<samueldr>
c74d: using variations of `find -L /Users/samuel/.nix-profile/ . -iname '*qt*' ` and verifying QT env vars, pretty sure nothing's bringing qt4 into my envs
<samueldr>
or my profiles
Drakonis has quit [Remote host closed the connection]
eqyiel has joined #nixos
rprije has quit [Remote host closed the connection]
rprije has joined #nixos
endformationage has quit [Quit: WeeChat 1.9.1]
simukis has joined #nixos
kyren has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
kyren has joined #nixos
<c74d>
samueldr: Hm. Thanks for checking so thoroughly.
simukis has quit [Ping timeout: 268 seconds]
justanotheruser has joined #nixos
<timclassic>
Hmm, how can I introduce new NixOS modules via an overlay?
<timclassic>
I'm searching but not coming up with anything.
<clever>
timclassic: modules can only be loaded by adding them to the imports section of configuration.nix
<clever>
overlays cant touch the module system
<timclassic>
Oh, interesting
Ridout has quit [Ping timeout: 272 seconds]
<timclassic>
Is there any convention for including modules in an overlay, such as using a nixos/ subdirectory?
<mdash>
you can go the other way and have an imported expr provide modules _and_ the overlay
rprije has quit [Read error: Connection reset by peer]
rprije has joined #nixos
jperras has joined #nixos
<timclassic>
mdash: Hmm, that's an idea
jperras has quit [Ping timeout: 256 seconds]
<kalbasit[m]>
when using home-manager as a nixos module (instead of the home-manager CLI), is it possible to access nixos config?
<kalbasit[m]>
I mean from within the `home-manager.users`
Guanin_ has quit [Ping timeout: 252 seconds]
Fare has joined #nixos
Izorkin has joined #nixos
heinrich5991 has quit [Ping timeout: 240 seconds]
Guanin_ has joined #nixos
heinrich5991 has joined #nixos
<ekleog>
timclassic: if you're building an overlay and modules for others' consumption, you may want to have a look at https://github.com/nix-community/NUR :)
<elvishjerricco>
kalbasit[m]: You can use the `config` variables from the NixOS module arguments.
<ekleog>
that's technically not working with overlays, though
<ekleog>
(but allows to add packages nonetheless, just not to replace a package used by nixpkgs -- thus requiring relatively less trust for enabling it)
<clever>
timclassic: oh, what i do, is i put my overlay, inside the module
<clever>
timclassic: this module adds qemu-userland options, and adds its own overlay to the cfg
<clever>
so its as simple as imports = [ ./qemu.nix ]; and the overlay is automatic
<timclassic>
clever: And so you only have to add it to imports and you're done?
<clever>
yep
<timclassic>
Nice, that is what I want here.
<timclassic>
Ah, I see, and it just gets appended to the list because it's a standard config option! Neat.
<timclassic>
clever: Thanks!
<timclassic>
ekleog: NUR looks cool! It's not exactly what I want here but I likely have some things to contribute.
<timclassic>
The world of Nix moves so fast, I can barely keep up.
<timclassic>
(In a good way, not in a JavaScript framework way)
<ekleog>
^
trevthedev has quit [Ping timeout: 265 seconds]
<kalbasit[m]>
elvishjerricco: so if you look at this PR https://github.com/kalbasit/system/pull/48/files#diff-73a616cb151a1960a82e8c44d7685a70R9 I have the nixos part just does `home-manager.users.kalbasit = import ../../home/cfg` inside of that `cfg/default.nix` I have `imports = [ ./i3 ]` which in turn calls another `import` this last import cannot find `networking` in `config`
<kalbasit[m]>
this suggest to me that it's nixpkgs.config not the config
hamishmack has joined #nixos
hamishmack has quit [Ping timeout: 272 seconds]
witchof0x20 has quit [Remote host closed the connection]
witchof0x20 has joined #nixos
logzet has joined #nixos
<elvishjerricco>
kalbasit[m]: That's because when you give something to `imports` in the home-manager submodule, `config` refers to the home-manager submodule, not the NixOS config.
<kalbasit[m]>
elvishjerricco: so what's the solution in this case?
Dedalo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Dedalo has joined #nixos
<elvishjerricco>
You have to get the NixOS `config` in there somehow. You could add an option to the `home-manager` module that would let you store it in the `home-manager` config, or you could literally just pass it down lambda style, or you could write all these modules as NixOS modules and just redo `home-manager.users = ...` in each of them.
Dedalo has quit [Client Quit]
heinrich5991 has quit [Ping timeout: 256 seconds]
trcc has joined #nixos
Fare has quit [Ping timeout: 240 seconds]
trcc has quit [Remote host closed the connection]
trcc has joined #nixos
heinrich5991 has joined #nixos
slyfox_ has joined #nixos
slyfox has quit [Ping timeout: 252 seconds]
heinrich5991 has quit [Ping timeout: 240 seconds]
heinrich5991 has joined #nixos
hyper_ch2 has joined #nixos
hamishmack has joined #nixos
drewr has quit [Ping timeout: 260 seconds]
c74d has left #nixos [#nixos]
orivej has quit [Ping timeout: 252 seconds]
sir_guy_carleton has quit [Quit: WeeChat 2.0]
roosemberth is now known as Orbstheorem
periklis has joined #nixos
mayhewluke has quit [Ping timeout: 260 seconds]
brodul has joined #nixos
Ariakenom has joined #nixos
klntsky has quit [Remote host closed the connection]
klntsky has joined #nixos
peel has joined #nixos
jperras has joined #nixos
kyren_ has joined #nixos
kyren has quit [Ping timeout: 265 seconds]
<{^_^}>
[nixpkgs] @danieldk opened pull request #45495 → pass: Fix dependencies on Darwin → https://git.io/fAIg8
mayhewluke has joined #nixos
jperras has quit [Ping timeout: 256 seconds]
rauno has joined #nixos
kyren_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
ThatDocsLady has joined #nixos
statusbot has quit [Ping timeout: 260 seconds]
[0__0] has joined #nixos
adamCS has quit [Ping timeout: 260 seconds]
johanot has joined #nixos
adamCS has joined #nixos
kyren has joined #nixos
Tucky has joined #nixos
<ldlework>
can I use functions inside of ${} in strings?
<ldlework>
like can I call a function that way
<ldlework>
something like " ... ${framerate 60} ... "
<ldlework>
ah why don't i try with repl
<johanot>
ldlework: as you may have found out already, the answer is yes you can :)
<ldlework>
cannot coerce integer to string
<ldlework>
rate = rate: "-framerate ${rate}"
<ldlework>
"... ${rate 60} ..."
<johanot>
${toString rate}
<ldlework>
aww
<ldlework>
rate already returns a string though
<ldlework>
oh
<ldlework>
nice!
