Foundry vm.readFile on broadcast/run-latest.json fails without fs_permissions in foundry.toml; decode addresses with stdJson.parseRaw

Symptom: a forge script that reads a previous deployment out of broadcast/<script>/<chainid>/run-latest.json dies with a bare script failed: and no useful message, or a cheatcode error about the path not being allowed.

Cause: Foundry sandboxes filesystem access. vm.readFile, vm.writeFile and the JSON helpers are denied unless the path is explicitly granted in foundry.toml:

fs_permissions = [{ access = "read", path = "./" }]

Then read and decode. stdJson.parseRaw returns ABI-encoded bytes, not a typed value, so it must be decoded — this is the second place people get stuck, because console.log(string(contractAddress)) prints garbage:

string memory json = vm.readFile(string.concat(vm.projectRoot(), "/broadcast/deploy.s.sol/11155111/run-latest.json"));
address deployed = abi.decode(stdJson.parseRaw(json, ".transactions[0].contractAddress"), (address));
bytes32 txHash = abi.decode(stdJson.parseRaw(json, ".receipts[0].transactionHash"), (bytes32));

Use vm.parseJsonAddress(json, key) when you want the typed helper instead of decoding by hand.

One correction worth carrying: there is no vm.envSet cheatcode. Foundry exposes vm.setEnv(string,string), and it only mutates the environment of the current forge process — it does not write your .env, so a value set in one forge script run is gone in the next. To hand addresses between separate runs, write them to a JSON/txt file (needs a write entry in fs_permissions) or export the variable in the shell and read it with vm.envAddress.

Also note run-latest.json is per chain id, and a broadcast that reverted still leaves a file, so assert the address is non-zero before using it.

Source: adapted from a Stack Exchange answer by SYED ASAD KAZMI (question by Blissful), https://ethereum.stackexchange.com/questions/162092/foundry-scripts-read-deployed-contract-addresses — CC BY-SA 4.0