About stability: Where we do use kernel/bootloader built with Armbian build system, it should be the same/similar in both projects. The patches we add are not many (yet) and specific, like fixing reboot on Odroid C2, LEDs on NanoPi NEO, adding some WiFi/BT drivers. We usually use the “current” branch, which is the same for Armbian’s images, i.e. the ones most prominently shown on their download pages. Only in case of RK3588 SoC SBCs we still use the “legacy” branch, while Armbian switched to “vendor” (based on newer Rockchip Linux 6.1 sources) with their primary images. That one however causes some udev-worker loop with constant ~5% CPU load on my Orange Pi 5 and 5 Plus, which is why we did not switch yet. Would make sense for Armbian to know about this, but if I report it, I likely get a page long blame why I dare to report a bug, wasting mine and the author’s time and nerves, so I leave it. Will have a closer look when I find time, but currently busy with issues on some other SBCs.
Since we provide certain (soon all) kernel upgrades, also for major kernel versions, earlier, our users are a larger number of ~early adopters, compared to Armbian where this is the case for freshly flashed images only, until their APT repository gets the update. Since issues can appear with those newer kernel builds, this is by times a stability downside our end. It could be a benefit for both projects, when we collected and analysed these issues as good as we can, and did an as complete bug report as possible, kept investigation, testing etc. So things had a chance to be fixed before being pushed to Armbian’s APT servers. However, due to mentioned attitude towards bug reports from our side (and bug reports in general), this potential benefit is sadly not claimed. We seem to have fundamental different opinions on whether bug reports are important/essential for the health of a project, or seen as a waste of time for developers only. And I am even on the Armbian GitHub orga blacklist, preventing me from contributing fixes for those solved within our community. I am drifting off …
With what we do in userland, can rarely ever have any impact on device stability: Default CPU governor schedutil
vs ondemand
(again) on Armbian + they do some IRQ affinity tweaks which, as far as I read into it, have no practical benefit on modern Linux anymore, but do not hurt either. But these are minor performance tweaks, which should only indirectly be able to affect device stability, respectively should not be able to cause instabilities, if the system otherwise runs within sane parameters, with good PSU, quality SD card etc.
This is exactly what I would suggest
, and how I do suggest on vendor forums. I actually want to make this more prominent on our download page. Something like “Kernel/bootloader powered by Armbian”, with a link to their donation page. Same for other such cases, like Fishwaldo for Star64 device tree and some drivers.