First of all, thank you for looking into this. I did a bugreport on MT: https://mametesters.org/view.php?id=7342 So somewhere in the changes from 0.179 to 0.180 lies the solution to the described problem. Just start „Asteroids“ in GDI mode and observe for example the rocks and the existing points in them. And then start it again with HLSL. In MAME 0.180 with HLSL, the points are gone.
> I'm game to convert them over to BGFX, but you're going to have to be more specific. > According to the project's Git history, I effectively didn't make any changes to MAME > during that cycle, but your friends Jezze and Calamity made a whole lot of them.
I think that the Calamity changes are fine, because it looks like a specific HLSL problem and Calamity did not change a thing there AFAIK. Yes, Jezze did a lot changes since 0.162, but he also left MAME after 0.182 AFAIK. Just to be clear here, this is a bug and nothing that Jezze wanted as a feature. It was on a list of „to do‘s“, but he left MAME before fixing this.
> For as much as it seems like you feel like their changes tend to be better than mine, > apparently there's something that they did wrong. So what is it? Can you narrow it > down to an exact commit from around 3,25 years ago where they broke something that > you really liked? > > I'm not asking just to be a dick, I really want to know where this apparent > regression in quality happened, so I have a point that I can work off of when porting > things over to BGFX, knowing safely that the code I'm porting over is the "right" > code.
I always liked what you did with HLSL, I mean more or less you invented it. I did not like the changes of user anikom15 or the status of BGFX some versions ago.