<ddellacosta>
so this is a bit of a long shot but: anyone have any idea how to get an ancient Edirol FA101 firewire audio interface working with a setup that connects it via firewire -> thunderbolt -> thunderbolt 3/usb-c?
jsgrant has quit [Remote host closed the connection]
<acowley>
Is anyone here familiar with using CUDA with nix?
ddellacosta has quit [Quit: WeeChat 2.2]
<acowley>
I have a jetson box on which I'd like to use nix as a package manager, but when I use the provided nvcc from within a nix-shell (or associated direnv), I get a broken executable (CUDA runtime isn't working, so segfaults or other random behavior).
<acowley>
So the problem probably happens with nvcc reaches out to g++, but I'm not sure what is going wrong. gdb just shows a segfault on an array write, for example, and ldd on the executable looks like a normal nixified executable.
jtojnar has quit [Quit: jtojnar]
<acowley>
The simplest answer is to not use nix on that box, but it's a fairly annoying environment since the answer to every problem is to reflash it so having an automated reproducible environment would be nice.
vk3wtf has joined #nixos
Supersonic has quit [Disconnected by services]
Supersonic112 has joined #nixos
Supersonic112 is now known as Supersonic
rprije has joined #nixos
<sveitser>
I had similar problems with CUDA nixos when I was missing `services.xserver.videoDrivers = [ "nvidia" ];` in my nixos configuration.
<acowley>
I have to use the provided Ubuntu on this machine, so full NixOS is out of the question. I wouldn't think such a service could impact nvcc in this way.
BombStrike has joined #nixos
maximiliantagher has quit [Remote host closed the connection]
maximiliantagher has joined #nixos
slack1256 has joined #nixos
maximiliantagher has quit [Remote host closed the connection]
maximiliantagher has joined #nixos
oida has quit [Quit: byez]
maximiliantagher has quit [Ping timeout: 245 seconds]
<`_>
which is a haskell package that reads jenkins plugins and outputs a nix expression
<`_>
now it would be nice not to run the binary and commit the expression, but have it as a nix function
<`_>
so that would require reflection of some kind, as we would need to run the binary at build time, and evaluate the resulting expression
<`_>
possible?
<acowley>
`_: Sounds like import from derivation (IFD)
<`_>
I don't understand
acarrico has quit [Ping timeout: 252 seconds]
vk3wtf has quit [Ping timeout: 252 seconds]
thomasjm has joined #nixos
jperras has joined #nixos
<acowley>
It's something too search for more info on: a long-standing wrinkle about what nix supports and what hydra allows. I often get wrong what would be IFD, so I might be wrong here, too.
<simpson>
I think IFD is allowed in nixpkgs, but you won't get it built by Hydra.
<`_>
what is IFD?
<acowley>
"import from derivation"
<simpson>
`_: IFD is quite what it says on the tin: Write a Nix expression into the Nix store as a file, and then use Nix's `import` on that file.
<`_>
I see
<`_>
that sounds easy enough
<`_>
I just want to use it on a project right now
<`_>
how does callCabal2nix work?
<`_>
does it use this IFD magic?
<clever>
yes
<`_>
ok going to try it
<thomasjm>
hello--i'm trying to use a custom store path (/nix_frozen instead of /nix) and as a result trying to build things from scratch. but autoconf-2.64 refuses to build with the error "configure: error: no acceptable m4 could be found in $PATH.". a lot of stuff (like git) depends on autoconf so I'm kind of stuck. i went pretty deep into this in nix-shell and found that the m4 nix has built is failing one of the tests in the configure
<thomasjm>
script. i'm using the most recent nixpkgs-unstable, does anyone know what's wrong? I also left some details on nixpkgs issue #40782
<clever>
thomasjm: Unode has been using /share/nix/store for his stuff, and may have fixed that already
sigmundv has quit [Ping timeout: 252 seconds]
vk3wtf has joined #nixos
<thomasjm>
the way I understand bbarker's original report is that he was trying to use a local checkout of the nixpkgs repo, so this is not related to using a different nix store (?)
BombStrike has left #nixos [#nixos]
<acowley>
Well that's weird. If I use nvcc without nix things on PATH to produce a working executable, ldd on that executable shows libstdc++.so as not found if I'm in a nix-shell, but the executable still works.
<acowley>
Without nix things on PATH, ldd shows the working executable linking to /usr/lib/aarch64-linux-gnu/libstdc++.so.6
<ekleog>
`_ acowley simpson: when it'll be available, recursive-nix could also be used for this, as far as I understand (but it isn't ready yet AFAIR)
<acowley>
But LD_PRELOAD'ing libstdc++.so.6 to either the nix library or the system one doesn't change anything.
<acowley>
The working executable is larger by 16KiB, too. Also weird.
<thomasjm>
clever: are there possibly any extra steps used to set up the build boxes that populate the nix binary cache? i'm wondering how anyone has managed to build this...
drakonis1 has quit [Quit: WeeChat 2.3]
<clever>
thomasjm: dont remember anything special that had to be setup
[0__0] has quit [Remote host closed the connection]
<simpson>
ekleog: Oh, of course, but that could be a very long time still!