<ldlework>
johanot: thanks!
<johanot>
ldlework: np
Izorkin has quit [Ping timeout: 268 seconds]
kyren_ has joined #nixos
rprije has quit [Ping timeout: 240 seconds]
<ldlework>
johanot: quality = min max target: "${minrate min} -b:v ${toString target} ${maxrate: max}";
<ldlework>
it's beautiful
<johanot>
:) stuff like that is why we hangout here
<ocharles>
ok, nix-store --restore foo.nar.bz2 works, but half the files are missing. I know they all exist on my Hydra server, so maybe there's a way to make them all valid
arahael1 has joined #nixos
<Dezgeg>
--restore doesn't restore crucial stuff like the dependency information
betaboon has joined #nixos
jperras has joined #nixos
<ocharles>
ah, nice, there is api/get-info on builds, and that gives me an outPath. Then I can just do nix-store -r "$outPath"
<ocharles>
I'm experimenting with building machines on Hydra, and having machines self update themselves :)
<adamantium>
the security implications of nixos ... i wonder, say i
<adamantium>
install virus, if that was more of a thing, in linux, ... Then I
<adamantium>
run nixos-switch later. If the virus did not reside in /home,
<adamantium>
would it likely be removed then?
<adamantium>
I am sorry !
phreedom has joined #nixos
<srhb>
adamantium: I think that's way too vague to really answer. Especially because our notion of "install" is a lot more nuanced than usually.
Zajcev has quit [Ping timeout: 252 seconds]
<srhb>
If you were to have a "virus" (or anything) in your environment.systemPackages, say, then it would get "installed" in the new system profile as well, unless you removed it from thence.
<srhb>
Same rules apply whether the package is malicious or not.
<srhb>
For more complex malicious things that are actually exploitative rather than just a thought experiment installed into some profile, the details are just insufficient to really give a good answer.
<adamantium>
oh ook, well that makes sense. I was hoping for a magic bullet, perhaps
<adamantium>
I suppose the good news is the ease of reinstallation, should a person feel worried about compromise
<elvishjerricco>
The fact that the nix store is read only and that we regularly purge it with nix-collect-garbage is a great assurance about security. It's obviously only a case of security through obscurity... but it does invalidate common attack vectors.
<hyper_ch2>
adamantium: also don't forget that the nix store is by default read-only.... you first have to make it rw before a virus - outside the nix packages - could even put anything in there
heinrich5991 has quit [Ping timeout: 256 seconds]
<adamantium>
we probably cannot uninstall sudo
<adamantium>
or
<adamantium>
hmm, well, I am just thinking about loud in here now. Sorry
<{^_^}>
[nixpkgs] @Mic92 merged pull request #45495 → pass: Fix dependencies on Darwin → https://git.io/fAIg8
<hyper_ch2>
or you can send it on the receivers end into a zfs pool
<elvishjerricco>
One nice thing about the nix store is that nix store paths are shared and only sent if not already present. This is determined via the hash of the path. I'm guessing ZFS's checksums aren't such strong cryptographic hashes? i.e. I'd assume they're there for cheap verification, not secure validation.
Izorkin has quit [Ping timeout: 268 seconds]
<adamantium>
i dont now
<adamantium>
i find zfs and nixos extremely powerful tools
<adamantium>
and work very well together for many purposes
<hyper_ch2>
(and zfs unstable has now native encryption)
<hyper_ch2>
zfs also support mirro and raidz
ThatDocsLady has quit [Remote host closed the connection]
<hyper_ch2>
where I used like 3 tools before: mdadm, luks/dm-crypt, ext4.... I can do it now all in zfs
ThatDocsLady has joined #nixos
<adamantium>
and omg nixos makes zfs easy
<hyper_ch2>
oh... 4 tools: + rsync
<adamantium>
try installing zfs on gentoo !!! then try nix
<hyper_ch2>
adamantium: if you want to setup a system with zfs, best to create custom nixos installer iso with zfs integrated
<hyper_ch2>
elvishjerricco: with clever's kexec you can build a kexec that you can scp to a remote server then run it and basically it will boot up nixos into ram, so that you then can do a nixos install from within the ram
<thekolb>
So I am trying to use nixops as the user deploy which can ssh root@target just fine via pubkey authentication, but when I invoke nixops it complains "root@target: Permission denied" and can’t connect. What am I missing this morning? O_O
<ldlework>
how do I prevent "${w}" from being interpolated as a nix variable thing
<ldlework>
i want ${w} in the string
<elvishjerricco>
ldlework: you the string but you don't want nix to build `w`?
<elvishjerricco>
ldlework: If you want the output path of `w` without building `w`, I don't think so. Evaluating `"${w}"` as a nix expression will yield the output path without building `w`, but building any derivation that uses that string in its build scripts will require building `w`, as interpolating a derivation will make it a context derivation
<srhb>
It seems it has been the intention for it to work for years now, but it has been broken multiple times.
mayhewluke has quit [Ping timeout: 264 seconds]
heinrich5991 has joined #nixos
mayhewluke has joined #nixos
<elvishjerricco>
srhb: Gah. The whole draw of ZFS to me is that I can discard so many other tools. It's very frustrating that I still have to use a separate swap system.
<srhb>
Yes, agreed.
<elvishjerricco>
srhb: Anyway thanks for the link
<hyper_ch2>
I don't use swap
<srhb>
I feel that random encryption of a swap partition is "okish" on NixOS because it has such a low mental overhead
<srhb>
But I still do want everything on the same system...
<hyper_ch2>
if I only could have /boot also as zfs dataset....
<srhb>
hyper_ch2: It's nice if you can avoid it, but I regularly build things that require 32GiB+ of memory
* srhb
wants a new laptop
<hyper_ch2>
srhb: :)
<hyper_ch2>
128gb ram ought to be enough for anyone
<srhb>
Sure... Sounds expensive :P
<elvishjerricco>
srhb: Yea... It really makes me feel like zfs is kind of an afterthought overall, not just for nixos. I guess that's to be expected but it's still really disappointing.
ericsagnes has quit [Ping timeout: 264 seconds]
<srhb>
I think it's safe to say that swap on zfs is a bit of an afterthought...
<elvishjerricco>
heh yea
<srhb>
The issues go back to 2012 and regularly are solved and broken again :P
<goibhniu>
did you get a laptop from tuxedo in the end hyper_ch2?
<ekleog>
it should just work if you put a derivation that puts it in [the right place] in buildInputs
<ekleog>
[the right place] being $out/include/libspeechd.h, if I can trust a find /nix/store -name '*.h'
heinrich5991 has joined #nixos
wirew0rm has joined #nixos
<genesis>
hum seems a bit complicated, btw ekleog do you know nix-locate ? :)
<elvishjerricco>
srhb: The lack of similar error messages with using ZFS as swap on other linux'es makes me think that it works fine on other distros. So does it seem reasonable that the minimalism of NixOS would be the reason that ZFS swap woudln't work?
