[arch-releng] btrfs fs can now be used "directly" in AIF

[C Anthony Risinger]

On Tue, May 17, 2011 at 3:18 PM, Dieter Plaetinck <dieter at plaetinck.be> wrote:
> We've discussed in detail why we should discourage users from doing this
> (although I actually only heard Anthony raise his concerns) - see
> http://mailman.archlinux.org/pipermail/arch-releng/2010-December/001341.html

jeez i yammer on a lot :-)  i ended up changing the structure up a bit
anyway, and mammoth waves of responsibility have prevented me from
finishing that hook :-( ... in time.

> It's turning out impractical to expect any kind of support for such advanced btrfs schemes (where we somehow enforce or even just support subvolumes) in AIF.
> So for now I just allow users to use the btrfs directly.
> See commit https://github.com/Dieterbe/aif/commit/281b49fd3dd0984805a9dc435d1a2c36c675eb4d
>
> This will at least enable people to actually use and test btrfs.

nice.  it should be ok -- even though the root subvol cannot be
snapshot, the biggest issue was that even trying to _move_ or reflink
the individual directories (/usr, /lib, etc) into an empty subvol
triggered a full on copy operation ... Chris has expressed that this
is a bug and would be fixed, but i haven't tested it recently to see
if it was taken care of.

i think i'll just warn the user on every boot for a few seconds that
they have dead directories in their btrfs root -- it doesn't start to
really be dead space until the reflinks start to break en mass (large
upgrades/etc).

C Anthony