Voxl-tflite-server build error

Maxim Temp

Hello there,
I have been using voxl-tflite-server v0.3.0 with my changes for a project I was working on. However, with your latest updates I wasn`t able to build this package and docker-cross gave me these errors:
I was installing the dependencies with the command:
./install_build_deps.sh apq8096 sdk-0.7
Also, when im installing dev and not sdk-0.7 it seems to build it, but then I encounter with these errors:
1)voxl-tflite-server: error while loading shared libraries: libtensorflow-lite.so: cannot open shared object file: No such file or directory
2) sed: /etc/modalai/voxl-tflite-server.conf: No such file or directory
This is not the first time i encounter with that problem and it seems that it happens every time when you publish new sdk. For example, when I was working during a project on the apq8096-camera-server (sdk-0.7) I couldnt deploy the package to the voxl because it downloaded automatically the new voxl-camera-server (becuase of your new sdk-0.8).
You prevent me every time from working on specific versions. You should understand that i wont copy my code each time you publish new sdk...

Hey @Maxim-Temp,

These issues are present as you were working with a dev version of voxl-tflite-server (v0.3.0) and voxl-tflite that both recently got merged to master.

Installing sdk0.7/sdk0.8 build deps will cause the build to fail because the tensorflow-lite version was upped to v2.8 from v2.2 , causing some file changes with its dependencies like absl.

Installing dependencies straight from dev will only work on voxl if you are also using the dev package repo and have the latest dev versions of voxl-tflite-server's dependencies, which from your error message "1)voxl-tflite-server: error while loading shared libraries: libtensorflow-lite.so: cannot open shared object file: No such file or directory" I can tell that you are not.

The easiest solution is upgrading your voxl-tflite package to the latest release from either the staging or dev repo on voxl, as your version (2.8.0-1 it seems) still provides a tensorflow-lite.a static lib where v2.8.3-1 (latest) provides the tensorflow-lite.so. This will not break any code/custom things you have done, and should add a few improvements over the version you have installed now that was only published to dev and has since been updated.

Maxim Temp

Thank you for your fast reply, now i understand the concept of dev. I want to try the new sdk-0.9 but i need to download the system image v4.0.0. I went to your downloads page but this image is not there. only the previous VOXL Platform 3.8.0-0.7 is there.

Maxim Temp

@Maxim-Temp Also please upload the voxl-cross docker image v2.2 .