Char-type is incorrectly sized

[vtjnash]

on windows, wint_t and wchar_t are declared unsigned short (Uint16)
http:https://msdn.microsoft.com/en-us/library/windows/desktop/aa367308(v=vs.85).aspx
and from _mingw.h:

typedef unsigned short wint_t;
typedef unsigned short wctype_t;

(somewhat amusingly, someone at microsoft apparently believes unsigned numbers can hold the value -1, as demonstrated by their documentation for this type. while not strictly wrong -- it can indeed hold the value 0xFFFF -- ok, yes it is wrong)

[mlubin]

Not key to the discussion, but I think -1 is a valid and idiomatic way in C/C++ to refer to the largest value an unsigned type can hold.

[JeffBezanson]

This is relevant if you want to add a Cwctype_t type, but otherwise not. One could point out that Char is not the same size as C's char as well. Julia makes its own decision about what Char means, which is that it represents a unicode code point. Which cannot fit in 16 bits, so windows loses there I'm afraid. Char should not be used as a ccall argument type (should be fixed in #2901).

[vtjnash]

Doc's need to be updated then: http:https://docs.julialang.org/en/latest/manual/calling-c-and-fortran-code.html#type-correspondences

[JeffBezanson]

Also, while there's nothing we can do about it, using unsigned short for wint_t is certainly wrong, since the whole point of wint_t is to be able to represent a -1 that is not a valid code points. All unsigned short values are valid code points.