<ottidmes>
kingemp: so if something has julia as a dependency it will be installed, but julia will not be available in the system environment (e.g. its /bin being on your PATH via a symlink to /run/current-system/sw/bin), only if it is added to environment.systemPackages this will be the case
init_6 has joined #nixos
hamishmack has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
pxc has quit [Ping timeout: 250 seconds]
jperras has quit [Ping timeout: 252 seconds]
<ottidmes>
kingemp: ahh, every package needs to follow the right structure /bin containing the binaries, /lib the libraries, /share the static additional files, otherwise they will not be picked up, but for most of the packages the build will be such that this is already the case
<init_6>
yep man FHS
<kingemp>
ottidmes, yes, the mkDerivation I posted does have a bin/ lib/ and share/ per usual
<ottidmes>
kingemp: yeah, but your question "Is there a reason why this isn't needed for most packages?" was what I was responding too, so it is needed in all packages, but for most packages this will work out due to it already being a standard of the build process, so if some packages have no extra lines of script to make it adhere to those directories, its only because they already conform to them out the box
<kingemp>
init_6, ah I thought you meant something beyond buildFHSUserEnv in nix manual
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/fpLHJ
judson has quit [Ping timeout: 246 seconds]
dbmikus has joined #nixos
<init_6>
kingemp: FHS is basic knowledge
<kingemp>
init_6, yes I understand FHS :)
<kingemp>
ottidmes, does the installPhase script or other phase set up these symlinks?
<ottidmes>
kingemp: ah, I think I see your problem
<kingemp>
:D
<ottidmes>
kingemp: I just tried your package, and noticed it was a file rather than directory, so what you need to do instead is: mkdir -p $out/bin; cp -r $src $out/bin/julia
<kingemp>
ottidmes, hm $src and $out are both directories
jperras has joined #nixos
<kingemp>
need bin etc include lib and share
<ottidmes>
kingemp: well, thats not whats inside that .tar.gz for me
<ottidmes>
kingemp: nope it is, weird, let me check
<kingemp>
ty!
<ottidmes>
kingemp: this is what I get from your derivation: -r--r--r-- 3 root root 88860959 Jan 1 1970 /nix/store/dysshwvjhnl97lgswdj0fj8zav01fm1b-julia-binary-1.0.2, so $src is probably not what you expect
maximiliantagher has quit [Remote host closed the connection]
hamishmack has joined #nixos
maximiliantagher has joined #nixos
rpg has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
maximiliantagher has quit [Read error: Connection reset by peer]
maximiliantagher has joined #nixos
<ottidmes>
kingemp: this works: cp -r . $out
<kingemp>
ottidmes, yes, ty that fixes the directory problem, oops! still not seeing symlink in /run/current-system/sw/bin/ though
<ottidmes>
kingemp: wut? I do: which julia ~> /run/current-system/sw/bin/julia
<ottidmes>
kingemp: you have it in: environment.systemPackages, right?
<kingemp>
wait think I know the issue
<kingemp>
ottidmes, was building not installing. ty very much!!
lassulus_ has joined #nixos
maximiliantagher has quit [Remote host closed the connection]
<ottidmes>
kingemp: $src contained the .tar.gz
<kingemp>
ahh
maximiliantagher has joined #nixos
<ottidmes>
kingemp: https://nixos.org/nixpkgs/manual/#ssec-unpack-phase "The default implementation of unpackPhase unpacks the source files listed in the src environment variable to the current directory.", hence in the installPhase the contents of the $src is now in the current directory
<ottidmes>
kingemp: it does not work, does it? I get "julia: No such file or directory", I checked, it probably needs more patching, since it references to the path: /lib64/ld-linux-x86-64.so.2, which does not exist on NixOS
brejoc has quit [Ping timeout: 245 seconds]
<kingemp>
ottidmes, yes, it only runs with a chrootenv, which I'm currently revising
aminechikhaoui has joined #nixos
mayhewluke has quit [Ping timeout: 245 seconds]
aminechikhaoui has quit [Remote host closed the connection]
maximiliantagher has quit [Read error: Connection reset by peer]
maximiliantagher has joined #nixos
<ottidmes>
I should experiment with buildFHSUserEnv as well, because just like what Julia does (as I read from some of the comment on your linked issue), I have a package that uses mono, which generates the binary on first start...
maximiliantagher has quit [Ping timeout: 252 seconds]
<kingemp>
ottidmes, yes definetely not my preference to go this route, but seems forced until julia ahead of time compilation matures / becomes deterministic
<colemickens>
It would be nice to ask ofborg to build the current version of a package. I have a PR and it doesn't build for aarch64, and I don't know if it's because of my change or because it has never built for aarch64.
emacsomancer has quit [Read error: Connection reset by peer]
<steveeJ>
is there a way around building twice when using `rustPlatform.buildPackage`? i.e., how can I know the cargosha256 upfront or ignore it altogether?
<andi->
steveeJ: well thats a problem with the buildRustPackage approach I believe... It does impure cargo vedoring in a thing that is supposed to have a fixed output... Which it might have most of the time... I don't think there is a nice way around that unless you execute the cargo too beforehand and then use the right hashing of the local folder(s)..
locallycompact has joined #nixos
nD5Xjz has joined #nixos
<locallycompact>
hi, when I try to use http://exodus.io from a tarball on nixos, I get a not very useful error.
