This directory contains tests that test the actual on-chain programs and CLI client, by uploading the programs to a local testnet, and using the CLI client to interact with them. This requires a local validator to be running.

The tests are scripts that exit with status code 0 when the test passes, or with a nonzero status code when the test fails.

In addition to the “black-box” tests here, there are also tests in the program/tests directory of the repository. Those tests do not require a local validator.

The following programs must be present in the local network:

• The SPL token program, which solana-test-validator includes by default.

The tests generate various key pairs to test with multiple accounts. These are stored in tests/.keys. They are not valuable or security-sensitive whatsoever.

It is possible to run all the scripts with --verbose to make them print more details when a called program exits with a nonzero exit code. This is disabled by default because some calls fail intentionally in the tests.

Set the NETWORK environment variable to https://api.testnet.solana.com or devnet respectively, to run the tests against testnet or devnet. Beware of rate limits and airdrop limitations on these networks.

coverage.py will compile the tests and the CLI with coverage instrumentation, and then run the unit tests, and the test scripts in this directory. It outputs lcov files that are used for codecov.io, and an html report that you can browse locally. Both are written to the coverage directory in the repository root.

Collecting coverage requires a nightly Rust compiler and a few other tools, see coverage.py itself for details.

Note that this script does not measure coverage of the solana_program_test tests in program/tests, as those tests run the BPF programs, and we don’t have a way to instrument those.