How to handle linked data with 'fake user' set #105786

Closed
opened 2023-03-15 11:15:22 +01:00 by Bastien Montagne · 11 comments

133dde41bb changed how 'fake user' flag is handled with linked data.

Previous behavior was a bug/inconsistency, by definition you cannot edit linked data, which means that its 'fake user' status is irrelevant (since it cannot be defined from the user side currently, only in the library). Therefore, it should be ignored, and the only way to decide whether to keep or not a linked ID is if it's actually used by some local data.

However, #103867 and #105687 show that this is causing issues in some cases, where users rely on the linked data's pre-defined 'Fake user' flag to keep their linked data in their production files, even if said data has no real user.

While not ideal, I think that for now we should start again considering 'fake user' flag for linked data as a real usage case.


Note: The 'fix' is trivial and safe for 3.5, but we need a quick decision on this:

diff --git a/source/blender/blenloader/intern/writefile.cc b/source/blender/blenloader/intern/writefile.cc
index 299735ccab0..eb4975bebfe 100644
--- a/source/blender/blenloader/intern/writefile.cc
+++ b/source/blender/blenloader/intern/writefile.cc
@@ -1141,6 +1141,16 @@ static bool write_file_handle(Main *mainvar,
            * asap afterward. */
           id_lib_extern(id_iter);
         }
+        else if (ID_FAKE_USERS(id_iter) > 0) {
+          /* Even though fake user is not directly editable by the user on linked data, it is a
+           * common 'work-around' to set it in library files on data-blocks that need to be linked
+           * but typically do not have an actual real user (e.g. texts, etc.).
+           * See e.g. #105687 and #103867.
+           *
+           * Would be good to find a better solution, but for now consider these as directly linked
+           * as well. */
+          id_lib_extern(id_iter);
+        }
         else {
           id_iter->tag |= LIB_TAG_INDIRECT;
           id_iter->tag &= ~LIB_TAG_EXTERN;
133dde41bb changed how 'fake user' flag is handled with linked data. Previous behavior was a bug/inconsistency, by definition you cannot edit linked data, which means that its 'fake user' status is irrelevant (since it cannot be defined from the user side currently, only in the library). Therefore, it should be ignored, and the only way to decide whether to keep or not a linked ID is if it's actually used by some local data. However, #103867 and #105687 show that this is causing issues in some cases, where users rely on the linked data's pre-defined 'Fake user' flag to keep their linked data in their production files, even if said data has no real user. While not ideal, I think that for now we should start again considering 'fake user' flag for linked data as a real usage case. ------------------- Note: The 'fix' is trivial and safe for 3.5, but we need a quick decision on this: ``` diff --git a/source/blender/blenloader/intern/writefile.cc b/source/blender/blenloader/intern/writefile.cc index 299735ccab0..eb4975bebfe 100644 --- a/source/blender/blenloader/intern/writefile.cc +++ b/source/blender/blenloader/intern/writefile.cc @@ -1141,6 +1141,16 @@ static bool write_file_handle(Main *mainvar, * asap afterward. */ id_lib_extern(id_iter); } + else if (ID_FAKE_USERS(id_iter) > 0) { + /* Even though fake user is not directly editable by the user on linked data, it is a + * common 'work-around' to set it in library files on data-blocks that need to be linked + * but typically do not have an actual real user (e.g. texts, etc.). + * See e.g. #105687 and #103867. + * + * Would be good to find a better solution, but for now consider these as directly linked + * as well. */ + id_lib_extern(id_iter); + } else { id_iter->tag |= LIB_TAG_INDIRECT; id_iter->tag &= ~LIB_TAG_EXTERN; ```
Bastien Montagne added this to the 3.5 milestone 2023-03-15 11:15:22 +01:00
Bastien Montagne added the
Type
Design
Module
Core
labels 2023-03-15 11:15:23 +01:00
Author
Owner

@brecht @dfelinto any opinion here?

@brecht @dfelinto any opinion here?
Member

For inspiration and consideration.

A detailed use-case for node groups

  • Add-ons (like animation nodes) uses node trees to store logic.

  • Logic is run based on a handler.

