> On Jan 28, 2019, at 3:45 AM, John Paul Adrian Glaubitz
> <glaub...@physik.fu-berlin.de> wrote:
>
> On 1/28/19 12:34 PM, Rick Thomas wrote:
>> I realize that this may require some programming, but would it be possible
>> to have it ask a question early on (maybe at or before the beginning of
>> partitioning) requesting the user to choose between yaboot and grub? Then
>> the partitioner would automatically create the necessary partition(s) and
>> “install boot loader” would automatically install the chosen boot loader
>> conditioned on the answer to the question…
>
> Yaboot is unmaintained upstream and does not support modern ext4 features. In
> order for Yaboot to work properly, you have to turn certain features in ext4
> off, otherwise it won't work and the boot fails.
>
> Unless someone picks up maintenance work on Yaboot and makes it work with
> modern ext4 versions, I don’t see any particular reason to keep Yaboot.
It’s not necessary to make yaboot work with modern ext4. It’s only necessary
to have a separate /boot partition which is ext2, as is the current procedure
when the user chooses LVM partitioning. Yaboot already requires a separate hfs
bootstrap partition, so I don’t think having a separate ext2 /boot is that
much of a stretch.
>> Are there other architectures where something like this might be useful?
>
> Why do you think should the installer support a boot loader that is known
> to be buggy and unmaintained? If users insist on using Yaboot, they can
> still install it manually. I do not see a point, however, to keep it in
> the archive.
When I asked about other architectures, I was asking if there are any cases
where there are two (or more) alternative boot loaders for a given architecture
and users might want to have a choice between them?
Yaboot may be buggy and unmaintained, but then so is the hardware it runs on.
PowerPC Macs haven’t had any updates that I know of in several years.
If the only known bug is that it doesn’t support modern ext4 features,
I have to say that’s something we know how to live with — a separate ext2
/boot partition fixes the problem.
I agree that migrating to grub has lots of advantages, and it’s something
we should do if we are serious about having a Debian port for powerpc
hardware. But at the present moment, grub isn’t working. All I’m
suggesting is that while we work on grub, we can make the transition
smoother by providing something for the time being that does work.
And don’t forget, there is a significant group of folks for whom yaboot,
for all of it’s problems, is a necessary part of using Debian on their
hardware.
> partman-ext3 still contains a work-around on powerpc [1] which I would
> like to get rid of. The workaround turns off 64-bit support in ext4
> and checksumming of metadata, both features are desirable to have
> these days.
If the reason for that work-around is solely to allow yaboot to work
with an ext[34] root, I think my proposal solves the problem. As long as
yaboot only has to deal with ext2, you can remove the work-around any
time it’s convenient.
> Adrian
>
>> [1]
>> https://salsa.debian.org/installer-team/partman-ext3/commit/f87dc92157262de1ad8dd3f2343436f08271b4dc
Rick
