Geometry Nodes: Point instance node provides deleted objects #89735

New Issue

user1 · 2021-07-08T14:50:00+02:00

user1 commented

2021-07-08 14:50:00 +02:00

System Information
Operating system: Linux-5.4.0-77-generic-x86_64-with-glibc2.27 64 Bits
Graphics card: GeForce GT 630/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 390.143

Blender Version
Broken: version: 3.0.0 Alpha, branch: arcpatch-D11813, commit date: 2021-07-07 09:31, hash: 8d222aa20d
Worked: (newest version of Blender that worked as expected)

Short description of error
After deleting a Suzanne object, the nodetree of another object still is able to call Suzanne.

untitled.blend

**System Information** Operating system: Linux-5.4.0-77-generic-x86_64-with-glibc2.27 64 Bits Graphics card: GeForce GT 630/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 390.143 **Blender Version** Broken: version: 3.0.0 Alpha, branch: arcpatch-[D11813](https://archive.blender.org/developer/D11813), commit date: 2021-07-07 09:31, hash: `8d222aa20d` Worked: (newest version of Blender that worked as expected) **Short description of error** After deleting a Suzanne object, the nodetree of another object still is able to call Suzanne. ![Bildschirmfoto zu 2021-07-08 14-42-18.png](https://archive.blender.org/developer/F10217119/Bildschirmfoto_zu_2021-07-08_14-42-18.png) [untitled.blend](https://archive.blender.org/developer/F10217121/untitled.blend)

user1 commented

2021-07-08 14:50:00 +02:00

Added subscriber: @user1

Fen commented

2021-07-08 15:37:09 +02:00

Added subscriber: @chemicalcrux

Fen commented

2021-07-08 15:37:09 +02:00

I believe this is intended, albeit surprising: deleting the object just unlinks it from the scene. It still exists, just without any users.

If you purge unused datablocks (check the File menu), the objects should go away.

I believe this is intended, albeit surprising: deleting the object just unlinks it from the scene. It still exists, just without any users. If you purge unused datablocks (check the File menu), the objects should go away.

Richard Antalik commented

2021-07-09 11:12:59 +02:00

Added subscriber: @iss

Richard Antalik commented

2021-07-09 11:12:59 +02:00

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

Richard Antalik commented

2021-07-09 11:12:59 +02:00

@chemicalcrux In this case, object will have user (nodes) and so it won't be marked as orphaned.

Not sure if this is intended really, because this changes behavior of what happens when object is deleted and it is used in some modifier for example:

if object is also used by geo nodes, it will still influence modifiers
if object is not used in geo nodes, it won't influence modifiers anymore.

@chemicalcrux In this case, object will have user (nodes) and so it won't be marked as orphaned. Not sure if this is intended really, because this changes behavior of what happens when object is deleted and it is used in some modifier for example: - if object is also used by geo nodes, it will still influence modifiers - if object is not used in geo nodes, it won't influence modifiers anymore.

Jacques Lucke commented

2021-07-23 14:25:52 +02:00

Added subscriber: @JacquesLucke

Jacques Lucke commented

2021-07-23 14:25:52 +02:00

I'm not sure about what behavior I'd expect tbh.

This bug is partially solved by the following diff. For some use cases it makes sense if the object would not be deleted though and when it is owned by the node group (e.g. when a node group asset "contains" an object).

diff --git a/source/blender/blenkernel/intern/node.cc b/source/blender/blenkernel/intern/node.cc
index e9608457896..ccb130f8704 100644
--- a/source/blender/blenkernel/intern/node.cc
+++ b/source/blender/blenkernel/intern/node.cc
@@ -282,7 +282,7 @@ static void library_foreach_node_socket(LibraryForeachIDData *data, bNodeSocket
   switch ((eNodeSocketDatatype)sock->type) {
     case SOCK_OBJECT: {
       bNodeSocketValueObject *default_value = (bNodeSocketValueObject *)sock->default_value;
-      BKE_LIB_FOREACHID_PROCESS(data, default_value->value, IDWALK_CB_USER);
+      BKE_LIB_FOREACHID_PROCESS(data, default_value->value, IDWALK_CB_NOP);
       break;
     }
     case SOCK_IMAGE: {
@@ -293,7 +293,7 @@ static void library_foreach_node_socket(LibraryForeachIDData *data, bNodeSocket
     case SOCK_COLLECTION: {
       bNodeSocketValueCollection *default_value = (bNodeSocketValueCollection *)
                                                       sock->default_value;
-      BKE_LIB_FOREACHID_PROCESS(data, default_value->value, IDWALK_CB_USER);
+      BKE_LIB_FOREACHID_PROCESS(data, default_value->value, IDWALK_CB_NOP);
       break;
     }
     case SOCK_TEXTURE: {
diff --git a/source/blender/modifiers/intern/MOD_nodes.cc b/source/blender/modifiers/intern/MOD_nodes.cc
index 5fa11ffdd10..516226c35ad 100644
--- a/source/blender/modifiers/intern/MOD_nodes.cc
+++ b/source/blender/modifiers/intern/MOD_nodes.cc
@@ -254,7 +254,7 @@ static void foreachIDLink(ModifierData *md, Object *ob, IDWalkFunc walk, void *u
       [](IDProperty *id_prop, void *user_data) {
         ForeachSettingData *settings = (ForeachSettingData *)user_data;
         settings->walk(
-            settings->userData, settings->ob, (ID **)&id_prop->data.pointer, IDWALK_CB_USER);
+            settings->userData, settings->ob, (ID **)&id_prop->data.pointer, IDWALK_CB_NOP);
       },
       &settings);
 }

