Do not execute txs on ZQ1 blocks

[86667]

Set ZQ1 block state root hash to zero when converting persistence. For proto networks with incorrect state root hashes currently we set the state root hash to zero upon fetching it.

With there now being blocks without any State in the db we return an error when an attempt to execute transactions on these blocks is made via an API call.

[github-actions]

Bencher Report

Branch	2054-historical-zq1-blocks-should-have-their-state-root-hash-set-to-zero
Testbed	self-hosted

Click to view all benchmark results

Benchmark	Latency	Benchmark Result milliseconds (ms) (Result Δ%)	Upper Boundary milliseconds (ms) (Limit %)
full-blocks-erc20-transfers/full-blocks-erc20-transfers	📈 view plot 🚷 view threshold	1,031.30 (-16.15%)	1,634.95 (63.08%)
full-blocks-evm-transfers/full-blocks-evm-transfers	📈 view plot 🚷 view threshold	434.77 (-10.56%)	697.44 (62.34%)
full-blocks-zil-transfers/full-blocks-zil-transfers	📈 view plot 🚷 view threshold	4,119.30 (+4.95%)	5,127.80 (80.33%)
process-empty/process-empty	📈 view plot 🚷 view threshold	10.32 (+12.89%)	11.51 (89.70%)

🐰 View full continuous benchmarking report in Bencher

[maxconway]

So I think we've still got some functions in the API, e.g. impl TxBlock in api/types/zil.rs, which return state root hashes by calling block.state_root_hash(), which gets a state root hash directly from the block header. Will this address that too or should we also add some calls to is_zq1_block there to zero the hashes if appropriate?

[86667]

So I think we've still got some functions in the API, e.g. impl TxBlock in api/types/zil.rs, which return state root hashes by calling block.state_root_hash(), which gets a state root hash directly from the block header. Will this address that too or should we also add some calls to is_zq1_block there to zero the hashes if appropriate?

the code has changed as discussed in a meeting since this comment was made but its still worth answering to confirm - yes these changes will change all ZQ1 blocks to have state root hashes of zeros.

[JamesHinshelwood]

@86667 Note that the final ZQ1 block shouldn't have a state root hash of zero since that's the only block where we do actually know the state.

[86667]

@86667 Note that the final ZQ1 block shouldn't have a state root hash of zero since that's the only block where we do actually know the state.

Ive pushed so that the final ZQ1 state is written to the empty block we write immediately after ZQ1 blocks, the kind-of genesis used as the high_qc. I think its easiest to understand if we say ZQ1 blocks always have no state and the state at block of conversion is at the block where ZQ2 begins. Would you agree?