Maniphest T101646

Geometry nodes "set position" node offset doesn't work if vertices are exactly at a metric grid coordinate
Closed, Archived
Actions

Assigned To
None
Authored By
Marius Eugen Mihai (mem)
Oct 6 2022, 3:49 PM
Tags
  • BF Blender
  • Nodes & Physics
Subscribers
Iliya Katueshenock (Moder)
Leon Schittek (lone_noel)
Marius Eugen Mihai (mem)

Description

System Information
Operating system: Windows-10-10.0.19044-SP0 64 Bits
Graphics card: NVIDIA GeForce RTX 3080 Laptop GPU/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 516.94

Blender Version
Broken: version: 3.3.1 Release Candidate, branch: master, commit date: 2022-10-03 11:17, hash: rBfb499916bb8f
Worked: Couldn't find a version where this worked (haven't tried versions before the change to geo nodes fields).

Short description of error
I've made a simple setup to randomize the vertex positions on a model. I'll attach a file with the node setup. Whichever vertices are at precise grid coordinates (up to cm level) don't get offset.

Exact steps for others to reproduce the error
I'll just attach an example file. The setup is nothing complicated, I'm just using three cubes to exemplify the issue. The center one is a default cube, the one on positive X has all the vertices slightly moved in edit mode so that all of them are not at precise grid locations, and the one on negative X has only one vertex slightly moved.

geonodes position bug.blend454 KBDownload

Related Objects

Event Timeline

Marius Eugen Mihai (mem) created this task.Oct 6 2022, 3:49 PM
Iliya Katueshenock (Moder) added a project: Nodes & Physics.Oct 6 2022, 5:54 PM
Iliya Katueshenock (Moder) added a subscriber: Iliya Katueshenock (Moder).

It's a noise problem. This is true for all parts of the blender. On integer values (all settings are multiples, not just the position), all amplitudes add up to 0.
You can fix this simply by changing something, such as the size, or introducing an altered position into the noise.

Leon Schittek (lone_noel) closed this task as Archived.Oct 6 2022, 6:08 PM
Leon Schittek (lone_noel) added a subscriber: Leon Schittek (lone_noel).

Thanks for the report, @Marius Eugen Mihai (mem) !

@Iliya Katueshenock (Moder) is right that this is a case of the noise just working out to be 0, exactly where the vertices are.
We had a similar reports for the cloud noise in the displacement modifier (T48814), which we archived. So I will close this report, as well.