bitcoin/src/script/standard.cpp

623 lines
24 KiB
C++

// Copyright (c) 2009-2010 Satoshi Nakamoto
// Copyright (c) 2009-2020 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 <script/standard.h>
#include <crypto/sha256.h>
#include <hash.h>
#include <pubkey.h>
#include <script/interpreter.h>
#include <script/script.h>
#include <util/strencodings.h>
#include <string>
typedef std::vector<unsigned char> valtype;
bool fAcceptDatacarrier = DEFAULT_ACCEPT_DATACARRIER;
unsigned nMaxDatacarrierBytes = MAX_OP_RETURN_RELAY;
CScriptID::CScriptID(const CScript& in) : BaseHash(Hash160(in)) {}
CScriptID::CScriptID(const ScriptHash& in) : BaseHash(static_cast<uint160>(in)) {}
ScriptHash::ScriptHash(const CScript& in) : BaseHash(Hash160(in)) {}
ScriptHash::ScriptHash(const CScriptID& in) : BaseHash(static_cast<uint160>(in)) {}
PKHash::PKHash(const CPubKey& pubkey) : BaseHash(pubkey.GetID()) {}
PKHash::PKHash(const CKeyID& pubkey_id) : BaseHash(pubkey_id) {}
WitnessV0KeyHash::WitnessV0KeyHash(const CPubKey& pubkey) : BaseHash(pubkey.GetID()) {}
WitnessV0KeyHash::WitnessV0KeyHash(const PKHash& pubkey_hash) : BaseHash(static_cast<uint160>(pubkey_hash)) {}
CKeyID ToKeyID(const PKHash& key_hash)
{
return CKeyID{static_cast<uint160>(key_hash)};
}
CKeyID ToKeyID(const WitnessV0KeyHash& key_hash)
{
return CKeyID{static_cast<uint160>(key_hash)};
}
WitnessV0ScriptHash::WitnessV0ScriptHash(const CScript& in)
{
CSHA256().Write(in.data(), in.size()).Finalize(begin());
}
std::string GetTxnOutputType(TxoutType t)
{
switch (t) {
case TxoutType::NONSTANDARD: return "nonstandard";
case TxoutType::PUBKEY: return "pubkey";
case TxoutType::PUBKEYHASH: return "pubkeyhash";
case TxoutType::SCRIPTHASH: return "scripthash";
case TxoutType::MULTISIG: return "multisig";
case TxoutType::NULL_DATA: return "nulldata";
case TxoutType::WITNESS_V0_KEYHASH: return "witness_v0_keyhash";
case TxoutType::WITNESS_V0_SCRIPTHASH: return "witness_v0_scripthash";
case TxoutType::WITNESS_V1_TAPROOT: return "witness_v1_taproot";
case TxoutType::WITNESS_UNKNOWN: return "witness_unknown";
} // no default case, so the compiler can warn about missing cases
assert(false);
}
static bool MatchPayToPubkey(const CScript& script, valtype& pubkey)
{
if (script.size() == CPubKey::SIZE + 2 && script[0] == CPubKey::SIZE && script.back() == OP_CHECKSIG) {
pubkey = valtype(script.begin() + 1, script.begin() + CPubKey::SIZE + 1);
return CPubKey::ValidSize(pubkey);
}
if (script.size() == CPubKey::COMPRESSED_SIZE + 2 && script[0] == CPubKey::COMPRESSED_SIZE && script.back() == OP_CHECKSIG) {
pubkey = valtype(script.begin() + 1, script.begin() + CPubKey::COMPRESSED_SIZE + 1);
return CPubKey::ValidSize(pubkey);
}
return false;
}
static bool MatchPayToPubkeyHash(const CScript& script, valtype& pubkeyhash)
{
if (script.size() == 25 && script[0] == OP_DUP && script[1] == OP_HASH160 && script[2] == 20 && script[23] == OP_EQUALVERIFY && script[24] == OP_CHECKSIG) {
pubkeyhash = valtype(script.begin () + 3, script.begin() + 23);
return true;
}
return false;
}