<Dezgeg>
no reason to believe that, it's all handled in the kernel
<Dezgeg>
maybe you have a non-trivial disk setup compared to what others have set up?
<elvishjerricco>
Dezgeg: My testing setup so far has been a single disk in a virtualbox environment, which has replicated the failure I've experienced in my single disk laptop.
<ekleog>
genesis: yeah, but to find whatever.h, it's just as easy to use find :)
<ekleog>
otherwise ,locate here works pretty well :)
trcc has joined #nixos
<srhb>
elvishjerricco: I doubt that.
<Dezgeg>
ok. I guess you need to try a matching kernel and zfs version to what others have done successfully
<elvishjerricco>
Yea, that sounds hard though :P
qknight has joined #nixos
<Dezgeg>
but personally I wouldn't simply do swap-on-ZFS... swap-on-anything-complicated is a minefield, e.g. swap-over-NFS had/has similar problems
vbernat has joined #nixos
philippD has joined #nixos
rprije has quit [Remote host closed the connection]
<hyper_ch2>
[12:57] <hyper_ch2> what's the current opinion on running swap on zfs/zvol? [12:59] <DeHackEd> avoid if possible, but if you have no other choice then make sure you configure it properly
heinrich5991 has quit [Ping timeout: 268 seconds]
heinrich5991 has joined #nixos
Fare has joined #nixos
<Mic92>
Who needs swap anyway.
hamishmack has quit [Ping timeout: 268 seconds]
<srhb>
All you rich people and your disk-sized RAM rigs. :-P
__monty__ has joined #nixos
<hyper_ch2>
swap-over-network :)
vidbina has quit [Ping timeout: 260 seconds]
rprije has quit [Remote host closed the connection]
<hyper_ch2>
Mic92: care to share some of your ram with us?
rprije has joined #nixos
<Mic92>
hyper_ch2: well, when you are low on RAM, zfs is not the best choice anyway.
<{^_^}>
[nixpkgs] @uri-canva opened pull request #45499 → bazel-deps: update to 2018-08-16 → https://git.io/fAIQ2
<Mic92>
Lisanna: from my understanding this should be enough, but I am not a laywer.
<ekleog>
Lisanna: first question is, do you want it for the package or for the module
<ekleog>
if you limit only the module people could still use the package
<Lisanna>
not sure what you mean "package" vs "module"
<sphalerite>
config.allowUnfree should be replaced by config.acceptedLicenses :p
<Lisanna>
I'm working through the packaging for Intel parallel studio
hamishmack has joined #nixos
<Lisanna>
eh, I'll just force the user to set a config param to be safe... they already have to set a config param to pass in their license file
<ekleog>
Lisanna: if you are making something in pkgs/, it's a package; if you're touching nixos/, it's a module :)
<Izorkin>
How to fix error nixop deploy - https://pastebin.com/DTf7Ligh Not work in nixos-18.09pre150245.ecd9d74d973
<Lisanna>
ekleog: oh, didn't know you weretalking about nixos modules
<Lisanna>
it's a package
<ekleog>
yeah, I wrongly assumed “option” meant “module option” :)
<Lisanna>
oh
<ekleog>
Lisanna: then, if there's a license file, then you can just assume that passing the license file means acceptance of the EULA, can't you?
<ekleog>
(this plus setting licence to unfree)
<ekleog>
sphalerite++
<{^_^}>
sphalerite's karma got increased to 11
<Yardanico>
are there any near plans to update nvidia driver to the latest version in unstable? :) with comparasion to the current nvidia version newest one has a ton of vulkan bugfixes, performance improvements and compiles with linux 4.18
<Lisanna>
*shrugs* if you ran it by hand that wouldn't be good enough
<ekleog>
oh.
<ekleog>
well, makes sense then :) though I'd intuitively think that EULA-based programs in nixpkgs usually use only license.unfree? don't use any, though, so… ^^
<ekleog>
Yardanico: neither vcunat nor abbradar appear to be here, so you could open an issue and tag them, I guess :) then, you can also have a look at pkgs/os-specific/linux/nvidia-x11/default.nix and try updating stuff and testing, if you tag them on a PR it'll likely be done faster :)
* ekleog
thinks potentially the website bans downloading from a non-whitelisted IP
<ekleog>
yeah, so that's it
<sphalerite>
oh right
<ekleog>
and likely the reason why oraclejdk is requireFile :þ
<{^_^}>
[rfcs] @CrystalGamma opened pull request #31 → [RFC 0030] Support ppc64(le) architecture → https://git.io/fAI5h
<ekleog>
Mic92: then, tbh the nixpkgs license has the usual disclaimer of liability, so I wonder who could be attacked if nixpkgs didn't properly mention the license
philippD has joined #nixos
Wharncliffe has quit [Quit: Lost terminal]
HarpoRoeder has quit [Remote host closed the connection]
periklis has quit [Ping timeout: 252 seconds]
<ldlework>
if a python package is really old
<ldlework>
how can I update it
<LnL>
ekleog: indeed
vidbina has joined #nixos
<ldlework>
I've got a simple default.nix for nix-shell
<Taneb>
Will nix-store --realise --add-root overwrite the root path if it already exists and is a symlink?
tokudan[m] has joined #nixos
<Dezgeg>
yes
<Taneb>
Thanks
<tokudan[m]>
finally... joining this channel from Matrix is not fun with nickserv registration enforced...
<gchristensen>
we can try turning off -rf and see if the spam is still problematic.
<Taneb>
gchristensen: I was still getting spam in another channel this morning
<tokudan[m]>
i don't care, i managed it now... but even after completing the registration I got kicked a few times ;)
<gchristensen>
I care
<gchristensen>
I don't like -rf. it makes it harder for people new to IRC and adds power to the "what about Slack?"-style questions.
<tokudan[m]>
I have a derivation that is roughly defines as follows: stdenv.mkDerivation rec { name="abc-${version}"; version="20180629-03:25"; ...}. my problem is that nix complains that the 03:25 is not valid in the derivation name, but i need that : to generate the correct url. Is there a simple function that removes invalid characters in names that I could throw in front of the version?
<gchristensen>
tokudan[m]: there is a substitute function but it is slow and not permitted in nixpkgs, in lib/strings.nix
<tokudan[m]>
any other suggestion how i could work around that, apart from removing the version from the name?
<aminechikhaoui>
I would prefer gitter over irc personnaly :D I know many won't share my preference though :)
<samueldr>
tokudan[m]: any tips for others facing issues joining from matrix?
