On Tue, Mar 16, 2004 at 06:40:39PM +1000, Andrew Pollock wrote: > Yeah, I'd like to have a fiddle with getting this support into d-i in time > for Sarge if possible. I'm not currently aware of a mainstream Linux > installer that does this, so we might score some points come review time... Some nice dangerous/complex things to do would be to mount the FS if it's something common like Reiser, XFS, Ext2/3 (basically all the FSes we support in partman) and see if they have /boot/grub or /etc/lilo.conf, mark them as kernels for the new bootloader in a generic way (see below), and use the appopriate root partitions etc. We can mount vfat, check for a WINDOWS/System32 directory, and mark that as an 'other' OS... the list goes on. Oh, and if the type is UFS, we can assume that the partition has been appropriately sliced internally for FreeBSD, and mark that also as other.. the list goes on. It's just a matter of seeing what's feasible and safe. (ie, what happens when you mount a trashed partition?) The test could be done very modularly just as they are done in packages like lintian. > Can anyone with better internal knowledge of d-i provide some ideas on which > udeb(s) would require some playing with? I'm assuming whatever installs grub > for a start, unless something can be done more centrally (which I doubt). I'm thinking that we can use a common format in a separate package to build lists of partitions and their corresponding OSes, then each *-installer package for bootloaders can process that output as necessary to write appropriate configuration entries. It should not be too hard once we have the means of obtaining that data. > regards > > Andrew -- Joshua Kwan
Attachment:
signature.asc
Description: Digital signature