Post-processing RPM build logs

I’m so happy to write this article. With Packit and Copr, we are improving the RPM ecosystem so much that we can work on User Experience (rather Developer Experience) more and more. Finally \o/

Context: Sorin recently reached out to us that we should improve readability of RPM build logs by highlighting the cause. Completely valid request. Although I saw much more in the ask, especially after Mirek recently nudged us to be creative with ideas what to work on next (= go big).

I completely support Sorin’s request. RPM build logs are a pain to process (no offense to RPM). ALL build logs are a pain to process. They are huge, unstructured, plaintext, cryptic and unfriendly. If you don’t have enough experience with building RPMs, they are a puzzle.

Pavel’s suggestion in the issue was spot on: we could start a tool that would do the heavy lifting of the log analysis. I like that.

So for my Red Hat day of learning I started learning machine learning. And thought I would get a ML PoC done in one day.

Nope. This article convinced me it’s not that easy. Thank you, Zebrium, whoever you are.

No need to bring ML when we can find patterns using regular expressions in the logs. Exactly what the mock script does.

So I copied all of that code and started implementing more cases.

https://github.com/TomasTomecek/rpm-loggy

What’s next?

Well, Christmas 2022. 🎄

I won’t have time to work on this tool fulltime, it was just a proof of concept.

I’ll bring it up in our group and see what to do about it.

Happy holidays!

comments powered by Disqus