Pass raw data with ContractLogicError

[kclowes]

What was wrong?

People frequently request the raw data to be sent back with the ContractLogicError.

Closes #2793, #2336

How was it fixed?

This PR sends back the raw data as an attribute on the error class without breaking the existing API.

Todo:

• Add entry to the release notes
• Documentation

Cute Animal Picture

[fselmo]

Oh I like this approach in general! The more information the better on exceptions 👌🏼.

I think we should be able to add this to the OffchainLookup exception also which would be some nice information to have available.

[kclowes]

@fselmo this assertion fails if it's run in isolation. the OFFCHAIN_LOOKUP_RETURN_DATA changes, but only fails that way on eth-tester, not with the geth integration tests. Do you have any ideas why the OFFCHAIN_LOOKUP_RETURN_DATA would be different when the whole file is run vs. run in isolation?

[fselmo]

Oh, yeah. The very first value that's in the OffchainLookup revert in the contract is the particular contract's address which will differ between the contract deployed via the geth fixture and the one for eth-tester. You'll just have to change the beginning of the constant value for OFFCHAIN_LOOKUP_RETURN_DATA for the eth-tester tests to match that.

It should only differ in that and I checked that:

>>> e.value.data[:10] + e.value.data[74:] == OFFCHAIN_LOOKUP_RETURN_DATA[:10] + OFFCHAIN_LOOKUP_RETURN_DATA[74:]
True

first 10: "0x" (2 indexes) + 4 byte function selector (8 indexes)
next 64 should be abi-encoded address value (32 bytes * 2 indexes per byte)
[74:] is the rest

so the differing 32 bytes should be indexes [10:74] in the string.

[kclowes]

Aha, thank you! I'll update the assertion so that it only asserts on e.value.data[:74] or at least skips the middle address section.

[fselmo]

Or just update [10:74] to match the address that deploys to the eth-tester fixture

[pacrob]

lgtm!