Re: Bug#396331: upgrade-reports: sarge to etch removes kernels
On Wednesday 01 November 2006 00:38, Kevin B. McCarty wrote:
> > Why *should* aptitude try to fix it?
>
> IMO, the benefits of users not having to see the scary kernel removal
> message, answer the right thing to it, and then have to re-run aptitude
> after explicitly marking the old kernel package to remain installed,
> outweigh the annoyance of having old kernel packages remain installed on
> their systems. If you disagree, though, I doubt I can say anything to
> change your mind. Then I guess the release notes should be adjusted to
> suggest the best order of operations.
Well, as a *user* I think my desired behaviour is clear: whenever a kernel is
upgraded, all previous kernels should be removed (actually, treated normally
as per aptitude rules) EXCEPT that the running kernel must always be retained
so that I have something to boot back in to if booting the new kernel does
not work. Ideally the old kernel should automatically disappear on the next
kernel upgrade (so I always have two kernels to choose from), although it
could disappear earlier than that.
I would not expect the upgrade to fail or abort unless dependencies mean that
this rule cannot be followed. In that case I would expect the release notes
to describe a series of actions which would allow me to work through the
dependency issue while making sure that I am never in the position of having
to boot a new kernel without being able to fall back to a previous one.
I don't care how that is achieved. Either aptitude can special case the
running kernel (but not all kernels) or the kernel package can do some magic
itself.
However, neither of the options being discussed here (removing the current
kernel or leaving all kernels installed forever) would be acceptable to me as
a user. Surely we can do better than that.
Graham
Reply to: