What is the encoding used by boost-asio error messages?

Question

I am using boost-asio and I want to process correctly the error message.

In this example, I made a typo (1278 instead of 127):

boost::system::error_code ec;
auto address=boost::asio::ip::address::from_string("1278.0.0.1",ec);
if(ec)
{
    auto text=ec.message();
    //Do Something with text but what is its encoding ?
}

I get an error message and it seems to be encoded in Windows 1252 (I am using Windows 7). So it seems that the encoding is the OS encoding.

However, I was unable to find any documentation that states this fact.

Are error message in boost asio encoded with the OS character set ?

It's not likely to be an ASIO question, but a [Boost.System](http://www.boost.org/doc/libs/1_54_0/libs/system/doc/index.html) question. So, it will come down to how the error category (which is ultimately responsible for the message strings) is implmented. — Dave S, Aug 28 '13 at 14:55
Just to let anyone copying this code know: the problem with this code is that it has an UB in it. Signature of the message() is `std::string message() const`, so it returns temporary, if you call c_str() on it then text will be valid only until the end of full statement. So i the line below its UB to use it. — marcinj, Jan 24 '20 at 20:30

score 5 · Accepted Answer · answered Aug 29 '13 at 09:00

After digging into boost on my system, I found that the included hpp files include an ipp file which in turn calls the OS functions. If there is an error, only the code of the error is known at this stage.

The real error message is formed when the message() function is called.

The implementation on windows calls FormatMessageA or FormatMessageW depending if BOOST_NO_ANSI_APIS is defined:

  std::string system_error_category::message( int ev ) const
  {
# ifndef BOOST_NO_ANSI_APIS  
    LPVOID lpMsgBuf = 0;
    DWORD retval = ::FormatMessageA( 
        FORMAT_MESSAGE_ALLOCATE_BUFFER | 
        FORMAT_MESSAGE_FROM_SYSTEM | 
        FORMAT_MESSAGE_IGNORE_INSERTS,
        NULL,
        ev,
        MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT), // Default language
        (LPSTR) &lpMsgBuf,
        0,
        NULL 
    );
    detail::local_free_on_destruction lfod(lpMsgBuf);
    if (retval == 0)
        return std::string("Unknown error");

    std::string str( static_cast<LPCSTR>(lpMsgBuf) );
# else  // WinCE workaround
    LPVOID lpMsgBuf = 0;
    DWORD retval = ::FormatMessageW( 
        FORMAT_MESSAGE_ALLOCATE_BUFFER | 
        FORMAT_MESSAGE_FROM_SYSTEM | 
        FORMAT_MESSAGE_IGNORE_INSERTS,
        NULL,
        ev,
        MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT), // Default language
        (LPWSTR) &lpMsgBuf,
        0,
        NULL 
    );
    detail::local_free_on_destruction lfod(lpMsgBuf);
    if (retval == 0)
        return std::string("Unknown error");

    int num_chars = (wcslen( static_cast<LPCWSTR>(lpMsgBuf) ) + 1) * 2;
    LPSTR narrow_buffer = (LPSTR)_alloca( num_chars );
    if (::WideCharToMultiByte(CP_ACP, 0, static_cast<LPCWSTR>(lpMsgBuf), -1, narrow_buffer, num_chars, NULL, NULL) == 0)
        return std::string("Unknown error");

    std::string str( narrow_buffer );
# endif
    while ( str.size()
      && (str[str.size()-1] == '\n' || str[str.size()-1] == '\r') )
        str.erase( str.size()-1 );
    if ( str.size() && str[str.size()-1] == '.' ) 
      { str.erase( str.size()-1 ); }
    return str;
  }

If message calls FormatMessageW, the string is narrowed back to the system default Windows ANSI code page (CP_ACP)

In all cases on Windows the result is the default ANSI Codepage.

This also appears to be the case for std::error_code – Pete Feb 13 '19 at 10:07 — Pete, Feb 13 '19 at 10:07

What is the encoding used by boost-asio error messages?

1 Answers1