First, since the book has moved forward to January, a Happy New Year
to everyone.
Now that Doug has come up with a patch to let rustc-1.37.0 build
with system llvm-9.0.1 (thanks!), most people should have no
problems with clang. However, my normal builds are optimized for
the machine I'm using, and on my more powerful machines they use
higher optimizations or extra security options.
I was testing the patch on a Ryzen 5 3400G using -O3 -march=native
-D_FORTIFY_SOURCE=2 -fstack-protector-strong in CFLAGS, CXXFLAGS and
also -D_GLIBCXX_ASSERTIONS in the CXXFLAGS.
Those have not given me any recent problems (some odd minor packages
were noted back in mid-year in my 'tuning' notes), but building
thunderbird died with a segmentation fault in llvm.
By a process of trial and error I established that without my
CFLAGS, CXXFLAGS it was fine. Then, picking at the details, I
established that llvm thought (correctly) that it was built on
znver1 (Ryzen1), and that both -march=native and -march=znver1
showed the problem, but that the less-specific -march=amdfam10 was
ok.
Then I dropped back to -j1 for the C and C++ parts and saw that it
was failing to build a file in the elfhack code. We've had problems
with elfhack on firefox over the years, it seems to be very prone to
breakage. Disabling it allowed by build to complete, so I've now
added it to the book's mozconfig, but commented because I'm sure not
many people will need to disable it.
If I was actively _using_ thunderbird, I'm sure that I would build
it with gcc and g++, use the patch for system graphite and system
harfbuzz, and probably not need to disable the elf hack. But that's
just a matter of preferences.
Regards,
ĸen
--
We've all got both light and dark inside of us.
What matters is the part we choose to act on.
-- Sirius Black
--
http://lists.linuxfromscratch.org/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page
