Re: GRUB fix for 1.1: Preferred approach for default grub version

[Tom Zanussi <tom.zanussi@intel.com>]

On Mon, 2011-10-03 at 13:40 -0700, Darren Hart wrote:
> 
> On 10/03/2011 12:07 PM, Tom Zanussi wrote:
> > On Mon, 2011-10-03 at 12:01 -0700, Richard Purdie wrote:
> >> On Mon, 2011-10-03 at 11:48 -0700, Darren Hart wrote:
> >>> When grub2 was merged int oe-core, it became the default (rather than
> >>> being explicitly requested by the 64b meta-intel BSPs). This breaks the
> >>> live install on 32b systems, like the n450 - installing grub v1.99 with
> >>> a grub v1 config.
> >>>
> >>> We have a couple options on how to fix this.
> >>>
> >>> 1) Set PREFERRED_PROVIDER_grub="0.97" in all 32b x86 machine configs
> >>> 2) Set DEFAULT_PREFERENCE="-1" in the grub_1.99.bb recipe which would
> >>>    affect all of oe-core.
> >>> 3) Some alternate approach which would be the equivalent of
> >>>    DEFAULT_PREFERENCE, but only for meta-yocto BSPs or the poky
> >>>    "distro".
> >>>
> >>> Given the time frame, I'd prefer to agree on the approach before writing
> >>> and testing patches. Preferences?
> >>
> >> I thought I'd been assured grub2 worked fine for our 32 and 64 bit
> >> platforms? :/.
> >>
> >> The correct question here is why doesn't grub2 work on 32 bit platforms?
> >>
> > 
> > I did test it with both 32-bit and 64-bit BSPs in meta-intel, so it
> > should work for the n450 too.
> 
> I checked out edison in poky in meta-intel and built core-image-minimal
> for n450. By default it builds GRUB 2, which after install boots to a
> GRUB prompt with no menu.
> 

I'm not even getting that far with edison - it fails at 'Searching for a
hard drive'.  I'd be happy to have it fail with no menu.

Can you send me the commit ids for meta-intel and poky that actually got
you that far, and if you could a serial capture of the install?

Thanks,

Tom

> Using the PREFERRED_PROVIDER method above with the edison branches
> results in a GRUB 1 menu after install.
> 
> --
> Darren
> 
> > 
> > Tom
> > 
> >> The answer might be "we need for time than is available in the 1.1
> >> release to fix it" but I'd at least like to see the answer to the
> >> question before we start with workarounds.
> >>
> >> Cheers,
> >>
> >> Richard
> >>
> > 
> > 
>