Re: [PATCH] Fix when installing on pationless but partionable medium

[Colin Watson]

On Sat, Jul 18, 2009 at 09:22:11PM +0200, Robert Millan wrote:
> On Sat, Jul 18, 2009 at 09:00:36PM +0200, Vladimir 'phcoder' Serbinenko wrote:
> > On Sat, Jul 18, 2009 at 8:42 PM, Robert Millan<address@hidden> wrote:
> > > I'm not sure there's much we can do about this.  Using heuristics sounds 
> > > like
> > > it will make the solution worse than the problem.  I don't care much about
> > > Microsoft filesystems, but I'd hate to see GRUB fail on a completely sane
> > > ext3 inside msdos label because it happened to look like FAT in raw disk 
> > > at
> > > the same time.
> > 
> > The approach proposed by Collin avoids such problems since correct
> > pc_partition_map is always detected as such.
> 
> I haven't looked at the source code, but what he said is we can determine if
> an MBR is valid by checking the bootable flag, and this is not always so.

If the bootable flag is neither 0 nor 0x80, then neither libparted nor
the Linux kernel will understand it as a DOS partition table. Is it
really all that helpful for GRUB to attempt to do so?

I've never heard of false positives with the libparted/Linux checks. Do
you have real-world examples of them failing?

-- 
Colin Watson                                       address@hidden