RETURNDATACOPY may check twice for EVM_INVALID_MEMORY_ACCESS

[ehildenb]

Right now there is an extra rule for RETURNDATACOPY which checks for invalid memory access, but this should already be caught by the #memory function.

We should remove the redundant check.

[dwightguth]

I have absolutely no idea what you're talking about. This rule isn't redundant at all and has nothing to do with the gas usage captured by #memory. It is directly part of ethereum/EIPs#211. Unlike all other COPY instructions, RETURNDATACOPY throws an exception if you try to access an offset in the return data buffer that doesn't exist.

If you remove this rule and side condition, tests are definitely going to fail.

Can we close this?

[ehildenb]

Please tone down your language. I noticed what looked like a redundant rule in the semantics and opened an issue for it, it's not a cause to get mad.

I'm leaving open until I've convinced myself that this is a non-issue. I open the issue so I can keep track of things I'm not prioritizing right now but want to get to eventually.

[ehildenb]

Indeed, the checks are for different cases. Thanks @dwightguth for explaining the difference.