Regression: My blend file crashes when rendering in Blender 3.4 upwards, because of an unchecked NULL in the code. (stack trace provided) #103031

New Issue

Chris Monachan · 2022-12-08T16:32:59+01:00

Chris Monachan commented

2022-12-08 16:32:59 +01:00

System Information
Operating system: Windows-10-10.0.22000-SP0 64 Bits
Graphics card: Intel(R) HD Graphics 620 Intel 4.5.0 - Build 31.0.101.2111

Blender Version
Broken: version: 3.4.0, branch: blender-v3.4-release, commit date: 2022-12-06 18:46, hash: a95bf1ac01
Worked: Any version prior to this

Short description of error
When rendering a large blend file in any build from 3.4 upwards I get a crash before the render starts. I have reproduced this in a debug build and can show where the issue is. I don't know the code so I am not sure why my blend file causes blender to get in this state and unfortunately it is commercially sensitive so can't share it.

Exact steps for others to reproduce the error
I load up my blend file and hit render. It fairly quickly hits an access violation.

The call-stack is as follows:

>	blender.exe!rna_Object_indirect_only_get(Object * ob, bContext * C, PointerRNA * view_layer_ptr) Line 208	C

 	blender.exe!Object_indirect_only_get_func(Object * _self, bContext * C, PointerRNA * view_layer) Line 3365	C
 	blender.exe!BL::Object::indirect_only_get(BL::Context C, BL::ViewLayer & view_layer) Line 65142	C++
 	blender.exe!ccl::BlenderSync::sync_object(BL::Depsgraph & b_depsgraph, BL::ViewLayer & b_view_layer, BL::DepsgraphObjectInstance & b_instance, float motion_time, bool use_particle_hair, bool show_lights, ccl::BlenderObjectCulling & culling, bool * use_portal, ccl::TaskPool * geom_task_pool) Line 227	C++
 	blender.exe!ccl::BlenderSync::sync_objects(BL::Depsgraph & b_depsgraph, BL::SpaceView3D & b_v3d, float motion_time) Line 613	C++
 	blender.exe!ccl::BlenderSync::sync_data(BL::RenderSettings & b_render, BL::Depsgraph & b_depsgraph, BL::SpaceView3D & b_v3d, BL::Object & b_override, int width, int height, void * * python_thread_state) Line 285	C++
 	blender.exe!ccl::BlenderSession::render(BL::Depsgraph & b_depsgraph_) Line 401	C++
 	blender.exe!ccl::render_func(_object * __formal, _object * args) Line 243	C++

The function where we access a null pointer is:

static bool rna_Object_indirect_only_get(Object *ob, bContext *C, PointerRNA *view_layer_ptr)
{
  Base *base = find_view_layer_base_with_synced_ensure(ob, C, view_layer_ptr, NULL, NULL);
  return ((base->flag & BASE_INDIRECT_ONLY) != 0);
}

Looking through the code a little, it can be expected that the fine function may return NULL, this isn't checked before looking at the flag. I have no idea the conditions in my file which would cause it to return NULL though as I don't know the code, but there is a code-path through it where it would return NULL if you look up the stack a little.

Hope this helps in resolving the issue, it still works fine in version of Blender Prior to 3.4

**System Information** Operating system: Windows-10-10.0.22000-SP0 64 Bits Graphics card: Intel(R) HD Graphics 620 Intel 4.5.0 - Build 31.0.101.2111 **Blender Version** Broken: version: 3.4.0, branch: blender-v3.4-release, commit date: 2022-12-06 18:46, hash: `a95bf1ac01` Worked: Any version prior to this **Short description of error** When rendering a large blend file in any build from 3.4 upwards I get a crash before the render starts. I have reproduced this in a debug build and can show where the issue is. I don't know the code so I am not sure why my blend file causes blender to get in this state and unfortunately it is commercially sensitive so can't share it. **Exact steps for others to reproduce the error** I load up my blend file and hit render. It fairly quickly hits an access violation. The call-stack is as follows: ``` > blender.exe!rna_Object_indirect_only_get(Object * ob, bContext * C, PointerRNA * view_layer_ptr) Line 208 C blender.exe!Object_indirect_only_get_func(Object * _self, bContext * C, PointerRNA * view_layer) Line 3365 C blender.exe!BL::Object::indirect_only_get(BL::Context C, BL::ViewLayer & view_layer) Line 65142 C++ blender.exe!ccl::BlenderSync::sync_object(BL::Depsgraph & b_depsgraph, BL::ViewLayer & b_view_layer, BL::DepsgraphObjectInstance & b_instance, float motion_time, bool use_particle_hair, bool show_lights, ccl::BlenderObjectCulling & culling, bool * use_portal, ccl::TaskPool * geom_task_pool) Line 227 C++ blender.exe!ccl::BlenderSync::sync_objects(BL::Depsgraph & b_depsgraph, BL::SpaceView3D & b_v3d, float motion_time) Line 613 C++ blender.exe!ccl::BlenderSync::sync_data(BL::RenderSettings & b_render, BL::Depsgraph & b_depsgraph, BL::SpaceView3D & b_v3d, BL::Object & b_override, int width, int height, void * * python_thread_state) Line 285 C++ blender.exe!ccl::BlenderSession::render(BL::Depsgraph & b_depsgraph_) Line 401 C++ blender.exe!ccl::render_func(_object * __formal, _object * args) Line 243 C++ ``` The function where we access a null pointer is: ``` static bool rna_Object_indirect_only_get(Object *ob, bContext *C, PointerRNA *view_layer_ptr) { Base *base = find_view_layer_base_with_synced_ensure(ob, C, view_layer_ptr, NULL, NULL); return ((base->flag & BASE_INDIRECT_ONLY) != 0); } ``` Looking through the code a little, it can be expected that the fine function may return NULL, this isn't checked before looking at the flag. I have no idea the conditions in my file which would cause it to return NULL though as I don't know the code, but there is a code-path through it where it would return NULL if you look up the stack a little. Hope this helps in resolving the issue, it still works fine in version of Blender Prior to 3.4