<colemickens>
andi-: I don't understand, running it twice is so that it errors the first time.
<colemickens>
so I think the ask i for something like `nix-prefetch-url --unpack` but for however buildRustPackage does cargo256
<colemickens>
that's presumably possible since the hash stays the same once it's wrong,fixed,updated-in-the-source.
<colemickens>
but I also don't know much about how buildRustPackage works, so I'm curious.
<colemickens>
locallycompact: I am stubborn and package everything for nixpkgs instead of one-off patching. Let me look it up, and see if htere's a Wiki entry or add one real quicl.
kyren has quit [Remote host closed the connection]
<locallycompact>
colemickens, if I wanted to do a full package of this is there a cookie cutter way to just say "these inputs, and then verbatim copy everything in the tarball" and then put on nixpkgs?
<colemickens>
locallycompact: depends, do you want to submit this upstream to nixos/nixpkgs ?
<colemickens>
locallycompact: or just something that will "just work" right now?
<locallycompact>
could submit yeah I use this on many machines
<colemickens>
if doing upstream, I'd recommend trying to build it from source. You can do `exodus-bin` but afaik it's generally preferred to build.
camsbury has quit [Ping timeout: 252 seconds]
<locallycompact>
oh I hadn;t even checked if this was open I'd always assumed it wasnt
<colemickens>
I don't think it is. Generally non-source packages are "whatever-bin" is all.
<locallycompact>
ok
<colemickens>
I'm seeing if I can whip something up.
<colemickens>
locallycompact: yes, but when it tries to patch the library references, it looks like it hits an internal trace line and seems to abort.
<colemickens>
I think even more work might be needed to get the packages dependencies to work.
<locallycompact>
it seemed to complete for me
<colemickens>
did it print a path at the end?
<locallycompact>
yes
<colemickens>
you could try `$(nix-build)/bin/Exodus` ?
Itkovian has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<Thra11>
R optionally depends on jdk. There's no free jdk on arm. Should I just exclude jdk on arm, or should I try to keep jdk as a dependency on arm *if* allowUnfree is true?
<locallycompact>
there's no /bin/Exodus in the store path
<locallycompact>
oh no there is
<locallycompact>
fish: “./Exodus” terminated by signal SIGTRAP (Trace or breakpoint trap)
<locallycompact>
I see
<colemickens>
Oh, I guess mine did run and failed.
<colemickens>
Yeah... :/
<colemickens>
Honestly, it would've been easier to get it going from source ;)
<locallycompact>
yeah
<locallycompact>
thanks for trying I appreciate it
<locallycompact>
colemickens++
<{^_^}>
colemickens's karma got increased to 1
voiceftp has quit [Read error: Connection reset by peer]
<colemickens>
np! you could try filing an issue and see if someone else can help.
voiceftp has joined #nixos
alphor has quit [Ping timeout: 252 seconds]
maximiliantagher has joined #nixos
<andi->
steveeJ: the problem with that is that it is tightly coupled with how cargo handles things... I've been trying to improve the buildRustCrate function and cargo does some crazy things becuase they have to support changing meanings of the Cargo.toml.. not sure how much heuristic is needed to properly parse a Cargo.lock. I tried building `spotifyd` with buildRustPackage yesterday and it failed at some stage of
<andi->
fetching deps.. When I compile it with `cargo build` it just works....
Itkovian has joined #nixos
<colemickens>
actually, I've had weirdness happen. I have a package in my overlay where all I did was regenerate the lock file, make a patch, and then all of the sudden it started building.
<colemickens>
and it wasn't a version mismatch or a cargo.lock format version problem either.
<colemickens>
(a Rust package, that is)
maximiliantagher has quit [Ping timeout: 268 seconds]
<colemickens>
if I build without the patch, I get "failed to load pkg lockfile".
<colemickens>
But the lockfile looks fine without the patch...
<colemickens>
and the patch doesn't seem to change much other than some timestamps and maybe an extra transitive dependency, but nothing that seemed like it would have fixed something causing that error.
<srhb>
Oh, sorry, I completely misunderstood as: I changed things why is it rebuilding :P
<colemickens>
yeah, I didn't explain very well that it was _failing_ before the regenerate :)
<srhb>
Did you change the cargosha though? :)
<colemickens>
just did again before replying :)
<colemickens>
thing is, it doesn't even get to checking the sha
<srhb>
I'm assuming that's the sha that describes the lockfile. Just wondering if you were accidentally using the old one.
<colemickens>
as in, I think the mechanics in buildRustPackage are truly failing to understand cargo.lock
<srhb>
How so?
<colemickens>
maybe there's some fixed behavior in newer versions of Cargo, thus regenerating made a better lock file.
<colemickens>
srhb: the package fails to build before it even complains that the cargo sha is wrong
ottidmes has joined #nixos
<srhb>
can I find the commit you expected to work but didn't without the patch somewhere?
<srhb>
colemickens: Why would the sha be wrong?
<srhb>
If you just left the old one in, there's no rechecking going on.
<colemickens>
srhb: I made it a bad one to trigger it
<srhb>
Ah
<colemickens>
srhb: the metadata.nix in the same directory points to the upstream (i3status-rust) commit that it is building against.
<colemickens>
and/or the patch probably has the base revision from when I originally generated it.
<srhb>
colemickens: So the upstream lockfile does indeed appear to be broken.
