: Simpler yet, just forgo builtins.tarball and refer to the path directly..
<eyJhb>
srhb: but wouldn't that require me to tar ALL the files in my local directory. I wnat to use e.g. `~/projects/nixpkgs` instead
<sphalerite>
samueldr: yeaaah I was hoping not to have to do any digging, but hey, I guess an opportunity to fix stuff in nixpkgs is an opportunity to fix stuff in nixpkgs
jungly has quit [Remote host closed the connection]
<srhb>
eyJhb: nixpkgs.config will only affect the pkgs imported by the module system itself, not your localPkgs, yeah.
<infandum>
I'm using emacs org mode. It wants to install the ZMQ module (through emacs in the buffer). I have zeromq, autoconf, automake, and pkgconfig installed, but it erroras out, saying I have possibly undefined macros of "AC_SUBST" etc. (AC_*) and missing some pkg-config macros (pkg-config package). Any ideas on what I should do?
rembo10 has joined #nixos
rcshm has quit [Read error: Connection reset by peer]
alex`` has quit [Ping timeout: 245 seconds]
<genesis>
i've a derivation #58367 with wrapGAppsHook and i'd like poedit find msgfmt ( gettext ) , how add such path ?
<genesis>
i don't find documentation about this wrapper to add option
tdammers has left #nixos ["WeeChat 1.6"]
<gchristensen>
hello Nix friends! I am thrilled to announce lorri, a project by Profpatsch, shlevy, Tweag, Target, and myself -- it aims to be a nicer nix-shell for your projects! check it out: https://www.tweag.io/posts/2019-03-28-introducing-lorri.html
<drakonis1>
ooooooh shiny
Ariakenom_ has quit [Ping timeout: 268 seconds]
<samueldr>
gchristensen++
<{^_^}>
gchristensen's karma got increased to 89
<drakonis1>
gchristensen++
<{^_^}>
gchristensen's karma got increased to 90
<drakonis1>
direnv is built with nix
<drakonis1>
how neat.
<drakonis1>
the page needs some updating tho
<drakonis1>
gchristensen: can it be shipped by default?
<infinisil>
gchristensen: That's awesome!
<gchristensen>
=)
<samueldr>
drakonis1: what page needs uptating, and how?
<drakonis1>
so good, but can we ship it instead of nix-shell?
<symphorien>
I want to test whether a C lib is not broken cross-compilation wise: what is the target in pkgsCross with the most stuff in the binary cache ?
<gchristensen>
is any of pkgsCross in the binary cache?
<symphorien>
I am trying with musl and not everything was built locally
<symphorien>
but musl is not a good indicator of whether cross is really working
edlin has joined #nixos
linarcx has quit [Ping timeout: 246 seconds]
Tucky has quit [Quit: WeeChat 2.2]
<edlin>
I'm trying to get virt-manager to work here, it isn't detecting KVM or any hypervisor. My configuration.nix is...
<genesis>
EdLin : do you not forget nixos-rebuild ?
<genesis>
(just ask)
<infandum>
How can I add to the stdenv.mkDerivation.buildInputs list instead of replacing it?
<edlin>
genesis: I did that.
<edlin>
nixos-rebuild switch
<srhb>
EdLin: I notice that you have some things in initrd.kernelModules that I would expect in kernel.modules. Like, say, kvm-ingel
<srhb>
intel*
<srhb>
EdLin: Is it loaded?
<eyJhb>
srhb: is there any way, that I can get nix-channels to use a local git too?
<srhb>
eyJhb: Not that I know, they're sort of orthogonal.
<samueldr>
EdLin: ls -l /dev/kvm, I think it will be in the group `kvm
<{^_^}>
[nixpkgs] @d10n opened pull request #58491 → sublime3: Fix python api locale → https://git.io/fjTuo
<samueldr>
you may need to add yourself to the kvm group if you intend to run it as your user
phreedom has quit [Remote host closed the connection]
<samueldr>
oh, disregard, it might also be rw-rw-rw-
phreedom_ has joined #nixos
<clever>
samueldr: on nixos, kvm is 666
<samueldr>
clever: ^ ninja'd you :)
agander has quit [Ping timeout: 244 seconds]
<eyJhb>
srhb: guess I have to setup my Git repo again, and then use that. It is easy to get single packages to work, but now I need to override how displaylink works as a videodriver...
