This removes a lot of the craziness that was added during the 0.21
rebase 240f03f586 (Bitcoin PR #17290)
as well as a lot of other upstream craziness. It's a very positive
change but will probably be a bear to review.
There are two places where I (purposely) nontrivially changed the
logic beyond what the upstream PR did:
1. When setting the change output positions, I used a uniform
distribution for all change outputs and also obeyed the user's
`change_position` choice. The old code was non-uniform and
had a bug where it would offset the change position.
For non-policy assets, I do not create zero-sized change
since we would just remove them later. For the policy asset,
I do sometimes create zero-sized change since the upstream
logic expects it to exist.
2. Rather than "reblinding whenever something changes" we blind
once for size/fee estimation purposes and then again at the
end after all potential adjustments have been made. The
resulting code should be closer to upstream.
We should really redo this code though because our use of the
`BlindDetails` structure is very confusing and stateful.
3. We consider whether or not we're using CT/CA when estimating
the size of change outpust. This is because we now use branch
and bound far more often (yay!) which selects coins such that
there will be no change output. This is implemented however
by adding a change output then dumping it to fees ... meaning
that if we dramatically overestimate the size of a change out
then we'll end up dramatically dumping too many coins to fees.
There are another couple small things, which I apologize for .. this
PR took me 2 hours to get compiling and 11 hours to get working(!!)
and some things got away from me.
One fun thing is that I changed CWallet::SelectCoinsMinConf to gate
an addition on coin_selection_params.m_subtract_fee_outputs ... this
is actually an upstream bug that I'll file whenever I get around to
producing a test that triggers it. (A bit hard as I have to gin up
a scenario where branch-and-bound fails to hit this codepath.)
Also, in test/functional/wallet_bumpfee.py I had to change the feerate
required to force selection to add another input from 500 to 800. I
don't know why and I don't care to figure it out.
This is a really non-trivial wallet refactor "effective value" which has been
in the Core queue for a pretty long time. There were a ton of merge conflicts
but they were mostly mechanical (all the benchmarks) or "take the Core side
of the diff" (most of wallet.cpp).
The two nontrivial things were:
* In `SelectCoinsMinConf` in wallet.cpp I had to slightly move the asset
filtering logic (but I didn't change it)
* In `OutputGroup::Insert' in coinselection.cpp I changed the effective
value computation to only consider policyAsset values
Instead of hijacking the effective_feerate to use the correct value
during coin selection, have OutputGroup be aware of whether we are
subtracting the fee from the outputs and provide the correct value to
use for selection.
To do this, OutputGroup now takes CoinSelectionParams and has a new
function GetSelectionAmount().
Although the CreateTransaction loop currently remains, it should be
largely unused. KnapsackSolver will now account for transaction fees
when doing its selection.
In the previous commit, SelectCoinsMinConf was refactored to have some
calculations become shared for KnapsackSolver and SelectCoinsBnB. In
this commit, KnapsackSolver will now use the not_input_fees and
effective_feerate so that it include the fee for non-input things
(excluding a change output) so that the algorithm will select enough to
cover those fees. This is necessary for selecting on effective values.
Additionally, the OutputGroups
created for KnapsackSolver will actually have their effective values
calculated and set, and KnapsackSolver will do its selection on those
effective values.
Lastly, SelectCoins is modified to use the same value for preselected
inputs for BnB and KnapsackSolver. While it will still use the real
value when subtracting the fee from outputs, this behavior will be
the same regardless of the algo used for selecting additional inputs.
The fees for transaction overhead and recipient outputs are now included
in nTargetValue instead of being a separate parameter. For the coin
selection algorithms, it doesn't matter that these are separate as in
either case, the algorithm needs to select enough to cover these fees.
Note that setting nValueToSelect is changed as it now includes
not_input_fees. Without the change to how nValueToSelect is increased
for KnapsackSolver, this would result in overpaying fees. The change to
increase by the difference between nFeeRet and not_input_fees allows
this to have the same behavior as previously.
Additionally, because we assume that KnapsackSolver will always find a
solution that requires change (we assume that BnB always finds a
non-change solution), we also include the fee for the change output in
KnapsackSolver's target. As part of this, we also use the changeless
nFeeRet when iterating for KnapsackSolver. This is because we include
the change fee when doing KnapsackSolver, so nFeeRet on further
iterations won't include the change fee.
5d4597666d Rewrite OutputGroups to be clearer and to use scriptPubKeys (Andrew Chow)
f6b3052739 Explicitly filter out partial groups when we don't want them (Andrew Chow)
416d74fb16 Move OutputGroup positive only filtering into Insert (Andrew Chow)
d895e98b59 Move EligibleForSpending into GroupOutputs (Andrew Chow)
99b399aba5 Move fee setting of OutputGroup to Insert (Andrew Chow)
6148a8acda Move GroupOutputs into SelectCoinsMinConf (Andrew Chow)
2acad03657 Remove OutputGroup non-default constructors (Andrew Chow)
Pull request description:
Even after #17458, we still deal with setting fees of an `OutputGroup` and filtering the `OutputGroup` outside of the struct. We currently make all of the `OutputGroup`s in `SelectCoins` and then copy and modify them within each `SelectCoinsMinConf` scenario. This PR changes this to constructing the `OutputGroup`s within the `SelectCoinsMinConf` so that the scenario can be taken into account during the group construction. Furthermore, setting of fees and filtering for effective value is moved into `OutputGroup::Insert` itself so that we don't add undesirable outputs to an `OutputGroup` rather than deleting them afterwards.
