On Sat, 2010-05-15 at 11:01 +0200, Richard Zidlicky wrote:
of course, and I do not think it is so hard to think of a sensible behaviour.
After each (semi)automatic change to grub/kernel conf as well as for the very first boot there should be a timeout as well as visible menu. Once the kernel did boot with default command line etc it would be safe to set the timeout to a small value - after asking the user.
More elaborate solution, there could be two config values - quicktimeout and safetimout. After kernel and config changes timeout would be changed to safetimout and once the kernel booted safely it could be reset to quicktimeout automatically.
Neat idea. But if a breaking kernel change somehow occurs without triggering the change to the safetimeout, we would not want the user to be completely stuck. I see two ways to address that:
- Make quicktimeout nonzero enough that the user has time to react.
- When grub attempts booting with quicktimeout, have it change to safetimeout. Then have an initscript that changes back to quicktimeout once booting has succeeded. Grub already has a "default boot entry" field in the stage2 image that can be written by boot commands for exactly this purpose; see the info docs. The same could be done for the timeout. (This would appear to be a common trick: my Dell Latitude D620's BIOS does the same thing with the power-on self test.)