Bugfix CI failing by forcing to retrieve the MKL package from the conda-forge channel

[Roy-Haolin-Du]

Description

• Our CI tests ran successfully across all Python versions (3.10, 3.11, 3.12) last week. However, this week the same code began failing specifically in Python 3.11 environments, while continuing to pass in 3.10 and 3.12. The failure manifested during the parameterization process where AmberTools' Antechamber component reported it could not find the libmkl_core.so.2 shared library.
• Resolving this issue with the following modification to our environment file(devtools/conda-envs/test_env.yaml):

dependencies:
  # Other dependencies unchanged
  - conda-forge::mkl>=2022.1.0  # Explicitly specify MKL from conda-forge channel

Close #41.

Status

• Ready to go
• Pass all CI test

Could you please review? @fjclark

[fjclark]

Perfect, thanks very much for getting to the bottom of this Roy!

Just for my understanding, why is it needed to explicitly specify conda-forge with conda-forge:: e.g. does mkl>=2022.1.0 not work?

[Roy-Haolin-Du]

Thank you, you’re right.
We've already set conda-forge as the highest priority channel, so specifying mkl >= 2022.1.0 works directly.
The tests have passed as well.
Can I go ahead and merge?
Thanks~

[fjclark]

Perfect, thanks Roy!