Re: [RFC 0/2] C++11 codecvt specializations.

[Rüdiger Sonderfeld <ruediger at c-plusplus dot de>]

Hello Tom,

On Monday 28 April 2014 08:49:23 Tom Tromey wrote:
> Also, some implementations of iconv cannot convert to or from the
> wchar_t encoding; though I imagine you can work around this with
> appropriate template specializations to use the <wchar.h> stuff.

It seems that the wchar.h conversion functions do not properly convert the 
encoding.  I used a few examples and they failed to convert non-ASCII 
characters in either direction even with a UTF-8 locale.

In any case the standard code conversion facilities ([locale.stdcvt]) are only 
defined with UCS2 and UCS4 in mind depending on the type.  So I guess this 
could be done through a template check and if wchar_t is two byte it's treated 
as UCS2.

Come to think of it, those conversion facilities are rather strangely defined:

From the standard:
| Elem is the wide-character type, such as wchar_t, char16_t, or char32_t.

...

| The facet shall convert between UTF-8 multibyte sequences and UCS2 or UCS4
| (depending on the size of Elem) within the program.

Does this mean codecvt_utf8<char16_t> is supposed to convert UTF-8 to UCS2 
because char16_t is two bytes or should it convert UTF-8 to the UTF-16 
representation of UCS4?  Looking at the code I would have expected the latter 
but the standard seems to be unclear.

Regards,
RÃdiger

btw. your FOSDEM talk about GDB was very interesting!