<srhb>
colemickens: It would probably be good to throw a --verbose at cargo vendor in fetchcargo.nix
<srhb>
Or at least check its exit status and do that if it fails
<srhb>
It would also be nice if buildRustPackage was able to pass fetchCargoArgs.
<srhb>
(Or something similar)
mayhewluke has quit [Ping timeout: 276 seconds]
mayhewluke has joined #nixos
<srhb>
colemickens: btw, it would be helpful if build.nix exposed some attribute that actually uses <nixpkgs> or takes nixpkgs as an argument.
<colemickens>
srhb: I was discussing that with zimbatm in an Issue. I'm reluctant to do that because the overlay doesn't actually work with 18.09, for example.
<srhb>
colemickens: Right, it's just for debugging
<colemickens>
Ah. Okay.
<colemickens>
ls
<srhb>
colemickens: (If you want to make changes to nixpkgs to test your overlay it's hard right now)
<srhb>
No biggie
<srhb>
colemickens: But yeah, I'm pretty sure that the vendored lockfile is just plain broken. :)
<colemickens>
Makes sense, appreciate the feedback.
<colemickens>
And the insight into the lock file :)
<srhb>
colemickens: The error message from cargo vendor --verbose is _really_ helpful and we ought to include that :P
maximiliantagher has joined #nixos
hydraz is now known as ^W
maximiliantagher has quit [Ping timeout: 240 seconds]
alienpirate5 has quit [Remote host closed the connection]
<steveeJ>
andi-: weird enough that I just changed the `rev` to `builtins.fetchGit` for the crate sources without changing the cargo hash and could install the drv without hicking up
zolk3ri has quit [Ping timeout: 256 seconds]
camsbury has quit [Ping timeout: 252 seconds]
<andi->
steveeJ: well probably because the depds didn't change? what `fetchCargo` does is call `cargo vendor` which then does a bunch of network and downloads all the things :/
<ottidmes>
I have a package that generates some files at startup, which will fail due to the directory being readonly, could I somehow use buildFHSUserEnv to workaround this? My current experiment is to just recreate the directory structure in my home folder and symlink all files, then I can use the correctly Nix build files, but still have them generate things in the a writable directory
<{^_^}>
[nixpkgs] @nyanloutre opened pull request #50177 → spaceship-prompt: init at 3.7.1 → https://git.io/fptIq
<lejonet>
Is there anything special I need to tell nixops about when deploying to a nixos host that is an arm machine? Because it seems like the perl binary it pushed to the machine is a 64-bit ELF lol
<lejonet>
(/nix/store/l6qc2aa8vf6sxpmi7cw28nrihhmmmvp5-perl-5.24.3/bin/perl: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /nix/store/hwwqshlmazzjzj7yhrkyjydxamvvkfd3-glibc-2.26-131/lib/ld-linux-x86-64.so.2, for GNU/Linux 2.6.32, not stripped)
<colemickens>
zimbatm: I'd definitely love to hear if upgrading helps, but based on what I've read it's nontrivial and I doubt they've fixed it.
<ottidmes>
hmm, I should look into that as well, I have a 4k display, but have been using it as is, and just zoomed e.g. vscode by zooming the application
<colemickens>
I was planning on trying to build firefox and chromium with wayland support but I'm having trouble getting a VM provisioned with the stats to build them, especially since I'm sure it's going to take lots of iterating and rebuilding.
<{^_^}>
[nixpkgs] @timokau pushed commit from @r-ryantm to master « flacon: 4.1.0 -> 5.0.0 (#50114) »: https://git.io/fptIF
<colemickens>
ottidmes: I think zimbatm is talking about a specific Wayland/Xwayland bug that we're seeing in Sway.
<zimbatm>
yes
<colemickens>
Code, etc, work well in GNOME though. Chrome has been able to sniff out GNOME's high dpi settings for a while.
<zimbatm>
which also made me discover that chrome isn't using wayland, even though the support was added in 50.0
<ottidmes>
colemickens: I gathered as much, but does only Wayland then have support for high DPI? I was just thinking, if there is some support for in xserver, I should probably try it, it would be better than doing it per application, like I have been doing so far
<zimbatm>
yeah the problem is that each toolkit is trying to do it's own thing regarding DPI scaling
<lejonet>
srhb: was going to prod you about what your thoughts about the future of ceph for nixos, the whole lts/current thing I mentioned on the PR
<{^_^}>
[nixpkgs] @hlolli opened pull request #50181 → add myself (hlolli) to maintainers list → https://git.io/fptt6
<srhb>
lejonet: Always up for discussion, however I'm not currently using ceph anymore, so better people to discuss it with is probably, among others, krav_ :)
<lejonet>
srhb: haha okay :P
<srhb>
I think following an LTS makes sense, though, but the current build systems is ungghhh and I wish we could properly pass our own deps all the way
camsbury has joined #nixos
<lejonet>
I just got up nixos on an odroid-hc2 which I'm going to use for a test cluster of ceph
<srhb>
lejonet: You're still using it?
<srhb>
ooo nice
hlolli has joined #nixos
<lejonet>
Yeah, the buildsystem is... :P
<lejonet>
srhb: yeah, I've got a cluster in prod and now setting up this for test (with a hope of adapting the prod cluster to use odroid-hc2s)
<srhb>
Ack, it's armv7?
<lejonet>
yeah...
