Can system_error be modified to add the numerical error code in the message?

[SoftdriveFelix]

We have logs that contain error messages in various languages like japanese, spanish, french, arabic, etc which causes our bug investigations to be difficult. Having a numerical error code would at least allow us to look up the message online.

Looking at the source, I think the following can be changed in the system_error file in the _System_error class:

Current:

STL/stl/inc/system_error

Lines 466 to 473 in 85e4b57

	static string _Makestr(error_code _Errcode, string _Message) { // compose error message
	if (!_Message.empty()) {
	_Message.append(": ");
	}
	_Message.append(_Errcode.message());
	return _Message;
	}

Suggested:

    static string _Makestr(error_code _Errcode, string _Message) { // compose error message
        if (!_Message.empty()) {
            _Message.append(": ");
        }
        _Message.append(to_string(_Errcode.value()));
        _Message.append(" - ");
        _Message.append(_Errcode.message());
        return _Message;
    }

When calling std::exception::what(), the error code would now be included.

This is just a proposition and I don't really mind the format itself. What's important is just the idea of adding the numerical code.

I wish it was possible to force FormatMessage to output english messages, but it doesn't work if the english package is not installed ☹️

Thanks,
Félix

[StephanTLavavej]

The error_code has both a numeric value and a named category, and I think both would need to be captured. It does seem reasonable to do.

[SoftdriveFelix]

Any format preferences on your side? boost::system::error_code::what() uses the following format Message [system:0] which I personally like.
If your team doesn't have the time to implement it I can submit a PR (or one of my teammate).

[StephanTLavavej]

Imitating their format seems like following reasonable precedent. (However, please don't look at their implementation while you do so, or your PR will become a derived work which adds additional headaches, even though Boost has a friendly license.)

I don't have the time to implement this (for context I am the only maintainer of the STL right now, keeping this jumbo jet in the air is all-consuming), but I can review a PR. Please follow codebase precedent and add reasonable test coverage. Thanks!