Storage in forked chain still not working correctly in latest beta

[area]

This is clearly related to #436 - there, it was not possible to set storage, but it seems like it's now too easy to delete storage on a forked chain!

Expected Behavior

Emptying a storage slot in a contract on a forked chain should not affect other contracts.

Current Behavior

Emptying a storage slot in a contract on a forked chain seems to empty that storage slot on all other contracts.

Possible Solution

Steps to Reproduce (for bugs)

Once again using:

pragma solidity 0.5.8;
contract ContractEditing {
  function setStorageSlot(uint256 _slot, bytes32 _value) public {
    uint x = _slot;
    bytes32 y = _value;
    assembly {
      sstore(x, y)
    }
  }
}

From a truffle console, connected to a ganache instance forking mainnet

a1="0x5346D0f80e2816FaD329F2c140c870ffc3c3E2Ef" // This is our contract
dai="0x89d24a6b4ccb1b6faa2625fe562bdd9a23260359" // This is the DAI contract

web3.eth.getStorageAt(a1, 7)
> '0x0000000000000000000000000000000000000000000000000000000000000003'
web3.eth.getStorageAt(dai, 7)
> '0x44616920537461626c65636f696e2076312e3000000000000000000000000000'

Looks good so far...

ce = await ContractEditing.new({from: f}) // Get f from web3.eth.getAccounts();
ce.setStorageSlot(7, "0xdeadbeef", {from:f})

web3.eth.getStorageAt(ce.address, 7)
> '0xdeadbeef00000000000000000000000000000000000000000000000000000000'
web3.eth.getStorageAt(a1, 7)
> '0x0000000000000000000000000000000000000000000000000000000000000003'
web3.eth.getStorageAt(dai, 7)
> '0x44616920537461626c65636f696e2076312e3000000000000000000000000000'

Again, looks good. Now let's empty the slot

ce.setStorageSlot(7, "0x00", {from:f})
web3.eth.getStorageAt(ce.address, 7)
> '0x'

Great!

web3.eth.getStorageAt(a1, 7)
> '0x'
web3.eth.getStorageAt(dai, 7)
> '0x'

Oh no...

Context

This obviously makes using a forked network in many meaningful ways impossible!

Your Environment

• Version used: Ganache CLI v6.8.1-beta.0 (ganache-core: 2.9.1-beta.0)

[area]

Not very familiar with the forking code, but it looks like this is because the database keeping track of which storage slots have been deleted on the fork is the same for all forked storage trees, but only indexes by key which appears to just be the storage slot.

[area]

I've made a fork that demonstrates a possible fix and modifies the forking deletion test to check for it.

[davidmurdoch]

If you want to go ahead and make a PR I'll pull it in!