Broken Shrinkwrap Modifier (and Bendy Bones) #73840

New Issue

Evan Biesaga · 2020-02-15T16:12:51+01:00

Evan Biesaga commented

2020-02-15 16:12:51 +01:00

System Information
Operating system: Windows 10 1909
Graphics card: Nvidia GeForce RTX 2070

Blender Version
Broken: 2.82, 2020-2-16
Worked: 2.81-adfe68e2025b-windows64 d1657b406e

Short description of error
The shrinkwrap modifier seems to have a completely different algorithm in this new version of Blender. The eyebrows, which I projected onto her face, worked perfect in 2.81, but is completely distorted in 2.82
2.81 on top, 2.82 on the bottom.
{F8341304}

File
clothes.blend

EDIT Upon closer inspection, it's a problem of both shrinkwrap AND bendy bones. For some reason, the roll of the bendy bones are really messed up.

**System Information** Operating system: Windows 10 1909 Graphics card: Nvidia GeForce RTX 2070 **Blender Version** Broken: 2.82, 2020-2-16 Worked: 2.81-adfe68e2025b-windows64 d1657b406e **Short description of error** The shrinkwrap modifier seems to have a completely different algorithm in this new version of Blender. The eyebrows, which I projected onto her face, worked perfect in 2.81, but is completely distorted in 2.82 2.81 on top, 2.82 on the bottom. {[F8341304](https://archive.blender.org/developer/F8341304/2.81.png)}![2.82.png](https://archive.blender.org/developer/F8341303/2.82.png) **File** [clothes.blend](https://archive.blender.org/developer/F8341288/clothes.blend) **EDIT** Upon closer inspection, it's a problem of both shrinkwrap AND bendy bones. For some reason, the roll of the bendy bones are really messed up.

Evan Biesaga commented

2020-02-15 16:12:51 +01:00

Added subscriber: @sagaevan

Evan Biesaga changed title from ~~Broken Shrinkwrap Modifier~~ to Broken Shrinkwrap Modifier (and Bendy Bones)

2020-02-15 16:33:17 +01:00

Cody Winchester commented

2020-02-16 05:13:53 +01:00

Added subscriber: @CodyWinchester

Cody Winchester commented

2020-02-16 05:13:54 +01:00

I have been looking into this and it seems this has nothing to do with the shrinkwrap modifier. It is the result of changes to the bbone roll calculation.

I have isolated the issue to the changes made in this patch https://developer.blender.org/rBd1657b406ed0f3df9e1690cd445702c6178ffd15

If I reverse the patch the provided file works as it did in 2.81a.

Also, the provided file has the bones using Absolute custom handles and the Start custom handle bone has its head at the same coord as the bbones head which causes roll issues.

I have been looking into this and it seems this has nothing to do with the shrinkwrap modifier. It is the result of changes to the bbone roll calculation. I have isolated the issue to the changes made in this patch https://developer.blender.org/rBd1657b406ed0f3df9e1690cd445702c6178ffd15 If I reverse the patch the provided file works as it did in 2.81a. Also, the provided file has the bones using Absolute custom handles and the Start custom handle bone has its head at the same coord as the bbones head which causes roll issues.

Ankit Meel commented

2020-02-16 05:33:26 +01:00

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

Evan Biesaga commented

2020-02-17 02:01:24 +01:00

Aaah! Sorry, I guess I just didn't understand bendy bones. Thank you so much!

Bastien Montagne commented

2020-02-17 15:22:45 +01:00

Added subscribers: @angavrilov, @mont29

Alexander Gavrilov was assigned by Bastien Montagne

2020-02-17 15:22:45 +01:00

Bastien Montagne commented

2020-02-17 15:22:45 +01:00

@angavrilov can you please check whether that behavior is expected/correct in current code? Or whether d1657b406 did introduce some regression? Thanks.

@angavrilov can you please check whether that behavior is expected/correct in current code? Or whether d1657b406 did introduce some regression? Thanks.

Alexander Gavrilov commented

2020-02-17 16:07:56 +01:00

In #73840#872701, @CodyWinchester wrote:
Also, the provided file has the bones using Absolute custom handles and the Start custom handle bone has its head at the same coord as the bbones head which causes roll issues.

Absolute handle with position that is the same as the bbone makes no sense. It's like expecting a reasonable result from dividing 0 by 0.

> In #73840#872701, @CodyWinchester wrote: > Also, the provided file has the bones using Absolute custom handles and the Start custom handle bone has its head at the same coord as the bbones head which causes roll issues. Absolute handle with position that is the same as the bbone makes no sense. It's like expecting a reasonable result from dividing 0 by 0.

Sybren A. Stüvel commented

2020-04-21 17:21:08 +02:00

Added subscriber: @dr.sybren

Sybren A. Stüvel commented

2020-04-21 17:21:08 +02:00

In #73840#873856, @mont29 wrote:
can you please check whether that behavior is expected/correct [...] Or whether d1657b406 did introduce some regression?

