On Fri, 2007-12-07 at 23:28 -0800, mdda wrote:
> What does the kickstart config you're using for the install
look like?
My idea is to fix up everything, but let the user make changes - and I've
found a way to make it work.
First : The (unchanged) Kickstart file :
http://www.nabble.com/file/p14226119/eee-live-install.ks eee-live-install.ks
Some pertinent points :
* I'm forcing the install partition to be ext2 (the rumour is that this will
wear down the Eee's internal flash least)
Is the partition on your live image ext2 also? If not, then I can
definitely see that leading to some of the weirdness you're describing.
* I found that the sizing of the (faked) partition within the ISO
(at
4096Mb) was too large for the Eee's 3.8Gb flash drive - and that the
minsizing thing wasn't behaving somehow (my ISO-building machine is fully
yum updated). So I fixed this hurdle by forcing the the size to (like)
2048Mb in livecd-creator.
davidz said something along these lines, but I did an install on
Thursday or so into a virtual disk that was < 4 gigs without problems.
And even with swap. So there's something weird going on with the Eee in
particular, but I'm not sure what. :-/
* To be on the safe side, I commented out the call to
self._resizeRootfs(anaconda, wait) in _doFilesystemMangling (this may have
been overkill, but it occurred to me that 'resize2fs' wouldn't be happy
resizing a mounted ext2 partition, so that might be causing problems)
The resize is done without the filesystem mounted, so this shouldn't be
needed
* Most significantly (?), since the writing of $SYSIMAGE/etc/fstab
seemed to
be failing because $SYSIMAGE/etc wasn't there, I brute-forced the following
inside doPostInstall in /usr/lib/anaconda/livecd.py :
def doPostInstall(self, anaconda):
self._doFilesystemMangling(anaconda)
_doFilesystemMangling ends up remounting the filesystems, so they should
be mounted at that point
Jeremy