CUDA builds are preferred over CPU builds

[traversaro]

Solution to issue cannot be found in the documentation.

• [X] I checked the documentation.

Issue

mamba create -n librealsense librealsense

results in the CUDA version being installed. I am not sure this is the indended behaviour.

Installed packages

.

Environment info

.

[traversaro]

Interesting, this is indeed the behaviour intended for jax and pytorch:

• https://github.com/conda-forge/jaxlib-feedstock/blob/46a1829880e7012e9a05bf8904ddbbc9e5e51375/recipe/meta.yaml#L4C1-L6C12
• https://github.com/conda-forge/pytorch-cpu-feedstock/blob/main/recipe/meta.yaml#L4
• https://github.com/conda-forge/tensorflow-feedstock/blob/2fdde9bd0977c9356ad192d684c8ca2751bf94ab/recipe/meta.yaml#L129

However, at the moment in librealsense we were trying to favor the CPU version over the CUDA one (as done in libarrow:

• https://github.com/conda-forge/arrow-cpp-feedstock/blob/752163190f2c02791addd4cc8c9f95790fab4b13/recipe/meta.yaml#L83

However, for some reason track_features seems to be working fine for libarrow, but not for librealsense, as the cuda version is installed by mamba, while the cpu version is preferred by conda.

So, to reformulate the two question we have are:

• Q1: What version should be preferred? The CUDA version or the CPU one?
• Q2: How do you indicate that a version is preferred? track_features or increase the build number of the preferred version?