<ekleog>
yeah :(
<ekleog>
but at least it should be allowed in nixpkgs when it gets done
<ekleog>
and we'll be able to get rid of all those pre-committed $lang2nix :D
justbeingglad has joined #nixos
justbeingglad has left #nixos [#nixos]
<acowley>
Well, I've got to pack it in for the time being. If anyone can think of how I can try to debug what happens to a system-installed nvcc when run in a nix-shell environment, please let me know :)
romanofskiWork has quit [Ping timeout: 245 seconds]
<johnw>
does anyone have any idea why setting stdenv = hiPrio (overrideCC llvmPackages_7.stdenv ccacheWrapper) for an expression has no effect? Neither is clang available, nor does cc refer to ccache
averell has joined #nixos
<johnw>
deep in the PATH I do find the ccache-wrapper, but I wonder why it's not earlier in the pATH?
<johnw>
when I have just stdenv = llvmPackages_7.stdenv, I see clang and clang++
maximiliantagher has quit [Remote host closed the connection]
maximiliantagher has joined #nixos
<PolarIntersect>
hmm
<PolarIntersect>
so I added xinit and it says "Permission denied: /dev/tty0"
<PolarIntersect>
I wonder if that's related or expected?
<PolarIntersect>
It doesn't seem to be an issue w/ the video driver, but the screens don't configure
<PolarIntersect>
like it doesn't see my screen somehow?...
<Lisanna>
there seems to be no end to my nixops-on-darwin troubles ):
<Lisanna>
seriously considering biting the bullet and setting up some kind of deploy from linux ):
<Lisanna>
what might be a good solution for this? I want to keep my nix files on my mac... some kind of docker container that has the local directories mapped?
<PolarIntersect>
You want to use a Linux docker container to cross-compile nix packages to darwin
maximiliantagher has quit [Ping timeout: 268 seconds]
<PolarIntersect>
?
hyper_ch2 has joined #nixos
<Lisanna>
PolarIntersect no, I am trying to use nixops to deploy from darwin to linux (nixos)
<Lisanna>
but I am met with endless evaluation errors
<PolarIntersect>
ah
<Lisanna>
I fundamentally do not trust the nixops code to be doing the right thing here.... so I think I'm going to give up and try to run nixops from linux
<PolarIntersect>
can read the code
<Lisanna>
?
<PolarIntersect>
Maybe I need to add myself to the tty group in order to use xinit? I'm already in video
<PolarIntersect>
seems dangerous though
<hyper_ch2>
good morning
<PolarIntersect>
clever: Do you know if there's a way to rebuild everything?
<PolarIntersect>
I feel like maybe something didn't build right, because this seems like a config issue
<PolarIntersect>
hi hyper_ch2 :)
<clever>
PolarIntersect: nix is pure and will rebuild things if config changes
<PolarIntersect>
clever: It's guaranteed files won't end up corrupt?
<PolarIntersect>
seems like /dev/tty0 isn't a real tty, it's a link to your current X TTY so that makes sense now -.-
<clever>
PolarIntersect: improper shutdowns can corrupt some files, but `nix-store --verify --check-contents` will detect that
<PolarIntersect>
clever: Oh! That's what I was looking for! Thank you :)
<hyper_ch2>
clever: why can improper shutdown corrupt files?
<hyper_ch2>
isn't /nix/store read only?
<clever>
hyper_ch2: ext4 has a nasty habbit of not saving data even after a sync() happens
<PolarIntersect>
:|
* PolarIntersect
is terrified
<clever>
just do clean shutdowns and youll be fine
<hyper_ch2>
I just use ext4 on /boot nowadays (and my android phone probably)
<PolarIntersect>
wow one was modified
<PolarIntersect>
how do you fixed one that was modified?
<hyper_ch2>
there's probably an option for it to fix it
<clever>
nix-store --repair-path /nix/store/foo
<PolarIntersect>
ooh thanks :)
<PolarIntersect>
nix is so cool ^.^
<clever>
yep
<PolarIntersect>
I wonder if that corrupt file was the issue... It was in the "system-units" package
<clever>
thats where all systemd units live
<PolarIntersect>
oic
<PolarIntersect>
so that was probably from me using "systemctl enable"
<PolarIntersect>
oops
<clever>
ah
<clever>
nixos normally make the store read-only to prevent such things
<PolarIntersect>
same issue though
<PolarIntersect>
weird
<PolarIntersect>
ya that's what I'd expect
<PolarIntersect>
but I guess it's a config dir so it's weird
<hyper_ch2>
well, this happens if you don't use zfs :)
endformationage has quit [Quit: WeeChat 1.9.1]
<PolarIntersect>
naw, this was pbcak
IRCsum has joined #nixos
<PolarIntersect>
hmm
<PolarIntersect>
how could X not detect my screens?
<PolarIntersect>
doesn't make sense
<clever>
PolarIntersect: have you done anything odd with the X config?
<hyper_ch2>
did you plugin the cables?
<PolarIntersect>
clever: nope
slack1256 has quit [Remote host closed the connection]
<PolarIntersect>
maybe my laptop's card isn't supported?
<PolarIntersect>
well I guess this isn't a video driver issue
<clever>
PolarIntersect: can you pastebin your configuration.nix ?
<clever>
PolarIntersect: any errors in `journalctl -u display-manager` ?
Ariakenom has joined #nixos
<PolarIntersect>
clever: "Screens found, but none have a usable configuration"
<PolarIntersect>
followed by "Fatal server error"
<PolarIntersect>
"no screens found"
<clever>
PolarIntersect: can you pastebin both the logs and the configuration.nix?
<PolarIntersect>
any clever tricks on how to do that from the CLI? sprunge.us has been down for months :(
<Lisanna>
joko me too... but I've always used NIX_CURL_FLAGS with nixos-install
<joko>
Hmmm
<jgt>
I know using Stack/Stackage/stack2nix solves this, but the advice I had in this channel yesterday was essentially "stack sucks; don't use it". That's fine, but conflicting packages means I can't do my job.
<Lisanna>
and if your proxy settings aren't *PERFECT* after the install, the daemon has no hope
<joko>
LOL
Guanin has joined #nixos
<Lisanna>
'NIX_REMOTE= nixos-rebuild switch' works
<Lisanna>
jesus.....
<Lisanna>
one of these days I'm going to get so fed up with all of these awful corner cases and am going to wage a full-scale war on the terrible usability of nix/nixos
Itkovian has quit [Read error: Connection reset by peer]
<joko>
Lisanna: btw, did you have any progress with avoiding downloading inputs for remote building?
<Lisanna>
joko yes
<__monty__>
jgt: How did you get duplicate packages?
erictapen has quit [Ping timeout: 244 seconds]
<joko>
Lisanna: oh, please, do share, a quick search in IRC history showed me nothing
<jgt>
__monty__: I *was* using stack2nix, and I never had this problem. stack2nix is currently broken though, so I'm trying to make do with cabal2nix and cabal-install.
<jgt>
cocreature: it's split into a few different files; I'll make a repo or something. One moment.
<__monty__>
jgt: Yeah but you don't ordinarily get multiple package versions through nix afaik. The packages in haskellPackages are mostly just a stackage snapshot afaiu.
<Lisanna>
joko and, to be honest, you should probably use nix-build instead of nix-store --realise
civodul has joined #nixos
<joko>
Lisanna: thanks
<Lisanna>
joko there is no real elegant way to do this though, unless you write a custom utility that links directly against libstore and stuff... separating out the command-line arguments meant for nix-build and those meant for nix-instantiate is kind of painful
<joko>
Maybe this should be provided by nix directly, it's quite helpful
<Lisanna>
like, nix-build-remote \<nixpkgs\> -A hello --arg localSystem '{ system = "x86_64-linux"; }'
<Lisanna>
wait, bad example
random_yanek has quit [Ping timeout: 246 seconds]
<Lisanna>
nix-build-remote \<nixpkgs\> -A hello -k
<Lisanna>
there
<Lisanna>
the first three arguments need to go to nix-instantiate, and the last argument needs to go to nix-build
<joko>
Would that work?
<joko>
(it also needs the remote)
<Lisanna>
oh, right, yeah,
<Lisanna>
include that too
<Lisanna>
but after the remote, the argparsing is a nightmare
<Lisanna>
but, you get the idea... yuou can type out all those commands manually if it's worth it
<jgt>
cocreature: I'm opening a shell with `nix-shell --attr env riskbook.nix`
<jgt>
and then I run `ghc-pkg list`
<__monty__>
joko: I think the reason you can't always do this is the remote may have a different nixpkgs. So a build you start locally for Attr might not match a build you start remotely for Attr.
<__monty__>
joko: And it's not always possible for the remote to get the right nixpkgs, if you're working on a local checkout for instance.
goibhniu has joined #nixos
romanofskiWork has joined #nixos
themattchan has joined #nixos
<themattchan>
hello
<themattchan>
i just installed nixos and am having trouble with things
<jgt>
cocreature: I have duplicates of resourcet and xml-conduit
<__monty__>
jgt: And which packages are duplicated? Maybe you still have stack-installed versions around and they don't match?
<jgt>
__monty__: see above :P
<themattchan>
where are the icons installed, and how can i change my cursor size? nothing is happening when i change the xresources file
<jgt>
__monty__: I haven't used stack directly. Not sure how that could have happened, but I could of course try cleaning everything and trying again. How would I do that?
<jgt>
__monty__: would I have to just nuke my nix store?
<__monty__>
jgt: Let's not do anything rash : ) I figured you were using stack before.
o1lo01ol1o has joined #nixos
<joko>
__monty__: no worries, one might use expressions which are not relying to NIX_PATH
<jgt>
__monty__: I have deadlines, so naturally I'm losing sleep and panicking :grimace:
<Lisanna>
__monty__ this issue doesn't concern evaluation
<Lisanna>
it's just an annoyance of remote building that the local host have *all* of the build-time dependencies that the remote builder is going to use
<Lisanna>
even though it doesn't need them
<Lisanna>
IMHO it should try to get them *only* if the remote builder is unable to get them itself
Itkovian has joined #nixos
<Lisanna>
(e.g., due to differences in connectivity)
manv has joined #nixos
<jgt>
cocreature: is that repo helpful?
