Merge bitcoin/bitcoin#27021: Implement Mini version of BlockAssembler to calculate mining scores

6b605b91c1 [fuzz] Add MiniMiner target + diff fuzz against BlockAssembler (glozow)
3f3f2d59ea [unit test] GatherClusters and MiniMiner unit tests (glozow)
59afcc8354 Implement Mini version of BlockAssembler to calculate mining scores (glozow)
56484f0fdc [mempool] find connected mempool entries with GatherClusters(…) (glozow)

Pull request description:

  Implement Mini version of BlockAssembler to calculate mining scores

  Run the mining algorithm on a subset of the mempool, only disturbing the
  mempool to copy out fee information for relevant entries. Intended to be
  used by wallet to calculate amounts needed for fee-bumping unconfirmed
  transactions.

  From comments of sipa and glozow below:

  > > In what way does the code added here differ from the real block assembly code?
  >
  >    * Only operates on the relevant transactions rather than full mempool
  >    * Has the ability to remove transactions that will be replaced so they don't impact their ancestors
  >    * Does not hold mempool lock outside of the constructor, makes copies of the entries it needs instead (though I'm not sure if this has an effect in practice)
  >    * Doesn't do the sanity checks like keeping weight within max block weight and `IsFinalTx()`
  >    * After the block template is built, additionally calculates fees to bump remaining ancestor packages to target feerate

ACKs for top commit:
  achow101:
    ACK 6b605b91c1
  Xekyo:
    > ACK [6b605b9](6b605b91c1) modulo `miniminer_overlap` test.
  furszy:
    ACK 6b605b91 modulo `miniminer_overlap` test.
  theStack:
    Code-review ACK 6b605b91c1

Tree-SHA512: f86a8b4ae0506858a7b15d90f417ebceea5038b395c05c825e3796123ad3b6cb8a98ebb948521316802a4c6d60ebd7041093356b1e2c2922a06b3b96b3b8acb6

src/test/fuzz/mini_miner.cpp

@@ -0,0 +1,192 @@
+#include <test/fuzz/FuzzedDataProvider.h>
+#include <test/fuzz/fuzz.h>
+#include <test/fuzz/util.h>
+#include <test/fuzz/util/mempool.h>
+#include <test/util/script.h>
+#include <test/util/setup_common.h>
+#include <test/util/txmempool.h>
+#include <test/util/mining.h>
+
+#include <node/mini_miner.h>
+#include <node/miner.h>
+#include <primitives/transaction.h>
+#include <random.h>
+#include <txmempool.h>
+
+#include <deque>
+#include <vector>
+
+namespace {
+
+const TestingSetup* g_setup;
+std::deque<COutPoint> g_available_coins;
+void initialize_miner()
+{
+    static const auto testing_setup = MakeNoLogFileContext<const TestingSetup>();
+    g_setup = testing_setup.get();
+    for (uint32_t i = 0; i < uint32_t{100}; ++i) {
+        g_available_coins.push_back(COutPoint{uint256::ZERO, i});
+    }
+}
+
+// Test that the MiniMiner can run with various outpoints and feerates.
+FUZZ_TARGET_INIT(mini_miner, initialize_miner)
+{
+    FuzzedDataProvider fuzzed_data_provider{buffer.data(), buffer.size()};
+    CTxMemPool pool{CTxMemPool::Options{}};
+    std::vector<COutPoint> outpoints;
+    std::deque<COutPoint> available_coins = g_available_coins;
+    LOCK2(::cs_main, pool.cs);
+    // Cluster size cannot exceed 500
+    LIMITED_WHILE(!available_coins.empty(), 500)
+    {
+        CMutableTransaction mtx = CMutableTransaction();
+        const size_t num_inputs = fuzzed_data_provider.ConsumeIntegralInRange<size_t>(1, available_coins.size());
+        const size_t num_outputs = fuzzed_data_provider.ConsumeIntegralInRange<size_t>(1, 50);
+        for (size_t n{0}; n < num_inputs; ++n) {
+            auto prevout = available_coins.front();
+            mtx.vin.push_back(CTxIn(prevout, CScript()));
+            available_coins.pop_front();
+        }
+        for (uint32_t n{0}; n < num_outputs; ++n) {
+            mtx.vout.push_back(CTxOut(100, P2WSH_OP_TRUE));
+        }
+        CTransactionRef tx = MakeTransactionRef(mtx);
+        TestMemPoolEntryHelper entry;
+        const CAmount fee{ConsumeMoney(fuzzed_data_provider, /*max=*/MAX_MONEY/100000)};
+        assert(MoneyRange(fee));
+        pool.addUnchecked(entry.Fee(fee).FromTx(tx));
+
+        // All outputs are available to spend
+        for (uint32_t n{0}; n < num_outputs; ++n) {
+            if (fuzzed_data_provider.ConsumeBool()) {
+                available_coins.push_back(COutPoint{tx->GetHash(), n});
+            }
+        }
+
+        if (fuzzed_data_provider.ConsumeBool() && !tx->vout.empty()) {
+            // Add outpoint from this tx (may or not be spent by a later tx)
+            outpoints.push_back(COutPoint{tx->GetHash(),
+                                          (uint32_t)fuzzed_data_provider.ConsumeIntegralInRange<size_t>(0, tx->vout.size())});
+        } else {
+            // Add some random outpoint (will be interpreted as confirmed or not yet submitted
+            // to mempool).
