So, I tried building LFS 6.2 on an Intel 945 board w/ a Pentium D processor... I can't even get through the initial toolchain because expect seems to want to find glibc 2.4 or higher, and the version used by the book is 2.3.x. So, naturally I attempt to 'integrate' the SVN version and the stable, only to find that the SVN fails as well, but for a different reason...

I cannot get glibc 2.4 or 2.5 to pass 'make check'. Both versions fail the cancelXX tests for NPTL... which of course leads to some more serious issues down the road, such as a broken gthread-2.0 library.

So, what I'm not understanding is why there's an apparent broken dependancy in the stable version of the book, and more importantly why I can't get glibc 2.4 or 2.5 to build without NPTL being broken?

Hi,

This all sounds a bit strange to me.

- Which chapter did give you the problem (you are probably referring to 5.7. Adjusting the Toolchain)?
- Which command gave the error?
- What is you base system and is up to specs?

The book (LFS 6.2) and all packages listed/installed are working and don't have dependency problems, the only thing that could be wrong: Your base system. I would advice using the LFS liveCD as a base system.

BTW: Mixing stable and SVN could give very strange results. One of them: dependency related issues.

There's not enough relevant information present to give you a more solid advice.

I'll go through the build process again and report the specific errors, and this most recent batch of problems was using the LFS LiveCD... Though, it's worth noting that the X86_64 LiveCD won't allow me to start a shell, despite the fact that virtually every prepackaged distro installs/runs their x86_64/AMD64 releases without issue (Debian Netinst has a problem detecting the Intel Pro100 type network adapter built in the 945 Express chipset.) So I'm using the x86 release, which doesn't bother me much... don't wanna have to deal with 64 bit compatability issues (no Flash ) anyway.

Something else I'm wondering, I built last weeks SVN on a P4 2.8GHz machine without problems, I would like to use the -march=prescott optimization when building the software for this machine (I was able to use '-O3 -march=pentium4' on that box for virtually all of the packages on the other one), but if I build on the P4 for use with the Pentium-D I'm having issues with I can't use the right optimizations... kinda defeats half the purpose of using LFS in my mind... Is there an 'easy' solution to this issue? I'd be happy to build on the P4 if I can move the stuff over.

Hi,

If this is the case, use the none 64 bit version of the LFS liveCD. Why do it the hard way....

You should not use any compiler optimization during chapter 5 (it's just a base for chapter 6). Chapter 6.1 talks about optimization and points to a hint. There's a possible risk when using optimization and the (small) benefits of using optimization don't outweigh the risks.

I don't agree. LFS is about learning linux, the way linux works/is set up and having a system that only has those things installed/running that you want. Being able to optimize a package when building it is nice (when it works) but is not one the goals of the LFS project.

My old system (now dead), failed alot of NPTL tests alongside others when building glibc. I vaguely remember a cancel something test failing too. Several gcc tests were failed also. Unfortunately it was a hardware problem. My hardware was slightly testy and wasn't exactly brand new (big chunky SocketA Athlon), and this may be the problem here. I became frustrated and ignored the errors every time and never had a problem with the finished system...