<joko>
Well, that's true, e.g. my Hydra has access to private Git repos that the workers do not
<Lisanna>
as far as I can tell the local host needs them so that it can upload them to the remote... if you use builders-use-substitutes then it doesn't need to upload all of them necessarily, but it will still try to acquire all of them locally first
<Lisanna>
even though it won't use them
o1lo01ol1o has quit [Ping timeout: 252 seconds]
thc202 has joined #nixos
<__monty__>
Lisanna: Oh, I didn't look at your snippet closely enough. Didn't realise you were generating the derivations locally and transferring them.
<Lisanna>
__monty__ yeah
<jgt>
__monty__: any idea how I should move forward?
<__monty__>
jgt: Not right away, still looking at it. Annoying that bitbucket doesn't seem to have code search.
maximiliantagher has joined #nixos
<__monty__>
jgt: I think the problem is the overrides pulling in other versions of resourcet and xml-conduit.
<jgt>
__monty__: ah crap; I'm sorry. Would it help you if I put it on GH too?
<jgt>
__monty__: where do you see that?
sigmundv has joined #nixos
<jgt>
__monty__: the overrides in riskbook.nix?
<manv>
hi, um how do I request for a package to be added to nixos? should I just create an issue on github, it's a straightforward application https://github.com/KDE/wacomtablet, i would do it myself but i am new to nixos
<jgt>
__monty__: i.e. esqueleto and mime-mail-ses?
<__monty__>
jgt: Not really anywhere. It's just a guess.
<domenkozar>
jgt: pong
<jgt>
domenkozar: hello!
<jgt>
domenkozar: I wanted to talk to you yesterday about stack2nix; I haven't been able to build it all week
<Guanin>
Hi, I asked a few days ago how to handle large .tar.gz files. I got stuck as / seems to be too small to handle both the .tar.gz and the extracted files. Is there a way around this?
<nschoe_>
Hi everyone, I'm running into a small, weird (to me) issue. I'm using this nixpkgs (https://github.com/airalab/airapkgs) for a project (nix-shell), and try to package package pcl_ros, but it fails with https://pastecode.xyz/view/942bd2aa. From lines 10 and 13 we can see there's an issue with a PATH: a ":" is missing. From what I understand it comes from package pcl_conversions (this is a dependency) but when I try to build pcl_conversion alone it works
<nschoe_>
And I'm a bit lost as to where to look to solve this.
<__monty__>
jgt: Oof, that doesn't look like fun. You may have to manually override versions until it works : / Maybe try a different nixpkgs revision? Maybe one that corresponds to a stackage LTS?
<jasongrossman>
Lots of upgrades have just turned up in nixos-unstable. Thank you all the people who worked on these!
<jgt>
__monty__: from that `cabal configure` output, it *seems* as though the conflicting packages come from amazonka
<jgt>
__monty__: how do I pick a revision that corresponds to a stackage LTS?
<__monty__>
jgt: Yes, the conflict is between yesod-static and amazonka.
<{^_^}>
[nixpkgs] @domenkozar pushed commit from @roberth to master « acsccid: init at 1.1.6 »: https://git.io/fpT8d
noam has quit [Read error: Connection reset by peer]
<{^_^}>
[nixpkgs] @Profpatsch opened pull request #49901 → CODEOWNERS: add mboes and Profpatsch for bazel packaging → https://git.io/fpT4L
smolboye_ has quit [Quit: WeeChat 2.2]
sigmundv has joined #nixos
smolboye has joined #nixos
hamishmack has joined #nixos
<Lisanna>
lol... so now I have a nixos virtualbox system running in the background with a shared nixops directory so I can edit nixops code on my mac natively and deploy with nixops running in the virtualbox
<Lisanna>
geez...
<{^_^}>
[nixpkgs] @domenkozar pushed commit from @roberth to release-18.09 « acsccid: init at 1.1.6 »: https://git.io/fpTBY
<{^_^}>
[nixpkgs] @timokau merged pull request #49888 → numpy: fix high cpu count bug → https://git.io/fpUoE
<{^_^}>
[nixpkgs] @timokau pushed 2 commits to staging-18.09: https://git.io/fpTBW
PolarIntersect has quit [Ping timeout: 252 seconds]
<jgt>
domenkozar: do you know why stack2nix might be hanging?
<rawtaz>
Lisanna: you probably should add context to that information
<jgt>
oh… handleStackConfig (remoteUri): Nothing
<Lisanna>
rawtaz the context is that I don't trust the "pkgs" nixos module input one single bit
<rawtaz>
thats not much of context
<rawtaz>
but whatever floats your boat
<Lisanna>
omfg
<Lisanna>
I knew it
<Lisanna>
nixpkgs.pkgs != { pkgs }
<Lisanna>
they are different sometimes
<rawtaz>
:<
<Lisanna>
The packages that I am passing in to nixpkgs.pkgs option is not always the same as the packages that are provided to modules in { config, pkgs, ... }:
<srhb>
Hmm, it should be the source of the pkgs argument.
<srhb>
But it is a bit convoluted. iirc nixops looks up <nixpkgs>, then /nixos is sourced from that, then nixpkgs.pkgs is /nixos/../
<moredhel>
Hi, I'm having setting up a Haskell development environment. I'm using cabal2nix and have a basic project that works if I go the cabal route. (ie. cabal install, cabal run).
<moredhel>
But I am wanting to start up a ghci repl and develop from there. So used `cabal2nix --shell . > shell.nix` and then jump into it. although ghc cannot find any of my dependencies.