+            auto outpoint = ConsumeDeserializable<COutPoint>(fuzzed_data_provider);
+            if (outpoint.has_value() && std::find(outpoints.begin(), outpoints.end(), *outpoint) == outpoints.end()) {
+                outpoints.push_back(*outpoint);
+            }
+        }
+
+    }
+
+    const CFeeRate target_feerate{CFeeRate{ConsumeMoney(fuzzed_data_provider, /*max=*/MAX_MONEY/1000)}};
+    std::optional<CAmount> total_bumpfee;
+    CAmount sum_fees = 0;
+    {
+        node::MiniMiner mini_miner{pool, outpoints};
+        assert(mini_miner.IsReadyToCalculate());
+        const auto bump_fees = mini_miner.CalculateBumpFees(target_feerate);
+        for (const auto& outpoint : outpoints) {
+            auto it = bump_fees.find(outpoint);
+            assert(it != bump_fees.end());
+            assert(it->second >= 0);
+            sum_fees += it->second;
+        }
+        assert(!mini_miner.IsReadyToCalculate());
+    }
+    {
+        node::MiniMiner mini_miner{pool, outpoints};
+        assert(mini_miner.IsReadyToCalculate());
+        total_bumpfee = mini_miner.CalculateTotalBumpFees(target_feerate);
+        assert(total_bumpfee.has_value());
+        assert(!mini_miner.IsReadyToCalculate());
+    }
+    // Overlapping ancestry across multiple outpoints can only reduce the total bump fee.
+    assert (sum_fees >= *total_bumpfee);
+}
+
+// Test that MiniMiner and BlockAssembler build the same block given the same transactions and constraints.
+FUZZ_TARGET_INIT(mini_miner_selection, initialize_miner)
+{
+    FuzzedDataProvider fuzzed_data_provider{buffer.data(), buffer.size()};
+    CTxMemPool pool{CTxMemPool::Options{}};
+    // Make a copy to preserve determinism.
+    std::deque<COutPoint> available_coins = g_available_coins;
+    std::vector<CTransactionRef> transactions;
+
+    LOCK2(::cs_main, pool.cs);
+    LIMITED_WHILE(fuzzed_data_provider.ConsumeBool(), 100)
+    {
+        CMutableTransaction mtx = CMutableTransaction();
+        const size_t num_inputs = 2;
+        const size_t num_outputs = fuzzed_data_provider.ConsumeIntegralInRange<size_t>(2, 5);
+        for (size_t n{0}; n < num_inputs; ++n) {
+            auto prevout = available_coins.front();
+            mtx.vin.push_back(CTxIn(prevout, CScript()));
+            available_coins.pop_front();
+        }
+        for (uint32_t n{0}; n < num_outputs; ++n) {
+            mtx.vout.push_back(CTxOut(100, P2WSH_OP_TRUE));
+        }
+        CTransactionRef tx = MakeTransactionRef(mtx);
+
+        // First 2 outputs are available to spend. The rest are added to outpoints to calculate bumpfees.
+        // There is no overlap between spendable coins and outpoints passed to MiniMiner because the
+        // MiniMiner interprets spent coins as to-be-replaced and excludes them.
+        for (uint32_t n{0}; n < num_outputs - 1; ++n) {
+            if (fuzzed_data_provider.ConsumeBool()) {
+                available_coins.push_front(COutPoint{tx->GetHash(), n});
+            } else {
+                available_coins.push_back(COutPoint{tx->GetHash(), n});
+            }
+        }
+
+        // Stop if pool reaches DEFAULT_BLOCK_MAX_WEIGHT because BlockAssembler will stop when the
+        // block template reaches that, but the MiniMiner will keep going.
+        if (pool.GetTotalTxSize() + GetVirtualTransactionSize(*tx) >= DEFAULT_BLOCK_MAX_WEIGHT) break;
+        TestMemPoolEntryHelper entry;
+        const CAmount fee{ConsumeMoney(fuzzed_data_provider, /*max=*/MAX_MONEY/100000)};
+        assert(MoneyRange(fee));
+        pool.addUnchecked(entry.Fee(fee).FromTx(tx));
+        transactions.push_back(tx);
+    }
+    std::vector<COutPoint> outpoints;
+    for (const auto& coin : g_available_coins) {
+        if (!pool.GetConflictTx(coin)) outpoints.push_back(coin);
+    }
+    for (const auto& tx : transactions) {
+        assert(pool.exists(GenTxid::Txid(tx->GetHash())));
+        for (uint32_t n{0}; n < tx->vout.size(); ++n) {
+            COutPoint coin{tx->GetHash(), n};
+            if (!pool.GetConflictTx(coin)) outpoints.push_back(coin);
+        }
+    }
+    const CFeeRate target_feerate{ConsumeMoney(fuzzed_data_provider, /*max=*/MAX_MONEY/100000)};
+
+    node::BlockAssembler::Options miner_options;
+    miner_options.blockMinFeeRate = target_feerate;
+    miner_options.nBlockMaxWeight = DEFAULT_BLOCK_MAX_WEIGHT;
+    miner_options.test_block_validity = false;
+
+    node::BlockAssembler miner{g_setup->m_node.chainman->ActiveChainstate(), &pool, miner_options};
+    node::MiniMiner mini_miner{pool, outpoints};
+    assert(mini_miner.IsReadyToCalculate());
+
+    CScript spk_placeholder = CScript() << OP_0;
+    // Use BlockAssembler as oracle. BlockAssembler and MiniMiner should select the same
+    // transactions, stopping once packages do not meet target_feerate.
+    const auto blocktemplate{miner.CreateNewBlock(spk_placeholder)};
+    mini_miner.BuildMockTemplate(target_feerate);
+    assert(!mini_miner.IsReadyToCalculate());
+    auto mock_template_txids = mini_miner.GetMockTemplateTxids();
+    // MiniMiner doesn't add a coinbase tx.
