Calling check_leaders too soon after set_candidates. This problem can be cleaned from the test by waiting after the cast to set_candidates/2.
Both these are potential bugs. However, riak_repl2_leader is used now, which changes the set_candidates/2 behaviour. So the intention is to ignore the EQC failure for release of develop-3.0. It was previously failing, there is also lots of pending change built up to be merged in 3.1 - EQC tests could be re-visited then.
The riak_repl_leader EQC tests will consistently fail. There appear to be two types of failure:
Calling set_candidates after toggle_type (although without calling toggle_type there are no candidates to test) which leads to a timeout waiting for helper_leader_node https://github.com/basho/riak_repl/blob/develop-3.0-292/eqc/repl_leader_eqc.erl#L281 as the helper is undefined - https://github.com/basho/riak_repl/blob/develop-3.0-292/eqc/repl_leader_eqc.erl#L438.
Calling check_leaders too soon after set_candidates. This problem can be cleaned from the test by waiting after the cast to set_candidates/2.
Both these are potential bugs. However, riak_repl2_leader is used now, which changes the set_candidates/2 behaviour. So the intention is to ignore the EQC failure for release of
develop-3.0
. It was previously failing, there is also lots of pending change built up to be merged in 3.1 - EQC tests could be re-visited then.FYI @ThomasArts @martincox