CI update to fix broken Docker image

[MatthiasZepper]

While troubleshooting an unusual sequencing run, I recently encountered an issue where both umi-tools dedup and UMICollapse crashed due to an extraordinarily high number of UMIs (>1e5) at a single position.

However, I recalled that you were working on another deduplication software, rumidup, and wanted to test, your tool could still handle this dataset. Unfortunately, when I tried pulling the Docker image from GHCR, I realized it no longer works:

/usr/local/bin/rumidup: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.33' not found (required by /usr/local/bin/rumidup)
/usr/local/bin/rumidup: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.32' not found (required by /usr/local/bin/rumidup)
/usr/local/bin/rumidup: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.34' not found (required by /usr/local/bin/rumidup)

The issue lies on my end, as two years ago I made the unfortunate decision to build it on the latest Rust image, which uses the current Debian stable release, but to run it on an image pinned to Debian Bullseye, which is now only oldstable. Hence, it is compiled against a newer GLIBC than available on the runner.

This PR proposes a fixed Dockerfile and upgraded GitHub Actions that include one to build binaries of your tool directly as release assets. I hope this helps and made separate commits for each, so you can git cherry pick from a different branch than master, which one(s) you like.

PS: Thank you so much for creating this tool! I hope you’ll release an official first version soon. Don’t forget to link the GitHub repository to Zenodo beforehand, so a citable record is automatically generated upon making a release on GitHub. If you’d like, feel free to adapt our Bioconda recipe from umi-transfer to rumidup to get it on Bioconda as well — I believe it should work with only minor adjustments!

[veldsla]

Awesome! The code test coverage is going to be pretty abysmal, but it will surely help us in pushing for a proper release. We actually are quite happy with the tool and use it in many production runs internally.

Thanks!

[veldsla]

New images have been created. I'm very curious to hear if you can process your dataset. Feel free to file an issue if it doesn't work!