<srhb>
moredhel: inside nix-shell shell.nix?
<Lisanna>
oh... nevermind
<Lisanna>
*I'M* insane
<moredhel>
srhb: that's currently what I'm using.
<Lisanna>
as usual
<srhb>
moredhel: That should work
<moredhel>
and my `cat $(which ghc)` seems correct.
<moredhel>
then I do `ghc src/Main.sh` and get compile errors
<moredhel>
s/sh/hs
<srhb>
moredhel: does runghc ./Setup.hs configure and runghc ./Setup.hs build work?
<srhb>
(That is, using Cabal the library)
nschoe_ has quit [Quit: Program. Terminated.]
mayhewluke has quit [Ping timeout: 252 seconds]
nschoe has joined #nixos
<moredhel>
yep, that works
<srhb>
I'm not sure whether you should be expecting ghc somesourcefile.hs to work without the build system.
<srhb>
Maybe someone else can weigh in.
<moredhel>
is that the recommended way of running the repl/building the project?
mayhewluke has joined #nixos
ottidmes has joined #nixos
<moredhel>
I tried with `cabal build` but that failed too because it was trying to use my home-dir deps. And even `cabal repl` failed wrt some missing .so
<srhb>
moredhel: PErsonally I just use cabal new-repl and co. in the nix-shell for my development cycle, then ditch cabal-install when I actually build it
<srhb>
(As in, the final build doesn't use cabal-install at all, but just nix-builds the project, which is similar to runghc ./Setup.hs build inside the nix shell)
<moredhel>
yeah, that makes sense. The build-process was pretty simple using Nix. I was just trying to use the same dependencies for my dev cycle too.
<srhb>
moredhel: Yes, it certainly should be!
ThatDocsLady has quit [Remote host closed the connection]
<srhb>
moredhel: I think you should just avoid cabal install at all costs for that.
ThatDocsLady has joined #nixos
<srhb>
moredhel: There's never a good reason to use it.
<srhb>
moredhel: By the logic that: If cabal-install has nothing installed itself, you only have your project specific dependencies, always.
<moredhel>
does that mean you never run a cabal update? so it can never download its own deps
<srhb>
moredhel: Yep.
<srhb>
moredhel: For throwaway shells we have better options anyway
<srhb>
eg nice aliases for nix-shell -p 'haskellPackages.ghcWithPackages (ps: with ps; [ lens aeson ])' --run ghci
ubert has joined #nixos
<infinisil>
Oh you know what would be cool
Lisanna has quit [Quit: Lisanna]
<srhb>
infinisil: What? :)
<moredhel>
which you stick in your bashrc?
<srhb>
moredhel: I have a shell function for it called gwpi that optionally takes the compiler version too, but yeah, something like that.
<infinisil>
Have a shell.nix like `{ args }: buildEnv { paths = [ (ghcWithPackages (p: map (n: p.${n}) args)) ]; }`
<infinisil>
Then be able to use nix-shell file.nix -- aeson lens
<MasseR>
infinisil: I have a function for that
<MasseR>
gwp () {
<srhb>
infinisil: Yeah. Probably using shellFor, actually.
<MasseR>
pkgs="$@"
<MasseR>
nix-shell -p "haskellPackages.ghcWithHoogle (pkgs: with pkgs; [$pkgs])" --run 'zsh'
<MasseR>
}
<infinisil>
Yeah but you could do many cool things if you get the arguments directly
<infinisil>
Probably
rprije has quit [Ping timeout: 240 seconds]
TweyIII has joined #nixos
<nschoe>
Anybody has issues with eigen3 when building? I seem to have issues with the ENGEIN3_INCLUDE_DIR.
TweyII has quit [Ping timeout: 252 seconds]
<TweyIII>
nschoe: Is it because you mistyped it? ;)
<nschoe>
TweyIII, ah ah, no ^^ Sorry for the typo in the chat, but that's not it.
<nschoe>
TweyIII, it seems there's a missing path separator: `CMake Error at /nix/store/cvmvm5v9dw3v15mjzh7hinpb0wd1zw3m-pcl_conversions-1.6.2/share/pcl_conversions/cmake/pcl_conversionsConfig.cmake:113 (message):
<{^_^}>
[nixpkgs] @domenkozar pushed 2 commits to release-18.09: https://git.io/fpT2n
<ottidmes>
infinisil: any idea why: pkgs.buildEnv { name = "ghc-env"; paths = [ ... ]; }, does not work, but: pkgs.stdenv.mkDerivation { name = "ghc-env"; paths = [ ... ]; }, works just fine in nix-shell, i.e. ghci is only present when using mkDerivation
<jgt>
does this command look valid? It appears to be hanging for me: nix-build -A haskell.compiler.ghc822 '<nixpkgs>' --no-build-output --no-out-link
<symphorien>
how can I evaluate a derivation as if I were on darwin ?
