New Boolean Solver: Doesn't cut an object (Suzanne) correctly when the mesh is open. #84493

New Issue

Adrian Thomet · 2021-01-07T15:13:01+01:00

Adrian Thomet commented

2021-01-07 15:13:01 +01:00

System Information
Operating system: Linux-5.8.0-7630-generic-x86_64-with-debian-bullseye-sid 64 Bits
Graphics card: GeForce GTX 1660 SUPER/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 455.38

Blender Version
Broken:

version: 2.92.0 Alpha, branch: master, commit date: 2021-01-06 23:54, hash: a9c607e60e
This does also happen with the stable Blender version 2.91.
Worked: ?

Short description of error
Suzanne doesn't get cut correctly and the mesh gets messed up. The bug happens, when the cutter cube passes through Suzanne's eyes.
It looks like the solver can't deal with an open mesh or multiple disconnected parts correctly.
When removing the separate eye part and filling the mesh so its closed, then the solver doesn't have a problem.
Using the old boolean solver, this bug doesn't seem to occur.

I made a short video to demonstrate the bug (Not from the blend file, that I uploaded!):

New_Boolean_Solver_Bug0001-0400.webm

Exact steps for others to reproduce the error
From Scratch:

Create a new blend file.
Add Suzanne into the scene.
Add a cube.
Optional: Put the cube into "Wire" mode to see the bug better.
Move the cube into and around Suzanne's eyes.
With the .blend file:
Move the cube called "Cutter" into and around Suzanne's eyes.
New Boolean Solver Bug.blend

Thanks!

**System Information** Operating system: Linux-5.8.0-7630-generic-x86_64-with-debian-bullseye-sid 64 Bits Graphics card: GeForce GTX 1660 SUPER/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 455.38 **Blender Version** Broken: * version: 2.92.0 Alpha, branch: master, commit date: 2021-01-06 23:54, hash: `a9c607e60e` * This does also happen with the stable Blender version 2.91. Worked: ? **Short description of error** Suzanne doesn't get cut correctly and the mesh gets messed up. The bug happens, when the cutter cube passes through Suzanne's eyes. It looks like the solver can't deal with an open mesh or multiple disconnected parts correctly. When removing the separate eye part and filling the mesh so its closed, then the solver doesn't have a problem. Using the old boolean solver, this bug doesn't seem to occur. I made a short video to demonstrate the bug (Not from the blend file, that I uploaded!): [New_Boolean_Solver_Bug0001-0400.webm](https://archive.blender.org/developer/F9554256/New_Boolean_Solver_Bug0001-0400.webm) **Exact steps for others to reproduce the error** *From Scratch:* 1. Create a new blend file. 2. Add Suzanne into the scene. 3. Add a cube. 4. **Optional:** Put the cube into "Wire" mode to see the bug better. 5. Move the cube into and around Suzanne's eyes. *With the .blend file:* 1. Move the cube called "Cutter" into and around Suzanne's eyes. [New Boolean Solver Bug.blend](https://archive.blender.org/developer/F9554245/New_Boolean_Solver_Bug.blend) Thanks!

Adrian Thomet commented

2021-01-07 15:13:01 +01:00

Added subscriber: @EnergieReaktor

Robert Guetzkow commented

2021-01-07 16:52:03 +01:00

Added subscriber: @rjg

Robert Guetzkow commented

2021-01-07 16:52:03 +01:00

If you enable Self this should work as intended.

If you enable *Self* this should work as intended.

Hans Goudey commented

2021-01-07 18:57:42 +01:00

Added subscribers: @howardt, @HooglyBoogly

Hans Goudey commented

2021-01-07 18:57:42 +01:00

Not always:

I don't think the new boolean method is always expected to work with non-manifold inputs. So I'd lean to calling this a known issue instead of a bug. But maybe there is a possible fix when one of the two inputs is manifold?

