Re: partman problems (sparc and elsewhere)

[Sven Luther <sven.luther@wanadoo.fr>]

On Wed, Dec 01, 2004 at 04:48:32PM +0200, Anton Zinoviev wrote:
> On Wed, Dec 01, 2004 at 02:25:02PM +0100, Sven Luther wrote:
> > 
> > Ah, no, it was a swap paritition previously, not an ext3.
> 
> Are you sure that it didn't contain ext3 signatures already? If it
> contained swap but partman didn't update the partition type then this
> must be some unknown bug.  (Well, it must be unknown for the present
> version of partman.  In the past partman would not update the partition
> type in this case.)

Mmm. Need to investigate, but anyway, the below should solve this reasonably.

> > > The solution (not nice) would be to write the partition tables
> > > unconditionaly.
> > 
> > Why is it not-nice ? It is a good solution, either that or have some way to
> > set it by hand.
> 
> Because:
> 
> 1. Partman may not write all partition tables - some of them may be for
>    example USB disk we are installing from

So what ? 

> 2. Consequently partman may try to write only the partition tables that
>    contain partitions that will be used somehow by the new Debian
>    (partitions with some file system, with swap or booting partition)

Well, the idea is to write the system type only for the partition we are
formating, not ? 

> 3. But if the user uses LVM or RAID partman has to write the partition
>    tables before it knows which of them will be used.  Latter they are
>    already in use and the kernel complains about changes in them

And ? You simply do the system flag writing at the same time you format the
partitions, provided they are not on a LVM or RAID device ? 

> Yes, these problems probably can be got round but not in a clear way.

Is the above not a clear way ? What would you find cleaner ? 

Friendly,

Sven Luther