+    assert(mock_template_txids.count(blocktemplate->block.vtx[0]->GetHash()) == 0);
+    mock_template_txids.emplace(blocktemplate->block.vtx[0]->GetHash());
+    assert(mock_template_txids.size() <= blocktemplate->block.vtx.size());
+    assert(mock_template_txids.size() >= blocktemplate->block.vtx.size());
+    assert(mock_template_txids.size() == blocktemplate->block.vtx.size());
+    for (const auto& tx : blocktemplate->block.vtx) {
+        assert(mock_template_txids.count(tx->GetHash()));
+    }
+}
+} // namespace

src/test/miniminer_tests.cpp

@@ -0,0 +1,477 @@
+// Copyright (c) 2021 The Bitcoin Core developers
+// Distributed under the MIT software license, see the accompanying
+// file COPYING or http://www.opensource.org/licenses/mit-license.php.
+#include <node/mini_miner.h>
+#include <txmempool.h>
+#include <util/system.h>
+#include <util/time.h>
+
+#include <test/util/setup_common.h>
+#include <test/util/txmempool.h>
+
+#include <boost/test/unit_test.hpp>
+#include <optional>
+#include <vector>
+
+BOOST_FIXTURE_TEST_SUITE(miniminer_tests, TestingSetup)
+
+static inline CTransactionRef make_tx(const std::vector<COutPoint>& inputs, size_t num_outputs)
+{
+    CMutableTransaction tx = CMutableTransaction();
+    tx.vin.resize(inputs.size());
+    tx.vout.resize(num_outputs);
+    for (size_t i = 0; i < inputs.size(); ++i) {
+        tx.vin[i].prevout = inputs[i];
+    }
+    for (size_t i = 0; i < num_outputs; ++i) {
+        tx.vout[i].scriptPubKey = CScript() << OP_11 << OP_EQUAL;
+        // The actual input and output values of these transactions don't really
+        // matter, since all accounting will use the entries' cached fees.
+        tx.vout[i].nValue = COIN;
+    }
+    return MakeTransactionRef(tx);
+}
+
+static inline bool sanity_check(const std::vector<CTransactionRef>& transactions,
+                                const std::map<COutPoint, CAmount>& bumpfees)
+{
+    // No negative bumpfees.
+    for (const auto& [outpoint, fee] : bumpfees) {
+        if (fee < 0) return false;
+        if (fee == 0) continue;
+        auto outpoint_ = outpoint; // structured bindings can't be captured in C++17, so we need to use a variable
+        const bool found = std::any_of(transactions.cbegin(), transactions.cend(), [&](const auto& tx) {
+            return outpoint_.hash == tx->GetHash() && outpoint_.n < tx->vout.size();
+        });
+        if (!found) return false;
+    }
+    for (const auto& tx : transactions) {
+        // If tx has multiple outputs, they must all have the same bumpfee (if they exist).
+        if (tx->vout.size() > 1) {
+            std::set<CAmount> distinct_bumpfees;
+            for (size_t i{0}; i < tx->vout.size(); ++i) {
+                const auto bumpfee = bumpfees.find(COutPoint{tx->GetHash(), static_cast<uint32_t>(i)});
+                if (bumpfee != bumpfees.end()) distinct_bumpfees.insert(bumpfee->second);
+            }
+            if (distinct_bumpfees.size() > 1) return false;
+        }
+    }
+    return true;
+}
+
+template <typename Key, typename Value>
+Value Find(const std::map<Key, Value>& map, const Key& key)
+{
+    auto it = map.find(key);
+    BOOST_CHECK_MESSAGE(it != map.end(), strprintf("Cannot find %s", key.ToString()));
+    return it->second;
+}
+
+BOOST_FIXTURE_TEST_CASE(miniminer_1p1c, TestChain100Setup)
+{
+    CTxMemPool& pool = *Assert(m_node.mempool);
+    LOCK2(::cs_main, pool.cs);
+    TestMemPoolEntryHelper entry;
+
+    const CAmount low_fee{CENT/2000};
+    const CAmount normal_fee{CENT/200};
+    const CAmount high_fee{CENT/10};
+
+    // Create a parent tx1 and child tx2 with normal fees:
+    const auto tx1 = make_tx({COutPoint{m_coinbase_txns[0]->GetHash(), 0}}, /*num_outputs=*/2);
+    pool.addUnchecked(entry.Fee(normal_fee).FromTx(tx1));
+    const auto tx2 = make_tx({COutPoint{tx1->GetHash(), 0}}, /*num_outputs=*/1);
+    pool.addUnchecked(entry.Fee(normal_fee).FromTx(tx2));
+
+    // Create a low-feerate parent tx3 and high-feerate child tx4 (cpfp)
+    const auto tx3 = make_tx({COutPoint{m_coinbase_txns[1]->GetHash(), 0}}, /*num_outputs=*/2);
+    pool.addUnchecked(entry.Fee(low_fee).FromTx(tx3));
+    const auto tx4 = make_tx({COutPoint{tx3->GetHash(), 0}}, /*num_outputs=*/1);
+    pool.addUnchecked(entry.Fee(high_fee).FromTx(tx4));
+
+    // Create a parent tx5 and child tx6 where both have low fees
+    const auto tx5 = make_tx({COutPoint{m_coinbase_txns[2]->GetHash(), 0}}, /*num_outputs=*/2);
+    pool.addUnchecked(entry.Fee(low_fee).FromTx(tx5));
+    const auto tx6 = make_tx({COutPoint{tx5->GetHash(), 0}}, /*num_outputs=*/1);
+    pool.addUnchecked(entry.Fee(low_fee).FromTx(tx6));
+    // Make tx6's modified fee much higher than its base fee. This should cause it to pass
+    // the fee-related checks despite being low-feerate.