<sphalerite>
tokudan[m]: did you use !storepass?
<gchristensen>
tokudan[m]: for me, I don't think it is so bad to retype the version.
<sphalerite>
tokudan[m]: because with that I think it should use SASL, making this not a problem?
<gchristensen>
I think it is an antipattern to go through heroic efforts to not type the version number twice
<samueldr>
(especially since it doesn't help with overriding, in the end)
<tokudan[m]>
samueldr: well, nickserv isn't exactly user friendly to use from matrix
<gchristensen>
and spoils grep
Fare has quit [Ping timeout: 255 seconds]
<tokudan[m]>
sphalerite: after registering, yes
<samueldr>
tokudan[m]: I'm asking mainly because I have never used matrix to join IRC; if people ask having tips handy is... handy!
heinrich5991 has quit [Ping timeout: 256 seconds]
<tokudan[m]>
gchristensen: yeah, i can do that. but having it automated reduces the chance of errors and if it's just ${version} turning into ${lib.removeInvalidChars version}, then it's worth it in my opinion
<gchristensen>
tokudan[m]: are you submitting it to nixpkgs?
<tokudan[m]>
samueldr: well, first thing: contact @freenode_NickServ:matrix.org send it the message "HELP"
<tokudan[m]>
gchristensen: no, it very likely wont end up there
<gchristensen>
tokudan[m]: oh! there is a builtin now!
<tokudan[m]>
samueldr: then register there and provide an email and password, check that email and send the verification code to nickserv. finally keep joining the room you want to join and if you get a message about getting kicked, just keep trying.
heinrich5991 has joined #nixos
<tokudan[m]>
gchristensen: nice :)
<tokudan[m]>
gchristensen: thanks, that's what I was looking for
<sphalerit>
tokudan: ah right so you hadn't registered at all
<sphalerit>
That explains it :p
<tokudan[m]>
Linus: never needed to before ;)
lawlesseel has quit [Remote host closed the connection]
johanot has quit [Quit: leaving]
rprije has quit [Ping timeout: 240 seconds]
lawlesseel has joined #nixos
Denommus has joined #nixos
heinrich5991 has quit [Ping timeout: 240 seconds]
Fare has joined #nixos
<ekleog>
gchristensen: there's a trick to allow matrix users but not matrix users
jD91mZM2 has joined #nixos
heinrich5991 has joined #nixos
<gchristensen>
oh there is?
<ekleog>
“replace `+r` with `+qe $~a *!*@gateway/shell/matrix.org/*`” (from matrix HS)
<ekleog>
didn't check what this exactly does
<ekleog>
errrr s/matrix/unregistered irc/2 in my first sentence, that didn't make sense at all
<{^_^}>
[nixpkgs] @ryantm opened pull request #45502 → treewide: remove mailing list references → https://git.io/fAIje
<{^_^}>
[nixpkgs] @Mic92 merged pull request #45472 → strace: check machine personalities for all non-x86 architectures → https://git.io/fAkyR
<ekleog>
it's been among the options proposed by ircops on #freenode
<gchristensen>
yeah, I know. it makes for an even worse experience for unregistered users
<ekleog>
“Garbee: yeah, that's +rf, that works as well, but you can also do +q $~a with a topic or entrymessage, because +r is a bit harder to bypass than +q for false positives” <- ~1hr ago on #freenode by an ircop
<ekleog>
then, it's your choice, I'm just evoking possibilities :)
Mateon1 has quit [Ping timeout: 272 seconds]
<gchristensen>
thank you :)
* ekleog
likes the redirection to #nixos-unregistered
inquisitiv3 has joined #nixos
<gchristensen>
me too, I think it is the least bad option
<inquisitiv3>
I'm trying to connect to my school's wireless network. According the instructions for Ubuntu I should browse to `/usr/lib/ssl/certs/` and choose the certificate `Thawte_Premium_Server_CA.pem`. The problem is that I can't find the certificate.
<{^_^}>
[nixpkgs] @xeji pushed commit from @CrystalGamma to staging « texlive-bin: disable mfluajit on ppc64(le) (#45475) »: https://git.io/fAIjj
heinrich5991 has quit [Ping timeout: 256 seconds]
mayhewluke has quit [Ping timeout: 240 seconds]
heinrich5991 has joined #nixos
Drakonis has quit [Ping timeout: 256 seconds]
Izorkin has quit [Ping timeout: 256 seconds]
<d1rewolf>
guys, in nix pills #12, discussing single repository pattern he states "With nixpkgs, we import it and then we peek derivations by accessing the giant attribute set. " What does he mean regarding "peek" here?
<mkaito>
the value of those attributes is the derivation that describes how to build something. so by referencing the attribute, nix will do the right thing.
<mkaito>
you get back a reference to said thing, and nix will build it as described in said derivation if necessary
<mkaito>
make sense?
IRCsum has quit [Ping timeout: 268 seconds]
<d1rewolf>
mkaito: I think so. thank you
<mkaito>
so what happened to make firefox-bin unbuildable? doesn't seem to have been a bump to firefox itself. it complains that it should not reference gcc.
<Pneumaticat>
hmm, I'm trying to package LizardFS 3.12, but it seems like they added instructions in https://github.com/lizardfs/lizardfs/blob/master/cmake/Libraries.cmake to call a `download_external` function to download more dependencies on build, which obviously fails on Nix :/
<Pneumaticat>
would anyone know of a good way around that? I guess I could try patching out those lines and manually downloading the files in the nix expression...
IRCsum has joined #nixos
jperras has joined #nixos
iyzsong has quit [Ping timeout: 256 seconds]
trcc has quit [Ping timeout: 256 seconds]
lawlesseel has joined #nixos
<sphalerite>
Pneumaticat: yes, that's what I'd recommend. Unless you can just plop the files in the locations that it would have downloaded them to and it will just accept them like that
<Pneumaticat>
no, it seems like it also does custom patching logic in `download_external`... wonderful
<Pneumaticat>
I doubt it'd accept just putting the files there, so I'll probably have to patch the script
Profpatsch has joined #nixos
<Profpatsch>
Ah, how I love wrappers that don’t pass through manpages and docs.
worldofpeace_ has quit [Read error: Connection reset by peer]
<ldlework>
Linking kitty/fast_data_types ...
<ldlework>
The package dbus-1 was not found on your system
<ldlework>
not sure where to get "dbus-1"
worldofpeace has joined #nixos
<d1rewolf>
". Any ideas?
