HotwheelsSisyphus

I'm looking into the testing... taking this as an opportunity to understand floats a bit better. But generally I figured one would want a behavior something like this: 1. if...

> something called the floating point epsilon My (limited) understanding is that epsilon is either used simply as a synonym for tolerance, or it refers to the machine epsilon /...

(pushed this mainly to test the changes I made, and lo, it doesn't do the thing it's meant to do.)

@dyfer thanks... yes, the issue is more that I don't have a decent build environment set up on the machine I'm currently on. So I can only test the primitives...

The tests pass now! I figure that the remaining CI checks probably don't have anything to do with my changes? But who knows. Will try to build it later on...

> I had assumed it would be the user’s responsibility Basically everything remains the user's responsibility, because afaik the license disavows "fitness for any particular purpose"... :P Nonetheless, it seems...

Yes, exactly. It can strictly speaking only be decided with respect to some use case. However, two points: 1. I’d argue that tolerance = 0 (which is the same as...

Short update: I've been thinking about this more than is strictly healthy, and the thing is, there isn't really an abstract solution for this. Obviously functions like `numpy.isclose` exist but...

Thanks for the review! I left the converstation re recBufSize open, just in case someone else wants to chime in.

> Why can't we just make addToTail on synths the same as after? Would this break anything? I hadn't thought of that! Easy to do I think, but a wider-ranging...