Well. In the end, I had to go back in time and take the faulty commit apart.
I still don't know why _exactly_ it broke everything, but at this point, I am ready to take the win and leave this ugly place without researching what these strange noise are that are coming out of the basement.

Yeah, there are some traps, but managing multiple shader versions is also painful. And these can still work differently on different platforms, too 🫠

So far it works! 🤞

Well, the geometry emerging from this memory bug is 3D 😅

Apart from that - it is mostly fine. So far I had no real deal breaking issues. It got much simpler after using only a single shader version variant:

#version 300 es
precision highp float;

... works both on desktop and in webGL.

You have a better explanation of a working theory than I do 😅

I released last Sunday the demo to my upcoming game #8-bitBot, a programming puzzlegame, on itch, running in browser: bsky.app/profile/zet2...

#cozygame #puzzlegame #programminggame

Well done, it looks really good 👍

Why do they look so tasty?!?

And how did you render those?

Brilliant idea 😂

Now I am tempted to make bot skins - but I have some more important tasks to do for the moment 🥲

I am more thinking about how to have a version that increments automatically. Like when triggering a Web build or a desktop build, that it increments a number in a file. Nothing big, just something that helps me identifying the last build/version quickly...

Yeah, it needs polishing. I got a bit distracted by preparing the itch.io upload and fixing the safari url open issues. There's no end to stuff that needs to get done 🙃

The technical challenge to integrate such a library is in my view more complicated than the problem itself. I spent over an hour trying to do it the "right" way by using a standard format.
Yes, it gives more peace of mind to use an established well working library, but it isn't cost free either...

So I wrote a "deflate" tool that recurses into the data file, concatenates all the data into a single file with a header per file (filename + size), compress it and apply a trivial encryption on top of it.
100 lines of code.
The download/unpack is another 50 lines of code...
Half an hour of time.