/** Test for "small positive integer" script opcodes - OP_1 through OP_16. */
static constexpr bool IsSmallInteger(opcodetype opcode)
{
return opcode >= OP_1 && opcode <= OP_16;
}
static constexpr bool IsPushdataOp(opcodetype opcode)
{
return opcode > OP_FALSE && opcode <= OP_PUSHDATA4;
}
static constexpr bool IsValidMultisigKeyCount(int n_keys)
{
return n_keys > 0 && n_keys <= MAX_PUBKEYS_PER_MULTISIG;
}
static bool GetMultisigKeyCount(opcodetype opcode, valtype data, int& count)
{
if (IsSmallInteger(opcode)) {
count = CScript::DecodeOP_N(opcode);
return IsValidMultisigKeyCount(count);
}
if (IsPushdataOp(opcode)) {
if (!CheckMinimalPush(data, opcode)) return false;
try {
count = CScriptNum(data, /* fRequireMinimal = */ true).getint();
return IsValidMultisigKeyCount(count);
} catch (const scriptnum_error&) {
return false;
}
}
return false;
}
static bool MatchMultisig(const CScript& script, int& required_sigs, std::vector<valtype>& pubkeys)
{
opcodetype opcode;
valtype data;
int num_keys;
CScript::const_iterator it = script.begin();
if (script.size() < 1 || script.back() != OP_CHECKMULTISIG) return false;
if (!script.GetOp(it, opcode, data) || !GetMultisigKeyCount(opcode, data, required_sigs)) return false;
while (script.GetOp(it, opcode, data) && CPubKey::ValidSize(data)) {
pubkeys.emplace_back(std::move(data));
}
if (!GetMultisigKeyCount(opcode, data, num_keys)) return false;
if (pubkeys.size() != static_cast<unsigned long>(num_keys) || num_keys < required_sigs) return false;
return (it + 1 == script.end());
}
TxoutType Solver(const CScript& scriptPubKey, std::vector<std::vector<unsigned char>>& vSolutionsRet)
{
vSolutionsRet.clear();
// Shortcut for pay-to-script-hash, which are more constrained than the other types:
// it is always OP_HASH160 20 [20 byte hash] OP_EQUAL
if (scriptPubKey.IsPayToScriptHash())
{
std::vector<unsigned char> hashBytes(scriptPubKey.begin()+2, scriptPubKey.begin()+22);
vSolutionsRet.push_back(hashBytes);
return TxoutType::SCRIPTHASH;
}
int witnessversion;
std::vector<unsigned char> witnessprogram;
if (scriptPubKey.IsWitnessProgram(witnessversion, witnessprogram)) {
if (witnessversion == 0 && witnessprogram.size() == WITNESS_V0_KEYHASH_SIZE) {
vSolutionsRet.push_back(std::move(witnessprogram));
return TxoutType::WITNESS_V0_KEYHASH;
}
if (witnessversion == 0 && witnessprogram.size() == WITNESS_V0_SCRIPTHASH_SIZE) {
vSolutionsRet.push_back(std::move(witnessprogram));
return TxoutType::WITNESS_V0_SCRIPTHASH;
}
if (witnessversion == 1 && witnessprogram.size() == WITNESS_V1_TAPROOT_SIZE) {
vSolutionsRet.push_back(std::move(witnessprogram));
return TxoutType::WITNESS_V1_TAPROOT;
}
if (witnessversion != 0) {
vSolutionsRet.push_back(std::vector<unsigned char>{(unsigned char)witnessversion});
vSolutionsRet.push_back(std::move(witnessprogram));
return TxoutType::WITNESS_UNKNOWN;
}
return TxoutType::NONSTANDARD;
}
// Provably prunable, data-carrying output
//
// So long as script passes the IsUnspendable() test and all but the first
// byte passes the IsPushOnly() test we don't care what exactly is in the
// script.
