[Clfs-support] 64 bit Source machine

[Joe Ciccone]

Ken Moffat wrote:
> On Tue, Mar 04, 2008 at 10:07:03PM -0500, Joe Ciccone wrote:
>   
>> We probably only see this for i?86 because this seems to be the only
>> time where people try to chroot. On most other hardware you'd typically
>> reboot. This really only needs to be addressed under i?86, mips32,
>> sparc32, and possibly ppc? I've thought this has needed to be in the
>> book for a while aswell, I just have never gotten around to making the
>> change.
>>     
>  As an inveterate chrooter, if I want to run straight ppc on my G5
> (ppc64) I use 'linux32', so I suppose the module would do the job -
> but, using only 32-bit userspace on ppc64 still needs a 64-bit
> kernel (and therefore a cross-compiler), so it isn't really supported
> - and now I've got a real 32-bit ppc I don't bother firing up the
> heat-intensive G5 for non-multilib.
>
> ĸen
>   
This works for x86 too. The reason I usually suggest the uname_hack
kernel module is because some older versions of distros didn't support
linux32, clfs-1.0.0 doesn't have the tools. I don't remember which
version of the kernel implemented personalities either. I can tell you
that on my latest development build, right after the upgrade of
util-linug-ng, setarch/linux32/linux64 are now included by default. The
uname_hack is a more general solution, the down side is that it effects
the entire system instead of just the chroot. Both of these methods
could and probably should be covered in the book.