dsd at laptop.org
Wed Nov 24 17:25:37 EST 2010
On 24 November 2010 08:26, Jerry Vonau <jvonau at shaw.ca> wrote:
> I've played around with the bootfw rpm, altering the install path to
> reflect where /boot really is on the filesystem while booted up. This
> has the advantage of being able to use rpm/yum to stage the upgrade of
> the firmware. Providing you have the power cable connected, once OFW
> looks in /boot it will look for bootfw.zip. I'm just putting the zip in
> the correct location to be found. I have a test re-roll of q3a50
> available from:
Unfortunately this isn't suitable for inclusion. At build time, the
partitioned layout is not setup, and it is not even known if the
resultant image will even be split into partitions. So this would
cause the build to fail.
Additionally, part of the olpc-update design requires the entire
crypto-signed filesystem to be available in one undivided place, so
the RPM must continue installing to the "unpartitioned area" with the
reliance that something else will handle partitioning later on.
This is all touched upon in /boot/README, and I'll document it more
detail (on the wiki) at some point soon.
The kernel RPM also faces the same situation.
And, in general OLPC advises against using yum and RPM on XOs. With
certain packages, doing this will override various tweaks put in place
by the build system. Sometimes the side effects will be subtle, and
other times they will be obvious.
It would be possible (but messy) to make bootfw and kernel RPMs "do
the right thing" when installed on live booted XOs, but it would be
fragile, and doesn't change the fact that general usage of yum/rpm is
likely to cause you problems at some point.
More information about the Devel