v0.9.1

Added

• vm.getEnv cheatcode to read environment variables as strings:

  function getEnv(string calldata key) external returns (string memory value);
  function getEnv(string calldata key, string calldata defaultValue)
      external
      returns (string memory value);

  The single-argument form reverts when the key is missing:

  Failed to get environment variable FOO as type string: environment variable not found
  

  The two-argument form returns defaultValue when the key is missing.

• Automatic .env loading from the project directory (defaults to the current
  working directory). Values are available to vm.getEnv. Existing process
  environment variables take precedence over .env.

• vm.fork cheatcode to create or select a remote chain fork:

  struct ForkConfig {
      uint32 retries;
      uint64 backoffMs;
      uint64 timeoutMs;
      uint64 rateLimit;
  }
  
  function fork(string calldata url, uint256 blockNumber) external;
  function fork(string calldata url, uint256 blockNumber, ForkConfig config)
      external;

  Campaigns always start as an empty sandbox. Call vm.fork in setup or
  action modifiers to opt into remote state. Multiple forks are cached and
  selected by (url, block). Local accounts (harness, deployer, vm.addr
  results) persist across switches. Remote state is isolated per fork, so the
  same address on two chains (e.g. a bridge on Ethereum and Polygon) keeps
  independent storage and balances. Coverage is keyed by bytecode hash, not
  address. Single-arg vm.fork defaults: retries 3, backoff 100ms, timeout
  30s, no rate limit (same as the former CLI defaults).

Changed

• RVM address is now derived from keccak256("ripfuzz cheatcode") instead of
  Foundry's hevm cheat code:

  // before (Foundry HEVM)
  0x7109709ECfa91a80626fF3989D68f67F5b1DD12D
  
  // after
  0x628dC59F11F72B611132eC40437F125ba1312F08
  

  Harnesses must point rvm at the new address (ripfuzz-std Harness already
  does this).

• ripfuzz run <HARNESS> accepts a bare harness name (Harness) or a full
  artifact id (src/Harness.sol:Harness). When multiple contracts share the
  same name, the command lists the matching full ids to choose from

• Upgraded solc dependency to v0.0.14

• Campaign directory IDs include seconds
  (.ripfuzz/campaigns/YYYY-MM-DD-HHMMSS-<uuid>/) so campaigns started in the
  same minute are easier to tell apart

• Fork mode is driven entirely by vm.fork in the harness. CLI flags
  --rpc-url, --rpc-block, --rpc-retries, --rpc-backoff,
  --rpc-timeout, and --rpc-rate-limit are removed. Single-arg
  vm.fork(url, block) uses built-in defaults (retries 3, backoff 100ms,
  timeout 30s, no rate limit). Override via vm.fork(url, block, ForkConfig)

• Removed the library helper Chain::fork_with_transport. Tests and campaigns
  create an empty sandbox and opt into remote state with vm.fork only.

• Removed the startup log for spawning the test chain (including empty-sandbox
  chain id, EVM version, block number, and timestamp). Empty vs fork is decided
  at runtime by vm.fork, so those defaults were misleading

Fixed

• vm.fork now applies the forked block's EVM SpecId (and matching mainnet
  gas params) to the active chain config. Previously only the former
  Chain::fork path did this, so harnesses that called vm.fork kept the
  empty-sandbox hardfork instead of the remote chain's hardfork at that height
  (opcodes, gas schedule, and blob base-fee fraction).

• Traces now surface calls to empty accounts clearly. A successful call with no
  bytecode shows ← [stop] (no code), and a parent empty revert that follows
  such a call decodes as no contract code at <address> instead of plain
  reverted. This makes --fail-on-revert failures actionable when a harness
  hits remote addresses without rvm.fork.