if (scriptPubKey.size() >= 1 && scriptPubKey[0] == OP_RETURN && scriptPubKey.IsPushOnly(scriptPubKey.begin()+1)) {
return TxoutType::NULL_DATA;
}
std::vector<unsigned char> data;
if (MatchPayToPubkey(scriptPubKey, data)) {
vSolutionsRet.push_back(std::move(data));
return TxoutType::PUBKEY;
}
if (MatchPayToPubkeyHash(scriptPubKey, data)) {
vSolutionsRet.push_back(std::move(data));
return TxoutType::PUBKEYHASH;
}
int required;
std::vector<std::vector<unsigned char>> keys;
if (MatchMultisig(scriptPubKey, required, keys)) {
vSolutionsRet.push_back({static_cast<unsigned char>(required)}); // safe as required is in range 1..20
vSolutionsRet.insert(vSolutionsRet.end(), keys.begin(), keys.end());
vSolutionsRet.push_back({static_cast<unsigned char>(keys.size())}); // safe as size is in range 1..20
return TxoutType::MULTISIG;
}
vSolutionsRet.clear();
return TxoutType::NONSTANDARD;
}
bool ExtractDestination(const CScript& scriptPubKey, CTxDestination& addressRet)
{
std::vector<valtype> vSolutions;
TxoutType whichType = Solver(scriptPubKey, vSolutions);
switch (whichType) {
case TxoutType::PUBKEY: {
CPubKey pubKey(vSolutions[0]);
if (!pubKey.IsValid())
return false;
addressRet = PKHash(pubKey);
return true;
}
case TxoutType::PUBKEYHASH: {
addressRet = PKHash(uint160(vSolutions[0]));
return true;
}
case TxoutType::SCRIPTHASH: {
addressRet = ScriptHash(uint160(vSolutions[0]));
return true;
}
case TxoutType::WITNESS_V0_KEYHASH: {
WitnessV0KeyHash hash;
std::copy(vSolutions[0].begin(), vSolutions[0].end(), hash.begin());
addressRet = hash;
return true;
}
case TxoutType::WITNESS_V0_SCRIPTHASH: {
WitnessV0ScriptHash hash;
std::copy(vSolutions[0].begin(), vSolutions[0].end(), hash.begin());
addressRet = hash;
return true;
}
case TxoutType::WITNESS_V1_TAPROOT: {
WitnessV1Taproot tap;
std::copy(vSolutions[0].begin(), vSolutions[0].end(), tap.begin());
addressRet = tap;
return true;
}
case TxoutType::WITNESS_UNKNOWN: {
WitnessUnknown unk;
unk.version = vSolutions[0][0];
std::copy(vSolutions[1].begin(), vSolutions[1].end(), unk.program);
unk.length = vSolutions[1].size();
addressRet = unk;
return true;
}
case TxoutType::MULTISIG:
case TxoutType::NULL_DATA:
case TxoutType::NONSTANDARD:
return false;
} // no default case, so the compiler can warn about missing cases
assert(false);
}
namespace {
class CScriptVisitor
{
public:
CScript operator()(const CNoDestination& dest) const
{
return CScript();
}
CScript operator()(const PKHash& keyID) const
{
return CScript() << OP_DUP << OP_HASH160 << ToByteVector(keyID) << OP_EQUALVERIFY << OP_CHECKSIG;
}
CScript operator()(const ScriptHash& scriptID) const
{
return CScript() << OP_HASH160 << ToByteVector(scriptID) << OP_EQUAL;
}
CScript operator()(const WitnessV0KeyHash& id) const
{
return CScript() << OP_0 << ToByteVector(id);
}
CScript operator()(const WitnessV0ScriptHash& id) const
{
return CScript() << OP_0 << ToByteVector(id);
}
CScript operator()(const WitnessV1Taproot& tap) const
{
return CScript() << OP_1 << ToByteVector(tap);
}
CScript operator()(const WitnessUnknown& id) const
{
return CScript() << CScript::EncodeOP_N(id.version) << std::vector<unsigned char>(id.program, id.program + id.length);
}
};
} // namespace
CScript GetScriptForDestination(const CTxDestination& dest)
{
return std::visit(CScriptVisitor(), dest);
}
CScript GetScriptForRawPubKey(const CPubKey& pubKey)
{
return CScript() << std::vector<unsigned char>(pubKey.begin(), pubKey.end()) << OP_CHECKSIG;
}
CScript GetScriptForMultisig(int nRequired, const std::vector<CPubKey>& keys)
{
CScript script;
script << nRequired;
for (const CPubKey& key : keys)
script << ToByteVector(key);
script << keys.size() << OP_CHECKMULTISIG;
return script;
}
bool IsValidDestination(const CTxDestination& dest) {
return dest.index() != 0;
}
/*static*/ TaprootBuilder::NodeInfo TaprootBuilder::Combine(NodeInfo&& a, NodeInfo&& b)
{
NodeInfo ret;
/* Iterate over all tracked leaves in a, add b's hash to their Merkle branch, and move them to ret. */
for (auto& leaf : a.leaves) {
leaf.merkle_branch.push_back(b.hash);
ret.leaves.emplace_back(std::move(leaf));
}
/* Iterate over all tracked leaves in b, add a's hash to their Merkle branch, and move them to ret. */
for (auto& leaf : b.leaves) {
leaf.merkle_branch.push_back(a.hash);
ret.leaves.emplace_back(std::move(leaf));
}
/* Lexicographically sort a and b's hash, and compute parent hash. */
if (a.hash < b.hash) {
ret.hash = (CHashWriter(HASHER_TAPBRANCH) << a.hash << b.hash).GetSHA256();
} else {
ret.hash = (CHashWriter(HASHER_TAPBRANCH) << b.hash << a.hash).GetSHA256();
}
return ret;
}
void TaprootSpendData::Merge(TaprootSpendData other)
{
// TODO: figure out how to better deal with conflicting information
// being merged.