<infinisil>
ottidmes: ahh right, buildEnv builds an environment, but nix-shell just enters a shell for how to build the buildEnv
<jgt>
oh I see the `--no-build-out` flag makes it silent
<infinisil>
Something like that
<ottidmes>
infinisil: ah make sense, i.e. the result of building the env, is an env where ghci would be present, while the result of building the derivation is well, in this case nothing, but the env to build it, contains ghci
<lukego>
Mic92: Thanks, you are right. Seems to be resolved with this patch to include root in the userns when --allow-new-privileges is enabled https://github.com/NixOS/nix/pull/2525
<{^_^}>
nix#2525 (by lukego, 3 minutes ago, open): Allow root uid/gid (0) in sandbox userns when allowNewPrivileges
<locallycompact>
if I ge trid of the overlay then it complains no pkg foo
<locallycompact>
no attribute foo
<locallycompact>
if I add overlay it complains infinite recursion error
Lisanna has joined #nixos
<ottidmes>
locallycompact: I think it has to do with your hello.nix, try to remove the with import <nixpkgs> there, by doing what you would normally do
<ottidmes>
locallycompact: does x86_build also contain a pkgs and lib attribute, otherwise you might want to use them instead, thats all I can think of without experimenting with the files myself
<jgt>
is ghc822 no longer in nixpkgs? I'm running this command `nix-build -A haskell.compiler.ghc822 '<nixpkgs>' --no-out-link` and I've been [I think] compiling GHC for a couple hours now…
<symphorien>
nixpkgs-unstable 4 days ago c9e13806267
<symphorien>
don't mix nixpkgs-unstable and nixos-unstable
<jgt>
oh sorry, I was looking at nixos-unstable
o1lo01ol1o has quit [Remote host closed the connection]
<jgt>
symphorien: am I up to date with nixpkgs-18.09-darwin?
<symphorien>
it's not the same commit, but it is difficult to say which commit is newer
periklis has quit [Remote host closed the connection]
periklis has joined #nixos
periklis has quit [*.net *.split]
locallycompact has quit [*.net *.split]
jasongrossman has quit [*.net *.split]
justanotheruser has quit [*.net *.split]
mayhewluke has quit [*.net *.split]
Izorkin has quit [*.net *.split]
IRCsum has quit [*.net *.split]
brejoc has quit [*.net *.split]
__Sander__ has quit [*.net *.split]
sg2002 has quit [*.net *.split]
xorkle has quit [*.net *.split]
hyper_ch has quit [*.net *.split]
av4h has quit [*.net *.split]
Cheery has quit [*.net *.split]
leothrix has quit [*.net *.split]
Xal has quit [*.net *.split]
dweller has quit [*.net *.split]
dmc has quit [*.net *.split]
MinceR has quit [*.net *.split]
drewr has quit [*.net *.split]
rodgort has quit [*.net *.split]
pyrtsa has quit [*.net *.split]
hiroshi has quit [*.net *.split]
amfl has quit [*.net *.split]
Guillaum has quit [*.net *.split]
infty has quit [*.net *.split]
LnL has quit [*.net *.split]
FireFly has quit [*.net *.split]
troydm has quit [*.net *.split]
ashkitten has quit [*.net *.split]
commander has quit [*.net *.split]
prusnak has quit [*.net *.split]
lejonet has quit [*.net *.split]
tg has quit [*.net *.split]
disasm has quit [*.net *.split]
delroth has quit [*.net *.split]
Unode has quit [*.net *.split]
phry_ has quit [*.net *.split]
ghasshee has quit [*.net *.split]
Mo0O has quit [*.net *.split]
liori_ has quit [*.net *.split]
tg has joined #nixos
__monty__ has quit [Quit: leaving]
commander has joined #nixos
sg2002 has joined #nixos
lejonet has joined #nixos
LnL has joined #nixos
IRCsum has joined #nixos
Izorkin has joined #nixos
hiroshi has joined #nixos
xorkle has joined #nixos
Unode has joined #nixos
hyper_ch has joined #nixos
leothrix has joined #nixos
Xal has joined #nixos
<jgt>
how do I use a different channel with nix-build? This doesn't work: nix-build -f channel:nixos-18.09 -A haskell.compiler.ghc844 '<nixpkgs>' --no-out-link -j0
crmlt has quit [Ping timeout: 252 seconds]
dmc has joined #nixos
FireFly has joined #nixos
<mdash>
jgt: nix-build '<nixos-18.09>' -A ....
erictapen has quit [Ping timeout: 245 seconds]
<mdash>
jgt: assuming you have a nixos-18.09 channel
<tilpner>
I don't think 844 is in 18.09?
<tilpner>
Oh wait, typo
<tilpner>
nix-build channel:nixos-18.09 -A haskell.compiler.ghc844
<tilpner>
^ works fine here
<jgt>
thanks
delroth has joined #nixos
crmlt has joined #nixos
mmercier has quit [Quit: mmercier]
<jgt>
can I set the channel in an environment variable? nixpkgs-unstable doesn't have ghc822, and I cancelled building it after three hours
<jgt>
basically I want to tell stack2nix to use a different channel when it runs nix-build
<tazjin>
in the override pattern (i.e. `override { oldAttrs: /* additionalAttrs */ }`), is there any way to *remove* an attribute from the original set?
<tazjin>
I attempted to set it to `null` but that doesn't seem to hel
<tazjin>
pp
<tazjin>
argh, *-p
Cheery has joined #nixos
<symphorien>
tazjin: override { new attrs } or overrideAttrs (oldattrs: { new attrs ? }) ?
