Inaccurate detection

[Chungmire]

Hello, I am quite intrigued by the concept of this project, but there appear to be some issues through my testing.

I have used the methods outlined to obtain the vphys data and turn it into a .tri file. I have minimally adapted the example to just cast between hard-coded position values which I obtain from cl_showpos 1. I used de_mirage for testing. In my examples, I'm using the same X and Y coordinates for the start and end position, but changing the Z coordinate. I am using this to detect if there is any ground surface I may collide with. An example is provided:

r_start = Vector(-148.04f, 368.25f, 30.00f);
r_end = Vector(-148.04f, 368.25f, -42.75f);

In the above example, z position 25.04 is where the ground actually is. It doesn't detect intersection until -42.75, however, which is 67.79 units away from the true surface. This discrepancy isn't consistent either. In 2 locations I tested, it detected at 63.87 units. Other examples detected at 62.99, 63.88, and 64.39 units away. I don't know much about mapping, but I have seen this method used in CSGO cheats for detecting surfaces, albeit with the game's traceray function. The suggestion to just find the largest discrepancy and always use it isn't sufficient as I need precision within the unit. I was wondering if it's not possible to do what I'm attempting to do with triangles alone.

[AtomicBool]

That command gives you position camera/chest, I can't remember but it can't be position for feet

[Chungmire]

Perhaps it is some pre-defined location in the body, but the camera angle doesn't change the position data. The data in cl_showpos1 is also the same as what you get from the memory location used in the example. What makes me think it actually is the feet however is that when you use "getpos" in the console it adds 60 to the Z component of what you can see in cl_showpos. When you use the command, you fall 60 units rather than being placed exactly on the surface. If you subtract 60 and use the command again, you are put exactly on the surface, AKA exactly where cl_showpos already said you were.

Regardless of where on the body it is, it is absolutely a constant value. My problem is that ground intersection is being detected at different amounts of units down from the grounded player's Z position (anywhere from 62-68 units down) which is contrary to how it should work. For example, if the player's Z position component is always 60 units above the ground, then when I'm standing on the ground it should always detect intersection 60 units below.

[AtomicBool]

triangles alone

you mean doing it with one custom triangle? It's moller trumbore algorithm, if you're interested in it you can ask gpt to code it or copy from source of this project.

[AtomicBool]

triangles alone

you mean doing it with one custom triangle? It's moller trumbore algorithm, if you're interested in it you can ask gpt to code it or copy from source of this project.

[JanEricNitschke]

@Chungmire you could try having a look at the triangles in source2viewer (https://valveresourceformat.github.io/) or use your own logic to visualize/inspect the produced triangle file