ci: the hardhat unit test suite is never run — 37 tests across 4 files are inert #369

Summary

make test-unit-hardhat exists, works, and globs every file in tests/unit/ — but no workflow ever invokes it. As of ac565f2 that leaves 37 tests across 4 files which pass locally and are never executed by CI.

Evidence

The target exists and is already glob-driven:

test-unit-hardhat: ## Run hardhat unit tests
	npx hardhat test ./tests/helpers/matchers/index.test.ts ./tests/unit/*.test.ts --no-compile

Nothing calls it:

$ grep -rn 'test-unit-hardhat\|hardhat test' .github/workflows/
(no matches)

ci.yml runs ten jobs, none of which touch TypeScript tests: rust-fmt, rust-deps, rust-lint, rust-test, rust-integration-test, proto, contracts-lint, contracts-build, contracts-test, docker-build.

Why this bites now

The glob means a new unit test file is adopted with zero wiring, which is convenient — but it also means freshly merged regression tests land directly in the inert set:

• fix(genesis): require positive validator voting power #346 (fixing bug(genesis): ValidatorManager accepts validator sets with no positive voting power #344) added 4 tests pinning "a validator set needs at least one positive voting power".
• fix(genesis): align ProtocolConfig schema with runtime consensus/fee checks #359 (fixing Genesis ProtocolConfig should reject zero blockGasLimit #321) added 4 tests pinning ProtocolConfig consensus/fee bounds.

Both invariants can now regress with CI staying green.

What the job actually needs — measured, including one trap

I tried to keep this to a Node-only job, and that does not work. Worth writing down, because the naive patch looks right and fails:

$ env PATH="${PATH-without-foundry}" make test-unit-hardhat
Error in plugin hardhat-foundry: Couldn't run `forge`. Please check that your foundry installation is correct.
make: *** [Makefile:177: test-unit-hardhat] Error 1

hardhat.config.ts imports @nomicfoundation/hardhat-foundry, and that plugin shells out to forge during config resolution — so it is required even with --no-compile, and even though none of these 37 tests touch a contract.

What it does not need is equally useful, since it keeps the job cheap. On a fresh --depth 3 clone of ac565f2 with all three contracts/lib/* submodule directories empty, no hardhat compile (typechain-types/ absent) and no contract artifacts of any kind:

$ npm ci && make test-unit-hardhat
  37 passing (58ms)

real	0m3.141s

So the job needs Node plus the forge binary, and can skip the git submodule update --init ... step that contracts-lint, contracts-build, contracts-test and docker-build all carry.

One caveat on my own measurement: the run above used forge 1.1.0-dev, whereas the repo pins v1.4.4 in .foundry-version. The patch below reads the pinned version the same way the existing jobs do, so CI would exercise 1.4.4 rather than what I happened to have.

Suggested patch

  ts-unit-test:
    name: TypeScript Unit Tests
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6.0.3

      - name: Setup Node.js
        uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e # v6.4.0
        with:
          node-version: 22
          cache: npm

      - name: Install Node dependencies
        run: npm ci

      - name: Read Foundry version
        id: foundry-version
        run: echo "version=$(cat .foundry-version)" >> $GITHUB_OUTPUT

      - name: Install Foundry
        uses: foundry-rs/foundry-toolchain@v1
        with:
          version: ${{ steps.foundry-version.outputs.version }}

      - name: Run hardhat unit tests
        run: make test-unit-hardhat

No submodule init and no compile step. I checked that this splices into ci.yml and parses cleanly — eleven jobs, with no needs edge so it runs in parallel with the Rust lanes.

Two things a patch alone will not cover:

1. Someone with admin rights has to add the job to the repo's required status checks, otherwise it runs but does not gate merges.
2. External fork PRs here do not get CI runs until a maintainer approves the workflow, so a fork-based PR for this cannot demonstrate itself green — the evidence has to come from a local run like the one above.

Related

Same class of gap, different suites: #248 (arcup and finalize-release shell suites never run) and #247 (arcup shell tests plus shellcheck), with #249 open for the finalize-release half. This is the TypeScript/hardhat instance of the same pattern and is independent of those.

Disclosure: I am osr21, external to Circle — an unaffiliated community contributor with no write access to this repository. This report is advisory only; the measurements above are from my own local runs and should be reproduced before acting on them.