Planning for Cross-LFS/Multi-Architecture 7.x Release

Matthew Burgess matthew at linuxfromscratch.org
Sat Apr 23 04:08:58 PDT 2005


Matthew Burgess wrote:
> Once a get a spare hour or so I'll take a look at this thread properly
 > to see what you lot have been so busy with :)

Firstly, thanks to everyone who participated in this discussion.  It's 
been a while since we've had such an in-depth, informative and well 
reasoned debate!

Secondly, there were some concerns regarding having to reboot into the 
target environment.  Note that this is a technical requirement because 
for the majority of host != target builds, the host won't be able to run 
binaries built for the target environment.  Obviously though, not 
everyone will be cross-building.  Therefore it would be prudent to 
provide the 'chroot' instructions where a reboot isn't required.  I 
believe the XML can be massaged into allowing the correct set of 
instructions dependent on whether or not the reboot is required. 
Failing that, we simply give folks a choice.  I realise this deviates 
slightly from the "linear" style that the book has followed up until 
now.  However, building LFS is complex as it is - even more so when 
considering the cross-building techniques.  I don't think asking the 
user a simple question of "is your host the same arch as your target 
box" is unreasonable.

I think the only other point of contention is the necessity of some 
additional tools (SSH, etc.) to facilitate building on some non-x86 
arches.  I believe these can be added by the maintainers of those 
arches, and surrounded by the necessary XML magic to prevent those that 
won't need them from seeing them.

Was there anything else I haven't addressed? If so please speak up now.

I think that the cross-lfs build method will offer benefits to all of 
us, in terms of preventing the host infecting our chapter 6 environment, 
and also by allowing the building of LFS on a wider range of 
architectures.  If the above two concerns are the only ones to come out 
of such an extensive change in the build method, I think we're doing 
alright :)

Regards,

Matt.



More information about the lfs-dev mailing list