Is state checkpointing in runBlock necessary?

[s1na]

runBlock checkpoints the state before processing the block and reverts on errors. I'm not sure if this is necessary. I think it should throw if there's any error.

The state transition unit in Eth1 is the transaction. There we should revert state on failure. But a block is only a collection of txes which themselves handle checkpointing. If something fails on the block level it's a bug in our code and it shouldn't revert the state.

Also had a quick look at geth and couldn't find this revert on block processing failure. Maybe I'm missing something, but if others agree we should remove this.

[ryanio]

Lines for reference:

https://github.com/ethereumjs/ethereumjs-vm/blob/259e66b8fc65b46351b60541035b8d4e75e38746/lib/runBlock.ts#L99-L107

[holgerd77]

On the block level uncle and miner rewards are attributed and the state updated accordingly. We experimented with removing checkpointing in #1017, but it turned out to be necessary. Will close.