strange conda update results on M1 iMac but not Intel MacBook Pro

[Joe Heafner]

This past week I created all new conda environments and I've run into a stubborn problem. I first created a pristine conda-forge environment with

conda create --name forge --channel conda forge

and from within the new forge environment, wrote that configuration to a file using

conda list --explicit > ~/conda-forge-pristine.txt

Then I created a pristine environment named jupyter using

conda create --name jupyter --file ~/conda-forge-pristine.txt

and then installed jupyter with

conda install -c conda-forge notebook

Then I saved this new configuration to a file using

conda list --explicit > ~/conda-jupyter-pristine.txt

Okay so next I created a new pristine vpython environment using

conda create --name vpython --file ~/conda-jupyter-pristine.txt

and then I populated the new vpython environment with

conda install -c conda-forge vpython.

I use this process of creating environments and exporting them so I can easily create them on my various computers. Except for hardware architecture, which conda handles for me, the process is identical on other machines.

VPython 7.6.5 seems to work just fine, but here is the problem. Every day I run a script to update all my conda environments. On my M1 iMac (and I presume the same thing will happen on my new M3 MacBook Air when I get it configured) when I update the vpython environment using

conda activate vpython
conda update -c conda-forge --all --strict-channel-priority

I get the following:

The following packages will be UPDATED:

libgfortran 5.0.0-13_2_0_hd922786_3 --> 5.0.0-12_2_0_hd922786_32

The following packages will be DOWNGRADED:

libblas 3.9.0-22_osxarm64_openblas --> 3.9.0-17_osxarm64_openblas
libcblas 3.9.0-22_osxarm64_openblas --> 3.9.0-17_osxarm64_openblas
libgfortran5 13.2.0-hf226fd6_3 --> 12.2.0-h0eea778_32
liblapack 3.9.0-22_osxarm64_openblas --> 3.9.0-17_osxarm64_openblas
libopenblas 0.3.27-openmp_h6c19121_0 --> 0.3.23-openmp_hc731615_0

and I approve the updates.

Then if I immediately run again

conda update -c conda-forge --all --strict-channel-priority

I get the following:

The following packages will be UPDATED:

libblas 3.9.0-17_osxarm64_openblas --> 3.9.0-22_osxarm64_openblas
libcblas 3.9.0-17_osxarm64_openblas --> 3.9.0-22_osxarm64_openblas
libgfortran5 12.2.0-h0eea778_32 --> 13.2.0-hf226fd6_3
liblapack 3.9.0-17_osxarm64_openblas --> 3.9.0-22_osxarm64_openblas
libopenblas 0.3.23-openmp_hc731615_0 --> 0.3.27-openmp_h6c19121_0

The following packages will be DOWNGRADED:

libgfortran 5.0.0-12_2_0_hd922786_32 --> 5.0.0-13_2_0_hd922786_3

and I approve the updates.

This cycle repeats on every update attempt and the vpython environment never stabilizes. This is NOT happening on the Intel MacBook Pro with an identical (aside from hardware architecture) conda environment. It's only happening on the Arm iMac. VPython 7.6.5 itself seems stable but the package situation doesn't. It's also not happening with other environments.

I've searched online and could not find instances of this particular problem. Any suggestions?



Joe Heafner
Sent from one of my Macs

[John]

Maybe you can try googling something like "github actions M1 iMac vs Intel MacBook Pro" to search for clues to help you.

Here is the build.yml that github actions uses to build vpython

https://github.com/vpython/vpython-jupyter/blob/master/.github/workflows/build.yml

It contains lines like the following for building for different platforms 

platform: [ubuntu-latest, macos-latest, windows-latest]

Doing a google search such as the one above will give you links like the following

https://medium.com/@shohruhs/setup-your-own-github-actions-self-hosted-macos-runner-on-apple-m1-mac-0ae367f57813

I am not a Mac user and don't know much about github actions but maybe you can sift through the google search results to find a potential solution to your problem.

John