Re: Bug#329422: No warnings when uninstalling kernel package
reassign 329422 kernel-package
On Thu, Sep 22, 2005 at 08:17:48AM -0400, Joe Mason wrote:
> On Thu, Sep 22, 2005 at 03:10:47PM +0900, Horms wrote:
> > Could you please send your reply to the bug,
> > email@example.com
> Oops, didn't realize Reply-To wasn't set. Here it is:
Thanks, an easy trap to fall into.
> On Thu, Sep 22, 2005 at 12:34:59AM -0400, Joe Mason wrote:
> > On Thu, Sep 22, 2005 at 12:20:52PM +0900, Horms wrote:
> > > The problem here is that Debian implicitly supports having
> > > multiple kernel packages and in fact no kernel packages installed
> > > to allow users to provide their own kernels.
> > >
> > > It might be prudent to document this "feature" somewhere,
> > > but I don't think its a bug in the kernel-image-2.6.8-2-686,
> > > or any other kernel package.
> > Perhaps apt should be able to notice when a kernel is being removed, and
> > either print a generic warning ("You're removing a kernel image! Please
> > make sure this is not the kernel you're actually running!" - this is my
> > preferred solution) or actually check uname -a and print a more specific
> > warning ("You're about to remove the kernel you're currently running -
> > this is only safe if you have set up your bootloader to load a different
> > kernel at the next boot") or even check the grub/lilo conf, although
> > that last is going a bit beyond the call of duty. That way at least a
> > new user who doesn't know what a "kernel-image" package is gets a
> > warning.
> > Documenting it in a way that isn't presented by apt when you actually
> > try to uninstall it wouldn't be very useful, I don't think.
> > As I said, apt still always gives the opportunity to type "I know what
> > I'm doing" if you know it's actually safe to uninstall the kernel
> > package, so I think printing the current danger warning wouldn't violate
> > the policy of allowing multiple kernel-images, although having a
> > slightly different error message would probably be clearer.
I think prompting the user in such a way is an excellent idea,
actually perhaps it already does but its not working for some reason.
In any case, that kind of functionality would almost certainly belong
in kernel-package. I am reassigning the bug accordingly so Manoj,
the kernel-package can add handle this as he sees best.