Krzysztof Kozlowski krzk@kernel.org wrote:
... and if you even asked your AI like I did now: "how linux kernel developers look at LLM contributions from beginners?" I got:
"Linux kernel developers view unverified AI-generated code from beginners with heavy skepticism and frustration due to a massive influx of low-quality submissions that overwhelm maintainers."
"Rejection of AI "slop": "
"Staging area bans: Certain high-traffic areas prone to beginner contributions, such as the kernel's staging drivers (drivers/staging/), have implemented strict policies forbidding LLM-generated patches ...."
Yet you did exactly what should not be done - were using LLM.
Best regards, Krzysztof
Hi Greg and Krzystof
I didn't use AI to generate this code. It took 5-6 hours.
I don't know why you said I used LLM to generate this such code.
Krzystof wrote:
So you received feedback from Greg, but tried on different part, hoping this time it will get accepted?
This looks similar to you previous approaches, e.g. for DT bindings, where you received comments and repeated the same thing immediately. Three or four time.
I generally write 2 or 3 patches and send them parallelly. I won't do this again.
I thought fixing some spaghetti code and changing mutex_lock/mutex_unlock to guard macros is fine.
Best Regards, Omer PALA