Small bug due to new kernels (since 2.6.39.1) where the m4 'make
check' has a test failure in test-readlink. Apparently the newer
kernels are returning EINVAL when readlink() is called with a null
string "". The previous behavior was returning ENOENT.
According to the second link below, the next version of m4 is expected
to contain a fix for the test to accept EINVAL to cope with the
kernel's change in behavior.
There is a debian patch referenced in the first link below.
Alternatively, this sed fixes the test as well:
sed -e \
'/ASSERT (func ("", buf, sizeof buf) == -1);/{n;s/(/((/;s/;/ ||
(errno == EINVAL));/}' \
-i.orig tests/test-readlink.h
The sed finds the line before the test we want to change, goes to next
line, then adjusts the test statement to include a second check.
sources:
http://old.nabble.com/Bug-630902%3A-m4%3A-FTBFS%3A-FAIL%3A-test-readlink-%28fwd%29-td31876292.html
http://lists.gnu.org/archive/html/bug-m4/2011-10/msg00000.html
Jonathan
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
