On Fri, 2003-06-06 at 04:39, David Laight wrote:
> On Thu, Jun 05, 2003 at 07:07:02PM -0400, felix zaslavskiy wrote:
> > One particular problem i had couple of days ago.
> > I was installing one a not so old computer with a new disk and the
> > question about BIOS geometry confused me because it claimed the disk had
> > less cylinders then what it actualy says on the disk. I know all about
> > that the BIOS geometry is fake now but for newbie people or who had not
> > read the installation guid in detail they would not know that.
> 
> That section does give far too much scope for shooting oneself in the foot.
> I fixed the equivalent code in fdisk so that it almost always gets the
> correct geometry, won't allow the user to input illegal geometries and
> gives resonably likely defaults (the true geometry is never right!)
>  
> > I said geometry was incorrect and picked the real geometry.  Ofcouse
> > then nothing worked.  The sysinstall continued and let me select the
> > partition sizes but the actualy write to disk failed.
> 
> I would havethought the writes suceeded, although you could easily
> be left with an unbootable disk.
I dont exactly remember what failed i know things didnt work, possible
the write succeded but the install of bootblock failed. I could  not
even run install a second time because it was reporting the bogus
partition from before so i did   dd if=/dev/zero of=/dev/wd0c  things
worked ok again.

> 
> The more recent netbsd boot code is less likely to be unbootable if the
> bios geometry is wrong.  It will use LBA reads in that case...
> 
> 	David