Testing

The repo has several test layers. CI runs all of them on every PR, so knowing how to reproduce each locally saves round-trips:

Rust tests

Contributor guide rule: any pallet or runtime change must come with unit tests covering its edge cases. This is how you run them.

All tests

SKIP_WASM_BUILD=1 cargo test --workspace

or, equivalently (the justfile exports SKIP_WASM_BUILD=1 for you):

just test

SKIP_WASM_BUILD=1 skips compiling the wasm runtime blob, which the unit tests don't need — without it every test run pays a multi-minute wasm build first. CI (check-rust.yml) runs the same thing with all features enabled:

SKIP_WASM_BUILD=1 cargo test --workspace --all-features

One pallet

SKIP_WASM_BUILD=1 cargo test -p pallet-subtensor

Pallet tests live inside each crate under src/tests/ (e.g. pallets/subtensor/src/tests/staking.rs), organized as modules of the library target.

One test (or a group of tests)

Pass a name filter — every test whose full path contains the string runs:

# Everything in the staking test module
SKIP_WASM_BUILD=1 cargo test -p pallet-subtensor tests::staking

# One test, with log output visible
SKIP_WASM_BUILD=1 cargo test -p pallet-subtensor test_add_stake_ok_no_emission -- --nocapture

# Exact match only (skip other tests whose names contain the same prefix)
SKIP_WASM_BUILD=1 cargo test -p pallet-subtensor -- --exact tests::staking::test_add_stake_ok_no_emission

Add RUST_LOG=debug before the command to see runtime log lines, and --release if a test is too slow in debug mode (epoch/consensus tests often are). scripts/test_specific.sh wraps a release-mode single-test run with these flags preconfigured.

Feature-gated tests

Some code only compiles under feature flags. The two you'll hit most:

# Benchmark tests (also what `just benchmarks` runs)
SKIP_WASM_BUILD=1 cargo test --workspace --features runtime-benchmarks

# The faucet extrinsic used by localnet
SKIP_WASM_BUILD=1 cargo test -p pallet-subtensor --features pow-faucet

When in doubt, mirror CI with --all-features.

Migration tests (try-runtime)

Storage migrations get two kinds of coverage. Unit tests live next to the other pallet tests (see pallets/subtensor/src/tests/migration.rs). To exercise a migration against real chain state, use try-runtime — this is what try-runtime.yml does on every PR against devnet, testnet, and mainnet state:

cargo install --git https://github.com/paritytech/try-runtime-cli --locked

See Repository scripts for the build and invocation commands to replay on_runtime_upgrade against a live endpoint or snapshot.

TypeScript integration tests

TypeScript tests are run with Moonwall. You will need Node (see ts-tests/.nvmrc) and pnpm:

cd ts-tests

# Use the correct Node version
nvm use

# Install pnpm
sudo npm i -g pnpm

# Install dependencies
pnpm i

# Run manual seal dev tests
pnpm moonwall test dev

# Run zombienet tests (one environment per suite)
pnpm moonwall test zombienet_staking
pnpm moonwall test zombienet_shield
pnpm moonwall test zombienet_coldkey_swap
pnpm moonwall test zombienet_evm
pnpm moonwall test zombienet_subnets

# If you have MacOS, you might need to run zombienet tests with sudo, because tmp folder
sudo pnpm moonwall test zombienet_staking

# Run smoke tests (against devnet, testnet, or mainnet)
pnpm moonwall test smoke_devnet
pnpm moonwall test smoke_testnet
pnpm moonwall test smoke_mainnet

Moonwall lets you also run the testing environment without performing any tests on it, as a method for you to manually test certain things:

# Dev tests in run mode
pnpm moonwall run dev

# Zombienet test with run mode
pnpm moonwall run zombienet_staking

Other suites

• Mainnet clone (clones/): clones live mainnet state, sudo-upgrades it to your runtime, and runs JS regression tests against it. Full walkthrough: Mainnet clone testing.
• Python SDK (sdk/python/): just sync && just check runs the same offline gates as CI; runtime metadata changes require regenerating the codegen bindings. Full walkthrough: Python SDK tests.
• eco-tests (eco-tests/): excluded from the cargo workspace, so run them from their own directory: cd eco-tests && cargo test. They pin the storage and RPC shapes the TAO.com ecosystem indexer consumes — see Eco-tests for what to do when one fails.