Implement method to submit package of transactions (submitpackage call)

[f321x]

Defines a protocol method blockchain.transaction.broadcast_package for protocol version 1.4.4 and implements it so clients can submit a package of transactions to the mempool using the bitcoin core submitpackage method. This allows clients to broadcast a package of transactions that would on their own not be accepted to the mempool (e.g. below purge fee parent and high fee child to bump)

[SomberNight]

Hmm. Not sure if we want to just forward the return value from bitcoind as-is.
The bitcoin core RPC docs say: This RPC is experimental and the interface may be unstable.
If we do want to just forward this result, I guess we too should mark the result as unstable (?).

[f321x]

I agree that forwarding the result could cause problems in clients if the e-x server upgrades its daemon to a new version and the result structure changes, while defining our own structure would limit the issues to the e-x server side which seems less problematic.

[f321x]

I added a verbose parameter so the user can decide if he wants the full response or only the package_msg and replaced txids. See PR spesmilo/electrum-protocol#1

[f321x]

Simple script to test the call:

import socket


def main():

    with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
        s.connect(("127.0.0.1", 51001))
        s.sendall(b'{"jsonrpc": "2.0", "method": "server.version", "params": ["", "1.4.4"], "id": 0}\n')

        data = s.recv(1028)
        s.sendall(b'{"jsonrpc": "2.0", "method": "blockchain.transaction.broadcast_package", "params": [["raw_parent_tx", "raw_child_tx"], false], "id": 0}\n')

        data = s.recv(1028)
        print(data)


if __name__ == '__main__':
    main()

[f321x]

Commit ff7c1a3 represents the protocol changes in commit spesmilo/electrum-protocol@7bc5f11

[f321x]

7de390f represents protocol pr status spesmilo/electrum-protocol@3e49ee8

[SomberNight]

Thanks! This looks good.
I added some comments but I will fix them myself in a follow-up.

[SomberNight]

no need to import traceback

Suggested change

	except ValueError:
	self.logger.info(f"error calculating txids: {traceback.format_exc()}")
	except ValueError as e:
	self.logger.info(f"error calculating txids", exc_info=e)

---

here is some boilerplate:

import traceback
import logging
import sys

logger = logging.getLogger(__name__)
logger.setLevel(logging.DEBUG)
console_stderr_handler = logging.StreamHandler(sys.stderr)
console_stderr_handler.setLevel(logging.DEBUG)
logger.addHandler(console_stderr_handler)

and now compare (in a fresh local python interpreter):

try:
    txids = [double_sha256(bytes.fromhex(tx)).hex() for tx in tx_package]
except Exception:
    logger.info(f"error calculating txids: {traceback.format_exc()}")

and

try:
    txids = [double_sha256(bytes.fromhex(tx)).hex() for tx in tx_package]
except Exception as e:
    logger.info(f"error calculating txids", exc_info=e)

and

try:
    txids = [double_sha256(bytes.fromhex(tx)).hex() for tx in tx_package]
except Exception:
    logger.info(f"error calculating txids", exc_info=True)

see https://docs.python.org/3/library/logging.html#logging.Logger.debug

[f321x]

Thanks, that's useful to know!

[SomberNight]

Let's either not expose the new method at all, or maybe expose it unconditionally by moving it ~6 lines higher into the big dict.

I don't want to gate it behind an arbitrary temporary version number. Once spesmilo/electrum-protocol#2 is done, we can gate it behind that version number along with the other changes.

Until it is documented in the spec, clients should not rely on this method existing at any version.

[cculianu]

What are these txids for? They are in reversed byte order versus normal RPC interface, so logging them isn't that helpful, because it's not like you can copy-paste them to an explorer to view the txns (unless you manually reverse them first).

Also, they are wtxids, and not normal txids, because you are hashing the raw txn which contains witness data, rather than the stripped txn...

[f321x]

Thanks, good catch. Yeah this is incorrect i remove it in this followup PR #319