Testing WideCharToMultiByte

[Bill Medland]

Has anyone ever done a proper investigation into the edges of 
WideCharToMultiByte?  I was quite surprised to see that given 
how important it is there is no test for it (or am I blind?)

I am investigating a potential off-by-one in msi and want to know 
very exactly what is supposed to happen in the edge cases (e.g. 
specifying a length of 0 or a length that does not actually 
include a terminating null character or ...).

If no-one else is involved then I am prepared to do a little 
investigation and write the tests but I guess we don't want to 
start adding tests like that at this late stage, or do we?