<srhb>
So no cache :<
<lejonet>
Exynos 5422
<lejonet>
Yeah, but I've been toying with the idea of a private cache anyway so, not too much of a bummer
* srhb
nods
<srhb>
Well it's a cool device :)
<lejonet>
it is indeed, and will drastically lower the cost per TB as long as it performs at least to the minimum of my needs
<lejonet>
(which I think it will)
dramforever has joined #nixos
<srhb>
Yeah, I'd expect so too!
<lejonet>
Because using odroid-hc2s I will up the network bandwidth for each OSD from 2 Gbps per 4 OSDs to 1 Gbps per OSD
<srhb>
10, right?
<srhb>
Oh, no, it's gigabit
<srhb>
But still!
<srhb>
Very nice :)
<lejonet>
Yeah, it'll be a theoretical increase of 100% for the cluster so :P
<lejonet>
With the bummer that I can't upgrade to 10 if needed :P
<lejonet>
(but I doubt I'll need it lol)
<srhb>
Indeed.
camsbury has quit [Ping timeout: 244 seconds]
<lejonet>
But I've heard that odroid might be working on a x86 board, with a little more goodies and such
<{^_^}>
[nixpkgs] @Mic92 merged pull request #50181 → add myself (hlolli) to maintainers list → https://git.io/fptt6
<lejonet>
now I just need to figure out how to convince my desktop to build stuff on the odroid when doing nixops instead of trying to do it locally :P
<gchristensen>
lejonet: maybe ping sphalerite about armv7
<dramforever>
Yeah that was what I mean by 'following along'
<{^_^}>
[nixpkgs] @Infinisil opened pull request #50183 → .github/CODEOWNERS: Add infinisil to idris-modules → https://git.io/fptqQ
<infinisil>
Ugh, the fact that it's "nixos" on NixOS but "nixpkgs" on non-NixOS is so annoying
<infinisil>
We should somehow get rid of this
<lejonet>
Gaaah, doesn't nixops know how to use SSH keys? I've setup my odroid-hc2 as a buildMachine for armv7l-linux and given it a ssh key and all, I can ssh with that key perfectly fine myself (no passphrase on key) but nixops doesn't manage to open up a ssh connection to it :(
<infinisil>
Nixops uses its own ssh kesy
<infinisil>
You can see them with `nixops --export`
<lejonet>
even if the buildmachine conf tells it to explicitly use something else?
<infinisil>
buildmachine conf?
<dramforever>
Also
<infinisil>
nixops isn't influenced by any NixOS options
<dramforever>
I don't think there's a easy way to get old versions of channels, is there?
<lejonet>
Well, it seems to follow the buildmachine conf, kind of, because it tries to connect to the machine in question when it is going to build stuff
<ottidmes>
is there any tool that says that the library named foo in Ubuntu or Arch is named bar in NixOS? I guess I could just try and see what is in those Ubuntu/Arch packages, and use nix-locate, any other way?
<dramforever>
channels.nix.gsc.io gives channel histories, but it's the... commit hash
Izorkin has joined #nixos
<gchristensen>
dramforever and what do you want instead? to download all the things?
<dramforever>
So I can get the store-paths.xz from it and download its closure
<gchristensen>
oh interesting
<dramforever>
So that I can populate a binary cache
<lejonet>
infinisil: even when I added nixops own key to authorized_keys on the machine, it fails to open a SSH connection O.o (which is why I think that it actually does listen to the nix.buildMachines conf from the host, when it actually wants to build someting)
<gchristensen>
makes sense
sonnenbloom has joined #nixos
<Thra11>
Does making a parameter optional (i.e. changing "foo" -> "foo ? null") trigger rebuilds even if foo is still supplied?
<dramforever>
Like, if the channel went A -> B -> C, and I try to clone the channel, I should probably also get B because some people might be still on B
Guanin has joined #nixos
oida has quit [Remote host closed the connection]
<dramforever>
The fact that store-paths.xz is not a closure really complicates the matter
<gchristensen>
oh?
<gchristensen>
hmm this bash script is getting a bit gnarly... making it better might require an actual programming language
<dramforever>
Exactly the same thing I did when I discovered that store-paths.xz is not a closure
wucke13 has joined #nixos
<dramforever>
Non-closure example: I just downloaded nixpkgs-unstable/store-paths.xz
<dramforever>
It has 65625 paths
<dramforever>
but the closure is actually 75322 paths
<gchristensen>
just extract-paths-from | xargs -P$(nproc) nix-store -r
<dramforever>
Oh
<gchristensen>
extract-paths-from is a fake command I invented
<dramforever>
I meant the 'download ~200k narinfos' part
<gchristensen>
ack
<dramforever>
I couldn't really download ~420GB files
<steveeJ>
is there a `sed` like function with regex support in nix?
<steveeJ>
I want to apply `sed -E 's,[^a-zA-Z0-9\.\+_\-]+,,g'` to a string
<dramforever>
(And I wouldn't know it would be ~426.7GB (sum of all 'FileSize's) if I didn't have the database :p)
<gchristensen>
dramforever: you can't download that much, but you want to?
<dramforever>
I can't do it on my laptop
<dramforever>
But if I can prove I can do this
<gchristensen>
ah :)
b has joined #nixos
<dramforever>
And get some information about it
<dramforever>
I can hopefully get some server maintainers here to run my script
<wucke13>
But they won't be included in the .desktop files of the ragarding applications (so that launching a program from a launcher starts the program with gksu/kdesu)?