Chris Monachan commented

2022-12-08 16:32:59 +01:00

Added subscriber: @Chris-Monachan

blender-admin commented

2022-12-08 16:32:59 +01:00

#103087 was marked as duplicate of this issue

blender-admin commented

2022-12-08 16:32:59 +01:00

#103010 was marked as duplicate of this issue

Iliya Katushenock commented

2022-12-08 16:55:36 +01:00

Added subscriber: @mod_moder

Iliya Katushenock commented

2022-12-08 16:55:36 +01:00

Is it in September builds? in 3.3, 3.4, 3.5?

Chris Monachan commented

2022-12-08 16:58:20 +01:00

It does not crash in 3.3.2 (2022-12-06 18:19), it does crash in the 3.4 Stable released this week, and in 3.5 (latest code from the repository which I used to build my debug build). Hope this helps.

Iliya Katushenock commented

2022-12-08 17:05:07 +01:00

It's just a test. I don't know this module, but I see recent commits. Usually bisect is made by people from the checker module, but since there is no project ...
In general here:
https://builder.blender.org/download/daily/
Open the archive (text below this week's links)
And try to find 2 versions between which it appeared.
The fastest way to jump far (for example)
https://builder.blender.org/download/daily/archive/blender-3.4.0-alpha+master.caf6225a3d01-windows.amd64-release.zip
If there are no problems, jump up, if not, a little below.
In general, I think it will take the longest to download all this, the search usually takes me no more than 10 minutes (when there are 150 downloaded versions)

It's just a test. I don't know this module, but I see recent commits. Usually bisect is made by people from the checker module, but since there is no project ... In general here: https://builder.blender.org/download/daily/ Open the archive (text below this week's links) And try to find 2 versions between which it appeared. The fastest way to jump far (for example) https://builder.blender.org/download/daily/archive/blender-3.4.0-alpha+master.caf6225a3d01-windows.amd64-release.zip If there are no problems, jump up, if not, a little below. In general, I think it will take the longest to download all this, the search usually takes me no more than 10 minutes (when there are 150 downloaded versions)

Iliya Katushenock commented

2022-12-08 17:08:22 +01:00

Potential caused by: D16004: ViewLayer: fix missing sync in object APIs (fixes #101128)
Missing check on NULL

Potential caused by: [D16004: ViewLayer: fix missing sync in object APIs (fixes #101128)](https://archive.blender.org/developer/D16004) Missing check on NULL

Iliya Katushenock commented

2022-12-08 17:09:19 +01:00

I think you should not look for a version, let's wait for the author's opinion, most likely he will be able to explain this without your project

Iliya Katushenock changed title from ~~My blend file crashes when rendering in Blender 3.4 upwards, because of an unchecked NULL in the code. (stack trace provided)~~ to Regression: My blend file crashes when rendering in Blender 3.4 upwards, because of an unchecked NULL in the code. (stack trace provided)

2022-12-08 17:12:22 +01:00

Pratik Borhade commented

2022-12-09 06:21:57 +01:00

Added subscribers: @aras_p, @PratikPB2123

Pratik Borhade commented

2022-12-09 06:21:57 +01:00

Changed status from 'Needs Triage' to: 'Needs Developer To Reproduce'

Pratik Borhade commented

2022-12-09 06:21:57 +01:00

@aras_p hi, can you take a look?
In e00f76c6a8 you'd removed the null check for *base in rna_Object_indirect_only_get

@aras_p hi, can you take a look? In e00f76c6a8 you'd removed the null check for `*base` in `rna_Object_indirect_only_get`

Aras Pranckevicius self-assigned this 2022-12-09 14:07:39 +01:00

Aras Pranckevicius commented

2022-12-09 14:07:39 +01:00

Changed status from 'Needs Developer To Reproduce' to: 'Confirmed'

Aras Pranckevicius commented

2022-12-09 14:10:22 +01:00

@Chris-Monachan yeah indeed the function seems to miss a

  if (!base) {
    return false;
  }

which was accidentally lost in e00f76c6a8 (sorry, my bad!). If you can confirm that this fixes the regression, I can commit that into the codebase properly.

@Chris-Monachan yeah indeed the function seems to miss a ``` if (!base) { return false; } ``` which was accidentally lost in e00f76c6a8 (sorry, my bad!). If you can confirm that this fixes the regression, I can commit that into the codebase properly.

Iliya Katushenock commented

2022-12-09 14:16:25 +01:00

Just in case it doesn't have a chance to test the build with this
And while doing this, you need to be sure that this is a working fix.
Can run a build bot?

Just in case it doesn't have a chance to test the build with this And while doing this, you need to be sure that this is a working fix. Can run a build bot?

Chris Monachan commented

2022-12-09 15:45:19 +01:00

@aras_p , yeah that does get me past that crash when I added that check to the function. Nicely done, thanks! It would be great if this fix could be merged back to the 3.4 branch as I use that for production work.