Exact Boolean Result is Wrong With Almost Coplanar Faces #82736

New Issue

Philip Holzmann · 2020-11-15T11:47:33+01:00

Philip Holzmann commented

2020-11-15 11:47:33 +01:00

System Information
Operating system: Windows 64 bit
Graphics card: Nvidia GTX 750 Ti

Blender Version
Broken: Blender 2.91 beta 9b54c81414, 2.92 alpha 194a57fd63

Short description of error
With the difference operation applied multiple times, if two meshes touch occupy almost the same space, the result may come out wrong.
This applies to both using the Self checkbox or not.

This bug has already been reported, and then closed here, with the note that the result is not a bug: https://developer.blender.org/T82047
I agree with the screenshot @howardt provided, but is different from what I observe.
Perhaps this is a platform dependent issue.

Exact steps for others to reproduce the error
Open the file. boolean_rotated_box2.blend

This is what I see:

This is what you should, if the boolean worked correctly.

**System Information** Operating system: Windows 64 bit Graphics card: Nvidia GTX 750 Ti **Blender Version** Broken: Blender 2.91 beta 9b54c8141480, 2.92 alpha 194a57fd631f **Short description of error** With the difference operation applied multiple times, if two meshes touch occupy almost the same space, the result may come out wrong. This applies to both using the *Self* checkbox or not. This bug has already been reported, and then closed here, with the note that the result is not a bug: https://developer.blender.org/T82047 I agree with the screenshot @howardt provided, but is different from what I observe. Perhaps this is a platform dependent issue. **Exact steps for others to reproduce the error** Open the file. [boolean_rotated_box2.blend](https://archive.blender.org/developer/F9235317/boolean_rotated_box2.blend) This is what I see: ![boolean_rotated_box2.png](https://archive.blender.org/developer/F9241248/boolean_rotated_box2.png) This is what you *should*, if the boolean worked correctly. ![rotated.png](https://archive.blender.org/developer/F9240889/rotated.png)

Philip Holzmann commented

2020-11-15 11:47:33 +01:00

Added subscribers: @howardt, @Foaly

Howard Trickey self-assigned this 2020-11-15 13:11:51 +01:00

Howard Trickey commented

2020-11-15 13:11:51 +01:00

I was looking at this yesterday. It is hard to debug cases like this and maybe I need a better debugging technique, because I have to painfully manually map out 3d topology of intermediate results where things are so close that you can't see what is going on by just looking. I've already spent 6 hours on this bug. I will continue to look at this today but may end up giving up.
One actual, tangential issue, that I can fix is that the rotation matrix has some numbers that are < 1e-6 in them, which should be 0 for exactly 180 degrees of rotation. By cleaning those small numbers out of the matrix, the answer looks like it should for the first three applications of the difference, but then goes somewhat wrong on the last one. I'm trying to track that down first, as maybe it will be simpler to find that bug and maybe it is the same one.

I was looking at this yesterday. It is hard to debug cases like this and maybe I need a better debugging technique, because I have to painfully manually map out 3d topology of intermediate results where things are so close that you can't see what is going on by just looking. I've already spent 6 hours on this bug. I will continue to look at this today but may end up giving up. One actual, tangential issue, that I can fix is that the rotation matrix has some numbers that are < 1e-6 in them, which should be 0 for exactly 180 degrees of rotation. By cleaning those small numbers out of the matrix, the answer looks like it should for the first three applications of the difference, but then goes somewhat wrong on the last one. I'm trying to track that down first, as maybe it will be simpler to find that bug and maybe it is the same one.

blender-admin commented

2020-11-16 02:24:59 +01:00

This issue was referenced by fbffff26e4

This issue was referenced by fbffff26e437ad87b644762b340dcddc7ddf16ab

Howard Trickey commented

2020-11-16 02:25:57 +01:00

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

Howard Trickey closed this issue

2020-11-16 02:25:57 +01:00

Howard Trickey commented

2020-11-16 02:29:29 +01:00

I found a real bug (an extra triangle was output that should not have been), and also made it so that a rotation of 180 does what the user expects for the boolean modifier exact version. So the result no longer looks like the picture above, but rather what I think it was originally hoped to look like.
I hope I didn't just move the problem around. If it does come back, I hope it will be with a simpler example.

I found a real bug (an extra triangle was output that should not have been), and also made it so that a rotation of 180 does what the user expects for the boolean modifier exact version. So the result no longer looks like the picture above, but rather what I think it was originally hoped to look like. I hope I didn't just move the problem around. If it does come back, I hope it will be with a simpler example.

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

Exact Boolean Result is Wrong With Almost Coplanar Faces #82736