In the past, MM INFO transmission was broken, so the tests were
written without expecting it. Now, the bug in osmo-msc has been
fixed.
With the config file enabling it, and the tests not expecting MM INFO,
a lot of them failed. Let's disable MM INFO for now until the tests
can deal with it
We recently introduced verification of the MCC/MNC, and the osmo-msc.cfg
MCC/MNC settings mis-matched those of MSC_Tests.ttcn resulting in lots
of test failures.
Debian 9.0 ships with titan 6.1.0 which for some unknown reasons
causes segfaults in the C++ part when parsing CTRL messages and
according to TITAN project is not supported anymore anyway.
So let's use a 6.3.x build, or whatever is the highest version in
the network:osmocom:latest feed instead.
In osmo-ttcn3-hacks, we have removed the *.control from the *.default
files, so let's add it here in the config files. You can now change
the tests to be executed by simply editing the config here.
the BSC configuration currently only offers A50 (no) encryption.
Since the TTCN3 tests also cover assignment tests with encryption
and osmo-bsc now supports multiple cipher settings at a time this
setting should be extended accordingly.
- Also enable A51 and A53 encryption.
Test TC_pdp6_act_deact_pcodns checks the 2 DNS values against matched
configuration. If osmo-ggsn.cfg doesn't have a 2nd DNS configured, it
won't sent the related PCO in the response, and the test will fail.
This way the address is reachable for ggsn system and can route the
packet correctly.
This commit fixes runtime failures of
GGSN_Tests.TC_pdp4_act_deact_gtpu_access when running inside the
container, since that test uses the DNS address to attempt a ping and
waits for a response. If the address is non reachable/routable from
osmo-ggsn tun, then the packet is dropped by the kernel and no ping
response is sent back.
Since October 2017 the default config name no longer has an underscore
but a dash in the filename. See osmo-sgsn.git Change-Id
If804da17a7481e79e000fe40ae0d9c4be9722e61
This uses osmo-ttcn3-hacks Change-Id
I3db452e24e5238aa05254d903739c64d202e61db, which introduces
some shared/common config file as well as per-testcase pcap file
generation.
Ideally we would want to launch a group of containers with their own
private network segment and use the same static IP addresses in those
isolated networks.
The stupidity of docker is requiring unique IPv4 addresses even on
isolated (!) networks. This means we have to manually give each of our
test setups a different subnet, and then we can at least run one
instance that test in parallel to at most one instance of each other
test.
If this weird reestriction about unique IPv4 addresses didn't exist,
we could start any number of test runs in parallel.