I don't believe what I just saw ...

[Adam Williamson]

On Fri, 2014-12-05 at 11:23 -0600, Ian Pilcher wrote:
> So after a bit of live CD jiggery pokery, I was able to run through an
> F21 RC5 installation.  I installed it in a separate logical volume,
> alongside my existing F20 installation.  /boot is a shared ext4
> partition on top of my IMSM RAID volume.  I still use GRUB legacy, so I
> told anaconda not to install a bootloader.
> 
> After the F21 installation completed, I was unable to boot back to F20.
> GRUB still worked, but the initramfs was unable to find the F20 root
> filesystem, which is an LV on top of a bcache device.
> 
> After a bit of investigation, it appears that the F21 installer rebuilt
> all of the F20 initramfs files on the shared /boot (and omitted the
> bcache dracut module).
> 
> Of all the anaconda weirdness I've experienced over the years, this has
> got to be the strangest thing I've seen.  In fact, I have a hard time
> believing the evidence of my own eyes.  Why would it do this?
> 
> Am I crazy?  Anyone else seen this?  Expected behavior?

https://bugzilla.redhat.com/show_bug.cgi?id=1074358

It is a pretty bad bit of behaviour, as I wrote in the bug I figure we
should either disallow shared /boot or try and filter the kernel list at
least a bit, for F22.
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net