if (internal_key.IsNull() && !other.internal_key.IsNull()) {
internal_key = other.internal_key;
}
if (merkle_root.IsNull() && !other.merkle_root.IsNull()) {
merkle_root = other.merkle_root;
}
for (auto& [key, control_blocks] : other.scripts) {
// Once P0083R3 is supported by all our targeted platforms,
// this loop body can be replaced with:
// scripts[key].merge(std::move(control_blocks));
auto& target = scripts[key];
for (auto& control_block: control_blocks) {
target.insert(std::move(control_block));
}
}
}
void TaprootBuilder::Insert(TaprootBuilder::NodeInfo&& node, int depth)
{
assert(depth >= 0 && (size_t)depth <= TAPROOT_CONTROL_MAX_NODE_COUNT);
/* We cannot insert a leaf at a lower depth while a deeper branch is unfinished. Doing
* so would mean the Add() invocations do not correspond to a DFS traversal of a
* binary tree. */
if ((size_t)depth + 1 < m_branch.size()) {
m_valid = false;
return;
}
/* As long as an entry in the branch exists at the specified depth, combine it and propagate up.
* The 'node' variable is overwritten here with the newly combined node. */
while (m_valid && m_branch.size() > (size_t)depth && m_branch[depth].has_value()) {
node = Combine(std::move(node), std::move(*m_branch[depth]));
m_branch.pop_back();
if (depth == 0) m_valid = false; /* Can't propagate further up than the root */
--depth;
}
if (m_valid) {
/* Make sure the branch is big enough to place the new node. */
if (m_branch.size() <= (size_t)depth) m_branch.resize((size_t)depth + 1);
assert(!m_branch[depth].has_value());
m_branch[depth] = std::move(node);
}
}
/*static*/ bool TaprootBuilder::ValidDepths(const std::vector<int>& depths)
{
std::vector<bool> branch;
for (int depth : depths) {
// This inner loop corresponds to effectively the same logic on branch
// as what Insert() performs on the m_branch variable. Instead of
// storing a NodeInfo object, just remember whether or not there is one
// at that depth.
if (depth < 0 || (size_t)depth > TAPROOT_CONTROL_MAX_NODE_COUNT) return false;
if ((size_t)depth + 1 < branch.size()) return false;
while (branch.size() > (size_t)depth && branch[depth]) {
branch.pop_back();
if (depth == 0) return false;
--depth;
}
if (branch.size() <= (size_t)depth) branch.resize((size_t)depth + 1);
assert(!branch[depth]);
branch[depth] = true;
}
// And this check corresponds to the IsComplete() check on m_branch.