<emily>
wucke13: my understanding is that running GUI apps as root is semi-"deprecated" these days, and applications should be written to elevate permissions only for certain operations with PolicyKit or whatever
<emily>
e.g., gnome's Disks app supports some partitioning/formatting runs as the user, not root
<emily>
honestly I don't think I've heard anyone mention gksu/kdesu in years
<emily>
I don't think Debian packages gksu even?
<ivan>
oops I did a nixos-rebuild switch --upgrade on a remote machine (incl a kernel upgrade) and that somehow took down wireguard and didn't bring it back up again
<dramforever>
gchristensen: Goal: With this mirror, all the official *.nixos.org stuff won't be required anymore
<wucke13>
emily: Ah hmm. Thanks :)
<gchristensen>
dramforever: right, but what do you want from the data API
<dramforever>
Seems like difference would be in the minutes range. Not a problem
mayhewluke has quit [Ping timeout: 272 seconds]
* lejonet
is starting to think that its less painful to setup a crossdev toolchain for armv7l than manage to convince the build daemon that it can connect to the machine
<gchristensen>
lejonet: have you looked at the logs of the remote builder?
<lejonet>
gchristensen: seeing as the ssh connection is dropped, by my desktop, before any auth even is done, I doubt there will be something there, but what logs would that be in? nix-daemon?
mayhewluke has joined #nixos
<gchristensen>
sudo journalctl -f, then do a build
<lejonet>
gchristensen: http://dpaste.com/15RMJFJ this is all I get, the error is related to how I forward my gpg-agent socket and shouldn't have any bearing on this
irdr has joined #nixos
<lejonet>
This is the error that nixops gives me "unable to open SSH connection to 'ssh://builder': cannot connect to 'builder'; trying other available machines..." before stating that it cannot build armv7l-linux itself, because its x86_64-linux
<gchristensen>
ah, are you on nixos?
<lejonet>
Yeah
<lejonet>
both machines are :)
<gchristensen>
you can't use an agent for this because ssh doens't happen as you
<lejonet>
Well thats the thing, I've told it to use its own key in buildMachines, but it ignores it
<gchristensen>
this log is from the target machine?
<lejonet>
Yes
<gchristensen>
and those connections are you?
<lejonet>
Yep
<lejonet>
and I know that its not using the supplied ssh key because that key is an ed25519 key, not RSA, RSA is the key from gpg-agent...
<gchristensen>
try becoming root and SSHing with that key
<lejonet>
nixops deploy -d entire-env --include odroid-hc2-0 is the command I'm using
<gchristensen>
ok, paste your /etc/nix/machines file?
<lejonet>
O.o switching back to this in buildMachines made it start working all of a sudden: http://dpaste.com/35R5P8S and literally the only difference now is where the key is placed, does the nix-daemon care about where its placed, and not just that it can access and use it?
<gchristensen>
"sshUser = root;" is fishy
<lejonet>
You don't say? ;)
Itkovian has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
acarrico has quit [Ping timeout: 268 seconds]
<lejonet>
This is just to get it to run at all, I really don't want it to ssh as root, just used that to rule out any type of permissions on the target host
<gchristensen>
gotcha
<gchristensen>
well the user has to be a Trusted User on the remote
<lejonet>
root is the only interactive user on the remote atm so
<gchristensen>
"Warning: Adding a user to trusted-users is essentially equivalent to giving that user root access to the system. For example, the user can set sandbox-paths and thereby obtain read access to directories that are otherwise inacessible to them."
<lejonet>
later on I want to create a builduser account and such
<lejonet>
maybe even lock it down with ForcedCommand through SSH if that is doable
<lejonet>
gchristensen: The thing is, the biggest problem behind this issues was the fact that I couldn't get any info from the command of what SSH command it tried to use, and what the exact error with connecting to the machine and such was (not even with --debug option to nixops deploy) so that was the main source of frustration for my part :P
<gchristensen>
for sure
<lejonet>
Because I couldn't even see if it used the right key or such
<lejonet>
I tried to strace it... but sifting through all that was... hard :P
<gchristensen>
ok time for me to go away from the computer a while. good luck y'all, have fun :)
<{^_^}>
[nixpkgs] @jfrankenau opened pull request #50191 → fusuma: init at 0.10.2 → https://git.io/fptOP
<lejonet>
gchristensen: Have a good one!
<lejonet>
gchristensen: Thanks for all the help :)
<ottidmes>
is there an easy way to find all actual executables (i.e. program binaries), whether they are marked executable or not? I now just have hardcoded some of them that were not executable and chmod'ed them, but is there a good alternative that does not need hardcoding?
grumble has quit [Quit: Need to reconnect so here's a random tip: The command "xrandr --output HDMI1 --set 'Broadcast RGB' Full" (you might need to replace HDMI1 with your monitor output name) might significantly improve contrast on external monitors. s/Full/Automatic/ to reset]
<Thra11>
gchristensen: I assumed the order didn't affect it, but it looks like you're probably right. Should I put the conditional in the middle of the list (messy), or just accept that it will trigger rebuilds?
<gchristensen>
Thra11: I wouldn't worry about 11-100 rebilds anyway
<gchristensen>
no, not worth the mess to avoid a few rebuilds.
