On Mon, 25 Jul 2005, Gerard Beekmans wrote: > Matthew Burgess wrote: > > Additionally, of course, cross-lfs is to be > > seriously considered at this point. I've not looked at Jim, Ryan, > > Opinions seem divided on this one. Should cross-lfs become part of the > mainstream book? In other words, will every LFS'er be building according > to the cross-lfs (toolchain) methodology even if they don't require it? > It seems a bit too much. >
Over the years, people have wanted to run LFS on different hardware, and many have succeeded. But, it would be nice to offer a reliable source of advice for the platform-specific variations. The hard part of cross-lfs is that it answers deeper questions along the lines of "I've got _a_ machine running a 'nix system, I'd like to put linux on this very different new box". My experience of cross-compiling is that there's a lot more to go wrong (and little chance of spotting many errors until you start to build natively), not least strangenesses in the configure script. Some of these oddities are already documented and worked around in cross-lfs, but there are others (e.g. I couldn't get coreutils to build df in the target phase - only noticed because my scripts use df to calculate space). Sometimes, cross-compiling is the best way, but I haven't yet seen any evidence that it's the best way to build for the same host already running linux. Ken -- das eine Mal als Tragödie, das andere Mal als Farce -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