rcshm has quit [Read error: Connection reset by peer]
rcshm has joined #nixos
edlin has joined #nixos
agander has joined #nixos
Erasmus has quit [Quit: END-AUMF]
<edlin>
I added myself to the kvm group, logged out, logged back in... virt-manager still isn't detecting my hypervisor
<edlin>
I'm also in the libvirtd group
<edlin>
and qemu.conf *should* be edited to have my user in it
endformationage has quit [Quit: WeeChat 2.4]
andi- has joined #nixos
<samueldr>
EdLin: /dev/kvm exists and `lsmod | grep ^kvm` has a module for your CPU vendor?
rcshm has quit [Read error: Connection reset by peer]
<edlin>
samueldr: /dev/kvm exists... and kvm and kvm_intel are both there
rcshm has joined #nixos
marler8997 has quit [Quit: Page closed]
rcshm has quit [Read error: Connection reset by peer]
<samueldr>
to help track whether it is libvirt doing something, or kvm not working, I tried by rmmod-ing kvm_intel, and using `nix-shell -p qemu --run 'qemu-kvm -enable-kvm'`
<samueldr>
when kvm is deactivated, qemu will not start
rcshm has joined #nixos
<samueldr>
(modprobe-ing it back made qemu-kvm work again)
<clever>
samueldr: the qemu-kvm bash script, simply tests for /dev/kvm, and conditionally passes -enable-kvm, so it wont fail like that
<samueldr>
clever: I tried and it failed it, so uh not sure exactly
<samueldr>
probably failed due to passing --enable-kvm
<clever>
samueldr: you also passed your own -enable-kvm, which isnt conditional
<samueldr>
just as if I had used qemu (not-kvm) directly
<edlin>
samueldr: all of that tested OK, with rmmod kvm_intel, it errors, and with it installed, it starts and looks for a pxe boot.
<edlin>
what could be wrong with libvirt then?
<samueldr>
"good", then kvm probably works on the machine and as your user, that's at least a chunk of the problem reduced
<samueldr>
good question, sorry, I don't have the knowledge to say with libvirtd :/
<edlin>
sigh.
o1lo01ol1o has quit []
<samueldr>
clever: let's say I used `qemu-kvm` as a platform-agnostic stand-in for qemu-system-x86_64 :) (same instructions would then work for aarch64 I guess!)
reinhardt has joined #nixos
<Taneb>
...I think I've messed up some of my overlays :(
o1lo01ol1o has joined #nixos
<{^_^}>
[nixpkgs] @LinArcX opened pull request #58493 → gandom-fonts: init at v0.6 → https://git.io/fjTzt
<clever>
nefix: you usually bake the paths into your package, either direct references, or prefix PATH with a bash wrapper
<nefix>
clever: not sure I've followed you here. The package I'm trying to build needs another package to be available when run. How can I specify this dependency in the .nix file?
<clever>
nefix: modify all of the binaries to be wrapped with a bash script, to put that package into PATH
<symphorien>
clever: I want to adapt the wafHook to cross, so I would need to set configurePlatforms to [], but by reading the code this can only be done at the nix level, any idea how I could do this ?
<{^_^}>
[nixpkgs] @Infinisil pushed commit from @bgamari to release-18.09 « nixos/gitaly: Run gitaly with procps in scope »: https://git.io/fjTzA
<clever>
symphorien: not sure how you would handle that
<nefix>
clever: could you provide me an example or point me the documentation? Thanks!
<clever>
nefix: search nixpkgs for wrapProgram
<nefix>
Thanks!
<gchristensen>
DigitalKiwi: =)
<gchristensen>
DigitalKiwi: I'm glad!
judson__ has joined #nixos
cromiumlake has joined #nixos
Glider_IRC_ has quit [Quit: Leaving]
cromiumlake has quit [Remote host closed the connection]
<catern>
the GC roots and direnv parts of lorri seem cool, but I'm not sure I understand the workflow that uses watch
<{^_^}>
[nixpkgs] @worldofpeace pushed 3 commits to release-19.03: https://git.io/fjTgv
<catern>
first off, I usually pin a version of Nixpkgs in my repos, so I have a consistent environment - although hydra and Nix itself don't do that - why not do that?
