Do not accept truncated function selectors.

[chriseth]

This prevents problems where function selectors end in zeros and similar things.

Especially, if you have a contract that only has a single function and this function's selector is 0x00000000, the fallback function was not triggered on empty data.

[pirapira]

Looks good to me.

[axic]

Is the interfaceFunction selector parser padding it with zero?

[chriseth]

No, the EVM is. The code is div(calldataload(0), 0x1000..0000)

[axic]

Right. I am inclined to say this may be worth considering as an entry on the bug list.

[chriseth]

We have scales on which we evaluate bugs for the bug list:

1. discoverability in tests, 2) ocurrence probability, 3) exploitability

This bug scores extremely positive on all three scales:

1. if you have a fallback function and you test it, you will notice that it does not work.

2. you have to have a contract with less than 5 external functions and one of them has to have a specific out of 2**32 names. If your fallback function is made to register ether transfers, then that function has to be non-payable.

3. I don't think this is exploitable, it can only be used to deceive users: You can fool them into calling a certain function on the contract although they "just" wanted to send ether to the contract.

[holiman]

I experimented with this for the uctf:

pragma solidity^0.4.11;

contract Foo{
    uint public result =0; 
    function Foo(){
        
        
    }
    function overdiffusingness(bytes a,uint256 b,uint256 c,uint256 d,uint256 e)  payable returns (uint)
    {
        result = 42;
        return 42;
    }
    
    function () {
        
        result = 13;
    }
    
}

I could find no fault in the solidity implementation.

[chriseth]

It turns out, there are even less cases where this can happen: If your contract has a fallback function, you are safe.

This means this can be used to call functions with signature zero if someone just wants to send ether to the contract, and this function also has to be payable.

[holiman]

I really like the method signature that I brute-forced. Too bad it was useless for the uctf competition though 😢

[pirapira]

I also tried two function signatures with the same hash, but Solidity refused that.

[holiman]

Yeah, that's also something I fiddled with for uctf.

[axic]

Should this be a key "verylow" or a user readable string?

[chriseth]

I would say both. Why can't keys have spaces?

[chriseth]

Weird, I don't get the test errors locally.