[buildd-tools-devel] Testing of schroot 1.3 unionfs support
Tim Abbott
tabbott at ksplice.com
Sun Aug 2 18:08:03 UTC 2009
On Sun, 2 Aug 2009, Roger Leigh wrote:
> On Sat, Aug 01, 2009 at 04:12:23PM -0400, Tim Abbott wrote:
> > On Sat, 1 Aug 2009, Tim Abbott wrote:
> >
> > > However, it immediately segfaults for block-device chroots (in my case,
> > > source chroots for LVM snapshot chroots). I'll debug a bit later today.
> >
> > As a first pass, below is a gdb backtrace from this.
>
> It turns out this was a misplaced assumption which broke when we added
> support for chroot facets. Briefly, the session code expected that
> if a chroot said it would support session cloning, clone_session()
> would always return a non-null session object. However,
> union-supporting chroot types (block-device, loopback) whose basic
> types did not support cloning would only clone if union support was
> active. This is a regression due to the facet change, since the
> session facet is always present; previously it always derived from
> chroot_session, but only selectivey called its get_session_flags
> method.
>
> This is now fixed. Now, all chroots (except plain) always allow
> session cloning, and this is always supported when SESSION_CLONE
> is set.
>
> If you can try again, I've fixed this as well as the mount options
> quoting in your previous message.
I tried it again, and it worked (I also did the same tests I did
yesterday, and they all worked as well).
-Tim Abbott
More information about the Buildd-tools-devel
mailing list