Sorry for late reply, still working through backlog... CC'ing because of that. On Thursday 23 November 2006 22:32, David Härdeman wrote: > >I think we have two options: > > > >1) Default to method "regular" > >If partman-auto/disk=/dev/* and partman-auto/method is not set, then > >default partman-auto/method to "regular". > > > >2) Error out if no method set > >If partman-auto/disk=/dev/* and partman-auto/method is not set, then > > show an error message and unset partman-auto/disk. > > > >I somewhat prefer 1) as it gives some backwards compatibility and 2) > >probably would mean adding a new error message which is undesirable > > for RC2. > > Said and done, patch attached for review. Hmmm. AFAICT the fact if method is set or not is currently one of the prime triggers to see if we want to do autopartitioning at all. That is why I suggested the additional check that partman-auto/disk should be preseeded with "/dev/..." as that will also indicate the intention to autopartition. I'd suggest moving the check partman-auto/disk up before the method check and maybe also checking that it has a valid device name. Hmmm again. I see that I made a wrong assumption that interactive auto-partitioning uses the same template as preseeded, but I now see that is not the case (p-a/select_disk versus p-a/disk). In that case, setting the default is fairly low risk and that means we could also just set the default in the template file instead of in the script. p-a/disk will then again be the primary trigger. It does make me wonder though at the hang reported in the initial report. Will try to reproduce that. Cheers, FJP
Attachment:
pgpPhhNaXXDYB.pgp
Description: PGP signature