<locallycompact>
TweyIII, I don't think that's what happened
<locallycompact>
when I tried to ./configure without cding into $src I just got a ./configure not found error
<locallycompact>
let me try again
<TweyIII>
locallycompact: Of course, it isn't
<TweyIII>
locallycompact: Derivation-building tricks: you can pass -K to nix-build to keep the build directory around after the build fails, so you can see what's going on in there
<locallycompact>
ok so there's nothing in there
<locallycompact>
so what's the right idea?
<TweyIII>
locallycompact: You can cp -r $src in there if you want
<TweyIII>
You just can't edit it in-place
<locallycompact>
but that folder is read only
crmlt has quit [Ping timeout: 240 seconds]
<locallycompact>
if I copy that in and cd in then it's the same, permission denied
<TweyIII>
locallycompact: You can chmod it once it's in your build directory, though
<locallycompact>
this feels really wierd
<locallycompact>
is this really the way to do this?
<symphorien>
you can compile nix from source with another path, or use a nix in proot to build a nix with another path, or use another computer with nix to build a nix in another path
crmlt has quit [Remote host closed the connection]
leothrix has quit [Remote host closed the connection]
<rpg>
symphorien: Thanks. Then it seems like those instructions are not very complete.
<symphorien>
I agree
johann__ has joined #nixos
endformationage has joined #nixos
jperras has quit [Quit: WeeChat 2.2]
<rpg>
symphorien: TBQH I don't even know what PRoot is when it's at home.
<symphorien>
The problem is that these three methods take a lot of time (compilation) so I could describe them in the wiki but I would not try the command I would write on the wiki
<rpg>
symphorien: It sounds like the PRoot method is likely the least painful for those of us who don't have sudo privs.
<symphorien>
it is the least painful but also unreliable
<jsgrant>
Anyone using NixOS as their primary development platform on the JVM? See almost exclusively people are using it for system-programming tier + haskell.
jperras has joined #nixos
<jsgrant>
I see a lot of Java-land and orbiting stuff is defined though, obviously; But can't seem to find really even any blog enteries or anything -- which I find a bit odd
<ottidmes>
jsgrant: I have done some Java and Scala projects on NixOS, but I just used the standard Java/Scala package managers, so I did not leverage Nix for this.
<manveru>
zimbatm: i can give you a hand with builtwithnix if you still need :)
<jsgrant>
ottidmes: Article seems like at the very least a spring-board, so tyvm -- sadly more than I was able to find already; Idk, maybe my searchfoo just sucks idk
<jsgrant>
I really haven't looked up the execution model enough on NixOS besides binaries are often not runnable sans non-standard paths; Really need to ... have pretty much am going all in (after a bunch of tepid dips the past 2 or-so years).
<zimbatm>
yis manveru !
drakonis has joined #nixos
<jsgrant>
Really plan to stay here too, unless Fedora Atomic/Silverblue (and recent CoreOS acq) makes MAJOR leaps and bounds and don't ever think it'll even get close to the declarative bit.
<zimbatm>
manveru: there is a PSD version that I should get soon, it might be easier to work on that
<zimbatm>
are you good at cutting PSDs into HTML?
<manveru>
that'd help :)
maximiliantagher has joined #nixos
* jsgrant
still has like a half-dozen issues he needs to resolve on graphical end though; Sans that, have been really happy with NixOS thusfar
<manveru>
but i can make the basic structure already... just would like the graphics
<ottidmes>
jsgrant: I can really recommend going all in, but only if you have enough time to learn how to build your own packages and such, because like you said, you might encounter some things no longer working due to non-standard paths. At least make sure that what you normally use is already there. If you are a JVM dev, then its quite easy to just take the basic tools from Nixpkgs, and for the rest work as normal (at
<ottidmes>
least as a fallback).
Ariakenom has joined #nixos
<manveru>
zimbatm: you wanna continue using jekyll?
<{^_^}>
[nixpkgs] @kimat opened pull request #49927 → openconnect-globalprotect: init at 7.08 → https://git.io/fpkYS
<zimbatm>
manveru: as long as it builds with nix I don't care ^_^
rpg has joined #nixos
<ottidmes>
jsgrant: what graphical issues? video driver?
<manveru>
alright :)
<zimbatm>
manveru: what framework do you have in mind?
<jsgrant>
Well, to be fair I've been a tiny tiny bit smarter recently and have stopped putting myself in the 'all ALL in' category as-much; Do now have a spare Fedora box (actually on now) and a spare Win10 bo even if I "really need it" -- feel a lot lot more confident now though to go 'all in' on all my day-to-day machines though when the fallback is not too rough
* jsgrant
has a long/hard history that has taken probably literally 100 times to learn from, that you never really go 'all in' just like 85%. :^P
rpg has quit [Client Quit]
orivej has quit [Ping timeout: 250 seconds]
<ottidmes>
jsgrant: Good choice, I did the same, still had Arch Linux on my laptop and a dual boot of Windows 7 (at the time), and actually needed to use Arch Linux as a fallback a few times when I was time constrained and just needed it to get it to work
rpg has joined #nixos
<jsgrant>
ottidmes: Weird stuff, like I have a default KDE Plasma session going and everything seems to work decently well -- but stuff like OBS (the screencasting software) has major screen tearing with present windows. Never had that issue on Arch or Fedora. Latte-dock is not persisstent through sessions, so I move the default to the left, remove the clock widget, and start pinning apps and when I restart the
<jsgrant>
session everything is back to restart. Also plasma complaining it doesn't have write access to my autostart folder (think in ~/.config/autostart) but seems to work anyways.
