- changed the logging level of some messages that should not appear by
default in a debug build
- most errors in persistWorkItem() are gracefully recovered from
- fixed the handling of out of order blocks
- dropped work items with fewer blocks than what we requested
- getBestBlockNumber(): log and allow exceptions from getBlockHeaders()
- obtainBlocksFromPeer(): moved here the check for block numbers being
in sequence (from nimbus/p2p/chain.nim)
- sendMsg(): log and catch exceptions in callers
- Adding peers to peer pool on incoming connection
- Exclude bootnodes from nodes to connect to
- Check if peer is already connected to before adding to peer pool
- Check if peer is still connected to before removing from peer pool
Enable by compiling with -d:p2pdump. A chronicles log file named
p2p_messages.json will be created in the working directory. This
file will be consumed by the upcoming Chronicles Tail GUI (more
details will be provided on the wiki of this repo).
Other changes:
* Removes the use of package_visible_types (only partially so far)
* Simplifies the new Snappy code a little bit
This also restores the old module structure of having separate
`peer_pool` and `rlpx` modules. This is made possible by the
new Nim package `package_visible_types` (please refer to its
README for an explanation).
Also introduces more error handling in the low-level RLPx routines.
All detected errors will result in immediate disconnection of the
corresponding peer, which may be detected in the protocols though
the `onPeerDisconnected` event handler.
In the next commit, I'm going to squash all the changes in the LES.
This commit just sets the directory tree to its final state, so the
diffs in the follow up commits are easier to review.
Please note that eth.nim was renamed to eth_protocol.nim, because
it's not advisable to have a module name that shares the name of
a protocol identifier (this creates hard to figure out ambiguity
errors)