[vbox-dev] Catastrophic bug in discarding of snapshots

[walt]

Hi.  I'm tracking vbox HEAD via svn and I've been bitten by this bug
a couple of times now.

When I have several snapshots taking up disk space I delete the oldest
one first (the one at the top of the snapshot list) because I have at
least one newer one that I know is good.  Everything seems to go fine
until I open vbox the next time, and now the virtual disk "with UUID
foo is unavailable -- file can't be found".

The backing file for the virtual disk is still there, but evidently
with a UUID number that's wrong, or maybe the file is corrupt, I
can't tell which.

So I delete the virtual machine and go to create a new one using
the same virtual disk, but when I do that, vbox just hangs forever,
using no cpu cycles.  When I kill vbox it leaves two stuck processes
behind, VBoxXPCOMIPCD and VBoxSVC --automate, which I also need to
kill -9 before I can restart vbox.

So far I can't find a way to reuse the disk image I still have.

I know there is a bug in qemu's qcow2 code that might account for
this problem -- if you are using that code.  Are you?

Thanks for any info you can give me.