Byzantium updates #5855

arkpar · 2017-06-16T12:12:15Z

Contains a number of small updates to byzantium EIPs.

arkpar · 2017-06-21T15:10:32Z

pirapira · 2017-08-09T12:25:00Z

Thank you.

debris · 2017-08-27T17:00:36Z

This pr hasn't been updated for > 28 days and I'm closing it as a part of pull request cleanup. If it's still relevant, please rebase and reopen it.

debris · 2017-09-14T09:43:55Z

json/src/vm/vm.rs

-			],
+			"logs" : "0x1dcc4de8dec75d7aab85b567b6ccd41ad312451b948a7413f0a142fd40d49347",
 			"out" : "0x",
+			"network" : "Fronier",


typo, Frontier

debris · 2017-09-14T09:45:14Z

json/src/state/test.rs

+	EIP158ToByzantiumAt5,
+	FrontierToHomesteadAt5,
+	HomesteadToDaoAt5,
+	HomesteadToEIP150At5


what does At5 mean?

Transition at block 5. This comes from json test definition.

debris · 2017-09-14T09:54:26Z

ethcore/wasm/src/runtime.rs

+			},
+			vm::ContractCreateResult::Reverted(gas_left, _) => {
+				trace!(target: "wasm", "runtime: create contract reverted");
+				self.gas_counter = self.gas_limit - gas_left.low_u64();


can it underflow?

I assume it can't because the gas was checked previously. Created branch above follows the same assumption.

a debug assertion could be useful

to be added by @NikVolf

debris · 2017-09-14T09:54:34Z

ethcore/wasm/src/runtime.rs

 				Ok(Some(0i32.into()))
 			},
+			vm::MessageCallResult::Reverted(gas_left, _) => {
+				self.gas_counter = self.gas_limit - gas_left.low_u64();


can it underflow?

I assume it can't because the gas was checked previously. Success branch above follows the same assumption.

I think I will change it to .checked_sub().expect("reason")

debris · 2017-09-14T09:57:07Z

ethcore/vm/src/schedule.rs

 	/// Kill basic accounts below this balance if touched.
 	pub kill_dust: CleanDustMode,
+	/// Enable EIP-86 rules
+	pub eip86: bool,


This property is always set to false. Is it intended?

Yes, it will be enabled in Constantinople

debris · 2017-09-14T10:03:54Z

ethcore/src/executive.rs

-					Ok((params.gas - cost, ReturnData::empty()))
+					Ok(FinalizationResult {
+						gas_left: params.gas - cost,
+						return_data: ReturnData::new(output.to_owned(), 0, output.len()),


isn't it dangerous to copy the output (via to_owned)? iirc, we've always avoided that at any cost

There is no way around it. output here a slice into callers memory. The built-ins don't have their own evm memory and the result has to be kept around after the call in return data buffer.

debris · 2017-09-14T10:07:33Z

ethcore/src/json_tests/executive.rs

 				let vm_output: Option<Vec<u8>> = vm.output.map(Into::into);
 				fail_unless(Some(output) == vm_output, "output is incorrect");
+				// TODO: verify logs
+				//fail_unless(Some(res.logs) == vm.logs, "logs are incorrect");


what about this TODO?

debris · 2017-09-14T10:09:16Z

ethcore/src/spec/spec.rs

 		schedule.have_return_data = block_number >= self.eip211_transition;
 		if block_number >= self.eip210_transition {
-			schedule.blockhash_gas = 350;
+			schedule.blockhash_gas = 800;


this value is also set here https://github.com/paritytech/parity/pull/5855/files#diff-368d67dfd535628837d23122934dc5c3R200 . imo, it should be a const

The latter function is used only in some tests and the value should not really be set there. I'll remove it

debris · 2017-09-14T10:10:45Z

ethcore/src/state/mod.rs

+		let eip658 = env_info.number >= engine.params().eip658_transition;
+		let no_intermediate_commits = (env_info.number >= engine.params().eip98_transition
+			&& env_info.number >= engine.params().validate_receipts_transition)
+			|| eip658;


it would be more readable (and efficient) to change the order of those checks and move eip658 || ... to the front

tomusdrw · 2017-09-14T09:25:33Z