wallet: RBF batch payments manager #9298
Conversation
ecdsa
force-pushed
the
batch_payments_manager
branch
from
b71b11f to
45c7543
Compare
ecdsa
force-pushed
the
batch_payments_manager
branch
13 times, most recently
from
a48e7ca to
5c7a247
Compare
ecdsa
force-pushed
the
batch_payments_manager
branch
13 times, most recently
from
3bdbf87 to
3ad4406
Compare
ecdsa
force-pushed
the
batch_payments_manager
branch
4 times, most recently
from
4a623fe to
0f27755
Compare
ecdsa
force-pushed
the
batch_payments_manager
branch
4 times, most recently
from
c065b4a to
689d93e
Compare
| cls.sign_tx(tx, swap) | ||
| return tx | ||
| self.add_txin_info(swap, txin) | ||
| txin.privkey = swap.privkey |
There was a problem hiding this comment.
see sign_transaction in wallet.py
| outputs=outputs, | ||
| password=password, | ||
| locktime=locktime, | ||
| BIP69_sort=False, |
There was a problem hiding this comment.
when sweeping htlc outputs with sighash single-anyonecanpay, the order must be preserved.
There was a problem hiding this comment.
oh right, the input-output pairings need to be kept for those.
we should have a unit test that would fail if this was set to True
| self.config.FEE_EST_DYNAMIC = False | ||
| self.config.FEE_EST_STATIC_FEERATE = 1000 | ||
|
|
||
| def test_claim_tx_for_successful_reverse_swap(self): |
There was a problem hiding this comment.
Why delete these tests?
Just because SwapManager._create_and_sign_claim_tx is being removed?
The tests could be adapted to test create_claim_txin instead. Maybe feed the txin to wallet.create_transaction along with some hardcoded values.
The point is to test that the txin is constructed correctly, but the easiest way of doing that is comparing a full tx.
There was a problem hiding this comment.
The point is to test that the txin is constructed correctly, but the easiest way of doing that is comparing a full tx.
Creating a full tx in order to check the construction of a tx input is indeed the easiest way, but it is completely overkill.
I think we should try to put a little bit more efforts in writing tests.
| # To make payments reorg-safe, we would need to persist more data and redo failed payments. | ||
|
|
||
|
|
||
| class TxEngine(Logger): |
There was a problem hiding this comment.
Could we put "RBF" or "batch" into the name? E.g. I find "RBFEngine" or "TxBatcher" more intuitive.
There was a problem hiding this comment.
ok let's call it TxBatcher
| outputs=outputs, | ||
| password=password, | ||
| locktime=locktime, | ||
| BIP69_sort=False, |
There was a problem hiding this comment.
oh right, the input-output pairings need to be kept for those.
we should have a unit test that would fail if this was set to True
ecdsa
force-pushed
the
batch_payments_manager
branch
8 times, most recently
from
b706662 to
22b3527
Compare
- delegate tx creation/broadcast to the wallet - callers use add_payment_output and add_sweep_input tests: - remove test_sswap.py: cannot be tested that way anymore - force regtests to use MPP, so that we sweep transactions with several HTLCs. This forces the payment manager to aggregate HTLC tx inputs.
- use separate pools, - fix future tx
- if coin control is active, disable batching - forward submarine swaps can be batched
ecdsa
force-pushed
the
batch_payments_manager
branch
from
9691a83 to
7a40399
Compare
note: this is not "reorg safe", meaning that there is no guarantee that all payments will end up in in the blockchain.
However, I believe this is "double send safe", meaning that we will never send a payment twice.