Replace use of standard IO from error handling

[theShmoo]

In
https://github.com/nlohmann/json/blame/develop/include/nlohmann/detail/output/serializer.hpp#L598-L599
and
https://github.com/nlohmann/json/blame/develop/include/nlohmann/detail/output/serializer.hpp#L504-L505
std::stringstream is used.

Use fmt instead. Instead of:

std::stringstream ss;
ss << std::uppercase << std::setfill('0') << std::setw(2) << std::hex << (byte | 0);
JSON_THROW(type_error::create(316, "invalid UTF-8 byte at index " + std::to_string(i) + ": 0x" + ss.str(), BasicJsonType()));

use:

JSON_THROW(type_error::create(316, fmt::format("invalid UTF-8 byte at index {}: {:#04x}", i, byte), BasicJsonType()));

[nlohmann]

(Removed the bug label)

The code parts you mentioned are just to create an exception message. We used std::snprintf before, but had some warnings in some exotic compiler IIRC. I think adding a thirdparty library here would be overkill, or did I miss anything?

[theShmoo]

Thanks for your fast response!

Aren't you already using fmt?

The problem is that we can't use IO streams in our code base (as we use your library in Intel SGX)

[theShmoo]

A while ago you helped me add the define JSON_NO_IO to your library.
maybe we could also use this define here?

Oh and I checked. You are not using fmt. Sorry. -> I will edit the title of the issue

[nlohmann]

Ah, now I understand. That's indeed a limitation. Maybe you can check what fmt is doing under the hood - I guess some printf magic.

[theShmoo]

I implemented a short version of their conversion.
They use (a more complex version) of something similar that this:
https://godbolt.org/z/9n48qPKa6

[nlohmann]

Awesome! I will have a look tonight. Is there a way I add the compiler you use in our CI? Then I would have caught these error myself.

[theShmoo]

That is very nice of you!

As Intel provides their own SGX version of the whole standard library I would not recommend to add this to your ci, I think this would not pay off.

I always check in on your new releases and try to use it in our Codebase. If I encounter some problems I will report it.