[Bug-Candidate]: `Cannot decode byte '\xe4': Data.Text.Encoding: Invalid UTF-8 stream`

[pcaversaccio]

Describe the issue:

Background: See https://github.com/pcaversaccio/snekmate/pull/239

Whenever I run my echidna tests, I see the following warnings:

Code example to reproduce the issue:

~$ git clone https://github.com/pcaversaccio/snekmate.git
~$ git checkout feat/echidna
~$ forge install
~$ pip install git+https://github.com/vyperlang/vyper.git@master
~$ forge build --build-info
~$ echidna test/tokens/echidna/ERC20Properties.sol --contract CryticERC20ExternalHarness --config test/tokens/echidna/echidna-config.yaml --crytic-args --ignore-compile

Version:

• Echidna: 2.2.3
• Slither: 0.10.2

Relevant log output:

[2024-04-23 15:44:45.45] Compiling test/tokens/echidna/ERC20Properties.sol... Done! (20.2442777s)
Analyzing contract: C:/Dev/Blockchain/snekmate/test/tokens/echidna/ERC20Properties.sol:CryticERC20ExternalHarness
[2024-04-23 15:45:09.77] Running slither on test/tokens/echidna/ERC20Properties.sol... Done! (35.8772859s)
Loaded 0 transaction sequences from echidna-corpus\reproducers
Loaded 0 transaction sequences from echidna-corpus\coverage
[2024-04-23 15:45:45.96] [Worker 0] Crashed:

Cannot decode byte '\xe4': Data.Text.Encoding: Invalid UTF-8 stream

Please report it to https://github.com/crytic/echidna/issues
[2024-04-23 15:45:45.96] [status] tests: 0/22, fuzzing: 0/10000, values: [], cov: 1560, corpus: 0
test_ERC20external_userBalanceNotHigherThanSupply(): passing
test_ERC20external_setAllowanceTwice(address,uint256): passing
test_ERC20external_burnFromUpdateAllowance(uint256): passing
test_ERC20external_transferFrom(address,uint256): passing
test_ERC20external_transferFromToZeroAddress(uint256): passing
test_ERC20external_constantSupply(): passing
test_ERC20external_setAllowance(address,uint256): passing
test_ERC20external_transferMoreThanBalance(address): passing
test_ERC20external_mintTokens(address,uint256): passing
test_ERC20external_transferToZeroAddress(): passing
test_ERC20external_transferFromMoreThanBalance(address): passing
test_ERC20external_zeroAddressBalance(): passing
test_ERC20external_transfer(address,uint256): passing
test_ERC20external_selfTransfer(uint256): passing
test_ERC20external_userBalancesLessThanTotalSupply(): passing
test_ERC20external_burnFrom(uint256): passing
test_ERC20external_transferFromZeroAmount(address): passing
test_ERC20external_burn(uint256): passing
test_ERC20external_transferZeroAmount(address): passing
test_ERC20external_spendAllowanceAfterTransfer(address,uint256): passing
test_ERC20external_selfTransferFrom(uint256): passing
AssertionFailed(..): passing

Unique instructions: 1560
Unique codehashes: 3
Corpus size: 0
Seed: 8944220378161519169

[2024-04-23 15:45:45.97] Saving test reproducers... Done! (0s)
[2024-04-23 15:45:45.97] Saving corpus... Done! (0s)

[ggrieco-tob]

Hey, this looks very cool. Do you know if the bug is related to hevm code or echidna?

[pcaversaccio]

Hey, this looks very cool. Do you know if the bug is related to hevm code or echidna?

I'm really not sure tbh - didn't have time to investigate properly.

[pcaversaccio]

and btw the decode error is also in the CI https://github.com/pcaversaccio/snekmate/actions/runs/8803366873/job/24161302034#step:19:21

[pcaversaccio]

I think I found partially the issue: Vyper adds the following signature to the init code which I use in the ffi command:

sig = b"\xa1\x65vyper\x83".hex()

See my comment here: https://github.com/vyperlang/vyper/pull/2860#issuecomment-1279717236. And \xa1 and \x83 are common decoding errors I see in that context. There might be some further method Ids that might not be decoded properly as well.

[elopez]

On an initial triage, it seems this happens due to Echidna calling VyperDeployer.deploy(....) with non-utf8-valid strings. The config has allContracts: true. valsArr in hevm ends up having invalid utf8 which then fails to decode. https://github.com/ethereum/hevm/blob/release/0.53.0/src/EVM.hs#L1564

Filtering the affected functions so echidna does not try to fuzz them seems to work around the issue:

filterFunctions: ["VyperDeployer.deployContract(string,string)", "VyperDeployer.deployContract(string,string,bytes)", "VyperDeployer.deployContract(string,string,bytes,string,bool)", "VyperDeployer.deployContract(string,string,string,bool)"]

This is a smaller repro of this issue:

interface Hevm {
  function ffi(string[] calldata) external returns (bytes memory);
}

contract TestFFI {
  address constant HEVM_ADDRESS = 0x7109709ECfa91a80626fF3989D68f67F5b1DD12D;

  function foo() public {
    bytes memory w = hex"c328"; // invalid utf-8

    string[] memory inputs = new string[](2);
    inputs[0] = "echo";
    inputs[1] = string(w);

    bytes memory res = Hevm(HEVM_ADDRESS).ffi(inputs);
  }
}

$ echo allowFFI: true > config.yml
$ echidna foo.sol --test-mode assertion --format text --contract TestFFI --config config.yml
[2024-04-23 18:04:34.90] Compiling foo.sol... Done! (0.775142s)
Analyzing contract: .../foo.sol:TestFFI
[2024-04-23 18:04:35.70] Running slither on foo.sol... Done! (0.81568s)
[2024-04-23 18:04:36.52] [Worker 0] Crashed:

Cannot decode byte '\xc3': Data.Text.Encoding: Invalid UTF-8 stream

Please report it to https://github.com/crytic/echidna/issues
[2024-04-23 18:04:36.52] [status] tests: 0/2, fuzzing: 0/50000, values: [], cov: 481, corpus: 0
foo(): passing
AssertionFailed(..): passing
...

[pcaversaccio]

@elopez confirmed, this was the issue. Thanks a lot! Fixed via https://github.com/pcaversaccio/snekmate/pull/239/commits/0d6889f3949aa89dcfdeadead5581fa395cb7d0c. Will close the issue accordingly.

[ggrieco-tob]

Uhm, is there something we should recommend to hevm to avoid this decoding error?

[pcaversaccio]

Uhm, is there something we should recommend to hevm to avoid this decoding error?

maybe raise this issue to them at least and get their thoughts on it?

[ggrieco-tob]

Sounds like a plan!