To facilitate fee calculation and effective value filtering during `OutputGroup::Insert`, `OutputGroup` now takes the feerates in its constructor and computes the fees and effective value for each output during `Insert`.
While removing `OutputGroup`s in accordance with the `CoinEligibilityFilter` still requires creating the `OutputGroup`s first, we can do that within the function that makes them - `GroupOutput`s.
ACKs for top commit:
Xekyo:
Code review ACK: https://github.com/bitcoin/bitcoin/pull/20040/commits/5d4597666d589e39354e0d8dd5b2afbe1a5d7d8e
fjahr:
Code review ACK 5d4597666d
meshcollider:
Light utACK 5d4597666d
Tree-SHA512: 35965b6d49a87f4ebb366ec4f00aafaaf78e9282481ae2c9682b515a3a9f2cbcd3cd6e202fee29489d48fe7f3a7cede4270796f5e72bbaff76da647138fb3059
OutputGroup will handle the fee and effective value computations
inside of Insert. It now needs to take the effective feerate and long
term feerates as arguments to its constructor.
Instead of having callers set the fees, effective values, and filtering
of outputs, do these within OutputGroups themselves as member functions.
m_fee and m_long_term_fee is added to OutputGroup to track the fees of
the OutputGroup.
If we find a solution which has no waste, just use that. This solution
is what we would consider to be optimal, and other solutions we find
would have to also have 0 waste, so they are equivalent to the first
one with 0 waste. Thus we can optimize by just choosing the first one
with 0 waste.
After this:
- `boost::optional` is no longer used directly (only through `Optional`
which is an alias for it)
- `boost/optional.hpp` is only included in one place
3fc20632a3 qt: Set BLOCK_CHAIN_SIZE = 220 (DrahtBot)
2b6a2f4a28 Regenerate manpages (DrahtBot)
eb7daf4d60 Update copyright headers to 2018 (DrahtBot)
Pull request description:
Some trivial maintenance to avoid having to do it again after the 0.17 branch off.
(The scripts to do this are in `./contrib/`)
Tree-SHA512: 16b2af45e0351b1c691c5311d48025dc6828079e98c2aa2e600dc5910ee8aa01858ca6c356538150dc46fe14c8819ed8ec8e4ec9a0f682b9950dd41bc50518fa
232f96f5c8 doc: Add release notes for -avoidpartialspends (Karl-Johan Alm)
e00b4699cc clean-up: Remove no longer used ivars from CInputCoin (Karl-Johan Alm)
43e04d13b1 wallet: Remove deprecated OutputEligibleForSpending (Karl-Johan Alm)
0128121101 test: Add basic testing for wallet groups (Karl-Johan Alm)
59d6f7b4e2 wallet: Switch to using output groups instead of coins in coin selection (Karl-Johan Alm)
87ebce25d6 wallet: Add output grouping (Karl-Johan Alm)
bb629cb9dc Add -avoidpartialspends and m_avoid_partial_spends (Karl-Johan Alm)
65b3eda458 wallet: Add input bytes to CInputCoin (Karl-Johan Alm)
a443d7a0ca moveonly: CoinElegibilityFilter into coinselection.h (Karl-Johan Alm)
173e18a289 utils: Add insert() convenience templates (Karl-Johan Alm)
Pull request description:
This PR adds an optional (off by default) `-avoidpartialspends` flag, which changes coin select to use output groups rather than outputs, where each output group corresponds to all outputs with the same destination.
It is a privacy improvement, as each time you spend some output, any other output that is publicly associated with the destination (address) will also be spent at the same time, at the cost of fee increase for cases where coin select without group restriction would find a more optimal set of coins (see example below).
For regular use without address reuse, this PR should have no effect on the user experience whatsoever; it only affects users who, for some reason, have multiple outputs with the same destination (i.e. address reuse).
Nodes with this turned off will still try to avoid partial spending, if the fee of the resulting transaction is not greater than the fee of the original transaction.
Example: a node has four outputs linked to two addresses `A` and `B`:
* 1.0 btc to `A`
* 0.5 btc to `A`
* 1.0 btc to `B`
* 0.5 btc to `B`
The node sends 0.2 btc to `C`. Without `-avoidpartialspends`, the following coin selection will occur:
* 0.5 btc to `A` or `B` is picked
* 0.2 btc is output to `C`
* 0.3 - fee is output to (unique change address)
With `-avoidpartialspends`, the following will instead happen:
* Both of (0.5, 1.0) btc to `A` or `B` is picked (one or the other pair)
* 0.2 btc is output to `C`
* 1.3 - fee is output to (unique change address)
As noted, the pro here is that, assuming nobody sends to the address after you spend from it, you will only ever use one address once. The con is that the transaction becomes slightly larger in this case, because it is overpicking outputs to adhere to the no partial spending rule.
This complements #10386, in particular it addresses @luke-jr and @gmaxwell's concerns in https://github.com/bitcoin/bitcoin/pull/10386#issuecomment-300667926 and https://github.com/bitcoin/bitcoin/pull/10386#issuecomment-302361381.
Together with `-avoidreuse`, this fully addresses the concerns in #10065 I believe.
Tree-SHA512: 24687a4490ba59cf4198ed90052944ff4996653a4257833bb52ed24d058b3e924800c9b3790aeb6be6385b653b49e304453e5d7ff960e64c682fc23bfc447621
Moves the current coin selection algorithm out of SelectCoinsMinConf
and puts it in coinselection.{cpp,h}. The new function, KnapsackSolver,
instead of taking a vector of COutputs, will take a vector of CInputCoins
that is prepared by SelectCoinsMinConf.
