lnpeer (and interface) response-handling-code should not run in the
network main_taskgroup as the remote could force an exception
to be raised and that would kill the whole network instead of just the peer
e.g. Alice sends upd1, upd2, upd3, commitment_signed, upd4, upd5.
Bob receives all of these; and sends a revoke_and_ack but there is
a disconnect before Alice receives the revoke_and_ack.
During reestablish, if Bob claims to have received the commitment_signed,
Alice must not replay the msgs before that; but she should replay upd4 and upd5.
In recv_rev() previously all unacked_local_updates were deleted
as it was assumed that all of them have been acked at that point by
the revoke_and_ack itself. However this is not necessarily the case:
see new test case.
renamed log['unacked_local_updates'] to log['unacked_local_updates2']
to avoid breaking existing wallet files
Even if we haven't signed them yet (did not send commitment_signed).
Alternatively, if they are not yet signed, we could discard them here,
like we do already for remote updates above (chan.hm.discard_unsigned_remote_updates).
One of these two options must be done, and before this commit we were not doing either.
"current" used to be "oldest_unrevoked"; and pending was "oldest_unrevoked + 1"
but this was very confusing...
so now we have "oldest_unrevoked", "latest", and "next"
where "next" is "latest + 1"
"oldest_unrevoked" and "latest" are either the same or are offset by 1
(but caller should know which one they need)
rm "got_sig_for_next" - it was a redundant sanity check, that really
just complicated things
rm "local_commitment", "remote_commitment", "set_local_commitment",
"set_remote_commitment" - just use "get_latest_commitment" instead
Replay un-acked local updates (including commitment_signed) byte-for-byte.
If we have sent them a commitment signature that they "lost" (due to disconnect),
we need to make sure we replay the same local updates, as otherwise they could
end up with two (or more) signed valid commitment transactions at the same ctn.
Multiple valid ctxs at the same ctn is a major headache for pre-signing spending txns,
e.g. for watchtowers, hence we must ensure these ctxs coincide.
- remove remote_commitment_to_be_revoked
- pass old ctns to lnsweep.create_sweeptxs_for_watchtower
- store the ctn of sweeptxs in sweepStore database
- request the highest ctn from sweepstore using get_ctn
- send sweeptxs asynchronously in LNWallet.sync_with_watchtower
- add functions to lnsweep
- lnworker: analyze candidate ctx and htlc_tx
- watchtower will be optional
- add test for breach remedy with spent htlcs
- save tx name as label
main_loop should dump traces of unexpected exceptions to log.
Coroutines/functions invoked inside main_loop should simply propagate it
up the chain.
Typical exceptions are handled in handle_disconnect without dumping the trace.
- fix bug in lnsweep: lnwatcher transactions were indexed by prev_txid
- add test for breach remedy with unsettled htlcs
- add timeout option to lnpay, and replace DO_NOT_SETTLE with SETTLE_DELAY
so that we can read intermediate commitment tx in regtest
SomberNight
ThomasV