Commit
af6ce893eb4fecce025d285a132e0d0fc3a4042c
by Pau Espin Pedrol
StatsD_Checker: Abort on expectancy failure with clear status message
Before this commit, a statsd expect failure would let the test continue the test for a while and then finally log a rellay generic timeout message, which makes really difficult to understand what's going on and why is the test failing.
Let's terminate the test early if a test expectation fails, so it's a lot easier to debug. Add an extra param to be able to use the API for test logic to check the state without failing/aborting. This is not really used anywhere so far but the API was built this way and I indeed think it may be handy in the future.
The test was passing when run alone, but was failing when run after other tests using other BSCs/RNCs, since osmo-msc learned about them and hence ran_peers.total was >1.
Commit
c50937d6371a7b203a52ec6c2daf52fe443f398f
by Vadim Yanitskiy
fixup: bts: create all components alive
Making the ConnHdlr and lapdm_CT components alive resulted in massive regressions. Specifically, trxcon started rejecting L1CTL connections with the following message:
L1CTL server cannot accept more than 1 connection(s)
Apparently, alive components do not close the L1CTL connection automatially, unless the L1CTL port is unmap()ped from the system explicitly. Thus, testcases spawning multiple alive components end up creating multiple L1CTL connections, which get rejected.
Let's revert those components that speak L1CTL back to normal (non-alive), while keeping the emulation components alive.
Change-Id: Id5edfa2514024d1fe83f6c4082e096044ce8e6fd Fixes: 713e7a86 ("bts: create all components alive")