<pie_>
can I add extra attributes to a derivation?
<catern>
second, is there some use case other than for <nixpkgs> that I'm missing?
<infinisil>
catern: I feel like watch should be used as a continuously running systemd service
<catern>
(that is to say, the feature seems cool but is it useful for other things?)
<nefix>
clever: so I can't wrap my head around adding font-awesome, since it's not a binary
<pie_>
i want to be able to do package and package.withPlugins [ package.plugins.someplugin ], and in this case id need to extend package with .plugins and .withPlugins
<gchristensen>
infinisil: indeed, that would be really nice -- and something which is documented in the repo lorri was developed in. those issues will move over soon
<catern>
but... having lorri being a continuously running systemd service sounds a lot like just a normal CI server
<simpson>
catern: I think that one has to experience the direnv-powered workflow first.
<gchristensen>
^
<gchristensen>
lorri is not CI, and does not build your project.
<simpson>
It's both really nice, and also has made me aware of nix-shell incantations.
<gchristensen>
you don't _need_ to run `lorri watch` all the time
<pie_>
im looking forward to not losing shells :V
<catern>
gchristensen: sure, I wasn't implying it did. but it does continuously build an artifact (the shell environment) whenever the inputs change, which is also what hydra does
<pie_>
however it is that i manage to do that
<catern>
but I don't want to get into that, I just want to understand the workflow that's based on using <nixpkgs>
<simpson>
gchristensen: FWIW I'm gonna watch lorri for a bit but it sounds too opinionated for me compared to the bare-bones direnv+Nix integration. `use nix` already *really* really good.
<gchristensen>
simpson: give it a try, file bugs for what seems opinionated.
<pie_>
gchristensen, nice
alex`` has joined #nixos
<pie_>
gchristensen, thats meeee lol
<catern>
I'll probably use lorri, because the GC roots and the faster direnv stuff are both excellent improvements (I do use direnv "use nix" a little right now, but I don't use it automatically because it can be slow)
<pie_>
my pain is known \o/
<catern>
but I'm still not clear on why one would use <nixpkgs> at all, rather than pin the repo to a specific version of Nixpkgs. this isn't really lorri-related question
<pie_>
not having actually looked at lorri, why isnt this stuff in upstream nix?
<gchristensen>
catern: even if you don't do that, your dependencies may change often through things like cabal file automation
ym555_ has quit [Quit: leaving...]
<pie_>
youll still lose your shells if you pin and gc i think
<catern>
gchristensen: aha :)
<catern>
so that's what I meant, "what is the other use case?"
<catern>
I don't write much Haskell, but that makes sense - what exactly does this "cabal file automation" do?
<simpson>
gchristensen: The entire concept feels opinionated; I often need to build PyPys and I prefer to have fine-grained control over that. If I make a change to my local nixpkgs, then lorri will rebuild nix-shells in other windows, won't it?
<gchristensen>
pie_: in a project like nix, it is hard to experiment because it is built right in to the tool.
<pie_>
gchristensen, so theoretically stuff *could* go back upstream
<{^_^}>
[nixpkgs] @LinArcX opened pull request #58497 → behdad-fonts: init at v0.0.3 → https://git.io/fjTg3
<{^_^}>
[nixpkgs] @Infinisil pushed commit from @xeji to release-19.03 « qtikz: fix build »: https://git.io/fjTgs
<gchristensen>
pie_: that is part of the mission!
<pie_>
\o/
<dminuoso>
catern: I do much haskell, I dont know what "cabal file automation" is either.
<pie_>
i get the space stuff and the name but i dont actually get what it has to do with the project
<gchristensen>
the Long Range Reconnaissance Imager (LORRI) is a camera on the New Horizons spacecraft, which took pictures of Pluto and its moons. lorri is a far-away view of what Nix could be.
<simpson>
gchristensen: I suppose that part of this is that I am committing an interesting sort of sin: I will put my local nixpkgs onto a certain branch, start a nix-shell based on that branch, then change to another nixpkgs branch and start another nix-shell or repl.
