Regular testing inside the Server QA Jenkins showed this
test as unreliable and only working in about 50% of the
cases.
Debugging revealed that the guest in our environment can
often take ~5 minutes to fully initialize.
Currently it tries 60 retries waiting 3 seconds each.
While debugging I also found that these connections hang
around in half closed state (CLOSE_WAIT and FIN_WAIT2)
being a (small) burden for the system.
So instead of modifying retries increase the individual
timeout.
This also helps if execution speed itself is slow and 3
seconds just would not be enough for a single connection
(unlikely but possible).
Therefore to make running the test reliable this change
increases the individual connection Timeout from 3 to 15
seconds.
Signed-off-by: Christian Ehrhardt <email address hidden>
Sort the testcases within architectures, as particularly the openjdk-8
build tests seem to have variance in ordering, leading to a lot of
gratuitous changes to look through in the diff output.
test-kernel-security: skip 32bit aslr tests on arm64
Need to figure out how to compile 32bit aslr on arm64, -m32 doesn't
do it. I suspect cross-compiling is involved, since it doesn't appear
that multiarch is supported by default on arm64.
test-kernel-security: update arm64/s390x status of tests on yakkety
The s390x kernel in yakkety 4.8 now has CONFIG_DEBUG_RODATA enabled, and
/proc/self/personality is no longer READ_IMPLIES_EXEC on arm64 (though
it looks like it's still set for pid one on arm64, so... sigh). Update
test-kernel-security tests to reflect those changes.