Re: partman problems (sparc and elsewhere)
On Wed, Dec 01, 2004 at 09:13:47PM +0200, Anton Zinoviev wrote:
> On Wed, Dec 01, 2004 at 03:36:14PM +0100, Sven Luther wrote:
> > > Or partition we are mounting without formating if its type doesn't
> > > correspond to the file system it contains.
> > Well, i don't think so, not automatically at least.
> (Lib)parted repairs the wrong partition types in all cases. In some
> cases partman will not write the partition table but if it has to for
> some reason then the partition type will be changed.
> > > At the time when some partition is being formatted (or simply sceduled
> > > for mounting) the partition table may already contain physical volumes
> > > of activated volume groups.
> > So what ? a raid or LVM partition is simply a partition with a given flag. The
> > fact that you modify the other partition on the partition table should in no
> > way affect any raid or lvm partitions you have,
> The kernel will complain that it can not reread the partition table.
> This complain is not dangerous and partman can probably safely hide it
> from the user but this already is not nice.
> Moreover the confirmation dialog contains a list of the changed
> partition tables. If we decide to write all partition tables to the
> disk just because they contain some formatted partition this means that
> they have to be included in this list. In most cases this list will
> contain all partition tables and because of that it will be less
I don't follow this. You read the partition table, modify the type for one
partition type, write it back. This should not change the rest of the
partitions, or it is a RC parted bug, and your duty is to fill a bug report
> I think I can take the partition type from the structures of parted but
> I wonder the format of these structures can change in future versions of
not withtout an api change. And anyway, we don't care, just automatically do a
set_system_type when formatting, and everything should be fine.