+    pool.PrioritiseTransaction(tx6->GetHash(), CENT/100);
+
+    // Create a high-feerate parent tx7, low-feerate child tx8
+    const auto tx7 = make_tx({COutPoint{m_coinbase_txns[3]->GetHash(), 0}}, /*num_outputs=*/2);
+    pool.addUnchecked(entry.Fee(high_fee).FromTx(tx7));
+    const auto tx8 = make_tx({COutPoint{tx7->GetHash(), 0}}, /*num_outputs=*/1);
+    pool.addUnchecked(entry.Fee(low_fee).FromTx(tx8));
+
+    std::vector<COutPoint> all_unspent_outpoints({
+        COutPoint{tx1->GetHash(), 1},
+        COutPoint{tx2->GetHash(), 0},
+        COutPoint{tx3->GetHash(), 1},
+        COutPoint{tx4->GetHash(), 0},
+        COutPoint{tx5->GetHash(), 1},
+        COutPoint{tx6->GetHash(), 0},
+        COutPoint{tx7->GetHash(), 1},
+        COutPoint{tx8->GetHash(), 0}
+    });
+    for (const auto& outpoint : all_unspent_outpoints) BOOST_CHECK(!pool.isSpent(outpoint));
+
+    std::vector<COutPoint> all_spent_outpoints({
+        COutPoint{tx1->GetHash(), 0},
+        COutPoint{tx3->GetHash(), 0},
+        COutPoint{tx5->GetHash(), 0},
+        COutPoint{tx7->GetHash(), 0}
+    });
+    for (const auto& outpoint : all_spent_outpoints) BOOST_CHECK(pool.GetConflictTx(outpoint) != nullptr);
+
+    std::vector<COutPoint> all_parent_outputs({
+        COutPoint{tx1->GetHash(), 0},
+        COutPoint{tx1->GetHash(), 1},
+        COutPoint{tx3->GetHash(), 0},
+        COutPoint{tx3->GetHash(), 1},
+        COutPoint{tx5->GetHash(), 0},
+        COutPoint{tx5->GetHash(), 1},
+        COutPoint{tx7->GetHash(), 0},
+        COutPoint{tx7->GetHash(), 1}
+    });
+
+
+    std::vector<CTransactionRef> all_transactions{tx1, tx2, tx3, tx4, tx5, tx6, tx7, tx8};
+    struct TxDimensions {
+        size_t vsize; CAmount mod_fee; CFeeRate feerate;
+    };
+    std::map<uint256, TxDimensions> tx_dims;
+    for (const auto& tx : all_transactions) {
+        const auto it = pool.GetIter(tx->GetHash()).value();
+        tx_dims.emplace(tx->GetHash(), TxDimensions{it->GetTxSize(), it->GetModifiedFee(),
+                                              CFeeRate(it->GetModifiedFee(), it->GetTxSize())});
+    }
+
+    const std::vector<CFeeRate> various_normal_feerates({CFeeRate(0), CFeeRate(500), CFeeRate(999),
+                                                         CFeeRate(1000), CFeeRate(2000), CFeeRate(2500),
+                                                         CFeeRate(3333), CFeeRate(7800), CFeeRate(11199),
+                                                         CFeeRate(23330), CFeeRate(50000), CFeeRate(5*CENT)});
+
+    // All nonexistent entries have a bumpfee of zero, regardless of feerate
+    std::vector<COutPoint> nonexistent_outpoints({ COutPoint{GetRandHash(), 0}, COutPoint{GetRandHash(), 3} });
+    for (const auto& outpoint : nonexistent_outpoints) BOOST_CHECK(!pool.isSpent(outpoint));
+    for (const auto& feerate : various_normal_feerates) {
+        node::MiniMiner mini_miner(pool, nonexistent_outpoints);
+        BOOST_CHECK(mini_miner.IsReadyToCalculate());
+        auto bump_fees = mini_miner.CalculateBumpFees(feerate);
+        BOOST_CHECK(!mini_miner.IsReadyToCalculate());
+        BOOST_CHECK(sanity_check(all_transactions, bump_fees));
+        BOOST_CHECK(bump_fees.size() == nonexistent_outpoints.size());
+        for (const auto& outpoint: nonexistent_outpoints) {
+            auto it = bump_fees.find(outpoint);
+            BOOST_CHECK(it != bump_fees.end());
+            BOOST_CHECK_EQUAL(it->second, 0);
+        }
+    }
+
+    // Gather bump fees for all available UTXOs.
+    for (const auto& target_feerate : various_normal_feerates) {
+        node::MiniMiner mini_miner(pool, all_unspent_outpoints);
+        BOOST_CHECK(mini_miner.IsReadyToCalculate());
+        auto bump_fees = mini_miner.CalculateBumpFees(target_feerate);
+        BOOST_CHECK(!mini_miner.IsReadyToCalculate());
+        BOOST_CHECK(sanity_check(all_transactions, bump_fees));
+        BOOST_CHECK_EQUAL(bump_fees.size(), all_unspent_outpoints.size());
+
+        // Check tx1 bumpfee: no other bumper.
+        const TxDimensions& tx1_dimensions = tx_dims.find(tx1->GetHash())->second;
+        CAmount bumpfee1 = Find(bump_fees, COutPoint{tx1->GetHash(), 1});
+        if (target_feerate <= tx1_dimensions.feerate) {
+            BOOST_CHECK_EQUAL(bumpfee1, 0);
+        } else {
+            // Difference is fee to bump tx1 from current to target feerate.
