Ben Sherman

What I'm not understanding is why would you provide the path `/` as an input to a process? What is your intent with that? Also not sure if it's related...

Not sure this is worth supporting. Normally we detect fusion symlinks by checking if the output file is also an input, which is cheap because the list of inputs is...

> should be the role of Fusion to concretize symlinks when referenced in the publishDir I assume you mean the process outputs? Because we are decoupling the publishing from the...

Well, if we use the workflow publish instead of publishDir then the task doesn't know which outputs will be published

Workflow publish (i.e. output) definition: https://github.com/nextflow-io/nextflow/pull/4784

I was going to summarize the hybrid solution we discussed on the call, but thinking through it now I realize that it won't work. The workflow output definition makes it...

Keep in mind that it only heeds to happen when the user creates such a link or forwards an input as an output. These cases are uncommon and can be...

In this case I recommend also moving `executor.perCpuMemAllocation` to `slurm.perCpuMemAllocation`.

For the `exec` operator, I would wait until the input/output file tracking is improved. Right now if you use it, it will work, but things like provenance tracking and automatic...

> I think you managed to hit the small window when the MultiQC bioconda build was broken Okay I deleted the conda environment for multiqc but still got the error...