<Thra11>
gchristensen: Ok, cool. Thanks!
<samueldr>
+1
<gchristensen>
(we rebuild so much that even big rebuilds are often not worth the mess to avoid...)
<ottidmes>
any reason to be explicit about .out, i.e. openssl vs openssl.out?
<gchristensen>
usually not
ninjin has quit [Quit: WeeChat 2.2]
illegalprime has quit [Ping timeout: 252 seconds]
<ottidmes>
and could I use buildFHSEnv to achieve the same as I am doing now with a service that does something like: cp -rs --no-preserve=mode $out/share/vscode/extensions/ms-vsliveshare.vsliveshare ~/.vscode/extensions (i.e. copies the complete directory structure and symlink files, so it can write new files in the directories)
<zimbatm>
colemickens: should we start a #nixos-wayland
freeman42x]NixOS has quit [Quit: Leaving]
<zimbatm>
i had a look at the xorg-xserver situation, it seems like people are already working on it
ma27 has quit [Ping timeout: 252 seconds]
maximiliantagher has quit [Ping timeout: 245 seconds]
<zimbatm>
i also checked sway nixos module, it seems like things could be moved the the nixos home-manager potentially as the setpcap are not necessary when using systemd-logind
<arianvp>
causing the changes to not be propagated
<gchristensen>
zimbatm: agreed :)
<arianvp>
is this a bug?
<arianvp>
I add a bridge to networking.bridges but it doesn't appear after nixos-rebuild switch
<zimbatm>
gchristensen: which bit? :_)
<arianvp>
only after I manually do "systemctl start network-setup.service"
<gchristensen>
zimbatm: not supporting 18.09
<zimbatm>
+
fresheyeball has joined #nixos
<fresheyeball>
hey out there
<fresheyeball>
I am building a haskell project with nix
<fresheyeball>
and want to enable cffi for aeson for performance reasons
ma27 has joined #nixos
<fresheyeball>
how can I do this?
<arianvp>
how does scripted networking even detect changes?
<arianvp>
fpletz:
johann__ has joined #nixos
<fresheyeball>
Can I set flags in the generic builder?
<fresheyeball>
configreFlags? or buildFalgs?
shabius has quit [Ping timeout: 252 seconds]
<Thra11>
worldofpeace: I didn't notice that you'd 'suggested changes', I just read your comment and pushed the changes you described, so now it won't show me your suggested changes. Was it just the commit message or were there other changes?
<fresheyeball>
c0bw3b_: I think it's different for the haskell builder
<c0bw3b_>
oh! I missed the 'Haskell' keyword! my bad..
<c0bw3b_>
don't know much about Haskell packaging
<fresheyeball>
no worries
<ottidmes>
is my understanding correct that if some executable has dynamic linked libraries they are the ones displayed when you use e.g. ldd? and it also load libraries on demand then I need to use a wrapper? since my packaged executable runs, but gives an error that libsecret cannot be loaded, which is not listed when doing ldd, and it was present in its rpath
shabius has joined #nixos
mahakala has joined #nixos
<mahakala>
hi all, I just installed nixOS for the first time ever... I've been planning to do this for 1.5 years and finally took the time. Anyway, there are some software I run that forces me to run unstable (I need wayland 1.16 instead of 1.15)... but I don't understand how to go about it. Can anyone point me in the right direction? :)
<gchristensen>
zimbatm: :)
shpx has joined #nixos
__monty__ has quit [Read error: Connection reset by peer]
<ixxie>
c0bw3b_: lol... I think I added that one to the wiki so ironic I forgot about it
__monty_1 has quit [Remote host closed the connection]
__monty__ has joined #nixos
<colemickens>
mahakala, In your PM you mention you specifically want to move to 'nixos-unstable'. That makes it sound like you want to change the channel your whole system is on. An important thing to understand what that means, the links ixxie and c0bw3b_ sent are good. Also, if you ask here or in an Issue regarding the overlay you'll have a better chance of getting help since I'm not going to be at a computer all day.
drakonis1 has joined #nixos
<c0bw3b_>
^^ it's clearer to newcomers I believe because it makes it clear you work with two channels and the default 'nixos' remains stable so you don't break your system
ixxie has quit [Remote host closed the connection]
<mahakala>
ok, looking into this now... thank you :)
drakonis_ has quit [Ping timeout: 260 seconds]
crmlt has quit [Ping timeout: 264 seconds]
camsbury has joined #nixos
hdeshev has quit [Quit: Leaving.]
zolk3ri has quit [Quit: Lost terminal]
dbmikus has joined #nixos
maximiliantagher has joined #nixos
camsbury has quit [Ping timeout: 260 seconds]
zarel has quit [Remote host closed the connection]
slyfox_ is now known as slyfox
maximiliantagher has quit [Ping timeout: 260 seconds]
<{^_^}>
[nixpkgs] @NeQuissimus pushed 4 commits to release-18.09: https://git.io/fptRs
<worldofpeace>
Thra11: That was the only change
<{^_^}>
[nixpkgs] @NeQuissimus pushed 3 commits to release-18.03: https://git.io/fptRc
<mahakala>
hm, I don
<mahakala>
I don't know what I've done wrong. I get "file 'nixos-unstable' was not found in the Nix search path"...
gammarray has joined #nixos
<yl[m]>
LnL: w.r.t nix-darwin users module, is it meant to bring up the users' user? I erased my old mac and brought up an `admin` account and used nix-darwin to create my `yl` user but I can't find it to login
gammarray has quit [Client Quit]
<yl[m]>
also no way to add the user to extra groups?
maximiliantagher has joined #nixos
sippeangelo has joined #nixos
<sippeangelo>
Quick NixOS newbie question: What do I do when nixos-build doesn't take my configuration changes into account?
