Fix #2248

[jonashaag]

Closes #2248

[jonashaag]

Hm this is bad the test is supposed to fail

[wolfv]

I think the conda version wasn't released on conda-forge yet, because of a conda-build bug.

[jonashaag]

Ah, I looked at the wrong place! https://github.com/jonashaag/mamba/actions/runs/4010134326/jobs/6886285496

[jonashaag]

OK, fixed: https://github.com/jonashaag/mamba/actions/runs/4013854780

[jonashaag]

Please squash merge

[jonashaag]

Hmm not sure what's going on

[jezdez]

I think the conda version wasn't released on conda-forge yet, because of a conda-build bug.

Yeah, this is ongoing

[jonashaag]

That's good because we haven't fixed compatibility yet! :-D

[jonashaag]

Isn’t the last commit just another way to write exactly the same thing?

[wolfv]

OK, I think I finally understand what's going on ... running mamba shell init probably overwrites what is written by micromamba in these files (e.g. stuff between # <<<<< mamba init <<<<). ugh. We should have used different delimiters from the start, too bad.

[jonashaag]

Another approach would’ve been to set $HOME to a temporary directory for the test.

[wolfv]

yes, that's true and is also something we do in the micromamba tests. I'd be fine with merging and releasing this though :)