Package Details: nvidia-container-toolkit 1.0.5-3

Git Clone URL: (read-only, click to copy)
Package Base: nvidia-container-toolkit
Description: NVIDIA container runtime toolkit
Upstream URL:
Keywords: docker nvidia nvidia-docker runc
Licenses: BSD
Conflicts: nvidia-container-runtime<2.0.0, nvidia-container-runtime-hook
Replaces: nvidia-container-runtime-hook
Submitter: jshap
Maintainer: jshap (kiendang)
Last Packager: jshap
Votes: 9
Popularity: 1.08
First Submitted: 2019-07-28 01:19
Last Updated: 2019-12-13 17:06

Pinned Comments

jshap commented on 2019-07-28 01:43

see the release notes here for why this exists:

tl;dr: nvidia-docker is deprecated because docker now has native gpu support, which this package is required to use. :)

Latest Comments

1 2 Next › Last »

hantian_pang commented on 2019-12-13 04:29

error log like this:

/usr/bin/ld: /home/pang/.cache/pikaur/build/libnvidia-container/src/libnvidia-container-1.0.5/deps/usr/lib/libelf.a(elf_scn.o): in function `_libelf_load_section_headers':

/home/pang/.cache/pikaur/build/libnvidia-container/src/libnvidia-container-1.0.5/deps/src/elftoolchain-0.7.1/libelf/elf_scn.c:72: undefined reference to `_libelf_fsize'

/usr/bin/ld: /home/pang/.cache/pikaur/build/libnvidia-container/src/libnvidia-container-1.0.5/deps/usr/lib/libelf.a(gelf_fsize.o): in function elf32_fsize': /home/pang/.cache/pikaur/build/libnvidia-container/src/libnvidia-container-1.0.5/deps/src/elftoolchain-0.7.1/libelf/gelf_fsize.c:37: undefined reference to_libelf_fsize'

/usr/bin/ld: /home/pang/.cache/pikaur/build/libnvidia-container/src/libnvidia-container-1.0.5/deps/usr/lib/libelf.a(gelf_fsize.o): in function elf64_fsize': /home/pang/.cache/pikaur/build/libnvidia-container/src/libnvidia-container-1.0.5/deps/src/elftoolchain-0.7.1/libelf/gelf_fsize.c:43: undefined reference to_libelf_fsize'

/usr/bin/ld: /home/pang/.cache/pikaur/build/libnvidia-container/src/libnvidia-container-1.0.5/deps/usr/lib/libelf.a(gelf_fsize.o): in function gelf_fsize': /home/pang/.cache/pikaur/build/libnvidia-container/src/libnvidia-container-1.0.5/deps/src/elftoolchain-0.7.1/libelf/gelf_fsize.c:56: undefined reference to_libelf_fsize'

/usr/bin/ld: /home/pang/.cache/pikaur/build/libnvidia-container/src/libnvidia-container-1.0.5/deps/usr/lib/libelf.a(libelf_ehdr.o): in function `_libelf_load_extended':

/home/pang/.cache/pikaur/build/libnvidia-container/src/libnvidia-container-1.0.5/deps/src/elftoolchain-0.7.1/libelf/libelf_ehdr.c:52: undefined reference to `_libelf_fsize'

/usr/bin/ld: /home/pang/.cache/pikaur/build/libnvidia-container/src/libnvidia-container-1.0.5/deps/usr/lib/libelf.a(libelf_ehdr.o):/home/pang/.cache/pikaur/build/libnvidia-container/src/libnvidia-container-1.0.5/deps/src/elftoolchain-0.7.1/libelf/libelf_ehdr.c:138: more undefined references to `_libelf_fsize' follow

jshap commented on 2019-11-06 06:43

@ruro no worries, I actually forgot that gcc-go exists. I was just trying to make things cleaner, the other flags should already be handling it anyways so it's really a non-issue :)

ruro commented on 2019-11-06 06:03

@jshap ah, yes, my bad. I thought, that I checked that I had the latest go, but apparently I must have misread the version.

Although, I am indeed not using Arch, but Manjaro, I think the issue is not that.

I have core/gcc-go 9.2.0-3 installed instead of community/go 2:1.13.4-1. Apparently, core/gcc-go provides go=1.12.2.

jshap commented on 2019-11-06 02:17

@ruro guessing not arch? since go 1.13 that flag is available :(

I'll remove it though.

ruro commented on 2019-11-05 23:20

Latest version doesn't build for me with the following error:

flag provided but not defined: -trimpath
usage: go build [-o output] [-i] [build flags] [packages]
Run 'go help build' for details.

Edit: removing line 33 of PKGBUILD seems to build just fine.

jshap commented on 2019-11-04 16:56

@ecly good catch, it was using go install but should have been using go build.

should be fixed now :)

ecly commented on 2019-11-04 14:32

This PKGBUILD is not compatible with go installations that use a local $GOBIN ENVVAR, as the resulting binary will be located in this directory when using go install, rather than the directory used in the PKGBUILD.

jshap commented on 2019-10-18 16:51

@darthdeus nvidia-smi reports the embedded version of cuda in the driver and not the installed toolkit version. so the version number is based on your nvidia kernel module version of the host, because really docker is mapping to that under the hood.

cuda toolkits are compatible on any driver higher than itself, so 9.0 will work on 10.1, but 10.1 won't work on 9.0, etc. so everything will be fine to continue using the 9.0 image.

you can check the toolkit version on the image with something like docker run --rm --gpus all nvidia/cuda:9.0-base cat /usr/local/cuda/version.txt

darthdeus commented on 2019-10-18 16:32

I'm not sure if I'm doing something wrong, but I have the regular cuda package installed from pacman, which has CUDA 10.1, and then tried to follow the wiki which says to run docker run --gpus all nvidia/cuda:9.0-base nvidia-smi, which also works, but the output says CUDA Version: 10.1, even though the image name has 9.0 in it.

Am I doing something wrong? I thought the container would have CUDA based on the image, and not on my external CUDA installation. Or is it just a problem of nvidia-smi?

jshap commented on 2019-09-11 13:40

@tndev yeah looks like the release was pulled for some reason? give me a bit to ask about it.

it makes sense that you cannot downgrade past runtime's v3.1.0 release as toolkit did not exist before that. Note that there is almost 0 difference for arch for toolkit between v3.1.0 (toolkit v1.0.1) and v3.1.3 (toolkit v1.0.4), so if you have to build this package at (this repos) commit 8378443f176f it should be fine.