General Rant about Frustrating Problem and Eventual Distro Hop

Software
, ,
1

Yesterday i distro-hopped a host of mine, but not because i just wanted to see if the “grass was greener” (which is my usual reason). This time i had to solve a real problem.

I’m writing a python script for work, and it’s using threads. The stack was…

Bare Metal → Fedora Server → KVM → Ubuntu 24.04 → Docker CE → Python Container

All of that was going very well. I was fiddling with code, running a docker compose up, finding my next problem, fixing it, looping. Until…

Docker compose started duplicating lines of text in it’s output. Because i was troubleshooting threading i couldn’t shrug this off. I wasted two days going around in circles with ChatGPT trying to find a solution. It didn’t seem to matter what information i gave to ChatGPT, it a) complemented me on being smart, insightful and good looking to ask the question, b) bashing out a fairly convincing page of explanation as to what the cause was and c) offering a solution that it was convinced would put an end to my misery. It was a similar rinse and repeat cycle to the one i wanted to be doing with Python in Docker, but completely unproductive.

My troubleshooting ended up with my Dockerfile ending with…


CMD [ "/bin/date" ]

…and i still got duplicate lines of output. about two out of three times though.

So atht ath point i gave up and replaced that stack with…

Bare Metal → Fedora Server → KVM → Debian 13 → Docker CE → Python Container

So far it’s all going very well. Fortunately i have a bunch of scripts that make light work of hopping distros.

I just wanted to shout that story into a post for the cathartic benefits. Thanks for listening.

Anybody tried Debian 13 yet? Anybody got enough experience with Ubuntu and Docker to have an opinion on what my original problem might have been?

Too good to throw out; too tricky to fix?
2

Hahah, that’s pretty annoying, especially when the AI builds you up with comments like “your FLOSS application is awesome! it will change the world and improve the lives of users everywhere!!”

Then the AI crashes, and you restart it but the context is lost and you’re just another looser hacker in your mums basement, again …

3

Can’t help on the Docker front, but I’ve tried Debian 13 in a few places over the last few weeks. All have been fresh installs, CLI only with no desktop environments - didn’t notice the difference from 12 for the very limited use cases so far. Other than migrating to Proxmox 9, I haven’t done an in place upgrade of any of my other Debian machines yet.

I can respect that - hope the shouting into the void was cathartic. Had a similar experience where ChatGPT and another model both “helpfully” hallucinated all these great ideas to help with a network configuration issue last week, none of which included real CLI commands that worked on the device in question. Nothing lost (other than the time taken), but it was my first and probably last experience seeing what AI had to say for these sorts of issues :joy:.

4

I’ve been using trixie in “production” for 6 months when I needed some bluetooth drivers that were not available in bookworm. I often run “testing” in a VM and have just installed a “forky”. My experience is that testing can be a little unstable initially but is pretty reliable in the second year of the release cycle.

I hope you are feeling better. It’s good to talk.

I think there is a temptation to just keep going and going with AI when you are not getting a solution to your problem. You feel like you’re just about there but the discussion just goes on and on. This probably relates to the fact that the word cluster associated with your question and follow ups are based on the distilled mass of all available text that is somewhat relevant. If your issue is a new edge case then the AI will not get you there.

Having said all that, the temptation to anthropomorphise the AI, particularly when it recognises what a great person you are, is strong.