Captain's Log: Stardate 77461.8

I've been working for a couple days on The Bug now. I haven't heard the weird distortion for a while, but this does not make me confident that it's gone.

Since I don't have a repro for The Bug, I decided that it was a good time to really make the fuzz testing I have super good, which is a nice thing to have anyway. I made a bunch of changes to make the fuzz tests generate more plausible random configurations, especially having it cover the input ranges of all the different parameters more thoroughly. This alone caught a bunch of small issues.

Then, I started adding all kinds of checks for simulation values that have gone inf / NaN, and found a number of them. This is something I knew was a problem, but had been putting off the work to fix it since it wasn't all that pressing. Now, I figure that I may as well start ruling out causes like this for The Bug, because I need to fix this anyway and maybe it will fix The Bug by accident.

I've hardened the GPU code quite a bit, looking for all the various code paths that could feasibly generate inf / NaN, and coming up with reasonable solutions for all the edge cases. There are a couple of remaining NaN cases in some of the newer code I've written, which is mildly promising in terms of the NaNs being somehow responsible for The Bug.

Once I've fully hardened all the GPU code against inf / NaN I will do an extended testing session where I try to get The Bug to show up. I'm not quite sure whether to hope that it does (and I can repro it), or doesn't (and thus I won't be certain it's gone)...