Native contract executor

[PearsonWhite]

Corresponding sequencer PR: NethermindEth/sequencer#19

PR: #2156

[xrvdg]

ContractExecutor::new does not do any caching to disk. The compile_and_load, load_compiled_contract and others needs to be reworked with ContractExecutor::save and ContractExecutor::load.

[PearsonWhite]

ContractExecutor::new does not do any caching to disk. The compile_and_load, load_compiled_contract and others needs to be reworked with ContractExecutor::save and ContractExecutor::load.

ContractExecutor's entrypoints are not public, so we must either:

1. Make them public. Then we can remove entry_points and Hash from NativeContractClassV1. (It looks like NativeContractClassV1 doesn't use PartialEq for anything). This would make NativeContractClassV1 a nothing more than a wrapper around ContractExecutor.
2. Create save/load functions for NativeContractClassV1 to save/load class hash. The de/serialization would de/serialize hash, entrypoints, and then use (executor.save/load).

[xrvdg]

ContractExecutor's entrypoints are not public, so we must either:

Those don't have to be public, it's handled by the load/save of ContractExecutor. The NativeContractEntryPoints are there to have a translation for starknet selector to function id. We are in talks with LambdaClass to have the interface different, but for now it is like this.

It looks like NativeContractClassV1 doesn't use PartialEq for anything

The PartialEq follows from the PartialEq on ContractClass which is due to a test.

[xrvdg]

Two points to address:

• caching has no effect if it's always invalidated
• format the code with cargo fmt.

and the minimum supported Rust version is now 1.80.1.

[xrvdg]

This will always compile and save the contract as the argument to unwrap_or is eagerly evaluated. Use unwrap_or_else instead.

[xrvdg]

Furthermore, load and save needlessly borrow the library_output_path.

[PearsonWhite]

This will always compile and save the contract as the argument to unwrap_or is eagerly evaluated. Use unwrap_or_else instead.

Changed to or_else so I could still use "?" operator. Had to specify the type though since it couldn't figure it out.

Furthermore, load and save needlessly borrow the library_output_path.

Fixed.

[xrvdg]

The change itself looks good and I've approved it. But before merging look into the following regression first.

Block 633333 fails with:

execute transaction #289: failed txn 1132083247661400189175640804679018936217786098286704013683829309015708481258 reason: L1 gas bounds (max amount: 1038, max price: 67469438047820) exceed balance (44664767982987702).

[PearsonWhite]

The change itself looks good and I've approved it. But before merging look into the following regression first.

Block 633333 fails with:

execute transaction #289: failed txn 1132083247661400189175640804679018936217786098286704013683829309015708481258 reason: L1 gas bounds (max amount: 1038, max price: 67469438047820) exceed balance (44664767982987702).

Noted. I think it's fine for now as the gas changes will break things until we have all the starkware changes in our repo.