grub-0.97 compile issues

Ken Moffat ken at
Tue Nov 1 12:58:47 PST 2005

On Tue, 1 Nov 2005, Stan Sander wrote:

> Basically I have been unable to get the configure for grub-0.97 to run 
> successfully.  I have hit two issues so far and want to make sure I'm not 
> shooting myself in the foot.  Both of these have seemed rather easy to fix, 
> but I'd appreciate your opinions to make sure I'm not setting myself up for 
> more trouble ahead.   Here is the first  problem:

  You are building x86_64-64 (pure64) ?  From the overall index.html page 
where you select your architecture: "This is experimental. Bootloaders 
do not work."

  I surmise that it _isn't_ as simple as changing -m64 to -m32, and your 
later problem with objcopy tends to confirm that.  For the moment, I 
guess you either have to use a pre-installed grub from a 32-bit or 
multilib system, or you have to build bin86 (still needed the patch last 
time I checked) and lilo.  All bootloaders are 32-bit at the moment, but 
lilo calls on bin86 to do its assembly.

> configure:3918: cmp -s conftest.old conftest
> configure:3921: $? = 1
> configure:3932: result: no
> configure:3935: error: GRUB requires a working absolute objcopy; upgrade your 
> binutils

> Ok, so I think the answer here is to set 
> grub_cv_prog_objcopy_absolute=yes in the config.cache file since we do 
> not have a cmp program that can handle 64 bit executables while I'm 
> still booted in the 32 bit environment.  Should I do this and then 
> change the configure options to point to a config.cache file?

  The problem here is the non-working objcopy, which is caused by not 
having the 32-bit objects for it to link to.  Pretending that you do 
have an acceptable objcopy is unlikely to get you to a bootable system. 
FWIW, you should have a perfectly good cmp installed in section 6.10, 
and indeed the cmp on your 32-bit host system is almost certainly good 

  das eine Mal als Tragödie, das andere Mal als Farce

More information about the cross-lfs mailing list