<zimbatm>
manveru: sounds good!
<zimbatm>
more nix
<manveru>
indeed :)
<zimbatm>
feel free to change the repo to your liking
<zimbatm>
I will ping you when I get the PSD, hopefully tomorrow
<jsgrant>
On i5 3000-gen & everything else renders fine, so really don't know on obs.
<manveru>
cool
<ottidmes>
jsgrant: I would scan your journal for problems, sudo journalctl -e -n 30000, and check for all errors and try and fix as many as you see, and if you can ignore one, document why in your configuration.nix as comments, that helped me fix and understand most of my problems.
<ottidmes>
jsgrant: And what driver are you using, since you mention only your CPU, do you have an intergrated GPU?
<jsgrant>
Yeah, no discrete graphics (will likely change in a few months though), but right now that box is using I assume whatever the standard Intel HD Graphics driver is; Will need to look into it more after this initial big (like 400gb) syncthing sync is done ... box is mostly unusable atm because of that
<jsgrant>
Also, thanks; Took a note about journalctl
<jsgrant>
I'm assuming they're mostly trivial issues and even stuff like latte-dock isn't a 'huge deal' just annoying at best, but yeah.
<jsgrant>
Everyonce and a while 'desktop settings' on right-click (on desktop) will crash the shell as well.
<jsgrant>
Just little things I haven't noticed in other distros, which I'm sure is some user-error but a lot of just needed refinement and maintance generally.
<goibhniu>
jsgrant: BTW you can run into weird problems if you install any Qt applications in a user profile
<goibhniu>
it's best to keep them all in your system.environmentPackages
<jsgrant>
goibhniu: Everything so-far has been system-side, but that's odd; Any idea why?
<clever>
goibhniu: environment.systemPackages
<goibhniu>
doh! thanks :D
<clever>
jsgrant: the qt libs conflict in ~/.nix-profile/lib/
<jsgrant>
clever: Ah, okay.
<clever>
goibhniu: how did you do those animated intro things for nixcon?
<goibhniu>
samueldr: created them ... using some kind of magic IIUC
<clever>
lol, was it just an avi being played and overlaid in obs?
<goibhniu>
I know he renders SVGs in some part of the pipeline
shabius has joined #nixos
<goibhniu>
yep, a video overlay
<clever>
ah
<goibhniu>
very tasty IMO
<clever>
but that also gives me another idea, it could also be done as a normal window capture, with some transparency
<clever>
then you could edit it in realtime
reinzelmann has joined #nixos
<clever>
you just need to alpha-key that window
<goibhniu>
I believe it takes quite a while to render each one
fooinabar has joined #nixos
<clever>
thats likely due to the video compression
erictapen has joined #nixos
<clever>
but if it was just a simple animation in an open window
<fooinabar>
clever: thanks! I think it works, since llvm is recompiling. FWIW, I do not want to install the package. I'm using nixpkgs to compile a project of mine. The link was the closest thing I found when searching.
Peetz0r has quit [Ping timeout: 252 seconds]
<clever>
fooinabar: ah, then you want something like (llvm.override { enableSharedLibraries = false; }) in your buildInputs
<fooinabar>
yap, I did: "myllvm = llvm.override { debugVersion = true; };" and added myllvm to buildInputs
<clever>
yep, that works
Peetz0r has joined #nixos
jsgrant has quit [Read error: No route to host]
fooinabar has quit [Quit: leaving]
<kimat>
kimat
kimat has quit [Quit: Leaving]
erictapen has quit [Ping timeout: 252 seconds]
<erasmas>
what's the best way to override gcc for one package? (it needs to be built with gcc 5)
apeyroux has joined #nixos
rpg has joined #nixos
rpg has quit [Client Quit]
<mdash>
erasmas: use "overrideCC stdenv gcc5" instead of "stdenv" for it.
<wpcarro>
I'm trying to declaratively cache my SSH passphrases using the `programs.ssh.startAgent = true` option, but SSH is still asking me for my passwords. Is there another option that I'm missing here?
halfbit has joined #nixos
erictapen has joined #nixos
rpg has joined #nixos
<tokudan[m]>
wpcarro: ssh agents don't store passwords, they keep public keys unlocked
<tokudan[m]>
wpcarro: try ssh-add -L to see which keys are in your keyring of the agent
jabranham has quit [Quit: ERC (IRC client for Emacs 27.0.50)]
jasongrossman has joined #nixos
rprije has joined #nixos
Ariakenom has quit [Quit: Leaving]
Lisanna has joined #nixos
<{^_^}>
[nixpkgs] @c0bw3b merged pull request #49954 → [18.09] nixos/gitea: fix mysql issue, add mysql socket auth, and add a nixos test → https://git.io/fpkMZ
<{^_^}>
[nixpkgs] @c0bw3b pushed 4 commits to release-18.09: https://git.io/fpkyp