Integer value problem

[mengguang]

I copyed the sample code and just change the value of c to UINT64_MAX, the assert failed.

#include <alpaca/alpaca.h>

int main() {

    struct MyStruct {
        char a;
        int b;
        uint64_t c;
        float d;
        bool e;
    };

    MyStruct s{ 'a', 5, UINT64_MAX, 3.14f, true };

    // Serialize
    std::vector<uint8_t> bytes;
    auto bytes_written = alpaca::serialize(s, bytes);

    // Deserialize
    std::error_code ec;
    auto recovered = alpaca::deserialize<MyStruct>(bytes, ec);
    assert((bool)ec == false);
    assert(recovered.a == 'a');
    assert(recovered.b == 5);
    assert(recovered.c == UINT64_MAX);
    assert(recovered.d == 3.14f);
    assert(recovered.e == true);
}

Is this a problem? My test environment is msvc2022.
I also tested the same code in Ubuntu 20.04.4 LTS and g++ version 9.4.0 , same result.

[marlowa]

This looks like a serious problem to me. I am hunting for a modern C++ library for serialise/deserialise and came across this one, but this bug would be a showstopper. The code I am working with employs structures where 64 bit unsigned integers take the maximum permitted value to indicate that the value is not there. I know that's horrible, but there's nothing I can do about it. But it is why this bug would need to be fixed before I could use the library.

[lrodorigo]

The variable length encoding is still totally broken.
However thanks to last merged pull requests (like #45), the fixed length encoding is properly working and it is architecture independent, I am using the library and it is working pretty well...

Just use

constexpr auto OPTIONS = options::fixed_length_encoding;
std::vector<uint8_t> bytes;
alpaca::serialize<OPTIONS>(data, bytes);

when serializing and deserializing