Current tests for the beacon API are of sub-optimal quality:
Very limited coverage
No assurance that all and future new endpoints are tested
Require import stubbing
Extensive use of brittle stubs: which require big diffs when modifying unrelated cost.
In my subjective opinion the current test suite has a very high cost of maintenance with very little return in ensuring the API is correct.
Proposed solution
Proper tests for a the beacon API should:
Have full coverage
Ensure programmatically that all endpoints current and future are tested
Be tested against a full BeaconChain instance without stubbing and with real data. We are already doing this for BeaconChain validation and such data is extremely fast to produce and to run tests against.
dapplion
commented
2 years ago
What is your question?
Current tests for the beacon API are of sub-optimal quality:
In my subjective opinion the current test suite has a very high cost of maintenance with very little return in ensuring the API is correct.
Proposed solution
Proper tests for a the beacon API should: