UInt64 Message (De)serialization code truncates the value

[sebandraos]

It appears that the serialization and deserialization code in UInt64.h truncates the value of the UInt64 to 4 bytes. This has no issues with values up to UInt32.Max (which is plenty big enough for most real world applications) but this appears to be a potential source of errors down the line. Unless of course I'm missing some trick being used to compress the UInt64 value into 4 bytes? But that would conflict with the offset line:

      offset += sizeof(this->data); // data is uint64 therefore size == 8

It's worth noting that the same is not true of the signed Int64.h, which does include all 8 bytes.

Here is a very quick sanity check I was using:

#include <iostream>

int main()
{
union
{
    uint64_t real;
    uint32_t base;
} u_x;
    u_x.real = 5294967295;
    std::cout << "u.real (size) = " << sizeof(u_x.real) << '\n';
    std::cout << "u.real = " << u_x.real << '\n';
    std::cout << std::hex << "u.real (hex) = " << u_x.real << '\n';
    std::cout << std::hex << "u.real[0] = " << ((u_x.real >> (8 * 0)) & 0xFF) << '\n';
    std::cout << std::hex << "u.real[1] = " << ((u_x.real >> (8 * 1)) & 0xFF) << '\n';
    std::cout << std::hex << "u.real[2] = " << ((u_x.real >> (8 * 2)) & 0xFF) << '\n';
    std::cout << std::hex << "u.real[3] = " << ((u_x.real >> (8 * 3)) & 0xFF) << '\n';
    std::cout << std::hex << "u.real[4] = " << ((u_x.real >> (8 * 4)) & 0xFF) << '\n';
    std::cout << std::hex << "u.real[5] = " << ((u_x.real >> (8 * 5)) & 0xFF) << '\n';
    std::cout << std::hex << "u.real[6] = " << ((u_x.real >> (8 * 6)) & 0xFF) << '\n';
    std::cout << std::hex << "u.real[7] = " << ((u_x.real >> (8 * 7)) & 0xFF) << '\n';
    std::cout << std::hex << "u.base (hex) = " << u_x.base << '\n';
    std::cout << std::hex << "u.base[0] = " << ((u_x.base >> (8 * 0)) & 0xFF) << '\n';
    std::cout << std::hex << "u.base[1] = " << ((u_x.base >> (8 * 1)) & 0xFF) << '\n';
    std::cout << std::hex << "u.base[2] = " << ((u_x.base >> (8 * 2)) & 0xFF) << '\n';
    std::cout << std::hex << "u.base[3] = " << ((u_x.base >> (8 * 3)) & 0xFF) << '\n';
    //std::cout << std::hex << "u.base[4] = " << ((u_x.base >> (8 * 4)) & 0xFF) << '\n';
    //std::cout << std::hex << "u.base[5] = " << ((u_x.base >> (8 * 5)) & 0xFF) << '\n';
    //std::cout << std::hex << "u.base[6] = " << ((u_x.base >> (8 * 6)) & 0xFF) << '\n';
    //std::cout << std::hex << "u.base[7] = " << ((u_x.base >> (8 * 7)) & 0xFF) << '\n';
    
union
{
    uint64_t real;
    uint32_t base;
} u_data;
    u_data.base = 0;
    u_data.base |= ((u_x.base >> (8 * 0)) & 0xFF) << (8 * 0);
    u_data.base |= ((u_x.base >> (8 * 1)) & 0xFF) << (8 * 1);
    u_data.base |= ((u_x.base >> (8 * 2)) & 0xFF) << (8 * 2);
    u_data.base |= ((u_x.base >> (8 * 3)) & 0xFF) << (8 * 3);
    uint64_t value = u_data.base;
    std::cout << "value = " << value << '\n';
}

Output:

u.real (size) = 8
u.real = 5294967295
u.real (hex) = 13b9ac9ff
u.real[0] = ff
u.real[1] = c9
u.real[2] = 9a
u.real[3] = 3b
u.real[4] = 1
u.real[5] = 0
u.real[6] = 0
u.real[7] = 0
u.base (hex) = 3b9ac9ff // missing 1 byte
u.base[0] = ff
u.base[1] = c9
u.base[2] = 9a
u.base[3] = 3b
value = 3b9ac9ff

Thanks,

Seb

[salkinium]

cc @strongly-typed

[sebandraos]

As a slight follow up:
This implemenation looks "correct" (per my initial post) and this one has the same implementation as here.

[salkinium]

I suspect there may be a patch missing somewhere upstream, as you found two different implementations.
I'm not sure how any of this works, ping @strongly-typed is the maintainer.