Conversation

Jonathan Corbet

The updated version of my plot of first-time contributors per kernel release. It seems this trend is not slowing down anytime soon. I wish I felt better about all of the stuff these folks are putting into the kernel...
3
14
26

@corbet well if it's any measure, this is the busiest cycle I've ever seen in mm and the review load has been unreal.

So I don't think 7.4 is going to be smaller :)

0
0
0

@corbet i'm sure it won't feel great when someone is asked to explain their changes months down the road and all you get back is a summary from a chatbot or, perhaps just as helpful, "idk my chatbot did it"

0
0
0

@corbet Do you have any sense of how much of it is new drivers vs changes to existing drivers vs changes to non-driver code? I'd like to think that the new contributors are concentrated in drivers, especially new drivers, but maybe not.

1
0
0
@cks I have not quantified it; perhaps I should. There are definitely new developers who show up with a new driver once a company decides to upstream it; I would guess there's a handful of them each cycle, just based on MAINTAINERS file changes.

As for the rest, a lot of changes are obviously focused on drivers just because that's a huge part of the kernel. But one thing that really characterizes the flood of new contributors is that they make changes all over the kernel. You'll see a previously unknown identity changing drivers, security modules, core networking code, and more, all in a single development cycle.

Essentially, if you're not concerned about really understanding the code you're working with, you can work all over, and a lot of the LLM-driven new contributors are doing exactly that.
0
0
0