mamba dependency changes mamba behavior

As advised previously, I only have mamba installed in the base environment, because otherwise mamba starts to behave differently from conda.

The best example is updating a different env from the current one:

conda activate py38
mamba update -n base --all

Only when the only available mamba is in the base env, this behaves as expected. If there’s a mamba installed in the env py38, then, with that activated, above command updates that env, and not base, despite the -n option used.

Now, for package development using mambabuild one requires boa and therefore mamba be installed in the env used for development.

Questions:

• Does this mean, there is no way to update other envs?
• Could boa not use/lookup mamba from the base env, to not break mamba’s correct functionality?

FYI, here’s my setup:

╰─ conda config --show-sources                                                            ─╯
==> /home/maye/miniconda3/.condarc <==
channels:
  - conda-forge

==> /home/maye/.condarc <==
channel_priority: strict
report_errors: True

==> /home/maye/miniconda3/envs/py38/.condarc <==
pinned_packages:
  - jupyter_client=6
  - nbconvert=5
channels:
  - conda-forge

─ mamba --version                                                                        ─╯
mamba 0.17.0
conda 4.10.3

[wolfv]

you just need to place a conda_build_config.yaml file next to your recipe, or include it using -m. This defines the build matrix, e.g.: https://github.com/conda-forge/mamba-feedstock/blob/9f9bece91d8f8c3c3c11db1326685ab4d715369f/.ci_support/linux_64_.yaml#L31-L36

[michaelaye]

i think there’s no CLI option to do the first build at a different than current environment, right?