Analy pure LFS =)

tchiwam tchiwam at sgo.fi
Sun Apr 6 17:30:02 PDT 2003


Hello,
	This might help some folks here, I am going off line for at least
a week if not 2. So here is the deal... I didn't updated my last Xcompiler
information for a long long time, so here is a new one.

	I tried to get all of the includes needed to compile gcc and glibc
on a foreign OS without copying anything from any host, (darwin and
Slowlaris56 for building). I have built a note / script to build my cross
compiler there www.lfs.org/~tchiwam/XCompilerLinuxGlibc

	The script is building 100% in user space, never needed superuser.
But it can be very destructive, be very carefull, is set the library, bin
and man path in my ~/xcompiler and build everything in ~/src. The
packages are taken from $SOURCES/package/${package-version}.tar.gz defined
in the beginning.

1st Kernel Headers
	I make the kernel headers 1st since gcc-core-3.2.2 needs some
headers to build and there is NO need for gcc to get the proper headers out of
the linux kernel. now I know cuz it works anywhere. (I think this is right
and best way to do)

2nd binutils
	It seems that binutils is a very nice a mature package (errm),
well it is not needing anything special in order to build itself, but
there is a libiberty.a generated in the host /lib. that in itself isn't
evil, but when someone build 2 or more cross compilers the files are
simply over written. This needs to be investigated.

3rd gcc-core pass1
	I did find the way to build JUST gcc at this point, so I use make
-k to get pass the libgcc. but this is enought to get us a bootstraping
gcc. In basic if you manage a normal make / make bootstrap here you are
using hosts includes ???

4th glibc pass1
	I try to gather as many includes as possible on the 1st pass using
make install-headers, then make install. I cannot yet say what else is
built.

5th gcc-core pass2
	Now that we have some glibc includes gcc is already going to know
more about our system, like version of glibc etc etc. The new headers will
be generated. It would be nice to examine the full output between
core-pass1 and core-pass2. Again failure is expected

6th glibc
	This is I suppose to be the glibc full build, I am very suspicous
about this build since the gcc-core pass2 is still not very complete,
another cycle of gcc-core and glibc might be needed (comparing binary
results to see if it is needed)

7th gcc-full built
	1st build gcc then the rest, the "rest" part might be better
delayed after another glibc build.


WHY?
	That is pretty simple it is very hard to spot the files taken from
the host, there is always few hidden connections. By do it this way, I am
90% sure it can't borrow headers from my platform without breaking stuff
badly (lilendian vs 64bit Big endian machine)

Big thanks to the LFS and PLFS troups, i hope this might prove usefull,
Ill be gone on thuesday, then Ill be back after the 14th.

Phil

-- 
Unsubscribe: send email to listar at linuxfromscratch.org
and put 'unsubscribe lfs-dev' in the subject header of the message



More information about the lfs-dev mailing list