On Thu, Mar 02, 2006 at 06:30:19PM +0000, Philip Armstrong wrote: > On Thu, Mar 02, 2006 at 07:04:31PM +0100, Geert Stappers wrote: > > > http://support.microsoft.com/kb/308424/ > > > > What I read from it, is that it has a "another disklabel" > > I now can imagine why partmen fails, but _I_ have no clue how to fix it. > > Indeed. Why MS couldn't just use Intel's GPT disklabel instead I have > no idea. NIH syndrome presumably. > > If there was a way to get partman to not use parted to detect the > partitions, but to just use the existing kernel detected ones in
What about fixing the parted support for these partition tables instead ? Should be rather trivial if you only want detection to work. > /proc/partitions (which of course would not be editable) then that > would at least allow pre-existing partitions to be used by the > installer. Perhaps partman is too closely intertwined with parted for > this to be possible -- I haven't looked. Partman is too intertwined with itself :) It is probably possible, but partman is a bunch of spagethi code which is difficult to deal with, well at least for me. I believe partman does already do its own stuff for some partition formats or filesystems, and in any case it does its own for lvm/raid. I guess it should be easy enough to create a partman-kernel or whatever plugin, which would allows read-only access to partitions the kernel knows about, and which would be a fallback when other method fail, instead of proposing to erase the whole disk like it happens now, like the case where the guy had a minimac and libparted issues a warning, and d-i/partman proposed to overwrite a new partition table because of that. This is indeed an important feature, not sure who is handling partman right now though. I would say it is even an RC bug, probably of critical severity, as it may cause unrelated data lose, but i will let other decide on that. Friendly, Sven Luther -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
