state_transition should use SignedBeaconBlock

[mratsim]

Missed interface change in #647

Normally an easy change (don't forget to update nbench #747 and the blockpool)

https://github.com/ethereum/eth2.0-specs/blob/v0.10.1/specs/phase0/beacon-chain.md#beacon-chain-state-transition-function

[JGcarv]

Hello @mratsim

Is anyone working on this yet? Can I tackle it?

[arnetheduck]

All yours! We work off the devel branch for PR's. Let us know if you start working on it.

[JGcarv]

Good! I'll start later today. Thank you

[arnetheduck]

the PR highlights an interesting issue: the invalid_block_sig test is detecting a randao mismatch, so it doesn't detect that we're not validating the block signature - it passes though it shouldn't - this is a problem, because it meant that our test suite did not detect that we were not processing the block correctly - it needs further investigation.

https://github.com/ethereum/eth2.0-spec-tests/tree/master/tests/minimal/phase0/sanity/blocks/pyspec_tests/invalid_block_sig

[JGcarv]

I submitted a new PR finish the remaining work on this ticket.

Also, regarding the issue with invalid_block_sig, my guess is because randao is evaluated before the block sig on the sanity tests, and it was outdated causing it to always fail.

But it would be good to double check once state_transition is completed and #780 is merged!

[arnetheduck]

Also, regarding the issue with invalid_block_sig, my guess is because randao is evaluated before the block sig on the sanity tests, and it was outdated causing it to always fail.

regarding this though, there are two possibilities: the randao is correct and if it wasn't for the signature, the test would pass - or the randao is broken, but nobody noticed because they have their sig check in order. I'd be curious to know which it is - if it's the latter with broken randao, it's not a good signature test really and we need to report it upstream.

[arnetheduck]

it should be possible to test this using https://github.com/protolambda/zcli

[tersec]

it should be possible to test this using https://github.com/protolambda/zcli

If the last commit really is v0.9.4 spec zrnt, then not yet. https://github.com/protolambda/zrnt appears to be on 0.9.4 as well.

[arnetheduck]

obsolete