Warnings with Visual Studio c++ (VS2015 Update 3)

[tksatware]

When using the library with uint64_t values, some specialisations throw al lot of warnings (4267 and 4293).
The resulting code works fine, but I like a clean build log :)

So I propose to add the following lines before the first namespace Definition (line 103):

#if defined(_MSC_VER)
#pragma warning(push)
#pragma warning(disable: 4267 4293)
#endif

and this at the end before the last #endif:

#if defined(_MSC_VER)
#pragma warning(pop)
#endif

This also could be solved with type casts, but I think they would make the code less readable.

[nlohmann]

Interesting - it seems as if MSVC treats number literals like 0x9a as size_t and warns about the conversion to unsigned char. I shall have a look whether there is a way to fix it without the pragmas.

[prsyahmi]

Adding another cast to all add_to_vector 8-bytes shift solve warning C4293

            case 8:
            {
                vec.push_back(static_cast<uint8_t>((static_cast<uint64_t>(number) >> 070) & 0xff));
                vec.push_back(static_cast<uint8_t>((static_cast<uint64_t>(number) >> 060) & 0xff));
                vec.push_back(static_cast<uint8_t>((static_cast<uint64_t>(number) >> 050) & 0xff));
                vec.push_back(static_cast<uint8_t>((static_cast<uint64_t>(number) >> 040) & 0xff));
                // intentional fall-through
            }

The 8-bytes add_to_vector however is never executed for string,object and array when targeting x86 on msvc since the size_type = size_t = 32bits. This (probably a dirty) fix works for me since I don't use large json data that exceeds 4GB.

[tksatware]

or more readable: case 8: { uint64_t n = number; // casts by itself vec.push_back(static_cast<uint8_t>((n >> 070) & 0xff)); vec.push_back(static_cast<uint8_t>((n >> 060) & 0xff)); vec.push_back(static_cast<uint8_t>((n >> 050) & 0xff)); vec.push_back(static_cast<uint8_t>((n >> 040) & 0xff)); // intentional fall-through } And it also happens on x86 code, of course, when you're storing large values like database Indices.

[nlohmann]

Both warnings to not occur any more. I think this can be closed, right?

[tksatware]

Confirmed! Thanks! Also on gcc 4.9.2, (raspian 4.9.2-10) where similar warnings occured.