That's why I moved the temp var and comparison into a separate function

Having a separate function is appealing, but not so straightforward to implement. It would be fine if we had an xs file to fiddle with, but the fact that the source file to be amended is a pd file (not an xs file) adds some complexity to the problem. I think it is possible to use the "separate function" approach - though it's actually "separate functions" (plural), as the templating dictates that we'll need separate functions to handle each of the different data types (ie byte, short, long, long long, double - not just float). There's also the issue of the second arg that gets supplied to the "separate function" - it could be a UV or an IV, not necessarily an NV, so we need to accommodate that as well (probably not difficult).

It's a much simpler solution if one instead just adds a dummy variable and turns off optimization - which seems to work quite well.

Cheers,
Rob