- the spec says response to getBlockHashes(from, max) should return all hashes starting from PARENT of from. This required major changes and results in much hackier code.
- Introduced a first round block request after peer introduces with current head, so that hashes can be linked to the head
- peerInfo records currentBlockHash, currentBlock, parentHash and headSection
- AddBlockHashes checks header section and creates the top node from the peerInfo of the best peer
- AddBlock checks peerInfo and updates the block there rather than in a node
- request further hashes once a section is created but then no more until the root block is found (so that we know when to stop asking)
- in processSection, when root node is checked and receives a block, we need to check if the section has a parent known to blockchain or blockPool
- when peers are switched, new peer launches a new requestHeadSection loop or activates its actual head section, i.e., the section for it currentBlockHash
- all tests pass
- AddBlockHashes ignores the first hash (just used to match getBlockHashes query) sends the rest as blocksMsg
- new test TestPeerWithKnownParentBlock
- new test TestChainConnectingWithParentHash
- adapt all other tests to the new scheme
...and make it a top-level function instead.
The original idea behind having EncodeMsg in the interface was that
implementations might be able to encode RLP data to their underlying
writer directly instead of buffering the encoded data. The encoder
will buffer anyway, so that doesn't matter anymore.
Given the recent problems with EncodeMsg (copy-pasted implementation
bug) I'd rather implement once, correctly.
