News for hackers. Stuff that matters.
Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by righthand on Monday December 23, 2024 @11:01PM (#42499652)
Aren’t debuggers effectively dead if everyone is just going to brute force a fix with an LLM?
Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by ehnto on Tuesday December 24, 2024 @12:16AM (#42499873)
One of the reasons I am not yet worried about LLMs is that this is not actually possible, at sufficient levels of domain complexity, and contextual complexity. The domain complexity that LLMs start failing is pretty low at the moment.

As well, as domain complexity increases so does the required amount of precision, and again I find LLMs start getting less precise when put in the deep end. You don't tend to find this kind of software in your "programmers programming". You tend to find this kind of software in industry, where large businesses consist of wild business rules built up over decades, and that has been half-codified into software. I just don't see LLMs swooping in to take that over.

Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by RobRivera on Monday December 23, 2024 @11:11PM (#42499684)
What is the best diet for trolls these days?
Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by metadat on Monday December 23, 2024 @11:38PM (#42499770)
This is a fair question, though it's probably still mostly the same as pre-LLM: spoiled data, heinous cackling laughter, various darkweb fungi, and occasional raw flesh.
Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by righthand on Tuesday December 24, 2024 @11:01AM (#42502748)
Ah yes the petty responses of the “bigger” “smarter” person.
Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by metadat on Tuesday December 24, 2024 @11:36AM (#42502969)
You're good in my book. Happy holidays. :)
Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by righthand on Monday December 23, 2024 @11:23PM (#42499719)
I’m sorry, but if LLMs are going to change the way we code, and the way we code is not even print debugging for junior developers. I know very few senior developers that use a debugger at all. I don’t understand why my question is trolling.

What is the best diet for those easily triggered?

Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by rfleury on Monday December 23, 2024 @11:38PM (#42499767)
Are you out of your mind? LLMs will require much more debugging, not less.

Separately, I have no idea which “senior developers” you know, but with or without LLMs, debugging is an immensely important part of development across the industry.

Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by jemmyw on Monday December 23, 2024 @11:39PM (#42499771)
I don't think LLMs have anything to do with it. I've rarely seen devs reach for a debugger until they really really need to. In web dev it can be pretty hard, you need a reproducible bug or production server access. I can probably count on one hand the number of times a bug occurred that stumped me to that point.

If anything, I can see LLMs making debuggers more used because they'll help you with getting setup and interpreting what is going on.

Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by ehnto on Tuesday December 24, 2024 @12:28AM (#42499899)
I am not trying to sell anyone on debugging, but honestly if you have a good debugger and it's easy to drop into, you'll find it way easier and faster than print debugging. I don't know why people put themselves through the pain of writing print all over the place when you can just, pause the program and see literally everything, all the way up the stack.
Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by jemmyw on Tuesday December 24, 2024 @12:47AM (#42499959)
There are different levels of debugger. I use language ones all the time. But this article is talking process debugging. If you dev native compiled apps then that's probably second nature. But if you do dev on a scripted or runtime language then it is, as I said in my previous comment, rarely applicable but occasionally very helpful.
Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by troupo on Tuesday December 24, 2024 @12:32AM (#42499910)
John Carmack: It still boggles my mind how hostile to debuggers and IDEs Silicon Valley VC- backed companies are

https://youtu.be/tzr7hRXcwkw?si=LX7LhGYUHmPAWVNJ

Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by troupo on Tuesday December 24, 2024 @02:05AM (#42500178)
> if LLMs are going to change the way we code

They are not

> the way we code is not even print debugging for junior developers.

what

> I know very few senior developers that use a debugger at all

I know very few, too, and it's their loss. Debugging is immeasurably more useful to understand and get to the root of the problem in a complex system than printf'ing and logging.

> What is the best diet for those easily triggered?

The best diet for you is to unplug from the hype trains on Twitter and HN.

Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by righthand on Tuesday December 24, 2024 @11:00AM (#42502744)
They did, as never before were people talking to a chat bot to have code generated. Seems different to me, but please more talking down and pettiness.
Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by norir on Tuesday December 24, 2024 @12:22PM (#42503301)
> Debugging is immeasurably more useful to understand and get to the root of the problem in a complex system

I try to write systems that aren't so complex that I need a debugger but get the same job done. And I choose not to work on systems already so complex that their usage is mandatory. I do not consider this a loss at all. Systems that I write can almost uniformly rebuild and rerun their tests in under a second. I can add (pretty) prints and see the results faster than I can switch to a different tab to run a debugger. I have never introduced or encountered a bug that required advanced debugging techniques besides getting a stacktrace from a crashing problem. YMMV.

Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by draven on Tuesday December 24, 2024 @03:06AM (#42500437)
Senior here, I started to use LLMs recently but I still use a debugger. Working with complex code it helps me find the difference between what I think should be happening and what really happens. I don't think a LLM could do better for this task.

Also, I use debuggers for print debugging: insert a breakpoint which doesn't suspend execution but prints something (perhaps conditionnaly.) That way no need to recompile the whole thing and restart it after adding a print statement.

Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by hun3 on Tuesday December 24, 2024 @08:22AM (#42501740)
Have fun brute forcing heisenbugs and LLM hallucinations
Re:Demystifying Debuggers, Part 2: The Anatomy of a Running ... by righthand on Tuesday December 24, 2024 @11:00AM (#42502738)
I use a debugger, tell that to all the people using LLMs.