Building a ARM EABI toolchain fails in ld-uClibc-0.9.28.so

[Peter S. Mazinger]

On Mon, 6 Mar 2006, Joseph S. Myers wrote:

> On Mon, 6 Mar 2006, Peter S. Mazinger wrote:
> 
> > I have looked at your patch and I like the way it is done. It seems to me 
> > that the only part that is missing (apart from not supporting all archs) 
> > is the --enable-clocale=uclibc feature. The math functions check is more 
> > correct then it is now, buildroot should benefit of it too. I haven't done 
> > a compile check, does your patch work w/ --enable-clocale=generic one 
> > (this should be picked up as default)?
> 
> Yes, it works with --enable-clocale=generic.
> 
> > Do I read the patch right? gcc -muclibc allows on a glibc host to use 
> > ld-uClibc.so.0, resp. gcc -mglibc on an uClibc host to use ld-linux.so.x?
> 
> Yes.  (This is most useful in conjunction with a custom multilib 
> configuration providing both glibc and uClibc multilibs with different 
> sysroots selected depending on the command-line options.)
> 
> > Would you provide an update patch that solves support for the other archs? 
> > If we put it into buildroot, it will get enough testing, so that it can be 
> > added upstream as well.
> 
> That patch <http://gcc.gnu.org/ml/gcc-patches/2006-02/msg01426.html> has 
> already gone in (to trunk and branches/csl/sourcerygxx-4_1).  The only 
> problem since reported is breakage on i?86-*-k*bsd-gnu (BSD with GNU libc) 
> <http://gcc.gnu.org/ml/gcc-patches/2006-03/msg00303.html>.

Thanks, I am looking to integrate the rest of 90_all_uclibc-conf.patch w/ 
your upstream patch.

should I do (libtool.m4 and almost all configure scripts):
linux-gnu*|linux-uclibc*)
	lt_cv_deplibs_check_method=pass_all;;

or add
linux-uclibc*)
	lt_cv_deplibs_check_method=pass_all
	lt_cv_file_magic_test_file=`echo /lib/libuClibc-*.so`
	;;
as done currently

A similar case is ltconfig, the detection of dynamic_linker="GNU/Linux 
ld.so"

I would opt for the first version, many, if not all linux-gnu* where 
changed to linux-*. It would be good to get rid of linux-gnu* at all, but 
I am not sure what the exceptions would be.

Peter

-- 
Peter S. Mazinger <ps dot m at gmx dot net>           ID: 0xA5F059F2
Key fingerprint = 92A4 31E1 56BC 3D5A 2D08  BB6E C389 975E A5F0 59F2