<d1rewolf>
guys, in the repl, I'm getting "error: derivation '/nix/store/dpzib6xygiz9mk1bkjfkghq5c19hm819-bootstrap-stage0-glibc.drv' has incorrect output '/nix/store/rhfaq9w1y1k16p80lsx923sps0b7hnzb-bootstrap-stage0-glibc', should be '/nix/store/n5jq7vlhyk6ggx51gfg4fb4lh6www720-bootstrap-stage0-glibc'
<sphalerite>
ldlework: add dbus to buildInputs
<sphalerite>
d1rewolf: what are you doing that gets you that error?
<d1rewolf>
sphalerite: s = (import <nixpkgs> {}), hit enter, and then on the next line, "toString s.i3"
{`-`} has joined #nixos
<sphalerite>
d1rewolf: that is very odd.
<sphalerite>
d1rewolf: it goes through all the paths in the store and checks them against the recorded hashes
<sphalerite>
i.e. checks for store corruption
zgrep has joined #nixos
<d1rewolf>
gotcha. sphalerite: is this oddness worth reporting or taking further, or should I just chalk it up to "don't use a 'nix repl' after an update if it was started before"?
shymega has joined #nixos
betaboon has quit [Quit: WeeChat 2.1]
<sphalerite>
it's worth reporting in principle, but you don't actually have a way of obtaining this behaviour consistently, do you?
<d1rewolf>
sphalerite: let me check
Lisanna_ has quit [Quit: Lisanna_]
<d1rewolf>
sphalerite: sadly, no
<sphalerite>
I suspected as much :/
<sphalerite>
it's especially weird that it gave a different error for different packages
<d1rewolf>
indeed :(
<sphalerite>
for all we know this could be a cosmic ray bit flip ;p
<{^_^}>
[nixpkgs] @bhipple opened pull request #45505 → pcre2: fetch src from HTTPS instead of FTP → https://git.io/fALtE
<d1rewolf>
sphalerite: indeed ;-) VALIS in action.
<d1rewolf>
can one print log messages to the console from a nix expression?
<sphalerit>
bruno2: dezgeg's cache covers natively-built stuff, not arm stuff
<bruno2>
ok, i'm fairly new to this, so what do you mean by natively-built?
<bruno2>
do you mean stuff built on an arm processor
<sphalerit>
> <@freenode_d1rewolf:matrix.org> can one print log messages to the console from a nix expression?
<sphalerit>
bruno2: built on ARM, not just for ARM
<{^_^}>
error: syntax error, unexpected '<', at (string):171:1
<sphalerit>
d1rewolf: yes, using builtins.trace
<sphalerit>
bruno2: yes, non-cross-built stuff
<sphalerit>
So in the case of ARM, stuff built on ARM
<bruno2>
so there's no way to get cached artefact built on x86 for ARM?
<bruno2>
I would have naively thought that the end result would be the same, but I guess the hash would be different?
<sphalerit>
You could set up your own binary cache :p
<sphalerit>
You'd still have to build it to get it in there of course
<d1rewolf>
are functions chainable? x.override{a=9;}.override{b=3;} reports an error stating that "attribute 'override' missing", but if I assign the result of the call "s=x.override{a=9;}", and then on a new line, "s.override{a=10;}", it works
<sphalerit>
But yeah I don't know of any cache providing x86-built ARM stuff
<sphalerit>
d1rewolf: yes, you just need some extra parentheses
<ldlework>
clever: let me tell you what I'm doing - I wanted to install the master version of the kitty terminal. so I went to nixpkgs master, got the latest derivation, and fixed it up as a module after changing the rev and sha256 - it worked! But then, I tried to do the same thing, as an overlay, thinking the only thing I need to change is the github rev and stuff, making the overlay much shorter - but nope, it
<ldlework>
complained about everything - i kept adding code from the original derivation and that got me further and further until this wrapPythonPrograms thing
<ldlework>
So while I could probably do what you just said, isn't there an easier way to just override the latest nixpkgs derivation with a new rev/sha256?
<ldlework>
Myrl-saki: oh sorry, what did you mean?
<Myrl-saki>
ldlework: If you have `nixpkgs.config.packageOverrides = { foo = ...; };` why does `imports = foo bar;` infinite loop?
<Myrl-saki>
ldlework: We encountered this even with the MWE I made, right?
<clever>
the imports key cant depend on config
<ldlework>
Myrl-saki: I'm lost
<infinisil>
Myrl-saki: all modules need to be imported in order to evaluate the config, but if the module list depends on the config, this ain't gun work!
<ldlework>
clever: so i moved the overlay into the unstable overlay
<{^_^}>
Use `,` to list all commands, `,foo = Foo!` to define foo as "Foo!", `,foo =` to undefine it, `,foo` to output "Foo!", `,foo somebody` to send "Foo!" to the nick somebody
<ldlework>
i made a writeScriptBin "ranger" which I want to always be called when "ranger" is invoked
<ldlework>
but when I call ranger, then, the path is such that, if ranger is called from within itself, then it doesn't invoke my writeScriptBin, it invokes the actual ranger
<clever>
Izorkin: command 9 was added on august 3rd, and greatly reduces memory usage
<clever>
Izorkin: you may need to downgrade the nix on the deployer side temporarily
<clever>
Izorkin: on the remote system, do `realpath /run/current-system/sw/bin/nix-store`, then on the nixops machine, add that bin dir to the start of `$PATH`, and the deploy should pass
<clever>
Izorkin: a bug should also be fixed on nix
trcc has joined #nixos
vmandela has quit [Quit: Leaving]
rauno has quit [Ping timeout: 265 seconds]
Cale has quit [Ping timeout: 260 seconds]
<Izorkin>
clever: working with downgrade nix to 2.0.4 on nixops machine
trcc has quit [Remote host closed the connection]
Cale has joined #nixos
<simukis>
what is the best way to generate nix for python packages?
<simukis>
notably this is not a library I’m looking to produce a `.nix` of, but rather an executable
qknight has quit [Quit: Lost terminal]
Fare has quit [Quit: Leaving]
Fare has joined #nixos
cfricke has joined #nixos
<{^_^}>
[nixpkgs] @Thra11 opened pull request #45510 → qtcreator: apply patch on aarch64 → https://git.io/fALaR
romildo has joined #nixos
<disasm>
Izorkin: if it's not something I'm planning on pushing upstream, I tend to use pypi2nix. Give it a requirements.txt and it gives you fairly good nix expressions for building it.
<Izorkin>
disasm: did not use pypi2nix
<clever>
simukis: ^^
<chreekat>
When nix-store --realise or nix-build runs, it starts with a message, 'Fetching [...]; Building [...]'. Is there any way to get that list of deps in a tree format (like nix-store -q --tree) without actually building the output?
<simukis>
I tried both python2nix and pypi2nix on the package I have and neither work
<hyper_ch>
who's running plex on a headless nixos server? I thought after installing plex I'd have to go to IP:32400/web to start the configuration
<simukis>
python2nix fails to run at all, and pypi2nix fails to find package version that seems to exist on pypi.
