sizeof() can return a platform-dependent result

[jmjatlanta]

As stated here: #995 (comment) , sizeof() can return a platform-dependent result. This can endanger consensus, if one node rejects a (block | transaction) due to size, but another accepts it.

In all code that limits or rejects due to size, a platform-independent size must be calculated.

Impacts

• API (the application programming interface)
• Build (the build process or something prior to compiled code)
• CLI (the command line wallet)
• Deployment (the deployment process after building such as Docker, Travis, etc.)
• DEX (the Decentralized EXchange, market engine, etc.)
• P2P (the peer-to-peer network for transaction/block propagation)
• Performance (system or user efficiency, etc.)
• Protocol (the blockchain logic, consensus, validation, etc.)
• Security (the security of system or user data, etc.)
• UX (the User Experience)
• Other (please add below)

Steps To Reproduce
tbd

Expected Behavior
Blocks or transactions should be accepted or rejected independent of the node platform.

CORE TEAM TASK LIST

• Evaluate / Prioritize Bug Report
• Refine User Stories / Requirements
• Define Test Cases
• Design / Develop Solution
• Perform QA/Testing
• Update Documentation

[jmjatlanta]

64bit Linux:
The size of a(n) int is: 4
The size of a(n) long is: 8
The size of a(n) long long is: 8
The size of a(n) unsigned long is: 8
The size of a(n) void* is: 8
The size of a(n) std::string is: 32

64bit macOS:
The size of a(n) int is: 4
The size of a(n) long is: 8
The size of a(n) long long is: 8
The size of a(n) unsigned long is: 8
The size of a(n) void* is: 8
The size of a(n) std::string is: 24

64bit WIndows 10
The size of a(n) int is: 4
The size of a(n) long is: 4
The size of a(n) long long is: 8
The size of a(n) unsigned long is: 4
The size of a(n) void* is: 4
The size of a(n) std::string is: 28

Related PR: bitshares/bitshares-fc#100

[jmjatlanta]

From what I can see, blocks are rejected if fc::raw::pack_size is greater than global_properties.parameters.maximum_block_size. Need to verify that pack_size reports that the same block is the same size regardless of platform, which I believe to be the case, but will investigate.

Transactions can also be rejected if fc::raw::pack_size is greater than global_properties.maximum_transaction_size. As above, we need to verify that packed_size reports the same number regardless of platform. See #1573

[abitmore]

E.G. code in memo.cpp:

bitshares-core/libraries/protocol/memo.cpp

Lines 76 to 90 in 173ff3d

	string memo_message::serialize() const
	{
	auto serial_checksum = string(sizeof(checksum), ' ');
	(uint32_t&)(*serial_checksum.data()) = boost::endian::native_to_little(checksum);
	return serial_checksum + text;
	}
	memo_message memo_message::deserialize(const string& serial)
	{
	memo_message result;
	FC_ASSERT( serial.size() >= sizeof(result.checksum) );
	result.checksum = boost::endian::little_to_native((uint32_t&)(*serial.data()));
	result.text = serial.substr(sizeof(result.checksum));
	return result;
	}

BTW the combination of a (uint32_t &) conversion and a native_to_little() call looks suspicious.