Note, this only partially solves the issue, because the manual user counting for object/collection sockets would still have to be updated.

I'm not sure about what behavior I'd expect tbh. This bug is partially solved by the following diff. For some use cases it makes sense if the object would not be deleted though and when it is owned by the node group (e.g. when a node group asset "contains" an object). ``` diff --git a/source/blender/blenkernel/intern/node.cc b/source/blender/blenkernel/intern/node.cc index e9608457896..ccb130f8704 100644 --- a/source/blender/blenkernel/intern/node.cc +++ b/source/blender/blenkernel/intern/node.cc @@ -282,7 +282,7 @@ static void library_foreach_node_socket(LibraryForeachIDData *data, bNodeSocket switch ((eNodeSocketDatatype)sock->type) { case SOCK_OBJECT: { bNodeSocketValueObject *default_value = (bNodeSocketValueObject *)sock->default_value; - BKE_LIB_FOREACHID_PROCESS(data, default_value->value, IDWALK_CB_USER); + BKE_LIB_FOREACHID_PROCESS(data, default_value->value, IDWALK_CB_NOP); break; } case SOCK_IMAGE: { @@ -293,7 +293,7 @@ static void library_foreach_node_socket(LibraryForeachIDData *data, bNodeSocket case SOCK_COLLECTION: { bNodeSocketValueCollection *default_value = (bNodeSocketValueCollection *) sock->default_value; - BKE_LIB_FOREACHID_PROCESS(data, default_value->value, IDWALK_CB_USER); + BKE_LIB_FOREACHID_PROCESS(data, default_value->value, IDWALK_CB_NOP); break; } case SOCK_TEXTURE: { diff --git a/source/blender/modifiers/intern/MOD_nodes.cc b/source/blender/modifiers/intern/MOD_nodes.cc index 5fa11ffdd10..516226c35ad 100644 --- a/source/blender/modifiers/intern/MOD_nodes.cc +++ b/source/blender/modifiers/intern/MOD_nodes.cc @@ -254,7 +254,7 @@ static void foreachIDLink(ModifierData *md, Object *ob, IDWalkFunc walk, void *u [](IDProperty *id_prop, void *user_data) { ForeachSettingData *settings = (ForeachSettingData *)user_data; settings->walk( - settings->userData, settings->ob, (ID **)&id_prop->data.pointer, IDWALK_CB_USER); + settings->userData, settings->ob, (ID **)&id_prop->data.pointer, IDWALK_CB_NOP); }, &settings); } ``` Note, this only partially solves the issue, because the manual user counting for object/collection sockets would still have to be updated.

Hans Goudey commented

2021-10-29 03:36:29 +02:00

Added subscriber: @HooglyBoogly

Hans Goudey commented

2021-10-29 03:36:29 +02:00

I also wouldn't necessarily call this a bug, since a node group owning an object seems like a valid situation. So personally I'd close this report-- we can change the behavior later on if we decide that's better.

Hans Goudey commented

2021-11-09 23:23:56 +01:00

Changed status from 'Confirmed' to: 'Archived'

Hans Goudey closed this issue

2021-11-09 23:23:56 +01:00

Hans Goudey commented

2021-11-09 23:23:56 +01:00

Since the expected behavior isn't obvious, I'm going to close this report. Maybe this behavior can be reconsidered as part of the fake user discussions (can't find the task right now though).

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

Geometry Nodes: Point instance node provides deleted objects #89735