regarding the dht, in the meeting swapping seqNo for ttl was discussed. After attempting to implement this change, the tests fail due to the lack of granularity, eg NodeA is instantiated, then NodeB is instantiated with a different ip/port and with NodeA as its previous record. The seqNo of both will be the same as more than a second hasn't elapsed, meaning that the record for NodeB would never been seen as newer.
Maybe the cheapest solution here would be to retain the seqNo field and add a ttl field that can be checked separately for expiration? Or if we only want one field, make it nanosecond resolution?
Swap all instances of Record with SignedPeerRecord.
Allow for `SignedPeerRecord`s to be updated by updating the first multiaddress in the `PeerRecord`. This also increments the `seqNo` in the `PeerRecord` only if the address was actually updated.
Merge the add/get providers messages with the discovery v5 messages inside of the discovery v5 code. Discovery v5 uses RLP encoding, while the add/get providers uses protobufs, as per the libp2p spec. The merge of these two uses RLP on the outer wrapper of the message, and a protobuf-encoded blob on the “inside” for the add/get providers messages only. Eventually this needs to change so that we use protobufs only.
There is still some code in the libp2pdht/dht directory, which is being read from the discovery v5 directory. Eventually these two should merge as well.
Add CI workflow to run tests.
Add code coverage workflow that publishes results to CodeCov.
Add CI and coverage badges to README.
Bump asynctest to allow for `setupAll` and `teardownAll`.
Add local coverage nimble task.
Combine tests in to a testAll file.
Project has been updated to use nimble as a build system. All required dependencies have been added.
All nim-eth mirrored files were added to a private folder in the libp2pdht module. A libp2pdht/discv5 module was added to alias the nim-eth modules (which will change over time).
Test have been updated to use status-im/asynctest. This PR uses a branch of asynctest that supports async suite before/after. This seemed like the only the tests would work without throwing gcsafe errors.
All tests working.
Protocol was actually made of two sub-protocols.
* a lower-half handling authentication, encryption,
key exchange, and request/response. This is now called
Transport.
* an upper-half handling DHT messages. This is still
called Protocol.
Separation of these two reduces dependencies and simplifies
modifications to the protocol.
Signed-off-by: Csaba Kiraly <csaba.kiraly@gmail.com>
