Re: [FYI] debian-installer for SONY PS3
Hi there,
Thanks for seeing my sites.
Joey Hess wrote:
> If Takeshi is interested in commit access to the d-i subversion repository,
> I feel that this easily justifies it. Then he could keep his patches in
> a branch and work on merging things in as appropriate.
>
> I've taken an initial look at the patch, and I feel the following hunks
> could be applied as-is without problems to d-i trunk, although we might
> prefer to wait until there is a post-etch branch to avoid
> destabilisation. Not that I feel there's much risk, most of these changes
> are minor.
Agreed. It's desirable to get things in the upstream repository
to provide other developers easy way to build a working image
and track changes.
If granted to have a dedicated branch, I'll keep it there.
> I don't understand the following changes:
Well, it contains ugly workarounds only for the experimental
release. They shouldn't be in the trunk.
> --- installer/build/pkg-lists/monolithic/powerpc.cfg (revision 43755)
> +++ installer/build/pkg-lists/monolithic/powerpc.cfg (working copy)
> @@ -8,5 +8,5 @@
> dhcp3-client-udeb -
> eject-udeb
> # Used by yaboot-installer
> -parted-udeb
> -os-prober
> +parted-udeb ?
> +os-prober ?
>
> As far as I can see, these should be available for the ps3 image too,
> so no need to mark them as optional?
Somehow d-i build fails in getting os-prober udeb, so I made it
optional.
> +++ packages/partman/partman-base/definitions.sh (working copy)
> @@ -771,6 +771,8 @@
> enable_swap () {
> local swaps dev num id size type fs path name method
> local startdir="$(pwd)"
> + # XXX
> + return 0
> # do swapon only when we will be able to swapoff afterwards
> [ -f /proc/swaps ] || return 0
> swaps=''
>
> If there's no swap on this machine, there are probably better ways to
> handle that. :-)
It's workaround for failure in processing swap partitions in
/lib/partman/mount.d/70basic. swapon $fs fails when the swap
already exists and is activated at prior mkfs phase.
It seems that non-devfs style device names like /dev/sda1 are
passed to mount.d scripts and it doesn't match any lines in
/proc/swaps (/dev/scsi/host0/bus0/target0/lun0/part1).
Probably it's ps3pf's kernel or something specific problem
because it didn't occur in i386 d-i.
> --- packages/debian-installer-utils/apt-install (revision 43755)
> +++ packages/debian-installer-utils/apt-install (working copy)
> @@ -30,7 +30,7 @@
> export DEBIAN_FRONTEND
>
> ERRCODE=0
> -log-output -t apt-install chroot /target apt-get -y --no-remove install $@ < /dev/null || ERRCODE=$?
> +log-output -t apt-install chroot /target apt-get -y --force-yes --no-remove install $@ < /dev/null || ERRCODE=$?
> chroot_cleanup
>
> if [ "$ERRCODE" != 0 ]; then
>
> This change seems odd, especially since the patch does not use apt-install.
To avoid interruption when accessing to my experimental apt
repository (http://ps3.keshi.org/debian/).
--
YAEGASHI Takeshi <t@keshi.org> <yaegashi@debian.org>
Reply to: