The sweeping changes in Linux are amazing, but they also break a lot, frequently, and the testing approach leaves a lot to be desired. I am old and a former embedded coder, I understand how the 90s developers feel about unit testing, and I am sympathetic to the idea that test-driven-development and heavy UT in modern teams is often completely out of control, but Linux is distinctly lacking in real testing of independent units in isolation. I am aware of kunit, kselftest, etc.
But you’re just reiterating that the kernel development itself lacks sufficient testing which necessitates this. Everyone is a volunteer; the result is as you’d expect.
Look, there’s nothing wrong with this. In reality all the people who are electing to run Linux without it actually having sufficient quality for their purposes are the actual problem, who’d which there are hundreds of companies that just don’t care. We created this situation by undervaluing quality vs cost.
But it’s also the reason it won’t be a success on the desktop. The only way that happens is if windows takes an even more substantial nosedive. Things are going to have to get pretty bad for that to happen.
I wasn't arguing that it was entirely false, but that the generalizations of "regressions are introduced all the time" and "Linux developers spend little to no time" can't be applied to the entire Linux ecosystem.
I wouldn't discredit your experience, but speaking of Linux kernel development specifically, my understanding as a user is that it's held to a very high standard of scrutiny. Rants from Linus are famous for drilling into people that "WE DON'T BREAK USERSPACE". If you consider how many thousands of lines of code are changed on a regular basis, it's an engineering wonder that it works as well as it does. In many ways Linux is a victim of its success.