Constraint Shrinkwrap does not applies, it reset all changes #94600

New Issue

Vladimir · 2022-01-03T13:31:15+01:00

Vladimir commented

2022-01-03 13:31:15 +01:00

System Information
Operating system: Linux-5.15.12-1-MANJARO-x86_64-with-glibc2.33 64 Bits
Graphics card: Mesa DRI Intel(R) HD Graphics 4600 (HSW GT2) Intel Open Source Technology Center 4.5 (Core Profile) Mesa 21.3.3

Blender Version
Broken: version: 3.1.0 Alpha, branch: makepkg (modified), commit date: 2021-12-31 20:32, hash: 6844304dda
Worked: (newest version of Blender that worked as expected)

Short description of error

Constraint (not modifier) Shrinkwrap does not applies, after click Apply it reset all changes:

Exact steps for others to reproduce the error

Open test file shrinkwrap.blend

**System Information** Operating system: Linux-5.15.12-1-MANJARO-x86_64-with-glibc2.33 64 Bits Graphics card: Mesa DRI Intel(R) HD Graphics 4600 (HSW GT2) Intel Open Source Technology Center 4.5 (Core Profile) Mesa 21.3.3 **Blender Version** Broken: version: 3.1.0 Alpha, branch: makepkg (modified), commit date: 2021-12-31 20:32, hash: `6844304dda` Worked: (newest version of Blender that worked as expected) **Short description of error** Constraint (not modifier) Shrinkwrap does not applies, after click Apply it reset all changes: ![Screenshot_20220103_152303.png](https://archive.blender.org/developer/F12786657/Screenshot_20220103_152303.png) **Exact steps for others to reproduce the error** Open test file [shrinkwrap.blend](https://archive.blender.org/developer/F12786659/shrinkwrap.blend)

Vladimir commented

2022-01-03 13:31:15 +01:00

Added subscriber: @Wovchick

Philipp Oeser commented

2022-01-03 16:33:07 +01:00

Added subscribers: @Sergey, @dr.sybren, @lichtwerk

Philipp Oeser commented

2022-01-03 16:33:07 +01:00

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

Philipp Oeser commented

2022-01-03 16:33:07 +01:00

Can confirm, Apply visual transform on the other hand seems to behave better.

I am trying to follow D5007: Fix #60971 apply visual transform not working / ddc19819f3, but its kinda for me hard to read the comments in relation to what was committed, so what is really necessary in the combo of BKE_constraints_make_evalob & BKE_constraints_solve. I assume this should always be fed with evaluated objects, and if I do the following, we indeed get proper results:

P2696: T94600_snippet



diff --git a/source/blender/blenkernel/intern/constraint.c b/source/blender/blenkernel/intern/constraint.c
index d284c32b1df..62e1c806dcd 100644
--- a/source/blender/blenkernel/intern/constraint.c
+++ b/source/blender/blenkernel/intern/constraint.c
@@ -5679,8 +5679,10 @@ bool BKE_constraint_apply_for_object(Depsgraph *depsgraph,
   bConstraint *new_con = BKE_constraint_duplicate_ex(con, 0, !ID_IS_LINKED(ob));
   ListBase single_con = {new_con, new_con};
 
+  Object *ob_eval = DEG_get_evaluated_object(depsgraph, ob);
+
   bConstraintOb *cob = BKE_constraints_make_evalob(
-      depsgraph, scene, ob, NULL, CONSTRAINT_OBTYPE_OBJECT);
+      depsgraph, scene, ob_eval, NULL, CONSTRAINT_OBTYPE_OBJECT);
   /* Undo the effect of the current constraint stack evaluation. */
   mul_m4_m4m4(cob->matrix, ob->constinv, cob->matrix);

context: quote from D5007:

It is unreliable to use BKE_object_where_is_calc on original object: it might depend on constraints which needs final geometry which does not exist in an original domain.

However if I look at other places with the combo of BKE_constraints_make_evalob & BKE_constraints_solve, these dont necessarily get evaluated objects/pchans either, so my understanding might be limited here...

