On 2025-11-23 21:15:49 -0600, David Wright wrote: > But what seems to be the main cause of it occurring is the new > addition of §4.8
The cause is bug #1091428, dkms depending on the kernel headers without declaring a dependency. The headers being autoremoved is the effect and it is not limited to this particular autoremove action. This is just the first opportunity for failure, and one you won't miss if you follow the instructions. > which doesn't give any warning of an undesired removal of > linux-headers-amd64 in this case and potentially many others Autoremove is not more dangerous than any other remove/purge action. Caution is always advised. I use autoremove frequently and it had never caused any undesired removals before. Perhaps that's why I didn't check carefully enough or I just failed to see the consequences. > the graphical failure is quite likely to be hit when first rebooting > into an upgraded trixie When upgrading from bookworm you likely still have the headers installed so the first kernel should be fine. The removal happens later and the problem is with the next kernel update after that. You might spot it then but you'll still need to find the cause. >a hint could be added at §5.1.30 about dkms's possible role, and the >need to manually install linux-headers-amd64 I'd add this to 5.2 "Items not limited to the upgrade process" (you still need to mind the headers if you (un)install dkms later) or even 5.4 "Known severe bugs". A package that can stop working at any time for no obvious reason is severely buggy IMO.
