Cross-LFS build method
jhuntwork at linuxfromscratch.org
Thu Jun 16 14:05:48 PDT 2005
It was suggested on IRC that instead of our current cross-lfs approach
(which is to cross-compile the temporary tools and build the final
system natively), that we instead simplify it a bit and build the
temporary tools on the host arch and cross-compile the entire final
system. This would put off the issue of getting the built system to the
target arch until the end of the book, and then we'd have a complete
system, and could *hopefully* use one official method for getting that
system onto the target machine.
One drawback is that we'd be cross-compiling every package (which we
don't currently do) and some would need to be hacked a bit to get that
to work. However, it can be done, and it has been done.
Another drawback would be all the work to bring the current cross-lfs
books into shape - but that's just part of the job, eh? ;)
On the surface, though, I like this suggestion for a couple of reasons:
1) If you chose to build on the host system because of its speed, you
get to use that speed for the entire build.
2) Doesn't break the book in the middle with a 'if ; then' choice.
Is there something major that's wrong with this suggestion that I'm not
seeing at the moment? Anyone else think of advantages or disadvantages?
More information about the lfs-dev