Has the behaviour of "back faces" in trianlge-based physics changed?

[Da5id Weatherwax]

It's been a LONG while since I made use of this "feature" but in the past, with a triangle-based physics mesh, you collided with a plane when approaching it from the side its normal faced, but could walk straight through it from its back face. I'm testing a build at the moment and it seems that you can no longer walk through the back faces of planar physics. Is this a known change? intended behaviour? bug?

[arton Rotaru]

It's true that you can slip through the backfacing side at certain spots, but it wasn't like that the physics shape was non-existent from that side per se. At least that's how I remember it.

There hasn't been done any changes to the physics engine since 2012.

[Chic Aeon]

The main reason that I have almost always used cube physics (except when learning the process of planes physics) is that the triangle physics was a bit "dodgy" for my liking. It worked in a sort of haphazard method and the walls were sometimes not as expected.   I agree with Arton though I don't remember them ever being "one way"  just "spongie".    If you aren't using the Linden viewer to upload you might test using the default viewer.  I suspect the results will be the same though. 

[Da5id Weatherwax]

Thanks, @arton Rotaru and @Chic Aeon - a previous proof of concept that never made it into a project, shortly after mesh hit the grid, had a skysphere that you could fly into from any angle but would block exit, so I know it did work sorta like I remember. Now I come to actually try and use it I find it not working. but I guess that the initial behaviour was a "misfeature" that has subsequently been coded out without anyone noticing. I'll just take the results of my testing and modify the project design accordingly. If anyone else had been using this feature at all extensively and found their builds breaking, there would have been a lot more replies to my question than just you two

BTW: I never upload meshes with anything but the latest version of the official viewer. I know Firestorm can also upload meshes but to know I always use official. Same as I will test RLV-Compliance of a script using those features against Marine's viewer rather than an implementation of RLVa. Excellent as the alternative code may be in either case, you test against the same reference as the doco was written to. Always. If something behaves differently in an alternate then it's not my code that aint compliant, yanno?

[Coffee Pancake]

23 minutes ago, Da5id Weatherwax said:

BTW: I never upload meshes with anything but the latest version of the official viewer. I know Firestorm can also upload meshes but to know I always use official. Same as I will test RLV-Compliance of a script using those features against Marine's viewer rather than an implementation of RLVa. Excellent as the alternative code may be in either case, you test against the same reference as the doco was written to. Always. If something behaves differently in an alternate then it's not my code that aint compliant, yanno?

All viewers that can upload mesh have jumped though some pretty serious hoops to get the required libraries from LL. We are prohibited from messing with it in any way at all. Your mesh will be identical regardless of viewer, although the upload dialog UI may be different.

As for RLV vs RLVa .. we do not make RLVa to marines reference specification (which comprises the wiki and execution in her viewer) and you will find considerable and intentional differences in how certain commands affect the user experience. RLVa has been bums in seats for over a decade at this point.

Test your code against the your end users environment. Anything less is akin to making a webpage for internet explorer and refusing to appreciate that chrome is the defacto standard, whatever the original W3C specifications might declare.

 

This post is best read with Netscape navigator 3.0

[Da5id Weatherwax]

5 minutes ago, Coffee Pancake said:

As for RLV vs RLVa .. we do not make RLVa to marines reference specification (which comprises the wiki and execution in her viewer) and you will find considerable and intentional differences in how certain commands affect the user experience. RLVa has been bums in seats for over a decade at this point.

Test your code against the your end users environment. Anything less is akin to making a webpage for internet explorer and refusing to appreciate that chrome is the defacto standard, whatever the original W3C specifications might declare.

That's fair enough. I always do a SECOND test against latest firestorm, which I consider the de facto RLVa reference and I'll tweak it if I see any behaviour I consider unacceptable.

However, to your webpage analogy, I think its not unreasonable for a developer to expect that something fully W3C compliant should work in any browser, and if it doesn't, that's a browser bug.

[Love Zhaoying]

Planes colliding into things..sweet Jehoshaphat!!

Are the "backside of triangles" like that old Disney joke about "the backside of water" that made it into the Jungle Cruise movie? 

[Quistess Alpha]

I tried it on a whim and you ~Can walk through the back face of a mesh if it's set up propperly: (only tested on the beta grid, Kokua viewer)

1. In blender create a single one-sided plane. (add-> mesh->plane)
2. Upload dialog, set the high LOD to the model, and use LOD above for all (it's 2 tris, you can't simplify that more in a sane way)
3. In the physics tab, select the model as the physics shape (this is necessary. not choosing a physics model won't work)
4. upload, rez the object. edit it, and in the features tab choose "prim" physics instead of the default convex Hull.
5. Now you can walk through the plane in one direction only. Neat!

Edit: Er actually there was some confirmation bias going on, It might not work 100% of the time, or at all. * more like you have 1 5~10 percent chance of going through the backface if you really don't believe in it.

Edited July 7, 2022 by Quistess Alpha