usrmove problem

[Tom H]

On Tue, Jan 31, 2012 at 3:25 PM, Adam Williamson <awilliam at redhat.com> wrote:
> On Tue, 2012-01-31 at 09:26 -0800, Adam Williamson wrote:
>> On Tue, 2012-01-31 at 10:44 -0500, Tom H wrote:
>>
>> > >> Also, the problem is in your initramfs, not the kernel itself.
>> > >>
>> > >> Harald?
>> > >
>> > > I was about to follow up. I tried rebuilding the rc6 kernel's
>> > > initramfs but there was no joy. I then managed to boot with the rc6
>> > > kernel using the rc4 initramfs.
>> > >
>> > > I've unpacked my the rc4 and rc6 initramfs's to take a look but have
>> > > had to go back to real work...
>> >
>> > Unpacked initramfs-3.3.0-0.rc1.git4.1.fc17.i686.img to "~/rc4" and
>> > initramfs-3.3.0-0.rc1.git6.1.fc17.i686.img to "~/rc6":
>> >
>> > [root at localhost ~]# find . -name 'libc.so.6'
>> > ./rc4/run/initramfs/lib/libc.so.6
>> > [root at localhost ~]#
>>
>> The initramfs is generated on-the-fly when the kernel is installed. So I
>> suspect the bug here is correctly described as 'When installing any
>> kernel package after doing the /usr move updates, the generated
>> initramfs will not contain /lib/libc.so.6' - i.e. the actual kernel
>> package in question doesn't matter, it's rather that any attempt to
>> generate an initramfs after doing the /usr move will fail. This is
>> obviously a significant problem, if I'm correct.
>
> Has anyone else who's done the /usr move tried to install a kernel after
> the move? If so, does it work, or do you hit the same problem Tom hit?

I hope that no one else hits it and that it's just some bad juju on my
side but it's happened with two different installs. I've also just
duped my VM and run dracut for the bootable kernel - and it's now
unbootable.

[root at localhost ~]# rpm -q dracut
dracut-014-77.git20120126.fc17.1.noarch