GeezerVison

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

❌ P1 high — uniswap/asyncuniswap.py:1235

Reopened review hold; awaiting explicit adjudication.

_{How this was verified: Reopened review hold: Fixed in 2d9d82b}

⚠️ P2 medium — tests/test_uniswap.py:343

The test now uses FeeTier.TIER_500 for DAI/USDC, but test_get_tvl_in_pool_on_chain and the other DAI/USDC tests still use TIER_3000, while test_close_position relies on state left by this test (get_liquidity_positions()[0]). Further, the test passes fee=fee (TIER_500) to make_trade, so the swap routes through the 0.05% pool, and the test uses anvil forked state from a live provider, so pool liquidity and tick state depend on the moment of the fork. The unchanged qty * 10 DAI->USDC swap (10 ETH worth of DAI) is the only sizing safeguard. If the 500 pool lacks depth at the fork block, the swap reverts or balance_1 > amount1 fails. The result is a flaky, non-hermetic test whose failure also cascades into test_close_position (IndexError on empty positions).

_{How this was verified: Read both test files; test_close_position indexes position_array[0] with no guard and depends on the mint in this test.}

⚠️ P2 medium — tests/test_uniswap.py:169

In tests/test_uniswap.py the ETH parameter was replaced with WETH in test_approve, removing the only case that approved the native-ETH sentinel address. The if client.version == 1 and ETH_ADDRESS == token0 skip guard just below now never triggers, since no parameter yields ETH_ADDRESS anymore. In the async file, by contrast, ETH was never present and WETH is simply added, so the two files now cover different token sets. The sync suite no longer exercises approve() with the ETH address, which was the behaviour that existed before, and the dead skip branch is misleading.

_{How this was verified: Checked the parametrize list and the guard at line 177 in tests/test_uniswap.py; no remaining parameter equals ETH.}

⚠️ P2 medium — tests/test_uniswap.py:343

Reopened review hold; awaiting explicit adjudication.

_{How this was verified: Reopened review hold: No reopening reply recorded.}

These P2 guard, heuristic, trade-off, or documentation claims are retained for judgment without opening review threads.

⚠️ P2 medium — .github/workflows/test.yml:12

The CI env block's comment says the secret points to a rate-limited Infura URL and CI must use a public endpoint. The PR flips the active public endpoint from eth.drpc.org to ethereum.publicnode.com, which is a free public RPC with its own rate limits, and anvil forks from it on every run across 3 matrix jobs (v2, v3, v4). Forking against a public node means every state read during tests is fetched remotely, so flakiness depends on that provider's availability and throttling. The change is a swap of one unreliable endpoint for another with no evidence it is more reliable, and the Foundry pin to 1.8.6 is unexplained, so future anvil updates are blocked.

• .github/workflows/test.yml — Swaps the active PROVIDER_MAINNET URL from eth.drpc.org to ethereum.publicnode.com (commenting out the other) and adds version: 1.8.6 to the foundry-toolchain step.
• tests/test_asyncuniswap.py — Adds WETH to the test_approve token parameters and changes the test_v3_deploy_pool_with_liquidity parameters to ONE_DAI amount0 and FeeTier.TIER_500.
• tests/test_uniswap.py — Replaces ETH with WETH in test_approve parameters and changes test_v3_deploy_pool_with_liquidity parameters to ONE_DAI amount0 and FeeTier.TIER_500.

_{@TimeToBuildBob review (own line) — fresh review · @TimeToBuildBob fix — a worker acts on the findings. Once per comment; 👀 = received.}