@Sergey: could you share your wisdom?
CC @dr.sybren

Can confirm, Apply visual transform on the other hand seems to behave better. I am trying to follow [D5007: Fix #60971 apply visual transform not working](https://archive.blender.org/developer/D5007) / ddc19819f3, but its kinda for me hard to read the comments in relation to what was committed, so what is really necessary in the combo of `BKE_constraints_make_evalob` & `BKE_constraints_solve`. I assume this should always be fed with evaluated objects, and if I do the following, we indeed get proper results: [P2696: T94600_snippet](https://archive.blender.org/developer/P2696.txt) ``` diff --git a/source/blender/blenkernel/intern/constraint.c b/source/blender/blenkernel/intern/constraint.c index d284c32b1df..62e1c806dcd 100644 --- a/source/blender/blenkernel/intern/constraint.c +++ b/source/blender/blenkernel/intern/constraint.c @@ -5679,8 +5679,10 @@ bool BKE_constraint_apply_for_object(Depsgraph *depsgraph, bConstraint *new_con = BKE_constraint_duplicate_ex(con, 0, !ID_IS_LINKED(ob)); ListBase single_con = {new_con, new_con}; + Object *ob_eval = DEG_get_evaluated_object(depsgraph, ob); + bConstraintOb *cob = BKE_constraints_make_evalob( - depsgraph, scene, ob, NULL, CONSTRAINT_OBTYPE_OBJECT); + depsgraph, scene, ob_eval, NULL, CONSTRAINT_OBTYPE_OBJECT); /* Undo the effect of the current constraint stack evaluation. */ mul_m4_m4m4(cob->matrix, ob->constinv, cob->matrix); ``` context: quote from [D5007](https://archive.blender.org/developer/D5007): > It is unreliable to use BKE_object_where_is_calc on original object: it might depend on constraints which needs final geometry which does not exist in an original domain. However if I look at other places with the combo of `BKE_constraints_make_evalob` & `BKE_constraints_solve`, these dont necessarily get evaluated objects/pchans either, so my understanding might be limited here... @Sergey: could you share your wisdom? CC @dr.sybren

blender-admin commented

2022-01-03 16:48:42 +01:00

This issue was referenced by 15e4d0f25d

This issue was referenced by 15e4d0f25d75b54d26101d9a3a65272f77241ec8

blender-admin commented

2022-01-03 16:48:42 +01:00

This issue was referenced by 1b1e947162

This issue was referenced by 1b1e9471625d195d418c995375a1cc6a0833031a

Vladimir commented

2022-01-04 21:35:58 +01:00

Hmm... Interested... if the object Cube join to Sphere (CTRL+J) then the constraint apply correctly! it means that somewhere it is already implemented in the code, it remains to find and refer to this method

Sergey Sharybin commented

2022-01-05 15:44:16 +01:00

@lichtwerk The other usages of BKE_constraints_make_evalobhappen from inside evaluation, so they are intended to operate on exact data they are given (which happened to be evaluated).

Not really sure why BKE_constraint_apply_for_object needs to access evaluated object (the matrices and such are copied from evaluated to original in the BKE_object_sync_to_original). Would be really nice to know what is it missing there to have proper behavior without requesting evaluated object.

@lichtwerk The other usages of `BKE_constraints_make_evalob`happen from inside evaluation, so they are intended to operate on exact data they are given (which happened to be evaluated). Not really sure why `BKE_constraint_apply_for_object` needs to access evaluated object (the matrices and such are copied from evaluated to original in the `BKE_object_sync_to_original`). Would be really nice to know what is it missing there to have proper behavior without requesting evaluated object.

Philipp Oeser commented

2022-01-05 15:56:51 +01:00

In #94600#1282908, @Wovchick wrote:
Hmm... Interested... if the object Cube join to Sphere (CTRL+J) then the constraint apply correctly! it means that somewhere it is already implemented in the code, it remains to find and refer to this method

Like I said, Object > Apply > Visual Transform is also fine

In #94600#1283298, @Sergey wrote:
@lichtwerk The other usages of BKE_constraints_make_evalobhappen from inside evaluation, so they are intended to operate on exact data they are given (which happened to be evaluated).

Not really sure why BKE_constraint_apply_for_object needs to access evaluated object (the matrices and such are copied from evaluated to original in the BKE_object_sync_to_original). Would be really nice to know what is it missing there to have proper behavior without requesting evaluated object.

Will see if I can continue digging in a bit...

> In #94600#1282908, @Wovchick wrote: > Hmm... Interested... if the object Cube join to Sphere (CTRL+J) then the constraint apply correctly! it means that somewhere it is already implemented in the code, it remains to find and refer to this method Like I said, `Object` > `Apply` > `Visual Transform` is also fine > In #94600#1283298, @Sergey wrote: > @lichtwerk The other usages of `BKE_constraints_make_evalob`happen from inside evaluation, so they are intended to operate on exact data they are given (which happened to be evaluated). > > Not really sure why `BKE_constraint_apply_for_object` needs to access evaluated object (the matrices and such are copied from evaluated to original in the `BKE_object_sync_to_original`). Would be really nice to know what is it missing there to have proper behavior without requesting evaluated object. Will see if I can continue digging in a bit...

Sergey Sharybin commented

2022-01-05 16:27:25 +01:00

@lichtwerk, actually, constraint might need an evaluated state of target's geometry (so object matrices is not enough for correct behavior). So your proposed patch is actually correct :)

Philipp Oeser commented

2022-01-05 17:11:29 +01:00

Since the example file does not actually have anything acting on its geometry, should I investigate further why there is something wrong with the matrices? Create a Diff? Commit straight?

Philipp Oeser self-assigned this 2022-01-05 17:52:22 +01:00

Philipp Oeser commented

2022-01-05 17:52:22 +01:00

Doing some more digging

Philipp Oeser commented

2022-01-10 12:04:14 +01:00

Changed status from 'Confirmed' to: 'Resolved'

Philipp Oeser closed this issue

2022-01-10 12:04:14 +01:00

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

Constraint Shrinkwrap does not applies, it reset all changes #94600