std::getline and UTF16 (LE) filestream - not working - c++11

Question

Well the following should work - I can't find a reason why it shouldn't:
std::fstream f;
std::string myoutput;
f.imbue(std::locale(f.getloc(), new std::codecvt_utf16<wchar_t, std::little_endian | std::consume_header>));
f.open("c:\\test.txt", std::ios::in);
std::getline(f, myoutput);
The code is executed on the following file (in hex - it should spell "hello world"):
FF FE 68 00 65 00 6C 00 6C 00 6F 00 20 00 77 00 6F 00 72 00 6C 00 64
00
The ultimate goal is to abstract the encoding away, always consider a file UTF-8 unless the first bytes are a BOM. Now above code would be executed after reading a BOM and noticing it is UTF-16. It should hence read UTF-16 file, and convert it to a utf-8 string.
However the std::getline does not ignore the BOM (easily fixed) but furthermore it does not respect the fact that UTF-16 uses 2 bytes. (And it stops after reading the first 3 bytes upon seeing "0" ).
Now of course I could use std::wfstream. But as I wish to "hide" the unicode type from the user for one thing all the "filestreams" are stored inside a container for referencing. So the signature of all those filestreams has to be equal - and be based on char and std::string

Answer 1

If you opened your file as basic_fstream<char>, you've already set both external and internal character width to 1 byte and the locale facet you're applying will never be used.
Either read into a string and apply wstring_convert twice, or apply wbuffer_convert to make the internal character width bigger, and then wstring_convert:
std::fstream f;
f.open("test.txt", std::ios::in | std::ios::binary);
std::wbuffer_convert<std::codecvt_utf16<wchar_t,
0x10ffff, // note your 2nd parameter was wrong
std::little_endian // or consume_header, not both
>> cvt1(f.rdbuf());
std::wistream wide_f(&cvt1);
std::wstring wstr;
std::getline(wide_f, wstr);
std::wstring_convert<std::codecvt_utf8<wchar_t>> cvt2;
std::string u8str = cvt2.to_bytes(wstr);
std::cout << u8str << '\n';