End2end tests base class.
@dataclass class RoundChecks()
Class for the necessary checks of a round during the tests.
name: is the name of the round for which the checks should be performed. event: is the name of the event that is considered as successful. n_periods: is the number of periods this event should appear for the check to be considered successful.
@pytest.mark.e2e @pytest.mark.integration class BaseTestEnd2End(AEATestCaseMany, UseFlaskTendermintNode, UseLocalIpfs)
Base class for end-to-end tests of agents with a skill extending the abstract_abci_round skill.
The setup test function of this class will configure a set of 'n' agents with the configured (agent_package) agent, and a Tendermint network of 'n' nodes, one for each agent.
Test subclasses must set
@classmethod def setup_class(cls) -> None
@classmethod def set_config(cls, dotted_path: str, value: Any, type_: Optional[str] = None, aev: bool = True) -> Result
Set config value.
def prepare(nb_nodes: int) -> None
Set up the agents.
def prepare_and_launch(nb_nodes: int) -> None
Prepare and launch the agents.
@classmethod def missing_from_output(cls, happy_path: Tuple[RoundChecks, ...] = (), strict_check_strings: Tuple[str, ...] = (), period: int = 1, is_terminating: bool = True, **kwargs: Any) -> Tuple[List[str], List[str]]
Check if strings are present in process output.
Read process stdout in thread and terminate when all strings are present or timeout expired.
happy_path: the happy path of the testing FSM.
strict_check_strings: tuple of strings expected to appear in output as is.
period: period of checking.
is_terminating: whether the agents are terminated if any of the check strings do not appear in the logs.
kwargs: the kwargs of the overridden method.
tuple with two lists of missed strings, the strict and the round respectively.
def check_aea_messages() -> None
Check that each AEA prints these messages.
First failing check will cause assertion error and test tear down.
Test that an agent that is launched later can synchronize with the rest of the network
- each agent starts, and sets up the ABCI connection, which in turn spawns both an ABCI server and a local Tendermint node (using the configuration folders we set up previously). The Tendermint node is unique for each agent
- when we will stop one agent, also the ABCI server created by the ABCI connection will stop, and in turn the Tendermint node will stop. In particular, it does not keep polling the endpoint until it is up again, it just stops.
- when we will restart the previously stopped agent, the ABCI connection will set up again both the server and the Tendermint node. The node will automatically connect to the rest of the Tendermint network, loads the entire blockchain bulit so far by the others, and starts sending ABCI requests to the agent (begin_block; deliver_tx*; end_block), plus other auxiliary requests like info , flush etc. The agent which is already processing incoming messages, forwards the ABCI requests to the ABCIHandler, which produces ABCI responses that are forwarded again via the ABCI connection such that the Tendermint node can receive the responses
number of agents with tendermint nodes
mandatory argument if n_terminal > 0
number of agents to be restarted
delay the termination event
how long to wait before restart
how long to check logs for
def check() -> None
Check pre-conditions of the test
def test_run(nb_nodes: int) -> None
Run the test.
@classmethod def teardown_class(cls) -> None
Teardown the test.