Validate stack items after operations

[axic]

This helped find some of the bugs already fixed, though it may be useful in the future too.

[axic]

This needs to be updated after #219 (or vice versa).

[jwasinger]

Hey @axic @Silur I've looked at this PR twice trying to figure out what the problem is. I'm thinking perhaps we can let this PR sit because the code it touches will become irrelevant after EWASM is implemented. Thoughts?

[axic]

I think this will work well after #219 is properly finished and this one rebased.

[jwasinger]

I took another look at this one. Still couldn't figure out what is causing the callback error that we are seeing here.

gotta love async....

[axic]

@jwasinger probably missed one of the functions since there were so many rebases over 1.5 years :)

[axic]

I think I should have fixed this now.

[hugo-dc]

Hi @axic , @jwasinger ,

I looked at this PR, I found out that SWAP opcode has this configuration:

//         name, baseCost, offStack, onStack, dynamic
0x90: ['SWAP', 3, 0, 0, false],

onStack is 0, swap opcode returns the swaped element https://github.com/ethereumjs/ethereumjs-vm/blob/master/lib/opFns.js#L483 , which has to be pushed again to the stack.

I changed onStack from 0 to 1 for the SWAP opcode, and the failed tests went from 868 to 43.

The remaining failing tests are related to the async opcodes (CALL, CALLCODE, and DELEGATECALL)

[axic]

@hugo-dc good find! I think though that swap is broken and it shouldn't return the new top stack item. It should just do the stack swap internally (since we don't have a better code base).

[axic]

@hugo-dc @jwasinger rebased after the other fix, this should work now?

[Silur]

isn't INTERNAL_ERROR is relevant in #219 where there was runState.vmError having a non enum value exception?

[hugo-dc]

@axic I confirm all state tests are passing after the fix 👍