+            BOOST_CHECK_EQUAL(bumpfee1, target_feerate.GetFee(tx1_dimensions.vsize) - tx1_dimensions.mod_fee);
+        }
+
+        // Check tx3 bumpfee: assisted by tx4.
+        const TxDimensions& tx3_dimensions = tx_dims.find(tx3->GetHash())->second;
+        const TxDimensions& tx4_dimensions = tx_dims.find(tx4->GetHash())->second;
+        const CFeeRate tx3_feerate = CFeeRate(tx3_dimensions.mod_fee + tx4_dimensions.mod_fee, tx3_dimensions.vsize + tx4_dimensions.vsize);
+        CAmount bumpfee3 = Find(bump_fees, COutPoint{tx3->GetHash(), 1});
+        if (target_feerate <= tx3_feerate) {
+            // As long as target feerate is below tx4's ancestor feerate, there is no bump fee.
+            BOOST_CHECK_EQUAL(bumpfee3, 0);
+        } else {
+            // Difference is fee to bump tx3 from current to target feerate, without tx4.
+            BOOST_CHECK_EQUAL(bumpfee3, target_feerate.GetFee(tx3_dimensions.vsize) - tx3_dimensions.mod_fee);
+        }
+
+        // If tx6’s modified fees are sufficient for tx5 and tx6 to be picked
+        // into the block, our prospective new transaction would not need to
+        // bump tx5 when using tx5’s second output. If however even tx6’s
+        // modified fee (which essentially indicates "effective feerate") is
+        // not sufficient to bump tx5, using the second output of tx5 would
+        // require our transaction to bump tx5 from scratch since we evaluate
+        // transaction packages per ancestor sets and do not consider multiple
+        // children’s fees.
+        const TxDimensions& tx5_dimensions = tx_dims.find(tx5->GetHash())->second;
+        const TxDimensions& tx6_dimensions = tx_dims.find(tx6->GetHash())->second;
+        const CFeeRate tx5_feerate = CFeeRate(tx5_dimensions.mod_fee + tx6_dimensions.mod_fee, tx5_dimensions.vsize + tx6_dimensions.vsize);
+        CAmount bumpfee5 = Find(bump_fees, COutPoint{tx5->GetHash(), 1});
+        if (target_feerate <= tx5_feerate) {
+            // As long as target feerate is below tx6's ancestor feerate, there is no bump fee.
+            BOOST_CHECK_EQUAL(bumpfee5, 0);
+        } else {
+            // Difference is fee to bump tx5 from current to target feerate, without tx6.
+            BOOST_CHECK_EQUAL(bumpfee5, target_feerate.GetFee(tx5_dimensions.vsize) - tx5_dimensions.mod_fee);
+        }
+    }
+    // Spent outpoints should usually not be requested as they would not be
+    // considered available. However, when they are explicitly requested, we
+    // can calculate their bumpfee to facilitate RBF-replacements
+    for (const auto& target_feerate : various_normal_feerates) {
+        node::MiniMiner mini_miner_all_spent(pool, all_spent_outpoints);
+        BOOST_CHECK(mini_miner_all_spent.IsReadyToCalculate());
+        auto bump_fees_all_spent = mini_miner_all_spent.CalculateBumpFees(target_feerate);
+        BOOST_CHECK(!mini_miner_all_spent.IsReadyToCalculate());
+        BOOST_CHECK_EQUAL(bump_fees_all_spent.size(), all_spent_outpoints.size());
+        node::MiniMiner mini_miner_all_parents(pool, all_parent_outputs);
+        BOOST_CHECK(mini_miner_all_parents.IsReadyToCalculate());
+        auto bump_fees_all_parents = mini_miner_all_parents.CalculateBumpFees(target_feerate);
+        BOOST_CHECK(!mini_miner_all_parents.IsReadyToCalculate());
+        BOOST_CHECK_EQUAL(bump_fees_all_parents.size(), all_parent_outputs.size());
+        for (auto& bump_fees : {bump_fees_all_parents, bump_fees_all_spent}) {
+            // For all_parents case, both outputs from the parent should have the same bump fee,
+            // even though only one of them is in a to-be-replaced transaction.
+            BOOST_CHECK(sanity_check(all_transactions, bump_fees));
+
+            // Check tx1 bumpfee: no other bumper.
+            const TxDimensions& tx1_dimensions = tx_dims.find(tx1->GetHash())->second;
+            CAmount it1_spent = Find(bump_fees, COutPoint{tx1->GetHash(), 0});
+            if (target_feerate <= tx1_dimensions.feerate) {
+                BOOST_CHECK_EQUAL(it1_spent, 0);
+            } else {
+                // Difference is fee to bump tx1 from current to target feerate.
+                BOOST_CHECK_EQUAL(it1_spent, target_feerate.GetFee(tx1_dimensions.vsize) - tx1_dimensions.mod_fee);
+            }
+
+            // Check tx3 bumpfee: no other bumper, because tx4 is to-be-replaced.
+            const TxDimensions& tx3_dimensions = tx_dims.find(tx3->GetHash())->second;
+            const CFeeRate tx3_feerate_unbumped = tx3_dimensions.feerate;
+            auto it3_spent = Find(bump_fees, COutPoint{tx3->GetHash(), 0});
+            if (target_feerate <= tx3_feerate_unbumped) {
+                BOOST_CHECK_EQUAL(it3_spent, 0);
+            } else {
+                // Difference is fee to bump tx3 from current to target feerate, without tx4.
+                BOOST_CHECK_EQUAL(it3_spent, target_feerate.GetFee(tx3_dimensions.vsize) - tx3_dimensions.mod_fee);
+            }
+
+            // Check tx5 bumpfee: no other bumper, because tx6 is to-be-replaced.