<catern>
One use case that I thought of with lorri that might be useful is if, say, my .nix refers to other projects I have checked out locally which I want to rebuild, it would be helpful if they automatically rebuilt and updated the environment as soon as I changed those other projects. although, I'd probably only want the rebuild to happen once I commit
MichaelRaskin has joined #nixos
<gchristensen>
simpson: then lorri would make that *easier*, actually, if you don't run `watch` continuously for each project.
<pie_>
gchristensen, :D
vmandela has quit [Quit: Leaving]
ym555 has joined #nixos
<gchristensen>
simpson: because then the evaluation from when nixpkgs was at the rev you liked, and the cached evaluation will continue being used
<catern>
But that doesn't seem too useful because, wouldn't I still need to block on waiting for the build to happen? Typically I make a change in project A then immediately switch to project B to rebuild against A and re-test, it doesn't seem lorri would speed that up?
<catern>
(again, lorri is super cool for the GC roots and direnv stuff, I am just trying to understand how "watch" could fit into my workflow)
<gchristensen>
depends, for example, you might realize "ok, building feature Foo, need dep Fizz..." add Fizz, hack on feature Foo, and by the time you're done working on the feature,it'll be ready.
<gchristensen>
if your cycle time is *edit one line of code* *try to build immediately* it won't speed things up, no
<simpson>
gchristensen: If I'm not gonna use the watch feature, though, then it sounds entirely superfluous to how I'm currently operating. (To say nothing of the fact that the reason that I can't always work on a single branch is because not enough of my stuff's going upstream, which is largely unrelated to my local tool choices.)
<gchristensen>
it definitely won't improve *every* possible workflow
<catern>
gchristensen: could you elaborate on the cabal integration that lorri helps with? I'm just trying to understand the workflow that lorri was designed against inside Target :)
<MichaelRaskin>
If I had a simpler setup, I would be interested in experimenting with lorri for managing user set of packages
andi- has joined #nixos
ajs124 has joined #nixos
nefix has quit [Ping timeout: 256 seconds]
<MichaelRaskin>
But I split my package set in quite a few layers so that heavy stuff doesn't block light stuff
<gchristensen>
nice
<MichaelRaskin>
This split is actually a few years old (and visible in my configuration repo)
dannx has joined #nixos
dannx has quit [Remote host closed the connection]
<edlin>
anyone here get libvirt/kvm/qemu/virt-manager working?
<edlin>
as I was debugging earlier, KVM seems to be operating right, but virt-manager isn't detecting my hypervisor.
ddellacosta has quit [Ping timeout: 246 seconds]
ddellacosta has joined #nixos
<pie_>
what do I need to do to get the contents of my somethingPackages to show up in the website's package index?
rprije has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @mweinelt opened pull request #58499 → [19.03] virtualenv: Update patch for 16.2.0 → https://git.io/fjTgd
rprije has joined #nixos
wolfman154 has joined #nixos
<infinisil>
pie_: recurseIntoAttrs
Ariakenom has joined #nixos
<pie_>
infinisil, for the websites package index or addign stuff to a derivation? the latter makes sense i gues
<pie_>
oh im dumb. i guess i can just use // externally
<infinisil>
No idea what you're talking about
<pie_>
infinisil, ok me neither
<pie_>
infinisil, what was yours a reply to
<infinisil>
The thing you asked?
<pie_>
then I dont follow
<pie_>
infinisil, pls expand
<infinisil>
By default nix-env and such don't recurse into attrs to find packages, but with recurseIntoAttrs you mark that attrset as one it should recurse into, so it can find packages within
<infinisil>
See `pkgs/top-level/all-packages.nix` for examples of recurseIntoAttrs
<pie_>
infinisil, derp i totally misread that as recursiveUpdate
infandum has quit [Ping timeout: 255 seconds]
Erasmus has quit [Read error: Connection reset by peer]
<gchristensen>
bgamari_: neededForBoot defaults to false, but that just ensures they're mounted super early. if you don't care if they mount, fileSystems.<name?>.options = [ "nofail" ] is a thing
<gchristensen>
that you want
<bgamari_>
gchristensen, great, thanks
<gchristensen>
you're welcome!
