On Sat, 7 Jan 2006, Dan Nicholson wrote:
Hi,
Recently some discussion has come up that the binutils and gcc built
in Ch. 6 are linking to the glibc in /tools. This was brought up by
Greg Schafer in this post
http://linuxfromscratch.org/pipermail/lfs-dev/2006-January/054970.html
I've done some analysis of the LFS build pre-utf8 and found that he's
right. Attached is a file showing some analysis I did that stems from
Greg's analysis at DIY. Links to Greg's original analysis and current
implementation are shown at the bottom of the attached file.
Fleshing out the sanity check shows that while the new dynamic linker
in /lib/ld-linux.so.2 is being used, linking is being done to
/tools/lib libraries and start files. This includes
/tools/lib/libc.so.6 which is obviously not what we want since we just
installed /lib/libc.so.6 in 6.11.
Thanks for doing the analysis. I suspect we are probably going to move
to something like Tushar's method documented towards the end of
http://bugs.linuxfromscratch.org/show_bug.cgi?id=1677
(don't let the bug's title mislead you).
I'm not close to starting testing for this, but I hope to get a round
tuit in a day or two. I'm also wondering if it will clear up some of
the 'transient' differences in my ICA analysis.
At least introducing Man-DB and Berkeley hasn't brought any new
differences, and Alexander's manconfig.h.in gives a better build IMHO.
Ken
--
das eine Mal als Trag?die, das andere Mal als Farce
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page