Stop masking Revert error as TransactionFailed

[kclowes]

What was wrong?

Original issue is here.

I'm not sure what the original rationale was, but the Revert error was getting turned into a TransactionFailed error. I want to test that Revert is raised in web3.py in this PR.

This is my first PR in eth-tester, so I'm not sure if there is something else that I need to do in py-evm, or what sorts of unintended consequences there will be in other libraries as this is a breaking change.

How was it fixed?

Removed TransactionFailed from the decorator that replaces exceptions.

Cute Animal Picture

[kclowes]

On second thought, this is not the change I need. Will open another one!

[carver]

I'm not sure what the original rationale was, but the Revert error was getting turned into a TransactionFailed error. I want to test that Revert is raised in web3.py in this PR.

FWIW, the rationale was to raise an exception that is not back-end dependent. eth-tester was trying to hide the pyethereum vs py-evm backends, so that the user can be agnostic. As py-evm has become more robust, this is maybe less important, and we can just make eth-tester like a convenience toolset with py-evm as the only explicit backend. Then we could pass through things like the underlying py-evm exceptions.

That's a chunk of work that's not done, and probably requires some discussion first.

[kclowes]

Aha, I see. Thanks for the clarification! Is pyethereum still in use? It looks from the readme like it's not...

[carver]

Nope. Really the only reason to keep the backend concept around is if we think it's important to support future backends (a world where py-evm becomes deprecated? or... some kind of c binding into a different evm engine?). I think the chances are small enough that I'm okay abandoning the backend idea. Though it would be nice to chat with Piper before actually digging into implementing that change.