Re: Parted, IA64 and ReiserFS

[Andrew Clausen]

On Wed, Sep 19, 2001 at 01:36:02PM -0500, address@hidden wrote:
> > OK, well, that should work now, so I'm applying the patch... :)
> 
> Yea!

:)

> Actually, (and maybe this is better done in 1.5.x), I'd prefer to see things
> like the partition flags be kept in PedPartition rather than in each disk
> type as they need them.  Likewise, ped_partition_{get/set}_flag() functions
> in disk.c can then operate on the PedPartition members (after calling
> partition_is_flag_available which is still needed).  Removes two functions
> from each disk label type, and forces the *_write() functions to do as you
> think and sync those into their on-disk structs.

Hmmm, that might be cleaner.  However, there might be constraints /
policies about setting flags.  Eg: only FAT and NTFS partitions can
be hidden.

Also, I would prefer to have a reflective-type thing (like EVMS or
GObject) that allows interrogation of flags, rather than the fat
interface with have (provide an interface for all flags, but disable
parts of it)

flags should be "type-specific", IMHO.

> Each ped_partition_new() can clear all the flags, *_partition_{new,read}()
> would change the flags to match what the disk type needs, and *_write()
> writes them out.

I'm not sure a partition_read() method would sit too well.
(Perhaps you mean disk_read()?)

Anyway, I think it's better to sanity check before writing.

> > A bigger problem is gpt_create().  In 1.5.x, I changed the 
> > semantics for
> > this a fair bit, so you can create new partition tables "in 
> > memory" without
> > touching the disk.  (so, there should be gpt_alloc() in 1.5.x that
> > returns a blank table)
> 
> Naah, not harder really.  gpt_alloc() allocates the pgpt, agpt, and ptes,
> zeros them, and fills in the normal header stuff.

OK, cool :)

Thanks,
Andrew