Development Journal of a Raytracer

I have also found the alignment problem with the raytracer I wrote about. One transformation was applied twice. I added that last October (I checked my source repository), I think as part of the Distributions. I have to test if they still work with the change I made today. The code I wrote back then was definitely wrong.

But I also found two more problems. First all objects created in the editor got the same names. If I create two planes I expect the names to be "plane 1" and "plane 2", at least that is what is supposed to happen. But they were both "plane 1", so I have to check why it doesn't work any longer, it worked not so long ago.

The second one is, that there is a problem with selections. I guess one of the changes I made recently had side effects. I even think I know what change it was. The old code was wrong, I removed some code I couldn't see any reason for being there. Now I guess I know what it was used for.

While working on the cat on the Weekend I had noticed a considerable slowdown. As I had never added the code to do incremental updates in the Free Edit Modifier, all the changes done with the modifier were applied in every single step. So I started to add support for doing the updates incrementally, the most important being moving vertices.

Then I did something completely different and forgot about it. Today I actually needed to move some vertices and noticed, that the movements were strange. It turned out, that the incremental applies interacted with another optimization. If a set of vertices is moved multiple times, only the compound transformation is recorded (and stored to disc). But if that compound transform is applied multiple times it obviously leads to wrong results. This bug is fixed now.