pie_ has joined #nixos
<bgamari_>
wolfman154, I'm not sure what copytoram does; I suspect it might copy the installation volume to a tmpfs or something along these lines
<bgamari_>
so in principle you could remove the installation volume
<bgamari_>
not sure though
<{^_^}>
[nixpkgs] @jokogr opened pull request #58500 → vscode-extensions.vscodevim.vim: init at 1.2.0 → https://git.io/fjT2b
muelleme has quit [Remote host closed the connection]
<ryantm>
Does it need to be in a channel to be on cache.nixos.org?
edlin has joined #nixos
<MichaelRaskin>
No
<ryantm>
So is the answer to d10n-work 's question "Right after Hydra builds it"?
<MichaelRaskin>
But channel delay is indeed indicative of rebuild delay
dckc has joined #nixos
<MichaelRaskin>
Technically yes
adetokunbo has joined #nixos
bodgix19 has quit [Ping timeout: 244 seconds]
<erictapen>
simpson: the builders option looks like this: 'ssh://root@10.137.12.1 - - 4 2'
<simpson>
d10n-work: If your actual question is "when can I tell my boss to expect an updated package?" then the answer is "your boss should consider compensating somebody for running a Hydra"
<erictapen>
simpson: the build does get distributed to the remote builder, but I want my laptop to only send and receive jobs, not to build anything
<dckc>
I'm getting: error: packages '...mercurial-4.9/lib/...' and '...tortoisehg-4.8.2/lib/...' have the same priority 5; use 'nix-env --set-flag priority NUMBER INSTALLED_PKGNAME'. What do I use for PKGNAME? "selector 'mercurial-4.9' matches no derivations" and likewise for "mercurial"
<simpson>
Ultimately you likely can't avoid having your laptop build *some* stuff that it itself will need.
<edlin>
when trying to start up an existing VM using the creation wizard process, virt-man says it can't create the QEMU log directory. Does virt-manager not work in nix?
edlin has quit [Remote host closed the connection]
<erictapen>
simpson: yep I read the manual. I guess it doesn't make sense to not build derivations with preferLocalBuild = true, but why shouldn't it be possible? From my understanding, the charm of Nix is that the build host doesn't influences the build product
<{^_^}>
[nix] @grahamc opened pull request #2749 → docs: document balancing cores and max-jobs → https://git.io/fjTaw
<simpson>
erictapen: I guess. From my POV, Nix is like Portage: builds are meant to be done locally and the cacheability of packages is a nice-to-have.
<genesis>
simpson : i still miss eix from my gentoo migration to nixos ;)
<simpson>
genesis: I usually use grep on my local nixpkgs checkout.
<genesis>
yeap me too but not same :')
<genesis>
eix is really a great tool, a day we have such.
<genesis>
btw gentoo is a dead cow.
<genesis>
i used it for 18 years !
<genesis>
(so i could speak about)
xxpor has joined #nixos
xxpor has quit [Remote host closed the connection]
<adamt>
Yeah, it seems like not-so-much have been happening the last few years. Too bad, I got fond memories of the fast penguin.
zupo has quit [Ping timeout: 245 seconds]
<genesis>
they did worse and worse
<gchristensen>
what determines the size of /run/user/UID? /run is controlled by boot.runSize (25% -- this matches what I'm expected) and /run/user/UID is 50% of that
<catern>
I vaguely remember that being a systemd thing
agander has quit [Ping timeout: 246 seconds]
<catern>
gchristensen: RuntimeDirectorySize?
baba__ has joined #nixos
baba__ has quit [Remote host closed the connection]
<catern>
Uh, that is thoroughly ungoogleable
<catern>
gchristensen: RuntimeDirectorySize, see logind.conf(5)
<gchristensen>
thank you!
<vaibhavsagar>
how hard would it be to build GCC 2.95 in nixpkgs?
<gchristensen>
oh dear
<catern>
gchristensen: you may also need to know that it's created by pam_systemd
<MichaelRaskin>
I don't remember wjat is the step for building old gcc, but I think you would need an indermediate gcc to build 2.95
<buffet>
hey, to the people using emacs: is there any neat way to get nix-shell to play together with eshell?