<Izorkin>
how to change extraConfig in linux kernel? linuxPackages_4_17 = pkgs.linuxPackages_4_17.overrideDerivation - not work, error - attribute 'overrideDerivation' missing
<clever>
Izorkin: linuxPackages_4_17 is a set, not a package, so override cant be used on it directly
<clever>
Izorkin: you need to override the kernel the set was made from
<clever>
Izorkin: nixpkgs will then use one of those kernels to generate pkgs.linuxPackages_4_17
ThatDocsLady has quit [Ping timeout: 256 seconds]
<hyper_ch>
how does plex handle tv shows that have no seasons but continuous numbering?
<gchristensen>
probably a better question for #plex
<clever>
gchristensen: ah, didnt think they would have a room here
<hyper_ch>
gchristensen: well, just though because there's so many experts here
<gchristensen>
yeah but even more experts theer
<clever>
hyper_ch: and there are 400 experts in #plex
<gchristensen>
clever++ :D
<clever>
thats nearly as many nixos users in here, lol
<disasm>
oh, nice, so you can just copy the hash from there and not run nix-prefetch-url to get it!
<hyper_ch>
I was so disappointed in dlna support.... the tv almost plays nothing when using dlna over the network... but whne putting same shows on usb stick the tv plays them
<clever>
hyper_ch: my cable box supports some weird form of upnp, but it only works with a specific ipad app, the android variant of the app does not work
<clever>
hyper_ch: and the ipad app streams the content, so if the ipad turns off at any point, the stream breaks
<clever>
the app also does not disable automatic shutoff, so the OS will try to shut off the ipad mid-stream :P
<hyper_ch>
well, got now appl tv 4k as well to have server (plex) -> apple tv 4k -> tv (hdmi)
<clever>
have fun trying to watch a movie!
jD91mZM2 has quit [Ping timeout: 272 seconds]
<hyper_ch>
still need to find free power plug, free ethernet slot and hdmi cable
<chreekat>
there's this enticing "WantMassQuery: 1" line in the standard /nix-cache-info document returned by nix-serve. What does that mean; how is it used?
<clever>
chreekat: when using something like `nix-env -qA`, it can check the binary cache to see if things are pre-built or not
<clever>
chreekat: and depending on your search terms, it may wind up checking every single package in all of nixpkgs
<clever>
which would result in over 10,000 requests for .narinfo files
<chreekat>
clever: That makes sense, but I do not see the connection to WantMassQuery. Does a value of 1 mean the server is happy to respond to 10,000 such requests?
<clever>
yeah
<clever>
and a value of 0 prevents nix-env from performing a DoS attack
* chreekat
nods
Judson1 has joined #nixos
<Judson1>
I'm contemplating fiddling with my pulseaudio configuration.nix stuff so that the service will restart on wake-from-sleep.
<Judson1>
(it breaks w/r/t bluetooth on my laptop)
<clever>
Judson1: note, that chromium looses the ability to capture any audio if pulseaudio has restarted
<Judson1>
I'm wondering if this is worth contributing back, or just my idiosyncracy?
<clever>
Judson1: and you must then restart chromium
<Judson1>
To *capture* audio.
<clever>
yeah, only capture
<clever>
it claims you lack a capture device
<Judson1>
Like, video chats you'll suddenly not be able to speak?
<clever>
exactly
<Judson1>
So that's why *that* happens.
<clever>
i figured it out on the 3rd day after needing to use hangouts for daily voice chats
<clever>
prior to that, i wasnt using capture in the browser enough to notice the patterns
<clever>
Judson1: see if you can fix it with unload-module and load-module via pactl instead
<clever>
and maybe unload the bluetooth BEFORE you go into standby
worldofpeace has quit [Remote host closed the connection]
<chreekat>
I want some service to be running when tests run for a package, and I want to clean it up even when things crash and burn. Is there any notion of 'catch' or 'finally' when it comes to build commands? Or a standard practice? (Maybe a shell trap?)
erickomoto has joined #nixos
<clever>
chreekat: if you are using multi-user nix, then the nix-daemon will kill every proccess in the build user after the build has finished
<erickomoto>
Hi! I am trying to get nix installed in a server owned by my supervisor and he's got me talking to the administrator. He mentions that to be extra safe he would like to install nix in its own partition with nosuid and nodev options. Has anyone tried this? And does it work? (I'm not very well versed on this so I am also asking him what exactly are we protecting against.)
<emily>
I'm pretty sure it'd work fine with nosuid
<emily>
I mean, programs that require setuid won't work, of course
<emily>
I run NixOS on a filesystem that doesn't allow setting capabilities and the only consequence is I have to "sudo ping" rather than ping etc., so I don't expect it'd cause a problem for normal Nix usage
<clever>
emily: nix also has the rule that there be no setuid binaries in the store
<clever>
so such things would already be broken on a normal nix install
<clever>
emily: what does `type ping` return?
<emily>
it's /run/wrappers/bin/ping. it's just that the actual setcap fails on nixos-rebuild.
<clever>
ah
<clever>
sudo should be equally broken then, it uses the same system
<emily>
I should probably stub it out so that it doesn't try to setcap stuff at all, but there's other failures I'd have to deal with too (e.g. it trying to remount stuff)
<emily>
clever: for whatever reason the filesystem these containers are on allows setuid but not setcap
<emily>
it's in a user namespace
<clever>
ah, weird
<emily>
perhaps giving random capacities to theoretically unprivileged executables rather than just making them fake-uid-0 is too scary, I dunno.
<emily>
ideally I'd tell NixOS to just setuid any binary that uses caps, but I haven't figured out any way to do that from my configuration.nix, it'd probably require nixpkgs patching.
<clever>
but the fake uid 0 can still use ping, which is a cabability
<emily>
yeah *shrug*
<emily>
I don't really understand it.
<clever>
erickomoto: in general, you can just mkdir /nix ; chown erickomoto /nix, and then do the entire install without ever gaining root
<clever>
erickomoto: and if you never gain root, you cant make setuid or device nodes, so both flags are pointless
<erickomoto>
Ok cool. Thanks emily and clever! I'm not really well versed in this area but I will relay the info. (I am literally searching for "what's suid"). [regarding chown erickomoto, admin wants to set multiuser install]
<clever>
erickomoto: ah, for multi-user, then you do need the daemon running as root, and i can see why he would want nosetuid then to be paranoid
<emily>
is there any obstacle to having it run as a non-root user?