We should leave it to @howardt to make the final decision here. I couldn't find a task listing known issues or bugs with the Exact boolean method, that might be helpful.

Not always: ![image.png](https://archive.blender.org/developer/F9554540/image.png) I don't think the new boolean method is always expected to work with non-manifold inputs. So I'd lean to calling this a known issue instead of a bug. But maybe there is a possible fix when one of the two inputs is manifold? We should leave it to @howardt to make the final decision here. I couldn't find a task listing known issues or bugs with the Exact boolean method, that might be helpful.

Robert Guetzkow commented

2021-01-07 19:25:55 +01:00

I agree that @howardt should decide on this. However, your screenshot seems to show that it is working, the "cutter" is subtracted from "Suzanne". Did I misunderstand the issue?

Philipp Oeser commented

2021-01-08 14:25:04 +01:00

Added subscriber: @lichtwerk

Philipp Oeser commented

2021-01-08 14:25:04 +01:00

Changed status from 'Needs Triage' to: 'Confirmed'

Philipp Oeser commented

2021-01-08 14:25:04 +01:00

In #84493#1089206, @rjg wrote:
However, your screenshot seems to show that it is working, the "cutter" is subtracted from "Suzanne". Did I misunderstand the issue?

Think the cutter is in wireframe display mode, so if its polys show, boolean acts like Union (where it should be Difference), so there is indeed something wrong.
Same is true the other way around, if the modifier is set to Union, it sometimes acts like Difference instead...

> In #84493#1089206, @rjg wrote: > However, your screenshot seems to show that it is working, the "cutter" is subtracted from "Suzanne". Did I misunderstand the issue? Think the cutter is in wireframe display mode, so if its polys show, boolean acts like `Union` (where it should be `Difference`), so there is indeed something wrong. Same is true the other way around, if the modifier is set to `Union`, it sometimes acts like `Difference` instead...

Robert Guetzkow commented

2021-01-08 14:42:11 +01:00

@lichtwerk You're right, I've been able to reproduce this now as well. This happens when the cutter starts to intersect the inner ear.

Howard Trickey commented

2021-01-08 16:48:08 +01:00

This is indeed a case where Exact Boolean cannot use its usual code and has to fall back to "maybe it will work" code, because Suzanne is not manifold, even with eyes removed.

I am currently working on trying raycasting as an alternative or replacement to the generalized-winding-number fallback that I use now. That is more in line with what Fast Boolean does (which, to be clear, also doesn't work in all cases). That may make the user happier here. But right now, this is in the "known issue" camp.

This is indeed a case where Exact Boolean cannot use its usual code and has to fall back to "maybe it will work" code, because Suzanne is not manifold, even with eyes removed. I am currently working on trying raycasting as an alternative or replacement to the generalized-winding-number fallback that I use now. That is more in line with what Fast Boolean does (which, to be clear, also doesn't work in all cases). That may make the user happier here. But right now, this is in the "known issue" camp.

blender-admin commented

2021-02-07 17:25:07 +01:00

This issue was referenced by 6f63417b50

This issue was referenced by 6f63417b500d0893e89fef1ecddb9ff345322e96

Howard Trickey commented

2021-02-07 17:33:14 +01:00

Changed status from 'Confirmed' to: 'Resolved'

Howard Trickey closed this issue

2021-02-07 17:33:14 +01:00

Howard Trickey self-assigned this 2021-02-07 17:33:14 +01:00

Howard Trickey commented

2021-02-07 17:35:54 +01:00

That latest commit make new boolean work in this case (though it is still the case that there are no guarantees when all operands are not volume-enclosing.

blender-admin commented

2021-11-22 17:47:46 +01:00

This issue was referenced by 55c82d8380

This issue was referenced by 55c82d8380ea3fd37a9d966fad10f42cc5b365d5

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

New Boolean Solver: Doesn't cut an object (Suzanne) correctly when the mesh is open. #84493