I've used this trick in the past:
g++ -c -DWCHAR_BIT_SIZE=32 foo.cpp
On some platforms it is 31 bit (sufficient for ISO 10646); but I use 32 even in that case, thus I'm including the fallow bit. WCHAR_MAX will be 0x7FFFFFFFu on those platforms.
g++ -c -DWCHAR_BIT_SIZE=16 -fshort-wchar foo.cpp
I then use the WCHAR_BIT_SIZE to alter code behavior. Absence of WCHAR_BIT_SIZE causes a #error.
HTH, --Eljay
