Messages

Wait, at 485 I must have forgotten to wrap lookahead_tiles_supersight[si].width into an abs. This means that, during redraws, the objects on the left of the camera (with width < 0)) will preserve more detail than the ones on the right. This is not super important since, especially with 80 tiles, you are extremely unlikely to overrun the 13K buffer and hence to need redraws. But it's still something that should be fixed.

the game chooses the low-polygon version under the condition (var_FC != 0). I could have patched away the corresponding if instruction, but I found it easier to fix the value of the variable to 0.

Can I put the value of the variable back where it was? If so, where and how?

I replaced the mechanism with something slightly more complex (see start of Part III). The relevant lines in frame.c are 484–490, 744 and 801–805.

Wow! I did not expect to hear about new Stunts discoveries on the orange site. Incredible work, and I can't overstate how delightful I found your in-depth and well-illustrated writeup. Very encouraging to see the restunts project resurrected, I'll have to relearn how to rebuild that relic.

Thanks @dstien ! But I'm really a dwarf on the shoulders of giants here: most of the work was done by you, @llm , @clvn and others. In reverse engineering I'm still an amateur. On the other hand I consider my prose pretty good and I'm learning to market my work well

Do we need to re-upload our replays or will the Stunts Oracle take care of that?

I love the first screenshot of the glitch gallery. It looks like a more delirious version of the ghost car artifacts.

Yes, same concept! It's the same way DOS encodes the file storage occupation in the FAT. I remember reading about that in a manual when I was a kid and being amazed by it. The first non-trivial data structure I ever learnt!

Can you make that number which sets the initial number of tiles to be introduced as a parameter when starting the game, like STUNTS.EXE 110 ... STUNTS.EXE 30 or read from a file?

It is technically possible, maybe in the future.

I am confident we won't need to wait other six months for the third and final chapter to be completed

Luckily I had enough writing momentum to complete the saga in a matter of days Here the grand finale:

https://marnetto.net/2025/08/13/broderbund-stunts-3

FWIW Google considers the page unsafe (https://transparencyreport.google.com/safe-browsing/search?url=https:%2F%2Fhyperionracingteam.rf.gd%2Ffsc25%2Findex.php&hl=en). I can ignore the fact, but Firefox presents me a giant crimson screen whenever I load the site.

Not sure what the problem is. You can try setting up the Google Search Console for your site (https://search.google.com/search-console/about), it might give you some hint about why your site was classified like that and should allow you to ask to be removed from the blacklist.

It it can be of consolation, I had the same experience with my blog.

In the wiki there's still the big version (RESTUNTS-10a1e10.EXE),  maybe somebody changes it

Now fixed, thanks.

(and puts all 3 versions)

I did not want to do that because that would basically commit me to update them every time I improve the original version, which is precisely what I wanted to avoid. But all the work you're putting in that is making me change my mind. I'll finish the final chapter of my writeup and then decide.

But really, it would be better if you set up the build system and learned to compile the thing yourself, with Wine. It's hard to understand its internals but not difficult to just run. I personally wrote the Linux part of the readme. Try it!

I just managed to install DosBOX-X and I can play with Stunts at 16000 cycles now (still SS40.EXE). Below are comparison pictures of all 4 versions (track ZCT007).

Very interesting comparison, it really gives an idea of what the different versions have to offer. But I must say the biggest insight for me is that the cloud pattern changes every time one reloads the replay (what I just verified in the original game). Cheap guys, it would have sufficed to save the seed of the random generator in the RPL file. It would have also helped with the problem of the opponent taking a different path at the bifurcations.

Ah, wonderful! I have tried for a while to get the perfect view where the loops align to form the Audi logo, but you framed it much better. Great work!

Other observations. The initial version, STUNTS*.exe, with an initial number of 110 tiles, has a size of 856493 B. The new versions: SS60.EXE - 255367 B, SS40.EXE - 255327 B, so 3.35 times smaller.

That's because you got the new 1.6 version, which is available but I have not yet announced since the only practical benefit is the reduced size (thanks to @llm's work). The official 1.6 version with 110 tiles, available in the wiki and on my site, has size 255467: the only difference are the data for the 50 extra tiles (4 bytes each).

With STUNTS*.exe the clouds flicker in track ZCT004, but with the others they don't. Also, SS60.EXE is much faster than STUNTS*.EXE but the rendering looks almost as good.

The cloud flickering is honestly bizarre, since my mod does nothing with that part of the code. Maybe one day I'll find the time to analyze the related routines. Anyway, if the lite version works better, it's good news. Maybe the engine decides whether to draw these ornaments depending on the time needed by the rendering. Just a vague hypothesis.

Here they are, the maximum amount of tiles capped to 40 and 60 respectively.

Thank you, with SS40.EXE I can now play the game with DosBOX at 11000 cycles, even with track ZCT004 and it's still a great improvement over the original game. Most importantly, the original 3D rendering had some bugs that you fixed.

Happy that you liked it! Looking forward to the Stunts page on your site!

Ok, fine. As long as I am not requested to keep them updated, I don't mind making a couple of detached versions. Here they are, the maximum amount of tiles capped to 40 and 60 respectively.

A teaser Soon™

I know that soundtrack! The film frame you posted in the shoutbox was not by chance!

I mean, SuperSight works well with DOSBox on new computers, and on native DOS on machines as slow as a 486. That seems to me a reasonable level of compatibility. Maintaining multiple variants is a chore and I will only do that if it significantly increases the number of machines that can run the game.