chapter 5.9 cocoon toolchain tests fail

For Junk Mail junk_mail at irishbroadband.net
Sat Dec 15 10:57:53 PST 2007


On Fri, 2007-12-14 at 03:25 +0100, George Tsalikis wrote: 
> I am reffering to the 3 tests in the end of the chapter.
> 2/3 of them fail and deleting the specs file corrected the issue.
> 
> Could the remnant specs spoil the whole system if not deleted?

I'm in trouble around the same spot, but I pass the tests, and get as
far as bash before it shows.
I'm trying uclibc-2.6

The toolchain fails to compile on hlfs-20051220 (I blame
gcc-3.4.4/glibc-2,3,6 as simply too old), and some instructions silently
misbehave in Fedora, like this one

echo "#undef STANDARD_INCLUDE_DIR
#define STANDARD_INCLUDE_DIR 0" >> gcc/config/linux.h ( which adds no
lines !)

So I have been criss-crossing a little (patch/sed/echo commands in
hlfs,  ./configure && make in Fedora), and I deserve what I get. But I
think I hit on this. 

I've ended up with i686-pc-linux-uclibc/ & i686-pc-linux-gnu/ - the
former from the embryo toolchain and the latter from the cocoon
toolchain (I know by the dates). That's OK until the end of the cocoon
build if I'm going glibc --> uClibc, which I am . But the pc-linux-gnu
stuff might then be removed with advantage on a glibc host. 

Also The cocoon gcc's specs file ended up
in /tools/lib/gcc/i686-pc-linux-gnu/4.1.2/(the embryo toolchain's gcc).
Again, I know by the dates.Adding --target=i686-pc-linux-uclibc to the
embryo configure fixes that. It's strictly untrue, but it helps.

But my cocoon compiler is fatally borked, and the adjustment of the
toolchain is not happening correctly :-(. Everything is copy and pasted,
but I'm not getting the right results. I hit this before with console
issues...

/Out of his depth. Water closes over his head... bubbles appear - then
silence. 


-- 
For Junk Mail <junk_mail at irishbroadband.net>




More information about the hlfs-dev mailing list