You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
189 lines
5.7 KiB
189 lines
5.7 KiB
// Copyright (c) 2009-2016 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 "core_io.h" |
|
|
|
#include "primitives/block.h" |
|
#include "primitives/transaction.h" |
|
#include "script/script.h" |
|
#include "serialize.h" |
|
#include "streams.h" |
|
#include <univalue.h> |
|
#include "util.h" |
|
#include "utilstrencodings.h" |
|
#include "version.h" |
|
|
|
#include <boost/algorithm/string/classification.hpp> |
|
#include <boost/algorithm/string/predicate.hpp> |
|
#include <boost/algorithm/string/replace.hpp> |
|
#include <boost/algorithm/string/split.hpp> |
|
|
|
CScript ParseScript(const std::string& s) |
|
{ |
|
CScript result; |
|
|
|
static std::map<std::string, opcodetype> mapOpNames; |
|
|
|
if (mapOpNames.empty()) |
|
{ |
|
for (unsigned int op = 0; op <= MAX_OPCODE; op++) |
|
{ |
|
// Allow OP_RESERVED to get into mapOpNames |
|
if (op < OP_NOP && op != OP_RESERVED) |
|
continue; |
|
|
|
const char* name = GetOpName((opcodetype)op); |
|
if (strcmp(name, "OP_UNKNOWN") == 0) |
|
continue; |
|
std::string strName(name); |
|
mapOpNames[strName] = (opcodetype)op; |
|
// Convenience: OP_ADD and just ADD are both recognized: |
|
boost::algorithm::replace_first(strName, "OP_", ""); |
|
mapOpNames[strName] = (opcodetype)op; |
|
} |
|
} |
|
|
|
std::vector<std::string> words; |
|
boost::algorithm::split(words, s, boost::algorithm::is_any_of(" \t\n"), boost::algorithm::token_compress_on); |
|
|
|
for (std::vector<std::string>::const_iterator w = words.begin(); w != words.end(); ++w) |
|
{ |
|
if (w->empty()) |
|
{ |
|
// Empty string, ignore. (boost::split given '' will return one word) |
|
} |
|
else if (all(*w, boost::algorithm::is_digit()) || |
|
(boost::algorithm::starts_with(*w, "-") && all(std::string(w->begin()+1, w->end()), boost::algorithm::is_digit()))) |
|
{ |
|
// Number |
|
int64_t n = atoi64(*w); |
|
result << n; |
|
} |
|
else if (boost::algorithm::starts_with(*w, "0x") && (w->begin()+2 != w->end()) && IsHex(std::string(w->begin()+2, w->end()))) |
|
{ |
|
// Raw hex data, inserted NOT pushed onto stack: |
|
std::vector<unsigned char> raw = ParseHex(std::string(w->begin()+2, w->end())); |
|
result.insert(result.end(), raw.begin(), raw.end()); |
|
} |
|
else if (w->size() >= 2 && boost::algorithm::starts_with(*w, "'") && boost::algorithm::ends_with(*w, "'")) |
|
{ |
|
// Single-quoted string, pushed as data. NOTE: this is poor-man's |
|
// parsing, spaces/tabs/newlines in single-quoted strings won't work. |
|
std::vector<unsigned char> value(w->begin()+1, w->end()-1); |
|
result << value; |
|
} |
|
else if (mapOpNames.count(*w)) |
|
{ |
|
// opcode, e.g. OP_ADD or ADD: |
|
result << mapOpNames[*w]; |
|
} |
|
else |
|
{ |
|
throw std::runtime_error("script parse error"); |
|
} |
|
} |
|
|
|
return result; |
|
} |
|
|
|
// Check that all of the input and output scripts of a transaction contains valid opcodes |
|
bool CheckTxScriptsSanity(const CMutableTransaction& tx) |
|
{ |
|
// Check input scripts for non-coinbase txs |
|
if (!CTransaction(tx).IsCoinBase()) { |
|
for (unsigned int i = 0; i < tx.vin.size(); i++) { |
|
if (!tx.vin[i].scriptSig.HasValidOps() || tx.vin[i].scriptSig.size() > MAX_SCRIPT_SIZE) { |
|
return false; |
|
} |
|
} |
|
} |
|
// Check output scripts |
|
for (unsigned int i = 0; i < tx.vout.size(); i++) { |
|
if (!tx.vout[i].scriptPubKey.HasValidOps() || tx.vout[i].scriptPubKey.size() > MAX_SCRIPT_SIZE) { |
|
return false; |
|
} |
|
} |
|
|
|
return true; |
|
} |
|
|
|
bool DecodeHexTx(CMutableTransaction& tx, const std::string& strHexTx, bool fTryNoWitness) |
|
{ |
|
if (!IsHex(strHexTx)) { |
|
return false; |
|
} |
|
|
|
std::vector<unsigned char> txData(ParseHex(strHexTx)); |
|
|
|
if (fTryNoWitness) { |
|
CDataStream ssData(txData, SER_NETWORK, PROTOCOL_VERSION | SERIALIZE_TRANSACTION_NO_WITNESS); |
|
try { |
|
ssData >> tx; |
|
if (ssData.eof() && CheckTxScriptsSanity(tx)) { |
|
return true; |
|
} |
|
} |
|
catch (const std::exception&) { |
|
// Fall through. |
|
} |
|
} |
|
|
|
CDataStream ssData(txData, SER_NETWORK, PROTOCOL_VERSION); |
|
try { |
|
ssData >> tx; |
|
if (!ssData.empty()) { |
|
return false; |
|
} |
|
} |
|
catch (const std::exception&) { |
|
return false; |
|
} |
|
|
|
return true; |
|
} |
|
|
|
bool DecodeHexBlk(CBlock& block, const std::string& strHexBlk) |
|
{ |
|
if (!IsHex(strHexBlk)) |
|
return false; |
|
|
|
std::vector<unsigned char> blockData(ParseHex(strHexBlk)); |
|
CDataStream ssBlock(blockData, SER_NETWORK, PROTOCOL_VERSION); |
|
try { |
|
ssBlock >> block; |
|
} |
|
catch (const std::exception&) { |
|
return false; |
|
} |
|
|
|
return true; |
|
} |
|
|
|
uint256 ParseHashUV(const UniValue& v, const std::string& strName) |
|
{ |
|
std::string strHex; |
|
if (v.isStr()) |
|
strHex = v.getValStr(); |
|
return ParseHashStr(strHex, strName); // Note: ParseHashStr("") throws a runtime_error |
|
} |
|
|
|
uint256 ParseHashStr(const std::string& strHex, const std::string& strName) |
|
{ |
|
if (!IsHex(strHex)) // Note: IsHex("") is false |
|
throw std::runtime_error(strName + " must be hexadecimal string (not '" + strHex + "')"); |
|
|
|
uint256 result; |
|
result.SetHex(strHex); |
|
return result; |
|
} |
|
|
|
std::vector<unsigned char> ParseHexUV(const UniValue& v, const std::string& strName) |
|
{ |
|
std::string strHex; |
|
if (v.isStr()) |
|
strHex = v.getValStr(); |
|
if (!IsHex(strHex)) |
|
throw std::runtime_error(strName + " must be hexadecimal string (not '" + strHex + "')"); |
|
return ParseHex(strHex); |
|
}
|
|
|