<emily>
and just having /nix be owned by that user
* emily
isn't sure why nix-daemon requires root
<clever>
emily: ive not tried a multi-user non-root install
<clever>
but a non-root setup means that you cant do automatic cleanup, which would solve the issue chreekat was asking earlier
<adamantium>
clever: after talking with you last week, i put /nix in a separate zfs pool, along with /boot/grub, but now i'm debating putting it all back in zroot/ROOT/nixos
<adamantium>
Talk me out of it ! I dont need to keep my backups for more than 3-7 days and i think a ROOT pool with all of nixos could add extra benefits
<clever>
adamantium: i keep nix on its own filesystem, within the same pool
<clever>
adamantium: and then i can turn snapshots off for just /nix
<erickomoto>
Thanks DigitalKiwi
<adamantium>
I'm sorry! It's all in the pool, I used the wrong word, just I mean all in one dataset.
<adamantium>
s/pool/dataset
<DigitalKiwi>
erickomoto yw. That'll give you a bit of an idea as a start
Denommus has quit [Remote host closed the connection]
<adamantium>
clever: will putting all in ONE dataset really take up //that// much more extra disk space when a snapshot happens on /nix after a nixos-switch? I don't forsee large changing of the nix-store often nor keeping zfs snapshots or nix store environemnts more than 7 days
<clever>
adamantium: the snapshot itself is basically free, but it means anything you delete with nix-collect-garbage will then not be deleted, for a week
<adamantium>
do /nix snapshots grow astronomically fast?
<adamantium>
alright
<adamantium>
well i will just try it, and see which method i prefer
<adamantium>
i think i will have to test it both ways
<adamantium>
i'm just much more comfortable with the traditional zfs on root situation.
Denommus has joined #nixos
<mkaito>
nix's error messages sure are something else. not only are they absolutely never even remotely helpful, they are quite often downright misleading. ┻━┻ ︵ヽ(`Д´)ﾉ︵ ┻━┻
<clever>
mkaito: can you pastebin one as an example?
<mkaito>
I'd have to sanitize the living bejeezus out of it or risk getting fired. I just wanted to vent to people who know my pain lol
<mkaito>
but basically, I'm fetching something with fetchGit, wrapping it in mkDerivation, calling that with callPackage, and according to nix, the result is not of type package. so (ノ ゜Д゜)ノ ︵ ┻━┻
<clever>
mkaito: why is it wrapped with mkDerivation?
<mkaito>
I tried without it, saw it was not of type package, so tried wrapping it in a derivation.
<mkaito>
the repo I'm fetching has a default.nix
<mkaito>
so... it should work
<clever>
how many args did you pass to callPackage?
<mkaito>
none
<mkaito>
or rather, an empty set
<clever>
ah, that should be good then
<clever>
it should return whatever the default.nix returns
<clever>
does that return a set? a package?
<mkaito>
actually, let me go make sure about that
<mkaito>
removed the wrapping mkDerivation, now it's complaining about calling a set and not a function, so I guess that derivation was necessary after all
<clever>
can you pastebin the nix expression?
<mkaito>
I seem to have hit a different snag now, somehow. let me follow this rabbit hole and I'll get back to you if that doesn't work out :)
<LnL>
type package, is this in the context of nixos modules?
<Watcher7>
Hey y'all. I'm attempting to install NixOS in a KVM guest, but I'm encountering an issue when `nixos-install` attempts to write grub to /dev/vda. I've got /dev/vda formated with a single ext4 partition, and I have those attributes defined in the NixOS configuration file. The target grub device is also "/dev/vda" in the NixOS configuration file. I've tried both virtio-scsi (device: /dev/sda) and virtio-blk
<Watcher7>
(device: /dev/vda) as device interfaces. The error is "error: cannot find a device for /boot/grub (is /dev mounted?)." and afterwards "installation of GRUB on /dev/vda failed". Any ideas?
<clever>
Watcher7: grub needs a partition table to install into the root of a disk
<Watcher7>
Oh so it's expecting its own boot partition in this case rather than just installing to the head of the disk?
<Watcher7>
Thanks.
<clever>
it doesnt need its own boot partition
<clever>
but it needs a gap between the partition table and the 1st partition
<Watcher7>
Oh, I've already done that.
<clever>
the alignment for the partitions ensures that gap
<clever>
so vda1 is ext4
<clever>
ah, misread your question a bit
<clever>
Watcher7: can you pastebin the output of lsblk ?
<drakonis__>
Watcher7, sup
chreekat has quit [Ping timeout: 256 seconds]
<Watcher7>
First partition starts at 2048. correct vda1 is ext4.
<Watcher7>
sure
<Watcher7>
drakonis__, hello.
<drakonis__>
vda1? isn't it normally sda1?
<Watcher7>
When using virtio-scsi, yes
<Watcher7>
I switched to virtio-blk.
<clever>
drakonis__: depends on which driver your using
<Watcher7>
clever, one more thing. Now it's telling me I'm failing an assertion that requires "boot.loader.grub.devices", but I've already got a device defined in my configuration.
<ij>
clever, Great idea! I've apparently screwed something, because nix is complaining about inf recursion
<clever>
Watcher7: check again, when you mounted vda1 onto /mnt, it hid all the old config
<Watcher7>
and modifying that option to be "devices" instead of "device", and using a list containing a single device
<Watcher7>
clever, right, I was aware that would happen so I upped a serial console and manually copied over the same confguration from pastebin
<Watcher7>
it's the same confguration as before
<Watcher7>
and modifying that option to be "devices" instead of "device", and using a list containing a single device doesn't make the assertion go away
<{^_^}>
[nixpkgs] @xeji pushed commit from @bhipple to master « scs: init at 2.0.2 (#45294) »: https://git.io/fALHX
<clever>
Watcher7: cat /mnt/etc/nixos/configuration.nix and throw it into a pastebin
__monty__ has quit [Quit: leaving]
jperras has quit [Ping timeout: 260 seconds]
<drakonis__>
Watcher7, arent you doing the thing most folks do with zfs, which is installing to a pool?
<Watcher7>
I--, you know what happened. I typo'd the directory when nano-ing over the old configuration.
<Watcher7>
Silly me.
<drakonis__>
jolly good.
<Watcher7>
drakonis__, I only use FreeBSD and ZFS for my home NAS. My virtualization server runs OpenSUSE with lvm+mdadm for its storage layer, with XFS on top for /, and VMs installed to logical volumes.
<drakonis__>
ah right
<drakonis__>
silly me, my experience with freebsd's irc denizens is that they run only freebsd
<drakonis__>
so i stand corrected
<drakonis__>
so, what're you trying to do with nixos atm, just experimentation?
<drakonis__>
also important to ask, VMs installed to logical volumes you say?
<Watcher7>
Also potentially using it as a wireguard appliance to replace some of my IPSec tunnels.
