fix(rpc): use gas_limit as gas_used if tx failed

[wetkeyboard]

closes #18871

set gas_used=gas_limit for the failed tx

[wetkeyboard]

@mattsse please take a look

[mattsse]

can you elaborate on why this fixes the issue?

unclear to me why gas used needs to be swapped with gaslimit here

maybe @rakita knows

[rakita]

Hm maybe the expectation for gas returned is to return the max spent gas for the Halt, as halt spends all the gas. For Revert and Success it is not gas limit.

[mattsse]

@rakita shouldn't res.gas_used always have the actual gas used by the tx?

so unclear why this is different here

https://etherscan.io/tx/0x2a4b68d823b4468df290508f8466feaa588e8127b5680c0ebc99d22b4b0fed69

[wetkeyboard]

Sorry, this fix is incorrect, I think it's a bug in revm, @rakita could you please take a look at bluealloy/revm#3100

[rakita]

This should be if res.result.is_halt() { gas_limit } else { res.result.gas_used() };

Revert version of return does not consume all the gas

[rakita]

@mattsse i think, the assumption is that Halt would take all the gas (This is how it is done for subcalls, taking all gas means using gas limit), but for first call we return used gas and not all the the spend gas.

Etherscan link that you shared shows last used gas, so maybe it is fine to leave as it is. Not sure if this Fail means REVERT opcode is called or it is Halt, diff is that REVERT would not spend all the gas.

It would be good to check how alloy/evm handles this,