<lassulus>
sippeangelo: thats odd, what exactly do you do?
ninjin has quit [Remote host closed the connection]
ninjin has joined #nixos
maximiliantagher has quit [Remote host closed the connection]
<sippeangelo>
lassulus, I've been trying to get the boot.initrd.network.ssh option to work, but it needs a host key file, and while the "hostECDSAKey" option worked the first time I built, if I change it to something else the build will fail when it still references the old file path which file doesn't exist on disk any more...
<sippeangelo>
Actually it's not the build that fails, but the activation of it, I guess...
maximiliantagher has joined #nixos
<lassulus>
if the activation fails grub/systemd-boot doesn't get updated and you will still boot the old config/initrd
<lassulus>
the sandbox usually can't access files outside of /nix/store
random_yanek has quit [Ping timeout: 245 seconds]
<lassulus>
but if you supply the files as paths they get copied into the store
<lassulus>
but maybe this is bad for secrets
<sippeangelo>
Yeah, this specific option seems to be some kind of special case where it uses initrd.secrets to make the file available, and it worked fine unless I change the path in my config, at which point it never actually updates my build with the new path
<sippeangelo>
hostECDSAKey is clearly different, but it still uses the old path
<lassulus>
sippeangelo: well you can write hostECDSAKey = /root/initrd-ssh-key; then it should work
<lassulus>
but initrd-ssh-key will end up in /nix/store
<sippeangelo>
Well it doesn't seem to matter what I put there any more, it just ignores it even when I change it!
random_yanek has joined #nixos
vidbina has quit [Ping timeout: 246 seconds]
<sippeangelo>
lassulus, I just wonder if there's any way to force it reevaluate my entire configuration from scratch? Delete some store path or the other, I don't quite understand how it all works internally.
<lassulus>
hmm, alright, I understand now, I will look into the source, normally no garbage-collection should be necessary
<lassulus>
can you tell me on which channel you are?
<sippeangelo>
Whatever the default is I guess!
<lassulus>
can you do nix-channel --list?
<ottidmes>
yay! I just finally got VSCode Live Share working on NixOS :) (confirmed with someone on Win10)
<lassulus>
(I'm not really sure what the default is right now :D )
<mahakala>
I've got support now for unstable packages, I think, but I cannot seem to get wayland-server 1.16 instead of 1.15. I have this "unstable" alias for nixos-unstable and "unstable.wayland" as systemPackage... any idea what I might be doing wrong? here is my configuration.nix: https://pastebin.com/veff4efB
<lassulus>
sippeangelo: can you paste the contents of /nix/store/9zh7brwrmggg8vyq540vqp1hspyz5rbv-append-initrd-secrets/bin/append-initrd-secrets ?
<yl[m]>
LnL: oh isHidden is default to true, interesting
<sippeangelo>
lassulus, Well right now I use this tool called morph that does some combination of nix-build and copy and activate under the surface, but I had the same problem when just running "nixos-rebuild switch" before. I "solved" this once by running "nix-collect-garbage --delete-old", but this also deletes my entire package cache... There must be SOMETHING I can delete to just make it rebuild that one "append-initrd-secrets" script,
<sippeangelo>
right?
Rusty1 has quit [Quit: Konversation terminated!]
<ottidmes>
sippeangelo: why not add some random stuff to the secrets, maybe that triggers it
aanderse has joined #nixos
dbmikus has quit [Ping timeout: 250 seconds]
goibhniu has joined #nixos
<lassulus>
hmm, maybe old generations still try to rebuild their initrd and try to get the secrets again? But I'm not sure and will have to look deeper into it
<lassulus>
I wanted to try out boot.initrd.network in the next week anyway, so maybe I will trigger the same error
<ottidmes>
lassulus: if you want, I have an openssh version of it
<lassulus>
ottidmes: I would try that out
<clever>
sippeangelo: nix-store --delete /nix/store/foo, and dont use force
<sippeangelo>
I just ran "nix-store --delete $(find /nix/store/ -name '*initrd-secrets*')" and this helped trigger a rebuild!
<lassulus>
ottidmes: cool, look good, why not upstream it?
rixed has joined #nixos
<rixed>
Hello there! Any up to date doc (I'm running nixos 18.09) on how to compile a kernel module? Everything I've copypasted from internet failed :/
rpg has joined #nixos
maximiliantagher has quit [Remote host closed the connection]
<ottidmes>
lassulus: because its not objectively better than dropbear, dropbear cost less in terms of space/memory, but I rather trust one implementation and space/memory is a non-issue for most servers
goibhniu has quit [Quit: Leaving.]
<mahakala>
colemickens: hm, been trying to get it work now but fail... I guess I don't understand enough how the nix language work or something...
