Run blockchain tests

[holgerd77]

This is now finally the usage of the mechanism running the blockchain tests on CircleCI by renaming the config files.

[holgerd77]

Have retriggered a test run on VM e24909e latest commit on master, ethereumjs-testing v1.2.4.

[holgerd77]

Result blockchain tests:

1..848
# tests 848
# pass  526
# fail  322

Too much.

[holgerd77]

Many of the failures above seem difficulty related. I tested:

node ./tests/tester.js -b --test='notxs_Byzantium'

This fails with Error: invalid POW, thrown from within the blockchain class and can be made explicit (normally hidden) by adding console.log(err) in the BlockchainTestRunner.js test class after blockchain.putBlock() (respectively in the callback).

This doesn't necessarily mean that the latest updates from the blockchain or block classes have introduced an error. There were lately various discussions about wrong difficulty values in the blockchain tests on tests Gitter, maybe these test cases are wrong (one could compare with an older snapshot blockchain test from ethereumjs-testing). I have also tested directly within ethereumjs-block in the node_modules folder, byzantium hardfork setting is passed correctly and the difficulty tests are all passing if one goes to the repo and checks.

[holgerd77]

Have investigated this further, this fails in ethashjs in Ethash.prototype._verifyPOW in the mixHash comparison, a.mix.toString('hex') === header.mixHash.toString('hex') evaluates to false, with the following values:

1639a35bab884e5d9b095bd860d6f776768b9be157275792a758e1e3909308e0
0000000000000000000000000000000000000000000000000000000000000000

The second one is coming from the mixHash field in the blockchain tests from ethereum/tests repo. I cross-checked an older version, before it had the correct hash, so this is definitely an error over on the test repo. Will file an issue there.

[holgerd77]

Have filed an issue here: ethereum/tests#530

[holgerd77]

Retriggered a test run after merging of #373, also to get an impression about changes in blockchain test run time after introduction of sealEngine usage (things should speed up).

[holgerd77]

Ok, this is now back to "normal" 60 failing tests (actually 6 less than before), see also #338 for comparison and context, some results:

1..848
# tests 848
# pass  788
# fail  60

Execution speed of test run on CircleCI is now down to a whopping 12:39 min, which is fantastic since this means we can now (after getting test failures down) run this on every PR CI run!!! ✨ 😄

[holgerd77]

Now obsolete to run periodic blockchain test runs through merging of #374.