Downcasting in difficulty calculation can lead to overflow

[CjS77]

The difficulty calculation has an overflow issue:

pub(crate) fn big_endian_difficulty(hash: &[u8]) -> Difficulty {
        let scalar = U256::from_big_endian(hash); // Big endian so the hash has leading zeroes
        let result = U256::MAX / scalar;
        result.low_u64().into()
    }

Here, scalar is the target for the mining algorithm. The smaller the target, the more difficult it is to find a hash with the requisite number of leading zeroes.

The definition of difficulty is TARGET_MAX / target. Once target falls below 2^192, the difficulty will overflow to zero, since we downcast to 64 bits by truncating the highest portion of the actual difficulty.

Now, this is possibly only a theoretical problem since we need a ridiculous hashrate before this become an eventuality. We will certainly have native 256-bit integers long before this happens.

However, to protect against weird things happening, it might be better to prevent the overflow anyway by returning u64::max if scalar <= 2^192.

[jorgeantonio21]

Hello, I am interested in tackling this issue. I included the suggested code change in a newly created PR:

#4090

I also want to add test coverage for this new functionality, but I am having some difficulty with it, which I describe in the PR. Any help is highly appreciate it.