Regression: behavior of Separate/Delete Geometry nodes changed in 3.3 #99667

New Issue

Jacques Lucke · 2022-07-13T13:38:06+02:00

Jacques Lucke commented

2022-07-13 13:38:06 +02:00

In the setup below, the output is an empty geometry in 3.3, while the mesh line is still there in 3.2. The old behavior is correct, since the Delete Geometry node should only delete faces here. When there are no faces, it shouldn't do anything.

This is caused by c4e11122c5. It's most likely an issue with the order of domain interpolation and selection inversion. Either that should be fixed by, or the commit should be reverted.

This was found in #99664.

delete_geometry_deletes_too_much.blend

In the setup below, the output is an empty geometry in 3.3, while the mesh line is still there in 3.2. The old behavior is correct, since the Delete Geometry node should only delete faces here. When there are no faces, it shouldn't do anything. This is caused by c4e11122c5. It's most likely an issue with the order of domain interpolation and selection inversion. Either that should be fixed by, or the commit should be reverted. This was found in #99664. ![image.png](https://archive.blender.org/developer/F13290218/image.png) [delete_geometry_deletes_too_much.blend](https://archive.blender.org/developer/F13290217/delete_geometry_deletes_too_much.blend)

Jacques Lucke commented

2022-07-13 13:38:06 +02:00

Added subscribers: @JacquesLucke, @HooglyBoogly, @wannes.malfait

Jacques Lucke changed title from ~~Behavior of Separate/Delete Geometry nodes changed in 3.3~~ to Regression: behavior of Separate/Delete Geometry nodes changed in 3.3

2022-07-13 13:39:04 +02:00

Wannes Malfait commented

2022-07-13 13:49:33 +02:00

Although #99664 (Delete Geometry node face mode behaves unexpectedly when used after Join Geometry) shows a bigger problem, I think this specific regression is caused by this check here:

  /* Check if there is anything to delete. */
  if (selection.is_single() && selection.get_internal_single()) {
    return;
  }

When the selection is empty, this doesn't return, while in the old version it did. So adding a check for empty selections would probably fix this:
(selection.is_empty() || (selection.is_single() && selection.get_internal_single()))

If #99664 were fixed, then this check wouldn't affect the outcome, just how long the node takes.

Although #99664 (Delete Geometry node face mode behaves unexpectedly when used after Join Geometry) shows a bigger problem, I think this specific regression is caused by this check here: ``` /* Check if there is anything to delete. */ if (selection.is_single() && selection.get_internal_single()) { return; } ``` When the selection is empty, this doesn't return, while in the old version it did. So adding a check for empty selections would probably fix this: `(selection.is_empty() || (selection.is_single() && selection.get_internal_single())) ` If #99664 were fixed, then this check wouldn't affect the outcome, just how long the node takes.

Jacques Lucke commented

2022-07-13 14:40:06 +02:00

Sounds reasonable, want to create a patch?