I think d1657b406 is a considerable improvement. I did some tests, and the old code did produce some dubious results. For example, in a case like this:

  float unit_matrix[3][3];
  unit_m3(unit_matrix);

  float vector[3] = {1.0f, 1.0f, 1.0f};
  float roll;
  mat3_vec_to_roll(unit_matrix, vector, &roll);

I would expect the roll to be zero, as the matrix performs no rotation at all. This is also the result now, but before d1657b406e it produced a roll of approximately -15 degrees. Another test case:

  const float rot_around_x[3] = {1.234f, 0.0f, 0.0f};
  eul_to_mat3(rotation_matrix, rot_around_x);

  const float between_x_and_y[3] = {1.0f, 1.0f, 0.0f};
  mat3_vec_to_roll(rotation_matrix, between_x_and_y, &roll);

Here the rotation matrix rotates 1.234 radians around the X-axis. Given that an input vector of (1, 0, 0) results in a roll of the full 1.234 radians, and an input vector (0, 1, 0) results in a zero roll, I would expect the answer for the in-between vector (1, 1, 0) to be in between these values as well. Neither the new code nor the old one produces exactly half the angle, but the new code is certainly closer, with respectively 1.111 and 0.572 radians.

In #73840#873963, @angavrilov wrote:
Absolute handle with position that is the same as the bbone makes no sense. It's like expecting a reasonable result from dividing 0 by 0.

To elaborate on this a bit more, the Custom Handles section in the Blender manual states about absolute handles:

The Bézier handle is controlled by the position of the head of the handle bone relative to the head of the current bone.

This means that there should be a positional difference between the bone positions, or the direction of the Bézier curve is undefined.
Since d1657b406e, Blender behaves differently in this undefined case, but I think that's not really a bug.

> In #73840#873856, @mont29 wrote: > can you please check whether that behavior is expected/correct [...] Or whether d1657b406 did introduce some regression? I think d1657b406 is a considerable improvement. I did some tests, and the old code did produce some dubious results. For example, in a case like this: ``` float unit_matrix[3][3]; unit_m3(unit_matrix); float vector[3] = {1.0f, 1.0f, 1.0f}; float roll; mat3_vec_to_roll(unit_matrix, vector, &roll); ``` I would expect the roll to be zero, as the matrix performs no rotation at all. This is also the result now, but before d1657b406e it produced a roll of approximately -15 degrees. Another test case: ``` const float rot_around_x[3] = {1.234f, 0.0f, 0.0f}; eul_to_mat3(rotation_matrix, rot_around_x); const float between_x_and_y[3] = {1.0f, 1.0f, 0.0f}; mat3_vec_to_roll(rotation_matrix, between_x_and_y, &roll); ``` Here the rotation matrix rotates 1.234 radians around the X-axis. Given that an input vector of `(1, 0, 0)` results in a roll of the full 1.234 radians, and an input vector `(0, 1, 0)` results in a zero roll, I would expect the answer for the in-between vector `(1, 1, 0)` to be in between these values as well. Neither the new code nor the old one produces exactly half the angle, but the new code is certainly closer, with respectively 1.111 and 0.572 radians. > In #73840#873963, @angavrilov wrote: > Absolute handle with position that is the same as the bbone makes no sense. It's like expecting a reasonable result from dividing 0 by 0. To elaborate on this a bit more, [the Custom Handles section in the Blender manual](https://docs.blender.org/manual/en/dev/animation/armatures/bones/properties/bendy_bones.html#custom-handles) states about absolute handles: > The Bézier handle is controlled by the position of the head of the handle bone relative to the head of the current bone. This means that there should be a positional difference between the bone positions, or the direction of the Bézier curve is undefined. Since d1657b406e, Blender behaves differently in this undefined case, but I think that's not really a bug.

Alexander Gavrilov commented

2020-04-22 11:27:29 +02:00

For a more practical answer, if you want handle controls that are located at the head/tail of the bbone, you probably want to use the Tangent mode. Absolute is just the default 'connected chain' mode that allows manually selecting the adjacent bone of the chain, instead of automatically using parent/child.

Sybren A. Stüvel commented

2020-05-12 11:24:04 +02:00

Changed status from 'Confirmed' to: 'Needs User Info'

Sybren A. Stüvel commented

2020-05-12 11:24:04 +02:00

@sagaevan I did a quick test with the left eyebrow, and the rig works a lot better with @angavrilov 's suggestion of setting the handle type to Tangent. It does require a flip ({key Alt+F} in edit mode) of the eyebrow control.L bone to work properly, though. Is this a solution you could work with?

Sybren A. Stüvel commented

2020-06-08 11:33:15 +02:00

Changed status from 'Needs User Info' to: 'Resolved'

Sybren A. Stüvel closed this issue

2020-06-08 11:33:15 +02:00

Sybren A. Stüvel commented

2020-06-08 11:33:15 +02:00

I have clarified the requirement (that there should be a distance between the bones) in the manual, see rBM6742.

I'll assume that the silence implies that this issue is indeed solved. If this is not the case, please open a new report with more detail about the issue.

I have clarified the requirement (that there should be a distance between the bones) in the manual, see rBM6742. I'll assume that the silence implies that this issue is indeed solved. If this is not the case, please open a new report with more detail about the issue.

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

Broken Shrinkwrap Modifier (and Bendy Bones) #73840