On 7 September 2014 07:58, Petter Reinholdtsen <p...@hungry.com> wrote:
> Control: tag 757818 + patch
>
> I've tested both '--yes' and '--wipesignatures n' in d-i now, and both
> work. With either, I am able to reinstall a Debian Edu profile
> without any problem. :)
>
> I am not quite sure which of these best fit d-i. I notice the Fedora
> people went with --yes in anakonda. Perhaps we should do the same?
> On the other hand, the old behaviour was to not wipe signatures, and
> to keep it we should add '--wipesignatures n'.
>
> Based on this, I recommend we use '--wipesignatures n' to keep the old
> behaviour. Here is a patch implementing the fix.
>
Heh, the old behaviour is exactly what we don't want =)
So what happens is that an "empty" block device is initialised, and a
new partition table is created, new partitions are created, and happen
to align as some old installation with matching default VG and LVM
metadata. We do not want to reuse it, but on the contrary wipe and
create a fresh. Thus I think we do want to use "--yes" option here.
Thanks a lot for digging into this.
Regards,
Dimitri.
> diff --git a/lib/lvm-base.sh b/lib/lvm-base.sh
> index 4e83d67..656893a 100644
> --- a/lib/lvm-base.sh
> +++ b/lib/lvm-base.sh
> @@ -477,7 +477,8 @@ lv_create() {
> lv="$2"
> extents="$3"
>
> - log-output -t partman-lvm lvcreate -l "$extents" -n "$lv" $vg
> + # Do not ask if signatures should be wiped, to avoid hanging the
> installer (BTS #757818).
> + log-output -t partman-lvm lvcreate --wipesignatures n -l "$extents"
> -n "$lv" $vg
> return $?
> }
>
>
> --
> Happy hacking
> Petter Reinholdtsen
>
>
> --
> To UNSUBSCRIBE, email to debian-boot-requ...@lists.debian.org
> with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
> Archive: https://lists.debian.org/20140907065847.ga14...@ulrik.uio.no
>
--
Regards,
Dimitri.
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