<drakonis__>
was looking into using a partition as a vm block
<Watcher7>
Yes, it's similar to `zvols`. With LVM I create a logical volume on my volume group, and then pass that to guests as a block device.
<drakonis__>
ah it needs lvm for that, alright
<drakonis__>
interesting trick
<Watcher7>
LVM is nice, I did a live migration from 4 LVM covered disks, to raid 10 on said same disks all with pv migrations and removal/replacement
<drakonis__>
not too keen on using freebsd anymore
sigmundv_ has quit [Ping timeout: 260 seconds]
ericsagnes has joined #nixos
<drakonis__>
wonder if there's an easy migration path to lvm
<Watcher7>
I use w/e I find to be the best tool for the job. NixOS's configuration method in this case I find attractive. I've become used to network appliances that have their own unified configuration that abstracts service configuration behind the scenes, and this seems to be that, but more general.
<Watcher7>
My curent NOS of choice lacks wireguard support, so I plan on sticking NixOS infront of it in the mean time.
<drakonis__>
NOS?
<Watcher7>
Network(ing) Operating System
<drakonis__>
huh i see
drakonis__ has quit [Remote host closed the connection]
<Watcher7>
clever, thanks for the assistance.
<clever>
Watcher7: yep
<clever>
Watcher7: #nixos-on-your-router may also be of interest to you
<infinisil>
That channel needs a smaller name lol
<infinisil>
#nixos-router or so
logzet has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @sterfield opened pull request #45516 → Bumping mono to 5.4.1.6 → https://git.io/fALdm
<{^_^}>
[nixpkgs] @LnL7 pushed 51 commits to staging-18.03: https://git.io/fALdB
erickomoto has quit [Ping timeout: 264 seconds]
Denommus has quit [Remote host closed the connection]
<Watcher7>
clever, ah interesting ok. Yeah. Tbh I was hoping NixOS had some form of zone-like abstraction for firewalling. The current distro I use for routing has zones which are really easy to audit.
<norfumpit>
so does nixos have any integration with dconf? im looking to use nix to configure gnome
Ariakenom has quit [Read error: Connection reset by peer]
<gchristensen>
nothing really, no
<ashkitten>
how can i get nvidia driver version 396 working? it's dropped 32-bit kernel support but still should build 32-bit libraries, right? but it outputs 64-bit libraries in the 32-bit lib paths
knupfer has joined #nixos
<ashkitten>
other distros seem to be fine with the change so i'm confused why it's stopped working
Wharncliffe has joined #nixos
Drakonis[m] has joined #nixos
<clever>
ashkitten: are you importing nixpkgs at any point?, how are you requesting a 32bit version?
<norfumpit>
gchristensen: how do yall do your gnome configurations then? im looking at services.xserver.desktopManager.gnome3.extraGSettingsOverrides
<gchristensen>
I don't :$
<clever>
same with xfce
sbdchd has joined #nixos
<ashkitten>
clever: i've modified the nixpkgs tree. it built 32-bit versions before, so i don't understand why it's not now?
<clever>
ashkitten: how are you telling nix to generate a 32bit build?
<ashkitten>
clever: i mean, i didn't even have to before, but i tried adding (pkgs.pkgsi686Linux.linuxPackages.override { libsOnly = true; }) to hardware.opengl.extraPackages32
<LnL>
:o
<emily>
norfumpit: you can set global dconf stuff in your configuration.nix.
<emily>
norfumpit: I can show you a snippet if you'd like.
<emily>
it's not, like, supported, but you can just put the right files into etc
<gchristensen>
emily: nice
<clever>
ashkitten: that override is setting a flag that doesnt exist
<ashkitten>
clever: the flag does exist tho
<emily>
I was going to use the gsettings override stuff but it looked like I'd have to enable all of gnome-on-X to get that, unless I just copy-pasted the code, which I was too lazy to do.
<clever>
ashkitten: not on linuxPackages
<norfumpit>
emily: please
<ashkitten>
clever: oh sorry i forgot a path element. linuxPackages.nvidia_x11_beta
<clever>
ashkitten: ah, that one i would expect to see working, but nixos should deal with that automatically when you have 32bit dri enabled
<ashkitten>
clever: exactly!
<ashkitten>
clever: and it builds the libraries, but they come out as 64-bit
taktoa has joined #nixos
taktoa has quit [Remote host closed the connection]
<emily>
(maybe this should go on the wiki or something.)
taktoa has joined #nixos
<emily>
norfumpit: also, dconf profiles are fancy.
erickomoto has joined #nixos
<norfumpit>
emily: oh i know about that one, but i kinda don't want to set this system-wide
<emily>
for instance, I bet you didn't know you could do this to avoid the binary database format: "A "service-db" line instructs dconf to place the binary database file for the user database in XDG_RUNTIME_DIR. Since this location is not persistent, the rest of the line instructs dconf how to store the database persistently. A typical line is service-db:keyfile/user, which tells dconf to synchronize the binary
<emily>
database with a plain text keyfile in $XDG_CONFIG_HOME/dconf/user.txt. The synchronization is bi-directional."
<emily>
(which might also help you in your quest)
<emily>
norfumpit: if you just want to like, configure gnome-terminal with nix, take a look at home-manager
<emily>
I think it has dconf integration of some sort
hamishmack has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
* emily
doesn't know what mechanism it uses.
<norfumpit>
emily: ill check that out thanks
tzemanovic has joined #nixos
<norfumpit>
home-manager seems to do it by running dconf load on a generated file
<emily>
oh, yeah, you also need to do "dconf update" in an activation script to use this, sorry :/
<emily>
I forgot how annoying it was to get this working.
Fare has quit [Ping timeout: 252 seconds]
<Drakonis[m]>
the patience of a saint
<norfumpit>
i think ill go with something based on home-managers gnome-terminal configuration
<qyliss^work>
looking at that dconf profile gave me a flashback
<Drakonis[m]>
the issue is perhaps the steam package
<ashkitten>
everything is the same except the version (and a bit in nvidia/generic.nix to account for there not being a separate 32bit version) but it's outputting 64bit libs
<ashkitten>
clever: does this look okay? feel like i might've missed some stuff but it works fine https://ptpb.pw/FmBN
<tertle||eltret>
Drakonis[m]: hopefully someone will make a .nix to accomidate the requirements for Proton
<clever>
ashkitten: probably, just make sure to also build the 64bit version, and confirm that one still works, and has 64bit libs
<ashkitten>
well i did a nixos-rebuild switch and everything is working (didn't reboot, but since everything still works that means it didn't rebuild the kernel module and the 64bit drivers are intact)
<adamantium>
i've decided to read the manual. first thing i noticed was chapter 7 recommends to source prefix/etc/profile.d/nix.sh -- how do i find my prefix ?