+            const TxDimensions& tx5_dimensions = tx_dims.find(tx5->GetHash())->second;
+            const CFeeRate tx5_feerate_unbumped = tx5_dimensions.feerate;
+            auto it5_spent = Find(bump_fees, COutPoint{tx5->GetHash(), 0});
+            if (target_feerate <= tx5_feerate_unbumped) {
+                BOOST_CHECK_EQUAL(it5_spent, 0);
+            } else {
+                // Difference is fee to bump tx5 from current to target feerate, without tx6.
+                BOOST_CHECK_EQUAL(it5_spent, target_feerate.GetFee(tx5_dimensions.vsize) - tx5_dimensions.mod_fee);
+            }
+        }
+    }
+}
+
+BOOST_FIXTURE_TEST_CASE(miniminer_overlap, TestChain100Setup)
+{
+    CTxMemPool& pool = *Assert(m_node.mempool);
+    LOCK2(::cs_main, pool.cs);
+    TestMemPoolEntryHelper entry;
+
+    const CAmount low_fee{CENT/2000};
+    const CAmount med_fee{CENT/200};
+    const CAmount high_fee{CENT/10};
+
+    // Create 3 parents of different feerates, and 1 child spending from all 3.
+    const auto tx1 = make_tx({COutPoint{m_coinbase_txns[0]->GetHash(), 0}}, /*num_outputs=*/2);
+    pool.addUnchecked(entry.Fee(low_fee).FromTx(tx1));
+    const auto tx2 = make_tx({COutPoint{m_coinbase_txns[1]->GetHash(), 0}}, /*num_outputs=*/2);
+    pool.addUnchecked(entry.Fee(med_fee).FromTx(tx2));
+    const auto tx3 = make_tx({COutPoint{m_coinbase_txns[2]->GetHash(), 0}}, /*num_outputs=*/2);
+    pool.addUnchecked(entry.Fee(high_fee).FromTx(tx3));
+    const auto tx4 = make_tx({COutPoint{tx1->GetHash(), 0}, COutPoint{tx2->GetHash(), 0}, COutPoint{tx3->GetHash(), 0}}, /*num_outputs=*/3);
+    pool.addUnchecked(entry.Fee(high_fee).FromTx(tx4));
+
+    // Create 1 grandparent and 1 parent, then 2 children.
+    const auto tx5 = make_tx({COutPoint{m_coinbase_txns[3]->GetHash(), 0}}, /*num_outputs=*/2);
+    pool.addUnchecked(entry.Fee(high_fee).FromTx(tx5));
+    const auto tx6 = make_tx({COutPoint{tx5->GetHash(), 0}}, /*num_outputs=*/3);
+    pool.addUnchecked(entry.Fee(low_fee).FromTx(tx6));
+    const auto tx7 = make_tx({COutPoint{tx6->GetHash(), 0}}, /*num_outputs=*/2);
+    pool.addUnchecked(entry.Fee(med_fee).FromTx(tx7));
+    const auto tx8 = make_tx({COutPoint{tx6->GetHash(), 1}}, /*num_outputs=*/2);
+    pool.addUnchecked(entry.Fee(high_fee).FromTx(tx8));
+
+    std::vector<CTransactionRef> all_transactions{tx1, tx2, tx3, tx4, tx5, tx6, tx7, tx8};
+    std::vector<int64_t> tx_vsizes;
+    tx_vsizes.reserve(all_transactions.size());
+    for (const auto& tx : all_transactions) tx_vsizes.push_back(GetVirtualTransactionSize(*tx));
+
+    std::vector<COutPoint> all_unspent_outpoints({
+        COutPoint{tx1->GetHash(), 1},
+        COutPoint{tx2->GetHash(), 1},
+        COutPoint{tx3->GetHash(), 1},
+        COutPoint{tx4->GetHash(), 0},
+        COutPoint{tx4->GetHash(), 1},
+        COutPoint{tx4->GetHash(), 2},
+        COutPoint{tx5->GetHash(), 1},
+        COutPoint{tx6->GetHash(), 2},
+        COutPoint{tx7->GetHash(), 0},
+        COutPoint{tx8->GetHash(), 0}
+    });
+    for (const auto& outpoint : all_unspent_outpoints) BOOST_CHECK(!pool.isSpent(outpoint));
+
+    const auto tx3_feerate = CFeeRate(high_fee, tx_vsizes[2]);
+    const auto tx4_feerate = CFeeRate(high_fee, tx_vsizes[3]);
+    // tx4's feerate is lower than tx3's. same fee, different weight.
+    BOOST_CHECK(tx3_feerate > tx4_feerate);
+    const auto tx4_anc_feerate = CFeeRate(low_fee + med_fee + high_fee, tx_vsizes[0] + tx_vsizes[1] + tx_vsizes[3]);
+    const auto tx5_feerate = CFeeRate(high_fee, tx_vsizes[4]);
+    const auto tx7_anc_feerate = CFeeRate(low_fee + med_fee, tx_vsizes[5] + tx_vsizes[6]);
+    const auto tx8_anc_feerate = CFeeRate(low_fee + high_fee, tx_vsizes[5] + tx_vsizes[7]);
+    BOOST_CHECK(tx5_feerate > tx7_anc_feerate);
+    BOOST_CHECK(tx5_feerate > tx8_anc_feerate);
+
+    // Extremely high feerate: everybody's bumpfee is from their full ancestor set.
