I know I can do the conversion manually like this:

// get function argument
char* NodeName = (char *)SvPV(ST(0),PL_na);

// get argument length
int nNodeNameLen = strlen(NodeName) + 1;

// alloc memory
LPWSTR wNodeName = (LPWSTR)malloc(nNodeNameLen * sizeof(WCHAR));

// do conversion
mbstowcs(wNodeName, NodeName, nNodeNameLen);

// Here's where I'd actually use the data
// I'm leaving out the part where I allocate widebuffer
someLibFunction(wNodeName, widebuffer, buffersize);

// de-allocate
free(wNodeName);

/*
    And now I'd have to do it all in reverse: 
    allocate a new char* buffer, convert the output
    back to char*, deallocate the wide buffer, put 
    the result back on the stack, and free the wide buffer.
*/
[download]

Yes, that's one way to do it, and it works fine. But it sure seems like a lot of work for every string, in every function, in every module. Even if I write a function to help with some of it, I would still have to free() the strings manually after I was done using them. That doesn't sound very perl-like. There has to be an easier way to do it... Does anybody know? I prefer to have something that handles memory allocation/deallocation automatically.

I've already looked at SvPVutf8() and sv_utf8_upgrade() but neither of these appear to do what I need. I've also tried passing in the string as unicode from the perl side, using pack/unpack or utf8 or Encode but that doesn't seem to work either. Is perl's notion of unicode not the same as "array of wchar_t"? Or am I confusing two different things?