<mahakala>
been trying different things, none that manages to build wlroots :/
orivej_ has quit [Ping timeout: 246 seconds]
<{^_^}>
[nixpkgs] @aanderse opened pull request #50210 → nixos/incron: added nixos test to ensure expected behaviour → https://git.io/fptua
<mahakala>
I'm looking at that readme.md with the info howe to use with 18.09, but when I add the text after with pkgs ;[...] I get syntax error :(
slack1256 has joined #nixos
jimbob123456789 has joined #nixos
jmeredith has quit [Quit: Connection closed for inactivity]
<sippeangelo>
ottidmes, So when I change boot.initrd.secrets to contain a bunch of junk, it does rebuild the append-initrd-secrets script, but still executes the old one!! https://pastebin.com/vEXbDS0k
<colemickens>
mahakala: see the comments in the PR, I have another small change/fix to it.
<colemickens>
someone else comments the same thing in the PR comments.
<jimbob123456789>
I want to build a make a package (overlay) that builds a dbus service. Can anyone recommend any examples I can use as a reference?
<mahakala>
colemickens: sorry, saw that just now... it seems to be compiling now :)
<colemickens>
cool!
<ottidmes>
sippeangelo: I think I know your problem
<ottidmes>
sippeangelo: You still have an active build of an old configuration, since you have been using `nixos-rebuild switch` all the time, but you are booted in an other configuration
maximiliantagher has joined #nixos
<ottidmes>
sippeangelo: in order for the old configuration to still work, NixOS needs to ensure the secrets file belonging to that old build still works
<ottidmes>
sippeangelo: easiest fix on top of my head, touch /home/imon/Workspace/Nix/nixos/profiles/initrd-ssh-key (make sure it exists), make sure your `sudo nixos-rebuild switch` succeeds, and then reboot, garbage collect, and it should not longer build the old one
<sippeangelo>
ottidmes, This makes a whole lot of sense, but the thing is that path never even existed in the first place, since it references a path on my other computer where the configuration was built, hence why I'm trying to change it. I have no idea how it accepted it in the first place if this is true, but I'll try this, thanks!
maximiliantagher has quit [Ping timeout: 268 seconds]
<sippeangelo>
ottidmes, It worked, thank you so much! NixOS is just too clever for me. I just had to go through recreating 6 or so files from previous failed attempts, including my junk ones xD
dbmikus has joined #nixos
<ottidmes>
sippeangelo: great to hear you managed to solve it :)
mayhewluke has quit [Ping timeout: 268 seconds]
mayhewluke has joined #nixos
<sippeangelo>
ottidmes, Thanks for the OpenSSH version too! I believe it would also help with this related issue for those using Nixops: https://github.com/NixOS/nixpkgs/issues/24727
<{^_^}>
#24727 (by mbrgm, 1 year ago, closed): Failing nixops deployment when using `boot.initrd.network.ssh`
<ottidmes>
sippeangelo: I read that it is caused by the different formats used, so probably. I have deployed with nixops while using those settings, so I am quite certain that the issue is not applicable to my module
freeman42x]NixOS has joined #nixos
zarel has quit [Quit: Leaving]
maximiliantagher has joined #nixos
<schmittlauch[m]>
Does someone here successfully use wireshark? I'm told to add myself to the "wireshark" group, but that one doesn't even appear in /etc/group
<clever>
sippeangelo: one sec
rpg has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<schmittlauch[m]>
probably the resulting question is: Is the `wireshark` group created automatically, either by the wirshark package itself or by the group appearing in `user.<name>.extraGroups`, or do I need to create it manually?
civodul has quit [Quit: ERC (IRC client for Emacs 26.1)]
<samueldr>
(though it won't add anyone to the `vboxusers` group, it is documented as required for some features)
jperras has joined #nixos
maximiliantagher has quit [Ping timeout: 272 seconds]
<schmittlauch[m]>
samueldr: a user still having to add themselves is ok. But at least the requred group should be created automatically by the package/ module
<schmittlauch[m]>
I could open an issue about that, shall I?
<clever>
i'm thinking a programs.wireshark.enable = true; which both installs wireshark, and creates the group and wrappers
<{^_^}>
[nixpkgs] @matthewbauer opened pull request #50212 → [wip] Add "exeWrapper" function → https://git.io/fptgV
<schmittlauch[m]>
clever: sounds good. Although I sometimes find the distinction between a module pulling in the package and the package itself a bit confusing
<samueldr>
packages will work outside nixos, modules are a nixos-only thing
<samueldr>
(if simplified, because I know about 10 people here are going to link things for using parts of the modules system outside nixos ;))
<schmittlauch[m]>
it's a bit of a discoverability problem. I first look for the package itself using `nix search`. So I should probably change that look-up order
Thra11 has quit [Ping timeout: 246 seconds]
<samueldr>
the main idea is that the packages cannot affect the system directly
<samueldr>
so it cannot create groups, users, change udev rules
<samueldr>
(among multiple others)
<schmittlauch[m]>
I got that, modifying system configuration in a package build expression is a side-effect we can not isolate
<{^_^}>
[nixpkgs] @markuskowa pushed 2 commits to pr-gnss-sdr: https://git.io/fptgP
<{^_^}>
[nixpkgs] @markuskowa pushed 0 commits to pull/50199: https://git.io/fptgM
<{^_^}>
[nixpkgs] @markuskowa pushed 0 commits to pull/50199: https://git.io/fptgy
<{^_^}>
[nixpkgs] @markuskowa pushed 0 commits to pr-gnss-sdr: https://git.io/fptgy