Re: [Surprise-dev] GNU Parted and Partition Surprise

[Andrew Clausen]

On Mon, Jul 23, 2001 at 02:16:01AM +0200, Jan Kratochvil wrote:
> On Mon, Jul 23, 2001 at 09:54:45AM +1000, Andrew Clausen wrote:
> ...
> > And you want to swap the order.  You need to do something like:
> > 
> >     AAAAaaBBBBbb
> >     AAAA  BBBBbb
> >     AAAABBBB
>                 ^^^^=fatal requirement!

Huh?

> >         BBBBAAAA
>         ^^^^=fatal requirement!

?

> >     BBBBbb  AAAA
> >     BBBBbbAAAAaa
> > 
> > Make sense?
> > So, you can't just commit to the final state immediately...
> > if you commit B to it's final state, it will clobber A.
> 
> Oh, sorry, OK, I got it now . So you need "md" device to rejoin and
> 'virtualize' the space (otherwise you would need "min(|A|,|B|)" of free 
> space).
> Such virtualization is done by our/Tom's General Converter. If you want it by
> kernel, use kernel features ("md")...

Oh, I see.  Right.  Interesting... never thought of that, very clever :)

> I don't want to dread you but the described functionality will be exactly the
> seme even with glib. :-)

I know, just someone else has solved the problem, not me.
Less code that *I* need to maintain.  Also, since it's "their job",
they'll probably have done it more elegantly than I can.

> Reference counting (for strong references), "destroy"
> signal disconnections (for weak references) are just the features of 
> glib/GTK+.
> No object implementation can simplify the principle of functionality.

Right.  But it provides the infrastructure.

> <divert>I was so kind that I took only the simple overlook, I haven't 
> explained
> you the partition-matching phase to synchronize object system state between 
> two
> Undo states. It has 3 layers of functionality, 6 cases for disk-space match,
> some temporary storage heaps etc. :-)))
> I only hope that I will never need to understand General Converter. (8+<=>)
> </divert>

I'm sure I'll have fun...

Andrew