Mesh Deform Modifier Vertex Count Error #110021

New Issue

Jaggz H · 2023-07-12T19:21:20+02:00

Jaggz H commented

2023-07-12 19:21:20 +02:00

System Information
Operating system: Debian Stable
Graphics card: Nvidia Geforce 3090

Blender Version
Broken: 4.0.0 Alpha, branch: main (modified), commit date: 2023-06-29 23:21, hash: f47eed749e, type: Release
Broken: 3.4
Worked: 3.3 LTS

Caused by 57ea827bfb

Short description of error
Mesh Deform has multiple bugs. This post addresses one:

Vertex counts change immediately upon Binding

Exact steps for others to reproduce the error

Load .blend
Click Unbind then Bind again
See vertex count mismatch error

** Additional notes: **

Apply Mirror modifier
Unbind and Bind again
See no verts_num mismatch error

**System Information** Operating system: Debian Stable Graphics card: Nvidia Geforce 3090 **Blender Version** Broken: 4.0.0 Alpha, branch: main (modified), commit date: 2023-06-29 23:21, hash: f47eed749eaf, type: Release Broken: 3.4 Worked: 3.3 LTS Caused by 57ea827bfb28eb697b74dbc606facbc133e10fab **Short description of error** Mesh Deform has multiple bugs. This post addresses one: * Vertex counts change immediately upon Binding **Exact steps for others to reproduce the error** 1. Load .blend 2. Click Unbind then Bind again 3. See vertex count mismatch error ** Additional notes: ** 1. Apply Mirror modifier 2. Unbind and Bind again 3. See no verts_num mismatch error

a.blend

126 KiB

Jaggz H added the

labels 2023-07-12 19:21:21 +02:00

Philipp Oeser commented

2023-07-13 08:46:28 +02:00

3.3 LTS seems fine, will check when behavior changed

Philipp Oeser added

and removed

labels 2023-07-13 08:46:35 +02:00

Pratik Borhade commented

2023-07-13 08:48:48 +02:00

Thanks for the report. Can confirm. As far as I can see, object only has 320 vertices. Not sure why it shows 342
I might be missing something. Will leave this report for others to triage.

edit: Did not refresh the page before commenting 😅

~~Thanks for the report. Can confirm. As far as I can see, object only has 320 vertices. Not sure why it shows 342 I might be missing something. Will leave this report for others to triage.~~ edit: Did not refresh the page before commenting 😅

Philipp Oeser commented

2023-07-13 09:29:09 +02:00

Seems the subd modifier is responsible for the vertex count mismatch (if Use Creases is used)

Seems the subd modifier is responsible for the vertex count mismatch (if `Use Creases` is used)

Philipp Oeser commented

2023-07-13 10:15:46 +02:00

Caused by 57ea827bfb

CC @HooglyBoogly

Caused by 57ea827bfb28eb697b74dbc606facbc133e10fab CC @HooglyBoogly

Philipp Oeser added

and removed

labels 2023-07-13 10:16:21 +02:00

Hans Goudey self-assigned this 2023-07-13 19:44:59 +02:00

Hans Goudey removed their assignment 2023-07-13 19:46:52 +02:00

Hans Goudey commented

2023-07-13 19:49:08 +02:00

This is because the mirror modifier deleted the crease before that commit. Now that it doesn't delete the crease, the results of the subdivision surface modifier are different. But that's just an unfortunate consequence of the changed (fixed) behavior. Though obviously we try to avoid it, sometimes changes in Blender will change the number of vertices generated. The way that interacts with binding is unfortunate but I don't think we can do anything about it in this case.

Hans Goudey closed this issue

2023-07-13 19:49:10 +02:00

Blender Bot added

and removed

labels 2023-07-13 19:49:45 +02:00

Philipp Oeser commented

2023-07-13 20:12:57 +02:00

Hold on, are you saying we cannot use this kind of modifier stack anymore?
Because it also doesnt work if you unbind/bind - or even delete and a re-add - the mesh deform modifier.

Hold on, are you saying we cannot use this kind of modifier stack anymore? Because it also doesnt work if you unbind/bind - or even delete and a re-add - the mesh deform modifier.

Philipp Oeser reopened this issue

2023-07-13 20:13:00 +02:00

Blender Bot added

and removed

labels 2023-07-13 20:13:02 +02:00

Hans Goudey commented

2023-07-13 20:33:10 +02:00

I don't know how it worked before then, or even how we expect the mesh deform modifier to work when there are "generative" modifiers that change the vertex count before it. Anyway, it seems like if that commit actually did change this behavior, the only reason is because the mesh deform modifier can't handle this situation, but it just happened to work before.

Jakub Jaszewski commented

2023-07-13 20:59:56 +02:00

Possibly related: #108819

Possibly related: https://projects.blender.org/blender/blender/issues/108819

Philipp Oeser commented

2023-07-14 10:44:41 +02:00

I don't know how it worked before then, or even how we expect the mesh deform modifier to work when there are "generative" modifiers that change the vertex count before it. Anyway, it seems like if that commit actually did change this behavior, the only reason is because the mesh deform modifier can't handle this situation, but it just happened to work before.

This should all work with generative modifiers before it.
The reason the commit broke the behavior is that 57ea827bfb (and later e5ec04d73c) changed the CustomData_MeshMasks in BKE_object_handle_data_update, but not in object_force_modifier_update_for_bind (this was still using CD_MASK_BAREMESH -- which is probably not appropriate anymore since moving so many relevant [to modifiers] data to attributes).

This not only affected the Mesh Deform modifier btw., but also Corrective Smooth and Laplacian Smooth (if Bind Coords are used)

Will put up a PR that fixes this by matching the masks in both places.

> I don't know how it worked before then, or even how we expect the mesh deform modifier to work when there are "generative" modifiers that change the vertex count before it. Anyway, it seems like if that commit actually did change this behavior, the only reason is because the mesh deform modifier can't handle this situation, but it just happened to work before. This should all work with generative modifiers before it. The reason the commit broke the behavior is that 57ea827bfb (and later e5ec04d73c78) changed the `CustomData_MeshMasks` in `BKE_object_handle_data_update`, but not in `object_force_modifier_update_for_bind` (this was still using `CD_MASK_BAREMESH` -- which is probably not appropriate anymore since moving so many relevant [to modifiers] data to attributes). This not only affected the `Mesh Deform` modifier btw., but also `Corrective Smooth` and `Laplacian Smooth` (if Bind Coords are used) Will put up a PR that fixes this by matching the masks in both places.

Philipp Oeser added

and removed

labels 2023-07-14 10:44:51 +02:00

Philipp Oeser self-assigned this 2023-07-14 10:44:55 +02:00

Philipp Oeser referenced a pull request that will close this issue

2023-07-14 11:28:02 +02:00

Fix #110021: Deform modifier binding can go wrong in certain stacks #110095

Hans Goudey added