Add E2E tests for geth and ganache dev clients

[cgewecke]

#2682 (original) and part of #3098 (E2E Proposal)

• Adds shell scripts for testing with geth and ganache development nodes (locally & CI)
• Combines coverage/coveralls measurement for E2E and unit tests in the Node 10 job.
• Adds 25-30 tests covering simple deployments, promievents and transaction signing
• Adds some small mock contracts and their compilation artifacts in test/sources

The tests run across these combinations.....

Client	instamine	automine 2s
geth --dev	(http)	(http, ws)
ganache-cli	(http, ws)

---

Things discovered by the test suite:

• confirmation handler over ws for ganache on deployment does not fire.
  • [Edit: ... was a bug in the tests].
• confirmation handler over ws fires each second for geth automine
  • [Edit: ... now believe this works correctly, have added assertions to test]
• ws for geth instamine is very flaky - possible race condition somewhere.

Tried to make this as simple as possible. I think the CI part will fit into 2.x without too many problems. It's just some small shell scripts.

I used this geth-dev-assistant wrapper to launch geth - it's small, all the APIs are exposed and you can pull recent images with it. I would be into publishing something like this in the web3-js org so anyone with permissions there can edit it as they wish and make necessary improvements.

To do in future....

• Will add Parity in a subsequent PR after this is reviewed and any required changes are made.
• More tests...it would be nice to get coverage up to 90% in 1.x ( like in 2.x )

[coveralls]

Coverage increased (+1.7%) to 84.651% when pulling ca45a8f on cgewecke:tests/e2e-nodes into 4178371 on ethereum:1.x.

[nivida]

Looks good to me! I only dropped some comments to have a more consistency code style.

[holgerd77]

This looks really like a great start, cool! 😄 How long is the additional test/CI run time with this (just out of interest)?

[cgewecke]

@holgerd77 The existing (non-e2e) tests are really fast.. >2500 tests in 40s.

E2E time for only ~30 tests looks like this (including client launch):

Client	Time
geth instamine (w/ skipped ws tests)	26s
ganache instamine	13s
geth automine 2s block interval	1m50s

So...it's slower.

To me it seems like the tests worth writing for E2E are mostly about checking that this complex transaction sending logic executes correctly for both instamine and longer duration blocks. And maybe also that the various client responses are formatted as expected.

[holgerd77]

@cgewecke Ah, thanks, that all sounds pretty fast to me, I am more used to count test time in minutes than in seconds, lol. 😛

[cgewecke]

Added some simple events tests

[cgewecke]

@nivida Have updated - this ready for another look-over...

(Rebased against 1.x to fix a package-lock.json merge conflict)