[blfs-dev] glibc errors in blfs packages on i686

Fernando de Oliveira famobr at yahoo.com.br
Thu Aug 29 04:00:22 PDT 2013

Em 28-08-2013 23:19, Bruce Dubbs escreveu:
> Fernando de Oliveira wrote:
>> Bruce, just replacing /lib/libc-2.18.so did no allow the machine to
>> reboot. I am a little too tired to go on now, could do wrong things.
>> Tomorrow morning, will backup and then try a complete replacement of
>> glibc, and the report back.
>> If you have the LFS book modified by tomorrow, I will test it tomorrow, too.
> I was wondering if something like that would happen.
> I've created a new branch, /LFS/branches/test, that has the modified 
> instructions.  You can use that to build a new lfs.
> I suppose that you could also try to just chroot from the host and 
> rebuild glibc as in Chapter 6 with the sed.  I think in that case you 
> might get away with:
> sed -i -e 's/static __m128i/inline &/' sysdeps/x86_64/multiarch/strstr.c
> mkdir -v ../glibc-build
> cd ../glibc-build
> ../glibc-2.18/configure    \
>      --prefix=/usr          \
>      --disable-profile      \
>      --enable-kernel=2.6.32 \
>      --libexecdir=/usr/lib/glibc
> make
> make install
> Nothing else should change.  The fallback would be to do a complete rebuild.
>    -- Bruce

Yes, that is what I did before, but with DESTDIR, then just replaced
libc-2.18.so and everything was fine.

But after icedtea-web-1.4 was built and installed, things went wrong
again. After some investigations, finally, found that ldconfig was
relinking libc.so.6 to /lib/libc-2.18.so-broken. Reason it did not
reboot is that I moved /lib/libc-2.18.so-broken out and forgot to relink
with ldconfig, so libc.so.6 was a broken link. Now, just with the
replaced libc-2.18.so, everything are back to what seems normal,
problematic machine is running right.

Will turn the attention to building jUNIT and ant, as they were the
initial triggers for the problem.

Then, instead of installing more packages for me or for tags, I will
navigate through the ones supposed to have problem and that should be
fixed, now. I can say glxgears was always running apparently right,
before, and is running ok, now.

Will start in parallel, a new 7.4-test build.

I will have breakfast, and come back. Bruce, I believe if it is OK my
tests in i686 and other people ones with perhaps other architectures,
this will become rc-2?


More information about the blfs-dev mailing list