Martin suggested: > > The way partconf mkreiserfs will cause mkreiserfs to ask questions, > > which will freeze partconf. > > Yes, I've been thinking about this. I think I will introduce templates > partconf/mkfs/<type> for this. Something like: > > Template: partconf/mkfs/reiserfs > Type: string > Default: %s <flags go here> %s > Description: internal thingy > > Where the first %s is the mkfs program and the second is the path to the > device.
Why not just make a policy that mkfs.* on d-i must all operate noninteractively (with the possibility of using stdio for debconf if they really need to ask a question), with a single argument that is the device to mkfs. Then mkfs.reiserfs would be a wrapper that calls the real binary, passing it any necessary flags. Or the binary itself could be modified in the udeb. Either way, no debconf database (abuse?) is needed, just some quick modifications of the udeb for reiserfs. I think this needs to be fixed soon. -- see shy jo
signature.asc
Description: Digital signature