Fix #112580: Limit Constraint with 'Affect Transform' not working properly in 'World Space' #112601

Germano Cavalcante · 2023-09-19T22:03:31+02:00

2023-09-19 22:03:31 +02:00

The problem is observed with the "Limit Distance" and "Limit Location"
constraints.

There is an incorrect usage of td->mtx and td->smtx when converting
TransData space from local to global.

In this case, the code is concatenating matrices instead of converting
the location component space.

Also, these matrices only inform the global transformation components
of rotation and scale. They do not include location.

Since the "Limit Distance" and "Limit Location" constraints only require
the location component, it is not necessary to convert the rotation and
scale components.

So, the solution is to convert the location component space instead of
concatenating matrices.

The problem is observed with the "Limit Distance" and "Limit Location" constraints. There is an incorrect usage of `td->mtx` and `td->smtx` when converting `TransData` space from local to global. In this case, the code is concatenating matrices instead of converting the location component space. Also, these matrices only inform the global transformation components of rotation and scale. They do not include location. Since the "Limit Distance" and "Limit Location" constraints only require the location component, it is not necessary to convert the rotation and scale components. So, the solution is to convert the location component space instead of concatenating matrices.

Germano Cavalcante added 1 commit 2023-09-19 22:03:41 +02:00

2738725e0e Fix #112580 : Limit Constraint with 'Affect Transform' not working properly in 'World Space'

The problem is observed with the Limit Distance and Limit Location
constraints.

There is an incorrect usage of td->mtx and td->smtx when converting
TransData space from local to global.

In this case, the code is concatenating matrices instead of converting
the location component space.

Since the Limit Distance and Limit Location constraints only require
the location component, it is not necessary to convert the rotation and
scale components.

So, the solution is to convert the location component space instead of
concatenating matrices.

Germano Cavalcante requested review from Philipp Oeser 2023-09-19 22:04:15 +02:00

Germano Cavalcante requested review from Alexander Gavrilov 2023-09-19 22:04:15 +02:00

Philipp Oeser requested changes 2023-09-20 09:40:15 +02:00

Dismissed

Philipp Oeser left a comment

Seems there is still something wrong:
if you take the armature object from #112580 and move it (so it is out of the origin), then go to posemode and translate the bone (with Affect Transform ON) it goes haywire. So it seems we still dont have full worldspace?

Seems there is still something wrong: if you take the armature object from #112580 and move it (so it is out of the origin), then go to posemode and translate the bone (with `Affect Transform` ON) it goes haywire. So it seems we still dont have full worldspace?

Germano Cavalcante added 2 commits 2023-09-20 22:34:35 +02:00

844409bba1 Merge remote-tracking branch 'origin/main' into fix_112580

9a5bf5c1e8 Also use the component's local matrix

Germano Cavalcante added 1 commit 2023-09-20 22:43:04 +02:00

139c6f4655 Fix duplicate rotation and scale transformation

Germano Cavalcante commented

2023-09-20 22:45:47 +02:00

In fact, since t->mtx and t->smtx are 3x3 matrices, they only have rotation and scale components.
The location offset needs to be obtained from somewhere else.

I believe tc->mat[3] should be enough.

In fact, since `t->mtx` and `t->smtx` are 3x3 matrices, they only have rotation and scale components. The location offset needs to be obtained from somewhere else. I believe `tc->mat[3]` should be enough.

Philipp Oeser commented

2023-09-28 13:59:18 +02:00

In fact, since t->mtx and t->smtx are 3x3 matrices, they only have rotation and scale components.
The location offset needs to be obtained from somewhere else.

I believe tc->mat[3] should be enough.

Sorry, I misread this comment (assuming there would be some upcoming commit still), will check again in a bit.
This would also have to target the blender-v4.0-release branch now.

> In fact, since `t->mtx` and `t->smtx` are 3x3 matrices, they only have rotation and scale components. > The location offset needs to be obtained from somewhere else. > > I believe `tc->mat[3]` should be enough. Sorry, I misread this comment (assuming there would be some upcoming commit still), will check again in a bit. This would also have to target the `blender-v4.0-release` branch now.

Christoph Lendenfeld approved these changes 2023-10-06 11:46:44 +02:00

Christoph Lendenfeld left a comment

this fixes the issue and works with armatures rotated and scaled
please rebase onto 4.0 the stabilizing branch so we can land it there

this fixes the issue and works with armatures rotated and scaled please rebase onto 4.0 the stabilizing branch so we can land it there

Christoph Lendenfeld commented

2023-10-06 16:59:13 +02:00

this might also solve #87219: Limit scale bone constraint: "affect transform" doesn't seem to work

this might also solve [#87219: Limit scale bone constraint: "affect transform" doesn't seem to work](https://projects.blender.org/blender/blender/issues/87219)

Philipp Oeser commented

2023-10-06 17:11:45 +02:00

Sorry, this slipped through again.
Afk now (so can't check), but just go ahead and merge this then, trust you on this

Sorry, this slipped through again. Afk now (so can't check), but just go ahead and merge this then, trust you on this

Philipp Oeser dismissed lichtwerk’s review 2023-10-06 17:12:14 +02:00

Germano Cavalcante merged commit 7e078b724a into main

2023-10-06 17:14:36 +02:00

Germano Cavalcante deleted branch fix_112580

2023-10-06 17:14:38 +02:00

Germano Cavalcante referenced this issue from a commit

2023-10-06 17:14:40 +02:00

Fix #112580: Limit Constraint with 'Affect Transform' not working properly in 'World Space'