Support isValidSignature(bytes, bytes)

[schmanu]

Some smart contract wallets including Safe use isValidSignature(bytes, bytes) instead of isValidSignature(bytes32, bytes).

For those this signature validator will not work as the magic value results in 0x20c13b0b instead of 0x1626ba7e

[Ivshti]

@schmanu the EIP was updated to use bytes32, bytes and that's it's latest version AFAIK?

Also, using bytes as input presumes that the function would take input that isn't the final digest, but might not be compatible with different schemes (eg with prefix vs without prefix, in other words hash(ethereum signed message+hash(input)) vs hash(input)). Unless it is used for the final digest.

Anyway, we have nothing against implementing this, I'm just curious what the rationale is. Furthermore whether the input is the final digest affects the implementation

[schmanu]

Hey @Ivshti, at the time the isValidSignaturecall was implemented in the safe-contracts the standard was still isValidSignature(bytes, bytes). That's why it is implemented in this way.

[Ivshti]

@schmanu can you give me an example of how it's used? are the first bytes the input to keccak256? Or is prefixing also done (Ethereum signed message)?