problem with 64 bit builds on ppc64

[Jasmine Iwanek]

64 bit builds fail threads checks in chapter 6 originally mentioned on lfs-dev:

​http://article.gmane.org/gmane.linux.lfs.devel/15036

changing in chapter 5.4.1: case $(uname -m) in

x86_64) mkdir -v /tools/lib && ln -sv lib /tools/lib64 ;;

esac to:

case $(uname -m) in

x86_64) mkdir -v /tools/lib && ln -sv lib /tools/lib64 ;; ppc64) mkdir -v /tools/lib && ln -sv lib /tools/lib64 ;;

esac

and chapter 6.5: case $(uname -m) in

x86_64) ln -sv lib /lib64 &&

ln -sv lib /usr/lib64 && ln -sv lib /usr/local/lib64 ;;

ppc64) ln -sv lib /lib64 &&

ln -sv lib /usr/lib64 && ln -sv lib /usr/local/lib64 ;;

esac

to:

case $(uname -m) in

x86_64) ln -sv lib /lib64 &&

ln -sv lib /usr/lib64 && ln -sv lib /usr/local/lib64 ;;

ppc64) ln -sv lib /lib64 &&

ln -sv lib /usr/lib64 && ln -sv lib /usr/local/lib64 ;;

esac

should resolve the problem.

[bdubbs@…]

This is the wrong place for this. First of all LFS-6.5 is over 5 years old. Second, LFS doesn't support ppc. It is only for x86 and x86_64. Did you want cross-lfs?

[Jasmine Iwanek]

Urm what?

I never mentioned LFS 6.5, I'm aware that LFS doesn't support ppc currently, but the fixes above would bring it significantly closer to being able to support it.

Chapter 6.5 != LFS 6.5.

If we dont want to provide support for ppc that's fine, but it's such a tiny change that I dont see why it shouldn't be added.

[Jasmine Iwanek]

Addendum:

Tests were done on LFS 7.5.

[ken@…]

Enjoy the issues along the way ;-) My last 64-bit build of LFS on ppc64 crapped out - I don't recall why, perhaps this would have solved it. A few more comments follow :

I proceeded by using linux32 to build 32-bit userspace, then separate versions of pass 1 binutils, gcc to build kgcc in /opt and always cross-compile the kernel. At that time, loadkeys from kbd-2 wasn't working - there was a fix for BE machines which went into kbd in approx the last week and hopefully will fix that one.

My G5 fan-heater is a desktop single processor 1800/900MHz with only 2GB RAM. I have _sometimes_ managed to build 32-bit recent versions of firefox, by passing ac_add_options --target=ppc (I think that 64-bit ppc firefox might be broken anyway, not sure), but the last time I tried it failed to build after between 5 and 6 hours and at that time debian was still using an old version (i.e. they hadn't fixed it either).

Summary - it will undoubtedly be an interesting journey, but well O/T for LFS.

[bdubbs@…]

My mistake on LFS-6.5.

However we don't want to mention ppc anywhere. That just opens a can of worms.

If you want, I can change the ticket from invalid to wontfix.

[Dylan Morrison]

Throwing in my 2 cents as the person who is actually trying to build the ppc64 system. Jasmine has been helping me along.

The reason I'm not using cross-lfs is well, I'm not cross-compiling, I'm compiling on the actual target system. This was always an experimental thing anyway, so I'm not chuffed that you guys don't want to support it, just figured I'd talk about issues I come across because they could potentially show underlying portability issues in general.

[Jasmine Iwanek]

Bruce: wontfix would be more appropriate, I totally appreciate that you don't want ppc in the main book, could I at least request a branch in svn to commit PPC stuffs to?