+    {
+        node::MiniMiner mini_miner(pool, all_unspent_outpoints);
+        const CFeeRate very_high_feerate(COIN);
+        BOOST_CHECK(tx4_anc_feerate < very_high_feerate);
+        BOOST_CHECK(mini_miner.IsReadyToCalculate());
+        auto bump_fees = mini_miner.CalculateBumpFees(very_high_feerate);
+        BOOST_CHECK_EQUAL(bump_fees.size(), all_unspent_outpoints.size());
+        BOOST_CHECK(!mini_miner.IsReadyToCalculate());
+        BOOST_CHECK(sanity_check(all_transactions, bump_fees));
+        const auto tx1_bumpfee = bump_fees.find(COutPoint{tx1->GetHash(), 1});
+        BOOST_CHECK(tx1_bumpfee != bump_fees.end());
+        BOOST_CHECK_EQUAL(tx1_bumpfee->second, very_high_feerate.GetFee(tx_vsizes[0]) - low_fee);
+        const auto tx4_bumpfee = bump_fees.find(COutPoint{tx4->GetHash(), 0});
+        BOOST_CHECK(tx4_bumpfee != bump_fees.end());
+        BOOST_CHECK_EQUAL(tx4_bumpfee->second,
+            very_high_feerate.GetFee(tx_vsizes[0] + tx_vsizes[1] + tx_vsizes[2] + tx_vsizes[3]) - (low_fee + med_fee + high_fee + high_fee));
+        const auto tx7_bumpfee = bump_fees.find(COutPoint{tx7->GetHash(), 0});
+        BOOST_CHECK(tx7_bumpfee != bump_fees.end());
+        BOOST_CHECK_EQUAL(tx7_bumpfee->second,
+            very_high_feerate.GetFee(tx_vsizes[4] + tx_vsizes[5] + tx_vsizes[6]) - (high_fee + low_fee + med_fee));
+        const auto tx8_bumpfee = bump_fees.find(COutPoint{tx8->GetHash(), 0});
+        BOOST_CHECK(tx8_bumpfee != bump_fees.end());
+        BOOST_CHECK_EQUAL(tx8_bumpfee->second,
+            very_high_feerate.GetFee(tx_vsizes[4] + tx_vsizes[5] + tx_vsizes[7]) - (high_fee + low_fee + high_fee));
+        // Total fees: if spending multiple outputs from tx4 don't double-count fees.
+        node::MiniMiner mini_miner_total_tx4(pool, {COutPoint{tx4->GetHash(), 0}, COutPoint{tx4->GetHash(), 1}});
+        BOOST_CHECK(mini_miner_total_tx4.IsReadyToCalculate());
+        const auto tx4_bump_fee = mini_miner_total_tx4.CalculateTotalBumpFees(very_high_feerate);
+        BOOST_CHECK(!mini_miner_total_tx4.IsReadyToCalculate());
+        BOOST_CHECK(tx4_bump_fee.has_value());
+        BOOST_CHECK_EQUAL(tx4_bump_fee.value(),
+            very_high_feerate.GetFee(tx_vsizes[0] + tx_vsizes[1] + tx_vsizes[2] + tx_vsizes[3]) - (low_fee + med_fee + high_fee + high_fee));
+        // Total fees: if spending both tx7 and tx8, don't double-count fees.
+        node::MiniMiner mini_miner_tx7_tx8(pool, {COutPoint{tx7->GetHash(), 0}, COutPoint{tx8->GetHash(), 0}});
+        BOOST_CHECK(mini_miner_tx7_tx8.IsReadyToCalculate());
+        const auto tx7_tx8_bumpfee = mini_miner_tx7_tx8.CalculateTotalBumpFees(very_high_feerate);
+        BOOST_CHECK(!mini_miner_tx7_tx8.IsReadyToCalculate());
+        BOOST_CHECK(tx7_tx8_bumpfee.has_value());
+        BOOST_CHECK_EQUAL(tx7_tx8_bumpfee.value(),
+            very_high_feerate.GetFee(tx_vsizes[4] + tx_vsizes[5] + tx_vsizes[6] + tx_vsizes[7]) - (high_fee + low_fee + med_fee + high_fee));
+    }
+    // Feerate just below tx5: tx7 and tx8 have different bump fees.
+    {
+        const auto just_below_tx5 = CFeeRate(tx5_feerate.GetFeePerK() - 5);
+        node::MiniMiner mini_miner(pool, all_unspent_outpoints);
+        BOOST_CHECK(mini_miner.IsReadyToCalculate());
+        auto bump_fees = mini_miner.CalculateBumpFees(just_below_tx5);
+        BOOST_CHECK(!mini_miner.IsReadyToCalculate());
+        BOOST_CHECK_EQUAL(bump_fees.size(), all_unspent_outpoints.size());
+        BOOST_CHECK(sanity_check(all_transactions, bump_fees));
+        const auto tx7_bumpfee = bump_fees.find(COutPoint{tx7->GetHash(), 0});
+        BOOST_CHECK(tx7_bumpfee != bump_fees.end());
+        BOOST_CHECK_EQUAL(tx7_bumpfee->second, just_below_tx5.GetFee(tx_vsizes[5] + tx_vsizes[6]) - (low_fee + med_fee));
+        const auto tx8_bumpfee = bump_fees.find(COutPoint{tx8->GetHash(), 0});
+        BOOST_CHECK(tx8_bumpfee != bump_fees.end());
+        BOOST_CHECK_EQUAL(tx8_bumpfee->second, just_below_tx5.GetFee(tx_vsizes[5] + tx_vsizes[7]) - (low_fee + high_fee));
+        // Total fees: if spending both tx7 and tx8, don't double-count fees.
