Re: forwarded message from Thierry Laronde

[Thierry Laronde]

On Mon, Jul 28, 2003 at 05:42:06PM +0200, address@hidden wrote:
>       From: Graeme Vetterlein <address@hidden>
> 
>       AB> (More precisely: The four primary partitions are always numbered,
>       AB> and always numbered in table order.  The logical partitions are
>       AB> numbered in chain order.  If the chain zigzags over the disk,
>       AB> then chain order will differ from sector order.
>       AB> Neither Linux nor DOS has any objection against zigzagging chains.)
> 
> 
>       The original problem is that W2K would not boot.
>       The boot.ini file on the W2K filesystem said:
> 
>       [boot loader]
>       timeout=20
>       default=multi(0)disk(0)rdisk(0)partition(1)\WINNT
>       [operating systems]
>       multi(0)disk(0)rdisk(0)partition(1)\WINNT="Microsoft Windows 2000
>       Professional" /fastdetect
> 
>       ...I'm no W2K expert, but that partition(1)  does not sound
>       like 4 or 3 ...I think when W2K says partition(1) it means
>       'the first non-empty partition' ... thus when extra partitions
>       are added it is no longer the first non-empty entry.
> 
> I agree entirely. From Q102873:
> 
>     * W is the partition number. All partitions receive a number except
>     for type 5 (MS-DOS Extended) and type 0 (unused) partitions, with
>     primary partitions being numbered first and then logical drives.
>     NOTE: The first valid number for W is 1.
> 
> Thus, adding partitions in a formerly unused slot will change
> Windows partition numbers.

Since with the different tries that you have made Graeme and the
different guesses (more or less wrong) that I have made I was a bit
lost, I have reread your first post and confer the GRUB source code.

In fact, [correct me if I'm one more time wrong] before "fixing" the MBR
grub attempted to boot but failed, and this is (after rereading the
first post) after the "fix" that you discovered that grub was booting
when asked to boot (hd0,1) that is the second partition (conform to the
second slot occupied by Windows partition after fixing).

So, contrary to what I thought, you didn't manage to boot from (hd0,1)
before the new MBR has put the partition in the new order ("logical"
ones if you want) and this is confirmed by the source code : no GRUB
numbers the primary partitions the same way as the others : in the order
of the partition table.

But, with the explanation given by yourself (Graeme) and Andries
Brouwer, the problem initially was that you have the obligation to hide
the newly created partitions before attempting to boot Windows since it
has put itself last but wants to be first.

Something like:

hide (hd0,1)
hide (hd0,2)
rootnoverify (hd0,3)
chainloader +1

should have worked. The second solution was to change the boot.ini to
partition number three, since there is still an empty one that is not
taken into account (while Linux and GRUB takes it into account).

So, I correct again myself : GRUB accessed (hd0,1) but after the fix,
that is GRUB numbers the partition the same way as others. 

I'm perhaps at fault, but GRUB is not, GNU/Linux neither, the installer
of Redhat in part because it has not circumvented Windows, but Windows
--- sorry Andries! --- is at fault since putting itself last and
installing itself as if it will be first is typically trying to create
a problem that will be put under the responsability of the newly
installed OS.

Thanks, I have learned things.
-- 
Thierry Laronde <address@hidden>
Site Debian Francophone (aka SDF) : http://www.debian-france.org/