I don't think filters are expected to function stably beyond the Nyquist anywhere. This could be grounds for a warning message or we could document it and leave it to the user to experiment with the noises that he can get that way. I like the last option.

Nobody is suggesting that we proof ChucK against blowing up in general (that would make for a very boring language indeed) but I really don't see why things would have to be as they are on Mac's. With a hard clip (using something new like dac.clip() ) all of the current behaviour could be had by turning the system's volume up to max, then setting dac.gain(). In addition a solution like that could give safety to those who might need/want it.

Can't we simply file a bug report against OSX about this? There is a dual issue here of OSX passing on floats in situations where they clearly make no sense and there being a problem with the convenience filters sometimes coming up with unstable sets of coefficients. These happen to interact in this way but they are quite separate, to me.

Yours,

Kas.