[lfs-dev] Failures in tests

Ken Moffat zarniwhoop at ntlworld.com
Tue Mar 3 12:08:55 PST 2015


Just in case anybody cares, this is a summary of the testsuite
failures I have had when building x86_64 on an AMD phenom x4 and a
SandyBridge i3 (identical failures), and building i686 on an AMD A4.

I will ignore the test results from bc - they always show loads of
rounding differences, as Bruce pointed out when I first looked at
that package (no, I never used to test it in BLFS).  Packages in
alphabetical order.

To be absolutely clear - testsuite failures only matter if they show
up a new problem, and even then they mostly do not matter.  For gcc,
the results look significantly better (only 18 failures on x86_64)
now that I have stopped running those tests in parallel (that used to
be ok, and the script formats the results nicely, but now there must
be races).

acl : I don't think I've got any access controls enabled in my
kernel, so I get 8 failures on each machine.

autoconf - Libtool test fails, as noted (one test fails due to
changes in libtool-2.4.3 and later).

eudev continues to report 1 errors, as it has done for some time,
but with no identifiable error.

gcc

 on i686, 4 failures in g++ - asan/asan_test.C and 3 in
 ipa/pr61160-3.C x3 (gcc itself is clean)

 on x86_64 (both AMD and intel) gcc has 2 failures in
 tsan/thread_leak1.c, and g++ has 18 failures - 16 in the torture
 tests and the same 2 in tsan/thread_leak1.c

glibc

 on i686, 4 failures math/test-ldouble nptl/tst-cleanupx4
 posix/tst-getaddrinfo{4,5}

 on x86_64 just posix/tst-getaddrinfo{4,5}

inetutils

 on i686 ping-localhost.sh failed (ok on x86_64, might be a kernel
 config issue).

libtool - Standalone Libltdl x4, and linking libltdl without
autotools

perl - cpan/Socket/t/getnameinfo.t dist/IO/t/io_multihomed.t
dist/IO/t/io_sock.t

Everything else was fine (well, I'm never quite sure about bash and
vim but their results looked normal).

Just in case anybody is in any doubt, everything I have tried so far
to use the finished systems is ok (still got a bit more to do,
and the i686 build is still struggling through libreoffice).

ĸen
-- 
Nanny Ogg usually went to bed early. After all, she was an old lady.
Sometimes she went to bed as early as 6 a.m.


More information about the lfs-dev mailing list