Enable state tests for constantinople

[cburgdorf]

What is wrong?

We currently do not run state tests against Constantinople.

How can it be fixed

• Pin the fixtures submodule to a newer version that contains the Constantinople tests (lookup which version geth is pinned against)
• Configure tests to run
• Make tests green

[cburgdorf]

I think I pushed this one as far as I could (see #1181) Both, the Py-EVM and ethereum/tests need to do some work before we can finish that.

[veox]

FTR, I'm trying to tease apart the issues, by first updating the fixtures, in PR #1224.

I've determined the last successful merge commit in ethereum/tests that py-evm still passes (test_state_fixtures in particular, since that's what's failing in PR #1181). Used git bisect run for that, with a script; but it basically amounts to:

source .virtualenv/*/bin/activate
cd fixtures
git checkout develop
git bisect start develop 47b09f42c0681548a00da5ab1c98808b368af49a --
git bisect run sh -c "cd .. && pytest --cache-clear --quiet --exitfirst --show-capture=no -k 'test_state_fixtures'"
git bisect log

I also tried every merge commit from current develop to db8ae9de (also in an automated manner), and they all have at least one failure.

---

I think it'd be best to update the fixtures first before enabling ConstantinopleVM. Will try to step through the state failures one-by-one, to see which end (py-evm or ethereum/tests) is the culprit.

---

EDIT: For ref, here's a gist of merge commits in ethereum/tests that have failures ATM.

[veox]

See #1224 (comment) for status of the approach after PR #1224 was merged.

There were seemingly changes in the test specification format, and it might not be prudent to try and "break through" those ethereum/tests commits without understanding the changes first; and whether they're still relevant, or can be "jumped over".

(FTR, I've laid off working on this.)

[carver]

Confirmed finished in #1579