Hello <!channel>! A quick nf-test (git CI) question while preparing a new pipeline release #CJDV0UWH2 (https://github.com/nf-core/bacass/pull/255/checks). • With Docker and Singularity the nf-tests are stable ✅; • With Conda, some tools fail (e.g.

NanoPlot

,

ToulligQC

) ❌ • I updated the affected `nf-core/modules`; this resolved some tool issues, but the output files still differ from those produced with Docker and Singularity. What’s the recommended way to proceed for releases? • Skip/allow-fail Conda in CI? (I am aware that sometimes Conda is fleaky) • Relax nf-test assertions (e.g., file exists/size/regex on metrics instead of MD5)? • Ignore assertion of files that changes between docker, singularity and conda Thank you in advance.

Relax the nf-test assertions is my preferred method

Thanks 🙂. (And sorry for the wide ping)

Sometimes it is not conda results are not stable but it is stable but differs slightly from docker. I also had often cases that there is tiny numeric difference diverging far out in the decimals.

Yes, exactly