Skip to content

Latest commit

 

History

History
83 lines (59 loc) · 4.25 KB

README.md

File metadata and controls

83 lines (59 loc) · 4.25 KB
Raw

Pytest

To run integration tests, just run pytest in root directory of repo, We use pytest to run integration tests. pytest discover test cases automatically. notably the python source files whose name starts with test_ is treated as test modules, and the functions in them whose name starts with test_ will be executed as test cases.

Pytest in macOS

To run pytest successfully in macOS, you need to run it under the nix-shell environment. There are two ways to run it in the nix-shell:

  • run nix-shell directly, you will need to assign the chain-maind binary build path (should be the PROJECT_ROOT/build) into the PATH env of the nix-shell, execute the make build, and then execute pytest.
  • run nix-shell ./. -A ci-shell, let the ci shell tobuild the binary automatically, and then execute pytest.
  • Also, set environment variable TMPDIR=/tmp in the user profile, to avoid the long path issue with the unix socket.

Fixture

We use pytest.fixture to setup/teardown testnet cluster, fixture can be defined at different scopes, we use a session scoped fixture for default cluster, we can override it with a module scoped one at test modules which require special testnet setup.

To use a custom cluster for your test case, just create module scoped fixture in your test module like this. you can put the custom cluster config file into the integration_tests/configs directory.

To write test case which depend on the default cluster, one only need to create a test function which accept the cluster parameter, then you get access to the ClusterCLI object automagically.

Concurrency

We use python-xdist to execute test cases in multi-processess in parallel, the implications are:

  • no memory sharing between test modules

  • session scope fixtures might get setup multiple times in different processes, be aware of system resource conflict like tcp port numbers. the default cluster fixture compute base_port like this: (100 + worker_id) * 100, be aware of this when choosing base_port when overriding the default cluster.

    pystarport's port rules

Markers

We can use markers to mark test cases, currently we use slow to mark test cases that runs slow (like slashing test which need to sleep quit long time to wait for blockchain events), we select or unselect test cases with markers. For example, passing -m 'not slow' to pytest can skip the slow test cases, useful for development.

Cluster api

cluster is an instance of ClusterCLI, which is used to interact with the chain. cluster.supervisor is used to access the embedded supervisord's xmlrpc service(api). for example:

# stop the chain-maind process of node2
cluster.supervisor.stopProcess('chainmaind-node2')
# start the chain-maind process of node2
cluster.supervisor.startProcess('chainmaind-node2')

# get address of specified account of node0
cluster.address('community')
# get the validator address of node2
cluster.address('validator', i=2, bech='val')
# call the "chain-maind tx bank send -y"
cluster.transfer('from addr', 'to addr', '1cro', i=2)

Temparary directory

We use the tmp_path_factory to create the data directory of test chain, pytest will take care of it's cleanup, particularly:

  • it's not cleaned up immediately after test runs, so you can investigate it later in development
  • it only keeps the directories of recent 3 test runs

For example, you can find the data directory of default cluster in most recent run here: /tmp/pytest-of-$(whoami)/pytest-current/chaintestcurrent/.

pystarport

The integration tests use pystarport to setup chain environment.