return branch.size() == 0 || (branch.size() == 1 && branch[0]);
}
TaprootBuilder& TaprootBuilder::Add(int depth, const CScript& script, int leaf_version, bool track)
{
assert((leaf_version & ~TAPROOT_LEAF_MASK) == 0);
if (!IsValid()) return *this;
/* Construct NodeInfo object with leaf hash and (if track is true) also leaf information. */
NodeInfo node;
node.hash = (CHashWriter{HASHER_TAPLEAF} << uint8_t(leaf_version) << script).GetSHA256();
if (track) node.leaves.emplace_back(LeafInfo{script, leaf_version, {}});
/* Insert into the branch. */
Insert(std::move(node), depth);
return *this;
}
TaprootBuilder& TaprootBuilder::AddOmitted(int depth, const uint256& hash)
{
if (!IsValid()) return *this;
/* Construct NodeInfo object with the hash directly, and insert it into the branch. */
NodeInfo node;
node.hash = hash;
Insert(std::move(node), depth);
return *this;
}
TaprootBuilder& TaprootBuilder::Finalize(const XOnlyPubKey& internal_key)
{
/* Can only call this function when IsComplete() is true. */
assert(IsComplete());
m_internal_key = internal_key;
auto ret = m_internal_key.CreateTapTweak(m_branch.size() == 0 ? nullptr : &m_branch[0]->hash);
assert(ret.has_value());
std::tie(m_output_key, m_parity) = *ret;
return *this;
}
WitnessV1Taproot TaprootBuilder::GetOutput() { return WitnessV1Taproot{m_output_key}; }
TaprootSpendData TaprootBuilder::GetSpendData() const
{
assert(IsComplete());
TaprootSpendData spd;
spd.merkle_root = m_branch.size() == 0 ? uint256() : m_branch[0]->hash;
spd.internal_key = m_internal_key;
if (m_branch.size()) {
// If any script paths exist, they have been combined into the root m_branch[0]
// by now. Compute the control block for each of its tracked leaves, and put them in
// spd.scripts.
for (const auto& leaf : m_branch[0]->leaves) {
std::vector<unsigned char> control_block;
control_block.resize(TAPROOT_CONTROL_BASE_SIZE + TAPROOT_CONTROL_NODE_SIZE * leaf.merkle_branch.size());
control_block[0] = leaf.leaf_version | (m_parity ? 1 : 0);
std::copy(m_internal_key.begin(), m_internal_key.end(), control_block.begin() + 1);
if (leaf.merkle_branch.size()) {
std::copy(leaf.merkle_branch[0].begin(),
leaf.merkle_branch[0].begin() + TAPROOT_CONTROL_NODE_SIZE * leaf.merkle_branch.size(),
control_block.begin() + TAPROOT_CONTROL_BASE_SIZE);
}
spd.scripts[{leaf.script, leaf.leaf_version}].insert(std::move(control_block));
}
}
return spd;
}
std::optional<std::vector<std::tuple<int, CScript, int>>> InferTaprootTree(const TaprootSpendData& spenddata, const XOnlyPubKey& output)
{
// Verify that the output matches the assumed Merkle root and internal key.
auto tweak = spenddata.internal_key.CreateTapTweak(spenddata.merkle_root.IsNull() ? nullptr : &spenddata.merkle_root);
if (!tweak || tweak->first != output) return std::nullopt;
// If the Merkle root is 0, the tree is empty, and we're done.
std::vector<std::tuple<int, CScript, int>> ret;
if (spenddata.merkle_root.IsNull()) return ret;
/** Data structure to represent the nodes of the tree we're going to build. */
struct TreeNode {
/** Hash of this node, if known; 0 otherwise. */
uint256 hash;
/** The left and right subtrees (note that their order is irrelevant). */
std::unique_ptr<TreeNode> sub[2];
/** If this is known to be a leaf node, a pointer to the (script, leaf_ver) pair.
* nullptr otherwise. */
const std::pair<CScript, int>* leaf = nullptr;
/** Whether or not this node has been explored (is known to be a leaf, or known to have children). */
bool explored = false;
/** Whether or not this node is an inner node (unknown until explored = true). */
bool inner;
/** Whether or not we have produced output for this subtree. */
bool done = false;
};
// Build tree from the provided branches.
TreeNode root;
root.hash = spenddata.merkle_root;
for (const auto& [key, control_blocks] : spenddata.scripts) {
const auto& [script, leaf_ver] = key;
for (const auto& control : control_blocks) {
// Skip script records with nonsensical leaf version.
if (leaf_ver < 0 || leaf_ver >= 0x100 || leaf_ver & 1) continue;
// Skip script records with invalid control block sizes.
if (control.size() < TAPROOT_CONTROL_BASE_SIZE || control.size() > TAPROOT_CONTROL_MAX_SIZE ||
((control.size() - TAPROOT_CONTROL_BASE_SIZE) % TAPROOT_CONTROL_NODE_SIZE) != 0) continue;
// Skip script records that don't match the control block.
