evm-testing

Add parameters to your test function and Foundry generates random inputs automatically. Use bound() to constrain values into valid ranges instead of vm.assume(), which discards runs and can fail the campaign after too many rejections.

Fuzz tests run one function with random inputs, while invariant tests call random sequences of functions across contracts and assert properties that must always hold. Invariant tests use handler contracts with ghost variables to track cumulative state.

vm.prank(addr) only affects the next external call, while vm.startPrank(addr) applies to all calls until vm.stopPrank(). Using prank when multiple calls need the same sender is the most common cause of tests passing when they should not.

Fork tests pull live mainnet state, so whale balances and contract storage change over time. Pin a block number in vm.createSelectFork, use a paid RPC provider to avoid rate limits, and enable RPC storage caching in foundry.toml.

Hardhat uses Mocha for test structure with describe/it blocks and Chai for assertions via @nomicfoundation/hardhat-chai-matchers. Jest matchers like toBe and toEqual do not exist; use expect().to.equal and revertedWithCustomError instead.

Common causes are handler functions not covering all entry points, depth set too low, ghost variables not updated after calls, or bound() ranges too narrow. Increase invariant depth and runs in foundry.toml and verify targetContract is set in setUp().