  • In motion graphics it is common to have a single node tree library that is linked in the scene using it. (Template)

  • When sending to a render farm this connection is being lost.

  • Work around is to store a reference to the node tree in the scene data. Although this is easy to implement, adding a automatic migration requires opening in Blender 3.4 and save, before opening in Blender 3.5.

For inspiration and consideration. *A detailed use-case for node groups* - Add-ons (like animation nodes) uses node trees to store logic. - Logic is run based on a handler. - In motion graphics it is common to have a single node tree library that is linked in the scene using it. (Template) - When sending to a render farm this connection is being lost. - Work around is to store a reference to the node tree in the scene data. Although this is easy to implement, adding a automatic migration requires opening in Blender 3.4 and save, before opening in Blender 3.5.

I think for at least 3.5 and 3.6 the right thing is to do is restore previous behavior to avoid breaking compatibility.

Longer term, I don't think it's correct for the node group use case to expect that link to remain. If it needs to retain the link there should be a user somewhere using a PointerProperty wherever is appropriate. To me that is not a workaround but the right way for a script to reference a datablock.

For the text datablocks, I think that should be part of a set of datablock types that are never auto deleted. But that's part of a bigger picture as discussed in #61209.

I think for at least 3.5 and 3.6 the right thing is to do is restore previous behavior to avoid breaking compatibility. Longer term, I don't think it's correct for the node group use case to expect that link to remain. If it needs to retain the link there should be a user somewhere using a `PointerProperty` wherever is appropriate. To me that is not a workaround but the right way for a script to reference a datablock. For the text datablocks, I think that should be part of a set of datablock types that are never auto deleted. But that's part of a bigger picture as discussed in #61209.

The proposed patch looks ok to me in isolation. Is that restoring exactly what it did before? Or are there differences still where it would now keep links to less or more data than before?

The proposed patch looks ok to me in isolation. Is that restoring exactly what it did before? Or are there differences still where it would now keep links to less or more data than before?
Author
Owner

Current proposed patch only affects linked data with 'Fake user' flag set. In that case, it may save more references to linked data than before:

  • before 133dde41bb (3.4 etc.), IDs marked as indirectly linked were not saved, regardless of their 'Fake User' flag values.
  • after 133dde41bb (current 3.5), only really used linked IDs are saved, regardless of their 'Fake User' flag values.
  • with proposed patch, all linked IDs with 'Fake User' set are saved, regardless of whether they are really used or not.

I think that the only way to restore exactly the same behavior as before would be to revert 133dde41bb. We would then get again into the initial issue that this commit tried to address: the fact that the direct vs. indirect linked status tags are more or less unreliable.

Current proposed patch only affects linked data with 'Fake user' flag set. In that case, it may save more references to linked data than before: * before 133dde41bb (3.4 etc.), IDs marked as indirectly linked were not saved, regardless of their 'Fake User' flag values. * after 133dde41bb (current 3.5), only really used linked IDs are saved, regardless of their 'Fake User' flag values. * with proposed patch, all linked IDs with 'Fake User' set are saved, regardless of whether they are really used or not. I think that the only way to restore exactly the same behavior as before would be to revert 133dde41bb. We would then get again into the initial issue that this commit tried to address: the fact that the direct vs. indirect linked status tags are more or less unreliable.

Discussed this with @mont29 offline, and we couldn't really come up with a safe solution for 3.5 better than the proposed patch.

It does have some possibility of ending up with indirectly linked fake user datablocks after deleting directly linked datablocks, that would have to be manually cleaned up by the user. But a proper solution is probably best left for a bigger change to fake user, or at least for after 3.5.

Discussed this with @mont29 offline, and we couldn't really come up with a safe solution for 3.5 better than the proposed patch. It does have some possibility of ending up with indirectly linked fake user datablocks after deleting directly linked datablocks, that would have to be manually cleaned up by the user. But a proper solution is probably best left for a bigger change to fake user, or at least for after 3.5.

There is no ideal solution (other than reverting the behavour back to 3.4 days). So I'm okay with the proposed patch.

