Remove leaf hashing in MerkleRoot function

[wemeetagain]

Description

Previously, the MerkleRoot function hashed the leaves provided as input.
This is in line with merkle hashing, generally speaking, however the spec's definition of the merkle_root function makes an optimization by assuming that the leaves are pre-hashed, and therefore not hashing the leaves.

This update removes the hashing of the leaf nodes from the MerkleRoot function and updates the MerkleRoot test.

See ethereum/consensus-specs#646 for clarification.

[CLAassistant]

All committers have signed the CLA.

[nisdas]

I would rather than not remove leaf hashing as we would still have to do it outside the function to generate the merkle root. Ex:

	
	for i, v := range values {	
		hashedValue := Hash(v)	
		values[i] = hashedValue[:]	
	}
       
        root := MerkleRoot(values)

I rather do the hashing in the function than outside the function and have to add boilerplate code whenever we have to calculate the merkle root.

[wemeetagain]

@nisdas

The thing is, these leaves are not just hashed directly and not hashed immediately before being put into MerkleRoot. They are currently stored, pre-hashed (I believe one by one over time), in the BeaconState.

To get specific, the MerkleRoot function is currently used only once, here, to hash state.LatestBlockRootHash32S.

state.LatestBlockRootHash32S are already hashed (they are block root hashes), and you would presumably need to gather the actual underlying blocks first if you wanted to keep the leaf hashing logic inside MerkleRoot.

The issue is less about whether or not to keep a MerkleRoot function that hashes the leaves, and more about this specific instance, where you most likely will need a function that returns a merkle root, but does not hash the leaves.

[nisdas]

@wemeetagain In that case it makes sense :) . Could you add a comment stating why we are not hashing the leaves in the function and are assuming that the leaves are pre-hashed

[nisdas]

LGTM , thanks for this !

[codecov]

Codecov Report

Merging #1653 into master will decrease coverage by 0.64%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #1653      +/-   ##
==========================================
- Coverage   71.49%   70.85%   -0.65%     
==========================================
  Files          98       97       -1     
  Lines        7010     6687     -323     
==========================================
- Hits         5012     4738     -274     
+ Misses       1553     1523      -30     
+ Partials      445      426      -19