More help needed - quad pan bug

[abhunkin]

I've simplified my "quad pan bug" example (attached). It now is very close
to the Ekman help pan example, and can be realized realtime or not on any
Csound capable of quad. (On Windows, this means CsoundAV and Csound5beta -
both of which I've tested.)

The problem is with the final optional value in pan, ioffset - which when
set to 1 is supposed to center the sound in the middle of quad space. It
doesn't.

The following line doesn't work - the final 1 is the offending ioffset:
a1,a2,a3,a4 pan asig, k2/2, k3/2, 2, 1, 1

OTOH, the line below line does what the above doesn't - by adding .5 to both
the x and y axes (and leaving, by default, ioffset at 0):
a1,a2,a3,a4 pan asig, (k2/2)+.5, (k3/2)+.5, 2, 1
This is the expected "quad-centered" behavior.

So, ioffset is not working correctly.

Running this .csd *should* result in equal max amps (about 9500) for all
four channels. The first pan line above does this; the second is *way* off.
(You don't need a quad system to hear this; just compute the file and look
at the max amps.)

One further weirdness: if like the original Ekman help example, you
substitute "a1" for the three "asig"'s (as it looks like you should, if
somewhat dangerously, be able to do), max amps for the quad channels come
out yet differently! This is a complete mystery to me.

Incidentally, I suspect that the same results would be had on Linux and Mac
systems capable of quad. I'd certainly be interested in knowing whether they
behave better or not.

Thanks for any light shed.

Art Hunkins