Complex bug with Eth prover and Eth client

[MaksymZavershynskyi]

Update: The cause is found and a simple fix should be submitted by EOD July 6, 2020.

Currently, transfer-eth-erc20-to-near command waits until Eth2NearClient catches up with the necessary block before proceeding with minting. However, it does not wait for the block to be confirmed, instead it simply waits for it to appear on the canonical chain: https://github.com/near/rainbow-bridge/blob/master/environment/commands/transfer-eth-erc20-to-near.js#L146

This bug went uncaught by the tests because of another bug described in here: #114 Specifically:

In Eth2NearProver contract the verification of the proof should be before the cross contract call to the client: https://github.com/near/rainbow-bridge/blob/c42defee097d2e4ac6d86777ae1c374e728402e1/libs-rs/eth-prover/src/lib.rs#L143

Fortunately, it is an easy fix, however a naive fix breaks the tests. This happens because of another bug in https://github.com/zmitton/eth-object that we are using (and probably shouldn't). When tests are run with Ganache the difficulty is set to 0, however eth-object incorrectly serializes Ethereum header with RLP, because 0 is a special case, see: https://ethereum.stackexchange.com/questions/67055/block-header-hash-verification Incorrect serialization results in incorrect header passed with the proof to the minting function, which causes it to crash.