<vaibhavsagar>
MichaelRaskin: can I use a newer GCC to build an old one?
<MichaelRaskin>
I think you can for small differences in version, but not for large ones
jabranham has joined #nixos
<vaibhavsagar>
wondering if it would be easier to get a VM of Ubuntu 8.04 or something like that
<vaibhavsagar>
I'm trying to build nhc98 and having no luck so far
<matthewbauer>
what version does it need?
<vaibhavsagar>
no idea, I tried with gcc48 (the oldest available in nixpkgs right now) and it didn't work
<MichaelRaskin>
Did you try forcing older standard?
<vaibhavsagar>
I have a good feeling about 2.95 though, that's apparently what the debian build used
<{^_^}>
[nixpkgs] @veprbl pushed commit from @jokogr to master « vscode-extensions.vscodevim.vim: init at 1.2.0 (#58500) »: https://git.io/fjTVr
<AlexRice[m]>
my only thought was that without specifying the -A flag, nix-env searches for the pname field in the derivation, which might be correctly set by default in the first, but not in the second because of the hyphen in the package name
the-kenny has quit [Quit: WeeChat 2.4]
the-kenny has joined #nixos
<AlexRice[m]>
though I don't have enough nix knowhow to check if that is what is happening
goibhniu2 has quit [Ping timeout: 255 seconds]
ddellacosta has quit [Ping timeout: 245 seconds]
<catern>
buffet: kyren: eshell should just automatically work fine if you use emacs-direnv, I believe
linarcx has quit [Quit: WeeChat 2.4]
goibhniu1 has joined #nixos
<ryantm>
AlexRice[m]: Does it work if you do `nix-env -i sgt-puzzles` ?
<catern>
btw, if you want a really good emacs integration, you should figure out how to have a project-specific exec-path, using project.el
<AlexRice[m]>
you need to put in nix-env -i sgt-puzzles-r20180429.31384ca
wraithm1 has quit [Remote host closed the connection]
civodul has joined #nixos
<kyren>
catern: oh that's good to know, I usually use emacs in a terminal so it hasn't been a big deal for me so I haven't tried it, but it's good to know in case I need to update shell.nix often and don't want to restart emacs
goibhniu1 has quit [Ping timeout: 250 seconds]
MichaelRaskin has quit [Ping timeout: 246 seconds]
ggherdov4 has quit [Remote host closed the connection]
silver_ has joined #nixos
<adamt>
I noticed several maven projects in nixpkgs using a script to fetch all dependencies. One of them is the chronos pkg. Does anybody know which script generated the list of urls to fetch?
<OmnipotentEntity>
My thoughts are that triggering cmake from within shell.nix is pretty jank, but I can't find a good way to hook into cmake better, because some of the paths don't seem to have support for environment variables.
<OmnipotentEntity>
it *works* I'm just worried that it's bad style.
<{^_^}>
[nixpkgs] @flokli pushed 2 commits to release-19.03: https://git.io/fjTrl
jabranham has quit [Quit: ERC (IRC client for Emacs 27.0.50)]
<OmnipotentEntity>
infinisil, thanks
<infinisil>
omnipotententity: And usually you'd add `cmake` to nativeBuildInputs, which will automatically write a correct `configurePhase` for you to execute
<OmnipotentEntity>
do you have documentation or an example of how to do that?
<infinisil>
Just add cmake to nativeBuildInputs :)
<OmnipotentEntity>
Should I be using the `cmakeFlags` variable?
<{^_^}>
[nixpkgs] @flokli pushed 2 commits to release-18.09: https://git.io/fjTr0
<{^_^}>
[patchelf] @edolstra pushed 0 commits to refs/tags/0.10: https://git.io/fjTrE
<infinisil>
omnipotententity: The best thing to do would be to make it actually build the derivation successfully with a nix-build
<OmnipotentEntity>
ok, consider me to be an idiot who has never done that before and only has had nix installed for about a month >_>, how do I do that?
