A contract should revert when no fn selector matches

[AlicanC]

ATM, when you specify an invalid function selector when calling a contract, the contract just ends up doing a RET $r0. A revert sounds better.

[SilentCicero]

@AlicanC this should be easy to implement and important for consistency.

[sezna]

We seem to have a disagreement here with this comment:

#97 (comment)

@adlerjohn -- what do you think? RVRT or RET with some error code that represents "no selectors matched"? If the latter, we should make an issue to standardize error codes.

[adlerjohn]

I personally prefer returning a Result or something. Then the caller can handle as it wants, instead of forcing the panic. That being said, that might involve the compiler knowing about the Result in the standard library, which maybe is a leaky abstraction?

RET $zero is definitely wrong though. So pending a Result, RVRT is fine.

[sezna]

I agree that using Result directly would be leaky, but what Rust does is it has a trait which types can opt into for converting program error types into standard out return codes. E.g., Result<_, _> will return a non-zero exit code if it is an Err.

We could do something like that, but unless we have a separate opcode or domain separator metadata, it will be impossible for the caller to know if the contract wanted to return that integer value as a genuine result of computation or as an error code.

There's a lot of ambiguity here, I think the cleanest and safest option would be RVRT. Thoughts?

[adlerjohn]

Cleanest is definitely RVRT for now. Can revisit in the future.