Use error strings in smart contracts

[czepluch]

Description

Since solidity 0.4.22 error strings are supported for revert and require:

pragma solidity ^0.4.22;

contract VendingMachine {
    function buy(uint amount) payable {
        if (amount > msg.value / 2 ether)
            revert("Not enough Ether provided.");
        // Alternative way to do it:
        require(
            amount <= msg.value / 2 ether,
            "Not enough Ether provided."
        );
        // Perform the purchase.
    }
}

Adding this to the smart contract will make it easier to debug where things go wrong in a function. It also makes the terrible work flow of removing requires to figure out which require is failing obsolete.

Tasklist

• SecretRegistry - no requires
• TokenNetworkRegistry (Add error messages to TNR #1498)
• TokenNetwork (TokenNetwork error messages #1499)
• Controllable (Allow disabling limits after deployment #1492)
• MonitoringService
• OneToN
• ServiceRegistry
• UserDeposit

[LefterisJP]

@loredanacirstea Do you think you can add them in this contract change (red eyes testnet 4)? I would say it's good to take advantage of the redeployment to include them.

[loredanacirstea]

Yes, added the milestone now.

[loredanacirstea]

Just some notes:

• revert error messages are not currently supported by web3.py - Return values for eth_call are truncated when using revert ethereum/web3.py#941. When they will be supported, in order to get them, one would need to eth_call the contract function and get the encoded error.
• there are EIPs in discussion about standardizing status codes for errors (similar to http), that can also be handled in the EVM - https://eips.ethereum.org/EIPS/eip-1066; This is worth following in my opinion, as it could improve the automatic interaction between smart contracts.

Aside from these mentions, I am ok with just using some error strings right now and re-evaluating them after standards are clarified, because I assume it will take a while.

[loredanacirstea]

The revert error strings cannot be implemented right now in a satisfactory manner, as commented in #213 (comment)
I would move this to the Ithaca milestone and address it after #215.

@LefterisJP ?

[LefterisJP]

@loredanacirstea Was looking at the Testnet 4 milestone and was wondering why this is not there yet. I understand now. Sure let's do this. It's not a requirement right now at all anyway.

[pirapira]

P2 because I heard error strings would have saved our time today.

[pirapira]

If we recover cooperativeSettle() and add a two-character comment on each of require, the runtime codesize will be 28835 bytes, which exceeds the deployable limit of 24576 bytes.

[pirapira]

Whether we optimize or not, any message that fits 32 bytes seems to compile into the samely sized runtime code. So no need to go for cryptic "a" "b" "c" style messages.

[pirapira]

I’ll just add somewhat recognizable comments under 32 bytes.

[pirapira]

This depends on #401.

[palango]

Just tried the old PR again, we're still running into code size issues.

Blocked by #401

[karlb]

See #1435 for more space saving.