Hi, I'm trying to fix the environment issue for phyloflash with conda. The environment can be create locally but fail on the github action with the following error:

>       ERROR conda.core.link:_execute(1014): An error occurred while installing package 'bioconda::spades-3.15.5-h95f258a_1'.
      >       Rolling back transaction:
      >       LinkError: post-link script failed for package bioconda::spades-3.15.5-h95f258a_1
      >       location of failed script: /home/runner/work/modules/modules/.nf-test/tests/769e57f0e7d8bc77bee19f97fef1c9d9/work/conda/env-27f8de4e77d219b103e2d1302f904167/bin/.spades-post-link.sh
      >       ==> script messages <==
      >       <None>
      >       ==> script output <==
      >       stdout:
      >       stderr: /home/runner/work/modules/modules/.nf-test/tests/769e57f0e7d8bc77bee19f97fef1c9d9/work/conda/env-27f8de4e77d219b103e2d1302f904167/etc/conda/deactivate.d/deactivate-gxx_linux-64.sh: line 68: CONDA_BACKUP_CXX: unbound variable

I've tried to specify additional libraries to match the singularity container, but I always came back to this error. Does anybody know how to solve this ? github.com/nf-core/modules/pull/12577

If I'm reading this correctly, it's because of the

deactivate-gxx_linux-64.sh: line 68: CONDA_BACKUP_CXX: unbound variable

The

process.shell

directive that we use across nf-core set's

-u

in addition to the

-e

and

-o pipefail

. That

-u

I think is what's resulting in the error. The normal process is to

set +u

around the command where it occurs. github.com/nf-core/modules/blob/…/main.nf#… It's hard for me to see from your code exactly where this occurring though. It looks like it is not in the script from that error message.

The issue appears at the conda environment creation level when the spades library is built.

I'm not sure how to solve that. I think setting the

process.shell

to not use

-u

would solve it, but that would affect every module, since I think the

.command.run

where the conda environment is being setup also uses the settings from that directive.

It seems the issue lies with the newest version of conda. Pinning it to v24 solve the issue. But I don't think we should pin it in the github action. I don't know what to do... Should I deactivate the conda test for this particular module ?

@Júlia Mir Any suggestions on what the preference would be?