Cure found for kernel updates

[Matthew Garrett]

On Wed, May 14, 2014 at 09:54:58PM +0200, Lennart Poettering wrote:
> On Wed, 14.05.14 19:47, Matthew Garrett (mjg59 at srcf.ucam.org) wrote:
> 
> > On Tue, May 13, 2014 at 10:27:54AM -0400, Matthias Clasen wrote:
> > 
> > > We really need to use the boot loader spec by default. Is there any
> > > reason why the support for it was added to grub, but not enabled ?
> > 
> > The full spec is incompatible with standard practices for the EFI system 
> > partition, 
> 
> Humm, is it? Is this about loving /boot/efi so much?

Basically. There's a bunch of places that would need to be changed, and 
doing it meaningfully involves doing that across multiple distributions. 
I mean, we *could*, but it's a pain.

> > and especially incompatible with the way we handle Mac 
> > hardware, so adoption is unlikely. 
> 
> Humm, what? the spec is used (in its gummiboot implementaiton) everyday
> on macs, not sure what you are referring to.

The bootloader needs to be on an HFS+ partition to appear in the OS X 
boot preferences.

> > The boot fragments are more 
> > attractive, but don't currently let us express the full set of 
> > configuration that we support (there's no way to specify chainloading 
> > another bootloader, for instance). These seem fixable, but nobody's 
> > currently fixing them.
> 
> The boot load spec stuff is not supposed to be necessarily exclusive. If
> you want stuff like chainloading or memory checking or whatever else,
> then i recommend simply doing that outside of the spec/drop-ins, and
> leave the drop-ins for kernels and EFI binaries only.

Then we have configuration split over two completely different formats. 
It's kind of unappealing.

-- 
Matthew Garrett | mjg59 at srcf.ucam.org