Tegra libreries are missing in docker build

[elad11]

hey so I saw a similar post about building docker and not seeing the tegra libreries (/usr/lib/aarch64-linux-gnu/tegra/) while build that is related to runtime nvidia.
i’ve set the default runtime to be nvidia at :

 cat  /etc/docker/daemon.json
{
    "default-runtime": "nvidia",
    "runtimes": {
        "nvidia": {
            "args": [],
            "path": "nvidia-container-runtime"
        }
    }

and you can see in :

docker info | grep -i runtime
 Runtimes: io.containerd.runc.v2 nvidia runc
 Default Runtime: nvidia

but still when trying to build a docker and ls to the tegra libreries I get :

 => ERROR [19/19] RUN /bin/bash -c "ls /usr/lib/aarch64-linux-gnu/tegra/"                                                                                                                             0.6s
------
 > [19/19] RUN /bin/bash -c "ls /usr/lib/aarch64-linux-gnu/tegra/":
0.574 ls: cannot access '/usr/lib/aarch64-linux-gnu/tegra/': No such file or directory
------
Dockerfile:167
--------------------

Any suggestions?

[AastaLLL]

Hi,

Could you tell us which base image you are using?
For Jetson, please try l4t-base for compatibility.

More, do you use JetPack 4 or JetPack 5?
On JetPack 4, the container mounts the library from the host while JetPack 5 includes the user-space libraries inside.

Thanks.

[elad11]

Hey i’m using the image nvcr.io/nvidia/l4t-tensorrt:r8.4.1.5-devel
and I have jetpack 5.0.2

[AastaLLL]

Hi,

The files are mounted when runtime and not available at building time.
Do you need these file when running docker build?

Thanks.

[elad11]

Hey,
Yes I need them during docker build.
Im running a project cmake build that needs an .so that is in those libraries.
Is there a solution?

[AastaLLL]

Hi,

We need to check this internally.
Will update more info with you later.

Thanks.

[AastaLLL]

Hi,

Could you share why you need the file at building time?
Suppose you can build the file natively and copy it into the container.

Thanks.