Another Glue network question

[JaydenDP]

Hi!

I have a glue network, and I'm editing the strength with some transfers, I have a pole hitting by geometry at an area with a strength of -1, but around that area is some weak Glue.

At the moment if I collide something with a -1 network, it just passes right through, shouldnt it astlease push the big chunk of -1 glued geo's? so that it collides and breaks the weak glue areas?

Any advice much appreciated, seems wrong to me :S Quicktime attached. And in the image, black network is -1.

Just to add, if i take away that strong area where the collision happens, the reaction is as expected.

Thanks

Jay

Strong_GLue_areas.mov

Edited November 26, 2013 by JaydenDP

[JaydenDP]

Here's the same but without that small area of infinitely strong glue:

GLue_Test_02.mov

[cwhite]

EDIT - double post.

glue_test.hip

Edited November 26, 2013 by cwhite

[cwhite]

I put together a similar scene with a mix of breakable and non-breakable glue and it worked as expected, so I'd suggest submitting a bug report with your file.

[JaydenDP]

I put together a similar scene with a mix of breakable and non-breakable glue and it worked as expected, so I'd suggest submitting a bug report with your file.

Thanks for taking the time cwhite!

I've edited your file to recreate my problem, I glued the bottom few pieces to the groundplane, leaving a chunch of weak glue above that, with the top of the pillar set to -1. If you look at this edited file it defo seems wrong :S ? Maybe I could submit this file?

Thanks

Jay

glue_test_broke.hip

[cwhite]

Yeah, I'd submit a bug for that. If you're gluing to a static object, the entire glued chunk should be treated as static, so there's likely an issue there.

[JaydenDP]

Thanks. Done:

Side Effects Support Ticket: #16437

[JaydenDP]

For those who are interested, I've had a response from Support:

"...When pieces are glued to the ground, the entire glued chunk has to be treated as a static object during the simulation. The Bullet engine doesn't perform collision detection between static objects while substepping, so the glue doesn't break and the two static objects pass through each other.

However, the Bullet solver does do a collision detection check at the start of the frame to help break the glue in situations like these. If you turn on wireframe mode, you'll see that a few glue bonds do break, but not enough to cause any pieces to fall off.

You can try adjusting the 'propagate_iteration' detail attribute on the constraint network geometry, and/or the Propagation Rate parameter to spread the impulses to more nearby glue bonds that can actually be broken. Alternatively, you can set up the glue network a bit differently to place some breakable bonds near the impact site..."

"...it's important to understand that glue is different from the other types of constraints that the Bullet solver supports. When you glue objects together, the solver actually creates a single object in the Bullet simulation, and that object has a compound collision shape containing all of the glued pieces' collision shapes. All of the pieces move as one object, until an impact occurs and breaks off some pieces. Therefore, if one of the glued pieces is static and cannot move, the glue object needs to be static or else any static child pieces could end up moving.

It is a limitation of the Bullet engine that it won't check for collisions between static objects, but that is also an important performance optimization in some cases. It may be possible to patch the Bullet library to avoid this behaviour for static glued objects.

It sounds like you might instead want to consider creating a pin constraint between your pieces and the ground, since that would let you constrain the pieces to the ground while keeping them as separate dynamic objects..."

Hope this help someone else who comes up against the same thing,

Thanks

Jay