There’s been a few times I’ve wanted to delve into a particular 3D graphics topic - one example being Constructive Solid Geometry.
I was shocked at the total lack of reproducibility, with claims on speed ups and so forth totally unverified. Even when code was available, it was often of very poor quality. It really astonished me quite how poor many of these papers were.
I'm certainly guilty of this. The code I wrote for my undergrad thesis over many sleepless nights was easily the worst code I've ever written and was too ashamed to release it. Since then, I've realized that there were fundamental bugs in the pseudocode I included in the paper that were not present in the original source. There are plenty of bugs in the source, but none too major. Effectively, the whole paper is garbage beyond the thinking behind it, but the source that (mostly) works is unreleased.
I would love to revisit it and correct things, but I've since moved on to the point where it would be difficult or impossible in many ways.
[1] https://www.glassner.com/portfolio/graphics-gems-series/