<OmnipotentEntity>
Hypothetically of course
erasmas has quit [Quit: leaving]
vk3wtf has quit [Ping timeout: 258 seconds]
<gchristensen>
not having done it before doesn't make you an idiot :)
<OmnipotentEntity>
of course not, me being an idiot makes me an idiot. They're independent variables :)
<OmnipotentEntity>
Is there a place to specify packages that are not used to build but are required to use the program? For instance, I need python3 and such to use certain features in this repo, but they are not required to compile it.
<OmnipotentEntity>
or can I just shove them in buildInputs without making people cringe?
<symphorien>
the best way is to patch the program to use /nix/store/fooo/bin/python instead of python
<symphorien>
this way, nix will detect the dependecy
<OmnipotentEntity>
Not my repo, and need to maintain compatibility
<AlexRice[m]>
wouldn't this break if python got updated?
<OmnipotentEntity>
The program doesn't use python, only the training pipeline does
<AlexRice[m]>
or I suppose the package would update when python does
<OmnipotentEntity>
a hard path to a random /nix/store folder seems brittle yeah
sanscoeur has joined #nixos
sanscoeur has quit [Remote host closed the connection]
<ryantm>
You don't patch it as a hard path, you patch it as something like ${pkgs.python}/bin/python
<AlexRice[m]>
think you can put ${python} instead of a hard path
<OmnipotentEntity>
And I just want to make sure that the user has python3 and tensorflow libraries installed when entering the directory. I know I can force them to be installed by putting them in buildInputs, but I'm not sure if that's the "correct" way.
GoMaD_ has joined #nixos
<OmnipotentEntity>
Because they're not actually used in the build.
<symphorien>
when using nix-shell ?
GoMaD_ has quit [Remote host closed the connection]
<OmnipotentEntity>
yes
<symphorien>
for a dev environment, yes, buildInputs is fine
<{^_^}>
[nixpkgs] @flokli pushed commit from @bgamari to release-19.03 « nixos/gitlab: Package gitlab-rails »: https://git.io/fjTry
Guanin_ has quit [Quit: Leaving]
<OmnipotentEntity>
ok awesome.
<OmnipotentEntity>
thanks for your help symphorien. And you too, AlexRice[m] gchristensen and infinisil
<gchristensen>
=)
<symphorien>
what I said above mostly applies to "properly packaged" build products
<symphorien>
:)
Anton-Latukha has quit [Quit: Leaving.]
<AlexRice[m]>
Bit of a random question and I'm not sure if it is related to nixos but does anyone have any idea why my trackpad scrolling sometimes breaks after a suspend?
<{^_^}>
[nixpkgs] @symphorien opened pull request #58509 → nixos: make hardware.enableAllFirmware enable *all* firmware → https://git.io/fjTrh
Xal has quit [Ping timeout: 246 seconds]
Xal has joined #nixos
kvda has joined #nixos
<OmnipotentEntity>
no idea AlexRice[m] I would assume that it's common Linux suspend weirdness. Nothing specific to NixOS I wouldn't imagine. Maybe check the usual places? ArchWiki and so on?
o1lo01ol1o has quit [Remote host closed the connection]
o1lo01ol1o has joined #nixos
jasongrossman has quit [Quit: ERC (IRC client for Emacs 26.1)]
o1lo01ol1o has quit [Read error: Connection reset by peer]
das_j has quit [Remote host closed the connection]
das_j has joined #nixos
<azazel>
gchristensen: looking at lorry, bu if it installs the dependencies in background, how can I know when it finished the process?
illegalprime has quit [Remote host closed the connection]
<Dagger>
net.ipv6.conf.eth0.accept_ra = 0 <-- well isn't that nice. I do a `nixos-rebuild` and something, somewhere disables IPv6 autoconf, so I lose access to the boxy
<Dagger>
y
<Dagger>
...
<Dagger>
well, it was a valid question that I was going to ask at some stage, but qemu grabbed my keyboard while I was trying to save a file in nano so please ignore the random 'y's
<Dagger>
is there any way of hunting down what's setting that sysctl? I do have networking.useDHCP=false but a) RAs aren't DHCP, b) that option actually disables dhcpcd which should mean one *fewer* thing screwing with the network config
erictapen has joined #nixos
justanotheruser has quit [Quit: WeeChat 2.2]
justanotheruser has joined #nixos
adetokunbo has quit [Quit: This computer has gone to sleep]