Hi

How do you people approach solving problems specifically on Linux?
Like at first read the docs and if not works then how do you  move forward?

Interested to hear your approach!!

Last edited by skyrim (2025-04-18 15:28:34)

My checklist usually looks like this.

- web search on the error message (~1/3 success rate [SR])
- web search on the problem (~1/4 SR)
- man page (~1/2 SR)
- if the man page is too long and complicated (ffmpeg, mpv, etc.), run up my local LLM and ask it (~1/4 SR)
- wiki.archlinux.org (~1/3 SR)
- wiki.gentoo.org (though arch wiki often links here) (~1/3 SR)
- look for alternative software -- I've even changed distributions at this point before (~1/5 SR)
- dive into the code or roll my own solution (or just give up if it's not a serious problem) (~1/10 SR)

A web search usually covers the wiki's too, but not always. My lowest success rates are with hardware/uefi errors.

I thought about adding a bullet for throwing a tantrum and cursing at the computer, but I'd have to put that between every other bullet.

"not" - I solve them like every other problem.
1. isolate the cause (98% of the task)
2. find a solution or workaround (the remaining 2%)

You solve problems - every! problem - by asking the right questions about it.
You can ask google (better than artificial idiocy) but you still need to figure the right questions first.

Can you elaborate on your approach of isolating the cause?

+1

I just recently re-watched the entire series. Still brilliantly hilarious.

OT: what seth said.

Read the output of the program.  Start it from a shell to see this.
Read your journal.

Nothing is too wonderful to be true, if it be consistent with the laws of nature -- Michael Faraday
The shortest way to ruin a country is to give power to demagogues.— Dionysius of Halicarnassus
---
How to Ask Questions the Smart Way

Specifcially start wide and then narrow down on the issue.
Eliminate variables as much as you can and try to establish a working baseline.
Don't create moving targets by changing a random set of variables at once - worse than blurring the relevant parameter (what change actually fixed it) is that you can hide a solution by introducing a second problem.

If you're lucky, you can skip this because the journal is yelling the cause at you - then see #2
The challenge here is to "know" which red line in the journal is the bad one - eg. when you see a backtrace, the evil line is typically neiter the first nor last one (often very generic calls) but one between.

Basically you have to be an expert ;-). Figuring things out by yourself itself is not very hard, but usually there's always one problem. Time is constrained, always.

Can you afford to sit on the same issue for months? Sure, you.will certainly figure it out. Will you not go to work because you have a few hobbyist issues for solve on your PC? Uhmm... probably not.

This is where it gets difficult for me, and tools and tech like Google and ChatGPT seem to come in handy, even though I actually rarely use them they do help a lot at times... I did use to use them a lot when I tried to get my foot out the door

Trace what I've done
Arch Wiki
DuckDuckGo
LLM if really necessary

I don't know if its the right approach, but avoiding problems has come down to really debating if I need a "feature" I'm looking at or if I'm just creating a project that wont add to my computer experience.

Simplicity is the ultimate sophistication

Being minimalist has helped me a lot with avoiding problems with my PC.

But, then, my PC is just so incredibly boring that I have to entertain myself via other means ;-(

I usually:

1. Search it on google (or whatever search engine you like)
2. If google doesn't help, search it specifically on reddit, other people probably had that specific problem in the past
3. Try to wrap my head around it for a bit, looking at documentation, wiki, FAQ, etc...
4. Ask on forum or reddit or irc describing in the detail what is the problem, what I already tried, and in general as much details as possible
5. Profit?

Feel free to send me an email if you need anything!

Absolutely love that you'll change your distribution *before* giving up if it's not a serious problem.  That is true Linux chad!

As to my way, I switched to Linux in February and chose Arch.  It has been generally brilliant and smooth, but there have been a few issues.  I kind of do the reverse of others.  I just dump the relevant error parts of the journal straight into a LLM, which sometimes gives me a more or less effective answer immediately, or gives me enough to solve it myself, but if not gives me search terms to then find a person who has solved the problem.  Not keen on AI in general but it is pretty useful for troubleshooting.

Last edited by invisibleaxeman (Today 20:50:57)