Docker and ubuntu aarch64 architecture installation

Orekit usage
1

Hi folks,

I am trying to set up a docker container to use Orekit alongside some other libraries, I’m using an M1 processor on my machine and some of these other libraries require that I user a docker image based around a linux aarch64 architecture. I am having some issues installing Orekit in the container as the install command doesn’t find a suitable Orekit version for my system configuration (I have a fresh OS install so I don’t think it is due to library clashes being present). I know that Orekit installs/runs with aarch64 as I am running it locally on my Mac. Is it the case that I would need to compile from scratch for linux aarch64? A minimum viable example of the Dockerfile is here:

FROM --platform=linux/aarch64 ubuntu:20.04
ENV DEBIAN_FRONTEND=noninteractive

RUN wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-aarch64.sh \
    && mkdir /root/.conda \
    && bash Miniconda3-latest-Linux-aarch64.sh -b \
    && rm -f Miniconda3-latest-Linux-aarch64.sh 

ENV PATH="/root/miniconda3/bin:${PATH}"

RUN conda install -y -c conda-forge Orekit=11.2

Thanks for the help.
/Paul

2

I got the install working manually using maven. Unfortunately, I was also using the python wrapper for which JCC does not support linux aarch64 so have been unable to compile the python wrapper. I guess this is maybe why there isn’t an image on conda-forge though if anyone else encounters the same issue.

3

Hi,

The lack of aarch64 is mainly because of the issues in building the package. JCC should work on arm but the thing is that the build stack at conda-forge uses cross compilation for building arm. JCC is built on generating wrapper code that is then compiled as a part of the JCC process and I have not managed to get this one to cross-compile for a different architecture. This is likely the reason that the mac build is struggling as well.

For arm I would until we solve this (which may take some time unfortunately) either try to build it locally on the platform, or use orekit_jpype, an experimental wrapper built on jpype instead.