Linus says that pointless fixes to Linux are getting out of hand and AI reviewers are making it worse

Before a new version of Linux sees a public release, it goes through several stages of release candidates. These are purely aimed at people who give kernel a spin, find bugs and report them so they can be fixed. Linus Torvald, the founder of Linux, noticed a strange increase in bug reports since version 7.0, which he quickly attributed to people using artificial intelligence tools to examine code and automatically report problems.

Last week, Linus took to the release candidate announcements to complain about how AI-generated bug reports were flooding secure channels when they were better off going through public channels. Now it seems that Linus is getting annoyed with people who dump any and all fixes they find on the maintainers, no matter how trivial.

Linus Torvalds is not happy about the departure of the Linux 7.1 release candidates

Things are a little too noisy for his liking

Linus sent a message Regarding the fifth release candidate of Linux 7.1. Linux kernels usually get seven release candidates before they’re released for general use, so if nothing game-breaking comes up in the next few weeks, it should be released soon.

Linus notes that 7.1 follows the same pattern as 7.0, with more changes to the kernel than usual. And, as with 7.0, it’s not because the kernel is in bad health; because error detectors in AI find and detect more problems. So much so that Linus is getting a little annoyed at being reported so many trivial issues. As he said:

I’m not entirely happy with it – a lot of it is completely pointless stuff for casual drivers, which obviously makes things a lot less scary, but at the same time, I’m not sure the confusion in rc5 time is worth it. These things are “fixes”, of course, but at the same time many of them are so trivial that I think they’d be better off in the Linux-next tree and merged during the merge window.

So I think I’m going to start being a little tougher on this kind of unnecessary panic late in the game. We should be looking for *regressions*. Non-critical fixes to long-standing issues are simply not appropriate for this late in the release cycle.

Later in the message, he said that “several of these series were triggered by AI code reviews” and that he would be stricter with those reviews in the future. He bases his position on the fact that even though trivial fixes have a low chance of causing more problems than they fix, it’s a non-zero chance, and he prefers that people take them to the Linux-next-tree to include them in the next kernel version.