Joinmarket v0.9.9: Tor Onion Messaging Changes and Several Bugfixes
JoinMarket is software to create a special kind of bitcoin transaction called a CoinJoin transaction. Its aim is to improve the confidentiality and privacy of bitcoin transactions.
Changes to directory node defaults
We have removed one of the directory nodes that was proving unresponsive, and added a new one. These of course are only defaults; this list may change quite often (more often than the IRC list used to, probably!). We will investigate ways to let users find other directory nodes than just this list, in future.
We also increase the timeout for making the initial connection, as onion service connections can sometimes be very slow.
Note that this is also related to the following topic (slowing down connection reattempts).
8ab2d50Increase directory node connection timeout to 60s
7aebed7Remove jmdirjmioywe2s5jad7ts6kgcqg66rj6wujj6q77n6wbdrgocqwexzid.onion directory node from default config
Slow down connection reattempts to directory nodes
Several users reported not only a failure to connect to directory nodes, but also a very rapid cycling, attempting to reconnect every couple of seconds. Here we fix that bug, and ensure that connection attempts back off exponentially, eventually slowing down to once per ~ 3 hours. This will clear up spam from the log, but also may help, if the problem is related to DOS limits in Tor (further investigation is required), once all users update to 0.9.9. This is one of the main reason users are encouraged to update immediately.
9a412d8Connections to dnodes slow down
New public orderbook
Provided by @kristapsk, added into README. As usual, users are advised that they can easily run their own orderbookwatcher on localhost, and this will be better in most situations.
34fa1b6Add link to my public orderbook mirror
Correct errors in transaction size estimation
This set of commits corrects pre-existing errors in estimation of transaction size, for nonstandard outputs. These errors were usually tiny and didn't matter, but caused problems for cases where a very low sats/vbyte was set by the user. In particular, the cases of spending to or from fidelity bond addresses (which are of type
p2wsh), or spending to taproot addresses (
p2tr), were not being calculated correctly.
We also added tests of our transaction size and weight estimation.
There can still be edge cases if the input or output type is unrecognized (most plausible case would be a customized script being spent, in a PSBT; it's not really possible for this to happen with Joinmarket's own coins), but the error in size (and therefore fee) estimation will still be small, even in these very exceptional cases.
bffad33Tx size estimation allows varied in, out types
Fixes to sendpayment
Two bugs in sendpayment were fixed:
Allowing sub-27K sats for non-coinjoin payments (our
DUST_THRESHOLD is not "the" Bitcoin dust threshold, it was deliberately set higher to account for certain estimation uncertainties related to coinjoin; but they are irrelevant for non-coinjoins):
d6d40dfAllow amounts below DUST_THRESHOLD (0.00027300 BTC) for non-cj direct sends
Fixes a bug where sending to a BIP21 URI fails (bug as described here).
9d9035bFix sendpayment without amount argument, BIP21 only
Switch to venv from virtualenv
venv is provided as standard in all versions of Python that we support, so we don't actually need the dependency
virtualenv. This removes it; there is no impact at the user level.
11ddec7docs: Remove now-confusing mentions of virtualenv
6e3b6ecDockerfiles: Use Python venv instead of virtualenv
8bbca60install: Use python3's venv module when available
Bugfixes and minor changes
This fix prevents directory nodes crashing in certain edge cases:
bfb0e25Account for missing nick fields in OnionPeer
This fix ensures fee randomness works even when a fee estimate can't be sourced from Core:
3d0f2d6 fix: randomize fallback transaction fee
2e44f00 Log IRC connection attempts when using SOCKS5 proxy or TLS too
2749da1 change payjoin default binding port
This was a rare to trigger, but pernicious bug: a user's wallet could show incorrect balances, if the RPC calls to the blockchain failed (the code in
sync_unspent was incorrectly ignoring this failure), this is no longer allowed:
0b34e0bAbort sync_unspent if blockheight RPC call fails
44aea32Fix typo in
b34b932Correction to docs: no coinjoin spends of FBs
4c53bb5Update link to RaspiBolt installation guide
Installation and dependencies
63d74bcCall num_cores() only after deps_install()
7d0ca22Update local Tor to 0.4.7.12
01dc08aUpdate local Tor to 0.4.7.13
e47f082Allow to run install.sh from any current working directory
46013d2chmod +x scripts/jmwalletd.py
8a25e3aPin remaining deps to specific versions, except for Qt and tests
22c13b0Remove unused code instead of commenting out
Mostly applies to the shell scripts used for installation, hence included here:
b928713 Add ShellCheck linter script
4f0eebc Apply all current shellcheck suggestions to rest of the scripts
f0b9872 Apply all current shellcheck suggestions to install.sh
bd5508c Minor improvements for question prompts and feedback
5b8ef40 jmbitcoin: eliminate dependency on urldecode
1e159df Update bencoder.pyx to 3.0.1
61b6400Tests: bind to any first free TCP ports instead of hardcoded ones
1370c12Use different jm_test_datadir for each local user
06c29c2Keep trying to receive websocket notification in test
3c8f247CI: Run tests on both Linux and macOS
c4d9b92tests: make setup fixtures optional.
980edd6CI: Add Python 3.11
7b4c42aRemove dead code
8d2a664Fix for recent Bitcoin Core versions and add type hints
b59fdcdRemove Python 3.6 (EOL), add 3.10
890dd50Bump pytest version to 6.2.5