if ((control[0] & TAPROOT_LEAF_MASK) != leaf_ver) continue;
// Skip script records that don't match the provided Merkle root.
const uint256 leaf_hash = ComputeTapleafHash(leaf_ver, script);
const uint256 merkle_root = ComputeTaprootMerkleRoot(control, leaf_hash);
if (merkle_root != spenddata.merkle_root) continue;
TreeNode* node = &root;
size_t levels = (control.size() - TAPROOT_CONTROL_BASE_SIZE) / TAPROOT_CONTROL_NODE_SIZE;
for (size_t depth = 0; depth < levels; ++depth) {
// Can't descend into a node which we already know is a leaf.
if (node->explored && !node->inner) return std::nullopt;
// Extract partner hash from Merkle branch in control block.
uint256 hash;
std::copy(control.begin() + TAPROOT_CONTROL_BASE_SIZE + (levels - 1 - depth) * TAPROOT_CONTROL_NODE_SIZE,
control.begin() + TAPROOT_CONTROL_BASE_SIZE + (levels - depth) * TAPROOT_CONTROL_NODE_SIZE,
hash.begin());
if (node->sub[0]) {
// Descend into the existing left or right branch.
bool desc = false;
for (int i = 0; i < 2; ++i) {
if (node->sub[i]->hash == hash || (node->sub[i]->hash.IsNull() && node->sub[1-i]->hash != hash)) {
node->sub[i]->hash = hash;
node = &*node->sub[1-i];
desc = true;
break;
}
}
if (!desc) return std::nullopt; // This probably requires a hash collision to hit.
} else {
// We're in an unexplored node. Create subtrees and descend.
node->explored = true;
node->inner = true;
node->sub[0] = std::make_unique<TreeNode>();
node->sub[1] = std::make_unique<TreeNode>();
node->sub[1]->hash = hash;
node = &*node->sub[0];
}
}
// Cannot turn a known inner node into a leaf.
if (node->sub[0]) return std::nullopt;
node->explored = true;
node->inner = false;
node->leaf = &key;
node->hash = leaf_hash;
}
}
// Recursive processing to turn the tree into flattened output. Use an explicit stack here to avoid
// overflowing the call stack (the tree may be 128 levels deep).
std::vector<TreeNode*> stack{&root};
while (!stack.empty()) {
TreeNode& node = *stack.back();
if (!node.explored) {
// Unexplored node, which means the tree is incomplete.
return std::nullopt;
} else if (!node.inner) {
// Leaf node; produce output.
ret.emplace_back(stack.size() - 1, node.leaf->first, node.leaf->second);
node.done = true;
stack.pop_back();
} else if (node.sub[0]->done && !node.sub[1]->done && !node.sub[1]->explored && !node.sub[1]->hash.IsNull() &&
(CHashWriter{HASHER_TAPBRANCH} << node.sub[1]->hash << node.sub[1]->hash).GetSHA256() == node.hash) {
// Whenever there are nodes with two identical subtrees under it, we run into a problem:
// the control blocks for the leaves underneath those will be identical as well, and thus
// they will all be matched to the same path in the tree. The result is that at the location
// where the duplicate occurred, the left child will contain a normal tree that can be explored
// and processed, but the right one will remain unexplored.
//
// This situation can be detected, by encountering an inner node with unexplored right subtree
// with known hash, and H_TapBranch(hash, hash) is equal to the parent node (this node)'s hash.
//
// To deal with this, simply process the left tree a second time (set its done flag to false;
// noting that the done flag of its children have already been set to false after processing
// those). To avoid ending up in an infinite loop, set the done flag of the right (unexplored)
// subtree to true.
node.sub[0]->done = false;
node.sub[1]->done = true;
} else if (node.sub[0]->done && node.sub[1]->done) {
// An internal node which we're finished with.
node.sub[0]->done = false;
node.sub[1]->done = false;
node.done = true;
stack.pop_back();
} else if (!node.sub[0]->done) {
// An internal node whose left branch hasn't been processed yet. Do so first.
stack.push_back(&*node.sub[0]);
} else if (!node.sub[1]->done) {
// An internal node whose right branch hasn't been processed yet. Do so first.
stack.push_back(&*node.sub[1]);
}
}
return ret;
}