Bug#473401: grub-installer: grub2 config for Windows partition does not boot
On Thu, May 01, 2008 at 02:46:58PM +0200, Frans Pop wrote:
> IMO the following should be checked _before_ we merge this patch:
> 1) does grub-probe produce the same output for the same input
Short answer is "no promises"; but I should ellaborate. Both methods are
basically flawed; they try to guess how the BIOS will number devices, which
can't be done reliably unless you ask the BIOS . Most of the time they
succeed, though, because all disks are part of the same bus (and there's
an "obvious" way to count them, shared by Linux and the BIOS).
When they're not part of the same bus, or when counting depends on races
during bus probing, both methods only succeed if you're lucky. Whether
they make the same choices or not is not really much relevant; only luck
 The long term solution to this problem will be to use UUIDs in GRUB
just like Linux does. GRUB 2 started doing this for the most relevant
disk accesses, although not everything is handled this way yet.
> 2) does it support _all_ the different device types that grub-installer
> supports, _including_ the recently added ones (vdX and xvdX) and for
> example multipath (/dev/mapper/mpath*); if not these should be added
> in grub-probe *first*
vdX and xvdX were added in #479056 and #456777. For Multipath you already
know about #483971 which is in need of some testing.
We also have one for IDA (my patch is at #483858, anyone can test?) and I2O
is supposedly supported but by looking at the code I suspect it might be
unhappy (please test!).
Not sure if there are others, but we didn't get reports about them (and I
don't like to add these blindly before we get someone to test patches).
> 3) does it also support gnu*, freebsd* and netbsd* host OS and, if not, does
> it return an error; if not we should keep that coded in grub-installer
These are supported.
<GPLv2> I know my rights; I want my phone call!
<DRM> What good is a phone call… if you are unable to speak?
(as seen on /.)