There is no ideal solution (other than reverting the behavour back to 3.4 days). So I'm okay with the proposed patch.
Author
Owner

Committed as a60f651502

Committed as a60f651502
Blender Bot added the
Status
Archived
label 2023-03-20 15:38:27 +01:00
Member

I wonder, could we get an option added to the purge operator (bpy.ops.outliner.orphans_purge) that ignores the fake user setting on linked data? Then we could have both behaviours sortof existing alongside each other.

I think we're gonna need something like that at the end of the current open movie production, when we try to make nice and tidy packaged files.

This option would probably be False by default, so add-ons that want the "old" behaviour would need to be updated, but that's still better than the old behaviour not being available at all.

Would be much appreciated!

I wonder, could we get an option added to the purge operator (bpy.ops.outliner.orphans_purge) that ignores the fake user setting on linked data? Then we could have both behaviours sortof existing alongside each other. I think we're gonna need something like that at the end of the current open movie production, when we try to make nice and tidy packaged files. This option would probably be False by default, so add-ons that want the "old" behaviour would need to be updated, but that's still better than the old behaviour not being available at all. Would be much appreciated!
Author
Owner

@mets need to create a task about proper fix for this topic actually, don't want to ship 3.6 with same half-broken situation. :(

@mets need to create a task about proper fix for this topic actually, don't want to ship 3.6 with same half-broken situation. :(
Author
Owner

Done as #106321.

Done as #106321.
Sign in to join this conversation.
No Label
Interest
Alembic
Interest
Animation & Rigging
Interest
Asset Browser
Interest
Asset Browser Project Overview
Interest
Audio
Interest
Automated Testing
Interest
Blender Asset Bundle
Interest
BlendFile
Interest
Collada
Interest
Compatibility
Interest
Compositing
Interest
Core
Interest
Cycles
Interest
Dependency Graph
Interest
Development Management
Interest
EEVEE
Interest
EEVEE & Viewport
Interest
Freestyle
Interest
Geometry Nodes
Interest
Grease Pencil
Interest
ID Management
Interest
Images & Movies
Interest
Import Export
Interest
Line Art
Interest
Masking
Interest
Metal
Interest
Modeling
Interest
Modifiers
Interest
Motion Tracking
Interest
Nodes & Physics
Interest
OpenGL
Interest
Overlay
Interest
Overrides
Interest
Performance
Interest
Physics
Interest
Pipeline, Assets & IO
Interest
Platforms, Builds & Tests
Interest
Python API
Interest
Render & Cycles
Interest
Render Pipeline
Interest
Sculpt, Paint & Texture
Interest
Text Editor
Interest
Translations
Interest
Triaging
Interest
Undo
Interest
USD
Interest
User Interface
Interest
UV Editing
Interest
VFX & Video
Interest
Video Sequencer
Interest
Virtual Reality
Interest
Vulkan
Interest
Wayland
Interest
Workbench
Interest: X11
Legacy
Blender 2.8 Project
Legacy
Milestone 1: Basic, Local Asset Browser
Legacy
OpenGL Error
Meta
Good First Issue
Meta
Papercut
Meta
Retrospective
Meta
Security
Module
Animation & Rigging
Module
Core
Module
Development Management
Module
EEVEE & Viewport
Module
Grease Pencil
Module
Modeling
Module
Nodes & Physics
Module
Pipeline, Assets & IO
Module
Platforms, Builds & Tests
Module
Python API
Module
Render & Cycles
Module
Sculpt, Paint & Texture
Module
Triaging
Module
User Interface
Module
VFX & Video
Platform
FreeBSD
Platform
Linux
Platform
macOS
Platform
Windows
Priority
High
Priority
Low
Priority
Normal
Priority
Unbreak Now!
Status
Archived
Status
Confirmed
Status
Duplicate
Status
Needs Info from Developers
Status
Needs Information from User
Status
Needs Triage
Status
Resolved
Type
Bug
Type
Design
Type
Known Issue
Type
Patch
Type
Report
Type
To Do
No Milestone
No project
No Assignees
5 Participants
Notifications
Due Date
The due date is invalid or out of range. Please use the format 'yyyy-mm-dd'.

No due date set.

Dependencies

No dependencies set.

Reference: blender/blender#105786
No description provided.