LFS and 64-bit hosts

Ken Moffat ken at linuxfromscratch.org
Mon Oct 1 15:24:08 PDT 2007


On Mon, Oct 01, 2007 at 01:43:58PM -0700, Dan Nicholson wrote:
> 
> As for the "for experienced and knowledgeable builders", I'm assuming
> that part of that knowledge would be that you don't need the LFS book
> to tell you what it can and can't do. What I want to guard against is
> the common new user who has no idea about the history of LFS. To me,
> the book should be completely clear about what is going to happen.
> 
> If we want to support another platform, it should be fully supported
> in the book. No relying on unstated "in the know" exceptions.
> 
 I don't have a problem with what people want to support, but I have
seen some discontinuity over the years between what the book
officially supported, and what platforms people were actually using.
Don't forget that people have built LFS on s390 (31 bits is enough
for anybody ;-), admittedly without needing *much* help.

 It's possible to say that LFS supports x86, or x86 and x86_64, or
whichever, but people will try to build on other architectures.
Sometimes, these are supported on clfs, other times they are not.  I
don't want to point everybody else at clfs if it is unlikely to
help (cross-compiling on new architectures tends to throw up new
problems).  So, even if the book remains at x86-only, I think the
notes on the toolchain and the linker are useful things to keep.

 From my viewpoint, it isn't about formal support (even on x86 there
is no guarantee that we can solve a problem) so much as providing
whatever help/tools/articles we can to enable people to get LFS
running on more machines.

>  I was thinking the Audience page would be appropriate, but I'd
> never read that until today :)
> 
 Appropriate, yes.  Likely to be read, no - I'v read that page in
the past, but as a user I would be very unlikely to read it unless
this was my first build, so if I'd already built for x86 I probably
wouldn't remember that that was the only supported platform .

 People omit everything which doesn't seem critical to their build,
with the result that they can go a long way before it all falls
apart.  Anything that improves their success-rate, or reduces their
pain, is good - even though failure and the pain of following
inappropriate instructions can sometimes assist the learning
experience.
> 
> I was sort of hoping just the top-level LFS page would take care of
> it, but I suppose it's needed in the version specific sections
> instead. Especially if/when x86_64 support is added.
> 
 Assuming that x86_64 will be added (ooh, a nice warm winter with
bootloader and usability («whaddya mean, I can't use flash plugins»)
flamewars, and maybe a bit of endianness heat ;-), we can revisit
the wording at that point.

 Joking aside, perhaps we should be thinking about supporting apple
on intel - some are 64-bit, I think, but they all have at least a
different bootloader.  Of course, if nobody has the hardware we
might be reliant on a 'contrib' chapter.  No, I'm not sure I like
that idea.

ĸen
-- 
das eine Mal als Tragödie, das andere Mal als Farce



More information about the lfs-dev mailing list