On 08/04/2018 11:58 AM, Xi Ruoyao wrote:
I just built a LFS system following system development (r11444). I note
several issues:
1. In 5.29 (temp Perl), change Perl configure line to
sh Configure \
-des \
-Dprefix=/tools \
-Dlibs=-lm \
-Dlocincpth=/tools/include \
-Dloclibpth=/tools/lib
The purpose is to avoid Perl to use "-I/usr/local/include" and
"-L/usr/local/lib" and grab headers or libs from host.
Reading the Configure script, what about just:
-Ulocincpth \
-Uloclibpth
to set these to 'undef'? (Lines 1745 to 1756)
These variables are fairly obscure. I found no explanation in the
tarball, but it was used for different architectures. A web search was
a little better:
https://perl5.git.perl.org/perl.git/blob/HEAD:/Porting/Glossary
2. In 6.20 (GCC), now we have 14 (instead of 6) libstdc++ tests
failing because we are running testsuite as root, and 2 tests failing
because of PR 86153
<https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86153>.
I'll change the wording. I believe six of the failures are duplicates.
3. In 6.54 (systemd), we can use two seds instead of the patch.
Just remove
#include <linux/stat.h>
lines and systemd will build normally. We know we are using
Glibc-2.28 so don't need the check in meson.build.
I'll let Douglas address this one.
4. In 6.56 (e2fsprogs) we should use `--enable-relative-symlinks`.
I asked a question about this on another post.
5. In 6.76 (vim), suggest to change the test command to
LANG=en_US.UTF-8 make -j1 test &> vim-test.log
then Test_terminal_composing_unicode would pass.
Confirmed. I'll update.
-- Bruce
--
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
