[PATCH] fix wide character literal use in some test cases

[Paolo Carlini]

Jan Beulich wrote:

>This was my first take, too. However, I wondered whether filling half
>of the space (for 32-bit wide chars) with zeroes wouldn't change
>behavior; after all the purpose of these memset()-s appears to be to put
>garbage in the whole memory range.
>  
>
You know, what is garbage and what is not is matter of taste ;)
Seriously, in such testcases any deterministic value != the values in
i_lit would do. In fact, any 16-bit  proper value will do, because i_lit
is made of wide-character literals (2.13.2/2) and in the C++ library we
don't really support targets encoding plain chars with more than 8 bits.
Well, otherwise, unlucky-unlucky worst-worst case, the test becomes
weaker, doesn't fail.

Paolo.