+        node::MiniMiner mini_miner_tx7_tx8(pool, {COutPoint{tx7->GetHash(), 0}, COutPoint{tx8->GetHash(), 0}});
+        BOOST_CHECK(mini_miner_tx7_tx8.IsReadyToCalculate());
+        const auto tx7_tx8_bumpfee = mini_miner_tx7_tx8.CalculateTotalBumpFees(just_below_tx5);
+        BOOST_CHECK(!mini_miner_tx7_tx8.IsReadyToCalculate());
+        BOOST_CHECK(tx7_tx8_bumpfee.has_value());
+        BOOST_CHECK_EQUAL(tx7_tx8_bumpfee.value(), just_below_tx5.GetFee(tx_vsizes[5] + tx_vsizes[6]) - (low_fee + med_fee));
+    }
+    // Feerate between tx7 and tx8's ancestor feerates: don't need to bump tx6 because tx8 already does.
+    {
+        const auto just_above_tx7 = CFeeRate(med_fee + 10, tx_vsizes[6]);
+        BOOST_CHECK(just_above_tx7 <= CFeeRate(low_fee + high_fee, tx_vsizes[5] + tx_vsizes[7]));
+        node::MiniMiner mini_miner(pool, all_unspent_outpoints);
+        BOOST_CHECK(mini_miner.IsReadyToCalculate());
+        auto bump_fees = mini_miner.CalculateBumpFees(just_above_tx7);
+        BOOST_CHECK(!mini_miner.IsReadyToCalculate());
+        BOOST_CHECK_EQUAL(bump_fees.size(), all_unspent_outpoints.size());
+        BOOST_CHECK(sanity_check(all_transactions, bump_fees));
+        const auto tx7_bumpfee = bump_fees.find(COutPoint{tx7->GetHash(), 0});
+        BOOST_CHECK(tx7_bumpfee != bump_fees.end());
+        BOOST_CHECK_EQUAL(tx7_bumpfee->second, just_above_tx7.GetFee(tx_vsizes[6]) - (med_fee));
+        const auto tx8_bumpfee = bump_fees.find(COutPoint{tx8->GetHash(), 0});
+        BOOST_CHECK(tx8_bumpfee != bump_fees.end());
+        BOOST_CHECK_EQUAL(tx8_bumpfee->second, 0);
+    }
+}
+BOOST_FIXTURE_TEST_CASE(calculate_cluster, TestChain100Setup)
+{
+    CTxMemPool& pool = *Assert(m_node.mempool);
+    LOCK2(cs_main, pool.cs);
+
+    // Add chain of size 500
+    TestMemPoolEntryHelper entry;
+    std::vector<uint256> chain_txids;
+    auto& lasttx = m_coinbase_txns[0];
+    for (auto i{0}; i < 500; ++i) {
+        const auto tx = make_tx({COutPoint{lasttx->GetHash(), 0}}, /*num_outputs=*/1);
+        pool.addUnchecked(entry.Fee(CENT).FromTx(tx));
+        chain_txids.push_back(tx->GetHash());
+        lasttx = tx;
+    }
+    const auto cluster_500tx = pool.GatherClusters({lasttx->GetHash()});
+    CTxMemPool::setEntries cluster_500tx_set{cluster_500tx.begin(), cluster_500tx.end()};
+    BOOST_CHECK_EQUAL(cluster_500tx.size(), cluster_500tx_set.size());
+    const auto vec_iters_500 = pool.GetIterVec(chain_txids);
+    for (const auto& iter : vec_iters_500) BOOST_CHECK(cluster_500tx_set.count(iter));
+
+    // GatherClusters stops at 500 transactions.
+    const auto tx_501 = make_tx({COutPoint{lasttx->GetHash(), 0}}, /*num_outputs=*/1);
+    pool.addUnchecked(entry.Fee(CENT).FromTx(tx_501));
+    const auto cluster_501 = pool.GatherClusters({tx_501->GetHash()});
+    BOOST_CHECK_EQUAL(cluster_501.size(), 0);
+
+    // Zig Zag cluster:
+    // txp0     txp1     txp2    ...  txp48  txp49
+    //    \    /    \   /   \            \   /
+    //     txc0     txc1    txc2  ...    txc48
+    // Note that each transaction's ancestor size is 1 or 3, and each descendant size is 1, 2 or 3.
+    // However, all of these transactions are in the same cluster.
+    std::vector<uint256> zigzag_txids;
+    for (auto p{0}; p < 50; ++p) {
+        const auto txp = make_tx({COutPoint{GetRandHash(), 0}}, /*num_outputs=*/2);
+        pool.addUnchecked(entry.Fee(CENT).FromTx(txp));
+        zigzag_txids.push_back(txp->GetHash());
+    }
+    for (auto c{0}; c < 49; ++c) {
+        const auto txc = make_tx({COutPoint{zigzag_txids[c], 1}, COutPoint{zigzag_txids[c+1], 0}}, /*num_outputs=*/1);
+        pool.addUnchecked(entry.Fee(CENT).FromTx(txc));
+        zigzag_txids.push_back(txc->GetHash());
+    }
+    const auto vec_iters_zigzag = pool.GetIterVec(zigzag_txids);
+    // It doesn't matter which tx we calculate cluster for, everybody is in it.
+    const std::vector<size_t> indeces{0, 22, 72, zigzag_txids.size() - 1};
+    for (const auto index : indeces) {
+        const auto cluster = pool.GatherClusters({zigzag_txids[index]});
+        BOOST_CHECK_EQUAL(cluster.size(), zigzag_txids.size());
+        CTxMemPool::setEntries clusterset{cluster.begin(), cluster.end()};
+        BOOST_CHECK_EQUAL(cluster.size(), clusterset.size());
+        for (const auto& iter : vec_iters_zigzag) BOOST_CHECK(clusterset.count(iter));
+    }
+}
+
+BOOST_AUTO_TEST_SUITE_END()