Qiime2 2022.8 can't be installed with mamba


I noticed that the CI for our Qiime2 plugin is failing for older versions now. In particular newer versions of conda don't seem to be able to install the environment for Qiime2 2022.8:

Run wget https://raw.githubusercontent.com/qiime2/environment-files/master/2022.8/release/qiime2-2022.8-py38-linux-conda.yml
--2023-03-02 19:16:39--  https://raw.githubusercontent.com/qiime2/environment-files/master/2022.8/release/qiime2-2022.8-py38-linux-conda.yml
Resolving raw.githubusercontent.com (raw.githubusercontent.com)...,,, ...
Connecting to raw.githubusercontent.com (raw.githubusercontent.com)||:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 9740 (9.5K) [text/plain]
Saving to: ‘qiime2-2022.8-py38-linux-conda.yml’

     0K .........                                             100% 31.5M=0s

2023-03-02 19:16:39 (31.5 MB/s) - ‘qiime2-2022.8-py38-linux-conda.yml’ saved [9740/9740]

Could not solve for environment specs
Encountered problems while solving:
  - package numpy-1.18.1-py36h4f9e942_0 requires libgfortran-ng >=7,<8.0a0, but none of the providers can be installed

See here for the full run.

I've also noticed that conda is good at installing the current version of stuff, but bad at installing old/historic software. I'm not sure how the Anaconda repo removes old packages, but longevity may be worth validating.

Also, that pin is very specific: libgfortran-ng >=7,<8.0a0


Hey @cdiener, another approach would be to install the metapackage, which should really end up being the same as the environment file, but happened to work for me just now (it's also possible I have a cached numpy that is letting this solve).

That said, the fact that your solver is trying to pull numpy-1.18.1-py36h4f9e942_0 (python 3.6!) means something has really gone off the rails already.

conda create -n <whatever you want> \
  -c https://packages.qiime2.org/qiime2/2022.8/passed/core/ \
  -c conda-forge \
  -c bioconda \
1 Like

Maybe it's because I am using mamba in the CI... I will try around a bit.

Yep I tested it, it's only an issue with mamba. Will be fine with conda. Sorry for that.