!fXpAvneDgyJuYMZSwO:nixos.org

Nix Data Science

289 Members
62 Servers

Load older messages


SenderMessageTime
1 Jul 2024
@benoitdr:matrix.orgbenoitdr
In reply to @ss:someonex.net

python app doesn't see the GPU

Usually that means you need to add a DT_RUNPATH or LD_LIBRARY_PATH somewhere, pointing at ${addDriverRunpath.driverLink}/lib

*

Could you please help me on the syntax ? The closest I can get is below and has no effect (in my shellHook) :

export LD\_LIBRARY\_PATH=${pkgs.lib.makeLibraryPath \[
pkgs.addDriverRunpath.driverLink
\]}
export DT\_RUNPATH=${pkgs.lib.makeLibraryPath \[
pkgs.addDriverRunpath.driverLink
\]}
13:30:45
@benoitdr:matrix.orgbenoitdr *

Could you please help me on the syntax ? The closest I can get is below and has no effect (in my shellHook) :

export LD_LIBRARY_PATH=${pkgs.lib.makeLibraryPath [
pkgs.addDriverRunpath.driverLink
]}
export DT_RUNPATH=${pkgs.lib.makeLibraryPath [
pkgs.addDriverRunpath.driverLink
]}
13:31:36
@ss:someonex.netSomeone (UTC+3)DT_RUNPATH is not an environment variable, but the first export seems like it should work13:36:18
@benoitdr:matrix.orgbenoitdrwell it doesn't ..13:37:16
@ss:someonex.netSomeone (UTC+3)Well now it's a matter of finding out how exactly it doesn't work:)13:41:17
@ss:someonex.netSomeone (UTC+3) Run the program with LD_DEBUG=libs, search for mentions of libcuda 13:41:35
@ss:someonex.netSomeone (UTC+3) ENsure that hardware.opengl.enable is set//run/opengl-driver/lib/libcuda.so exists 13:41:49
2 Jul 2024
@brodriguesco:matrix.orgBruno Rodrigues

I'm trying to package positron here is my attempt: https://github.com/b-rodrigues/nixpkgs/blob/positron/pkgs/by-name/po/positron/package.nix
but I get this error message when trying to build

b-rodrigues@localhost:~/Documents/github_repos/nixpkgs(master)$ nix-shell -I nixpkgs=. -p R rPackages.dplyr positron
this derivation will be built:
  /nix/store/qh59xjk5cz4dq44y0jixma219sr9xb0f-Positron-2024.06.1-27.drv
building '/nix/store/qh59xjk5cz4dq44y0jixma219sr9xb0f-Positron-2024.06.1-27.drv'...
Running phase: unpackPhase
tar: ./usr/share/positron/chrome-sandbox: Cannot change mode to rwsr-xr-x: Operation not permitted
tar: Exiting with failure status due to previous errors
dpkg-deb: error: tar subprocess returned error exit status 2
error: builder for '/nix/store/qh59xjk5cz4dq44y0jixma219sr9xb0f-Positron-2024.06.1-27.drv' failed with exit code 2;
       last 4 log lines:
       > Running phase: unpackPhase
       > tar: ./usr/share/positron/chrome-sandbox: Cannot change mode to rwsr-xr-x: Operation not permitted
       > tar: Exiting with failure status due to previous errors
       > dpkg-deb: error: tar subprocess returned error exit status 2
       For full logs, run 'nix log /nix/store/qh59xjk5cz4dq44y0jixma219sr9xb0f-Positron-2024.06.1-27.drv'.

08:58:35
@brodriguesco:matrix.orgBruno Rodriguesit's probably something simple that I'm missing08:58:52
@brodriguesco:matrix.orgBruno Rodrigues I removed the unpackPhase now and simply use dpkg -x in the installPhase, but I get the same issue 09:27:10
@jb:vk3.wtfjbedoMight be the suid bit10:03:10
@jb:vk3.wtfjbedo The store doesn’t allow suid so packaging won’t be straightforward 10:11:28
@jb:vk3.wtfjbedo I might be a bit old fashioned but suid for a text editor seems ludicrous 10:12:14
@brodriguesco:matrix.orgBruno Rodriguesit’s not possible to change the permissions during the unpackPhase for example?10:25:07
@brodriguesco:matrix.orgBruno Rodriguesmaybe I can look into this: https://github.com/NixOS/nixpkgs/blob/88d829e52cfbeee71d81704ce28f5b439f6dea16/nixos/modules/security/chromium-suid-sandbox.nix#L1410:28:13
@brodriguesco:matrix.orgBruno RodriguesI've reached the limits of my knowledge. Likely it will have to be built from source then, but I'm too happy on spacemacs to spend the time trying to do that :D11:45:05
@kupac:matrix.orgkupacIt's in experimental phase, so we don't have to rush with the packaging imo. We can file a bug report upstream about the suid and wait for it to run its course.13:40:38
@kupac:matrix.orgkupac* It's in experimental phase, so we don't have to rush with the packaging imo. You can file a bug report upstream about the suid and wait for it to run its course.13:40:54
@janik0:matrix.org@janik0:matrix.org left the room.13:54:17
@brodriguesco:matrix.orgBruno Rodriguesyou're right, I wanted to try to package it more as a learning exercise. Well it was worth it, because I've learned about suid!14:25:03
@brodriguesco:matrix.orgBruno RodriguesRedacted or Malformed Event14:28:26
@brodriguesco:matrix.orgBruno Rodrigues
In reply to @kupac:matrix.org
It's in experimental phase, so we don't have to rush with the packaging imo. You can file a bug report upstream about the suid and wait for it to run its course.
but would you say it's an issue? they're likely doing that for sandboxing
14:41:32
@alexoo:matrix.orgAlexo joined the room.16:16:22
3 Jul 2024
@anjannath:matrix.org@anjannath:matrix.org joined the room.18:01:08
4 Jul 2024
@monadam:matrix.orgmonadam joined the room.00:25:49
6 Jul 2024
@anjannath:matrix.org@anjannath:matrix.org left the room.04:50:52
@jeroenvb3:matrix.orgjeroenvb3 joined the room.23:43:32
@jeroenvb3:matrix.orgjeroenvb3Hi there, could someone point me to up to date documentation for getting python to work with cuda? I have tried some things but I don't know what the best/official way is that I should start with. Thanks.23:54:01
7 Jul 2024
@ss:someonex.netSomeone (UTC+3) import <nixpkgs> { config.cudaSupport = true } 00:59:29
@1h0:matrix.org@1h0:matrix.org left the room.08:53:13

Show newer messages


Back to Room ListRoom Version: 6