Remesh-bug.crash #92355

New Issue

Paolo Furani · 2021-10-19T22:18:24+02:00

Paolo Furani commented

2021-10-19 22:18:24 +02:00

System Information
Operating system: Linux-4.15.0-159-generic-x86_64-with-glibc2.27 64 Bits
Graphics card: Mesa DRI Intel(R) HD Graphics 5500 (BDW GT2) Intel Open Source Technology Center 4.6 (Core Profile) Mesa 20.0.8

Blender Version
Broken: version: 3.0.0 Alpha, branch: master, commit date: 2021-10-18 21:38, hash: d7b231baa8
Worked: (newest version of Blender that worked as expected)

Short description of error
Hi when I try to use the remesh command on this object, Blender crashes

crash-report

# Blender 3.0.0, Commit date: 2021-10-18 21:38, Hash d7b231baa87e
bpy.context.space_data.context = 'DATA'  # Property
bpy.context.object.data.remesh_mode = 'QUAD'  # Property
bpy.data.window_managers["WinMan"].(null) = False  # Property
bpy.data.window_managers["WinMan"].(null) = True  # Property
bpy.data.window_managers["WinMan"].(null) = True  # Property
bpy.ops.object.quadriflow_remesh(use_mesh_symmetry=False, use_preserve_sharp=True, use_preserve_boundary=True)  # Operator
QuadriFlow: Remeshing completed  # Info
bpy.ops.object.editmode_toggle()  # Operator
bpy.ops.object.editmode_toggle()  # Operator
bpy.ops.object.delete(use_global=False, confirm=False)  # Operator
Deleted 1 object(s)  # Info
bpy.context.space_data.recent_folders_active = 0  # Property
bpy.ops.object.hide_view_set(unselected=False)  # Operator
bpy.ops.object.select_all(action='SELECT')  # Operator
bpy.data.window_managers["WinMan"].(null) = "bool-obj"  # Property
bpy.ops.object.move_to_collection(collection_index=0, is_new=True, new_collection_name="bool-obj")  # Operator
Objects moved to bool-obj  # Info
bpy.data.scenes["Scene"].(null) = True  # Property
bpy.data.scenes["Scene"].(null) = False  # Property
bpy.ops.object.convert(target='MESH')  # Operator
bpy.context.space_data.params.filename = "remesh-bug.blend"  # Property
Saved "remesh-bug.blend"  # Info
bpy.ops.outliner.item_activate(deselect_all=True)  # Operator
bpy.ops.outliner.delete(hierarchy=True)  # Operator
Saved "remesh-bug.blend"  # Info
bpy.context.object.data.remesh_mode = 'QUAD'  # Property
bpy.ops.object.quadriflow_remesh()  # Operator

# backtrace
/home/pafu/Software/blender-3.0.0/blender(BLI_system_backtrace+0x20) [0xaf71120]
/home/pafu/Software/blender-3.0.0/blender() [0x109e04b]
/lib/x86_64-linux-gnu/libc.so.6(+0x3f040) [0x7fb4c2537040]
/home/pafu/Software/blender-3.0.0/blender(AUD_Device_getRate+0x3d) [0x9ce251d]
/home/pafu/Software/blender-3.0.0/blender(BKE_sound_scene_playing+0x20) [0x118d760]
/home/pafu/Software/blender-3.0.0/blender(wm_event_do_handlers+0xa6) [0x149b416]
/home/pafu/Software/blender-3.0.0/blender(WM_main+0x20) [0x1491620]
/home/pafu/Software/blender-3.0.0/blender(main+0x31e) [0xf875ae]
/lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xe7) [0x7fb4c2519bf7]
/home/pafu/Software/blender-3.0.0/blender() [0x109a9ac]

# Python backtrace

Exact steps for others to reproduce the error

Open .blend file
{nav Object Data Properties > Remesh > Quad > QuadriFlow Remesh}
Refer image for the settings
{F11295360}
{F11295382}

**System Information** Operating system: Linux-4.15.0-159-generic-x86_64-with-glibc2.27 64 Bits Graphics card: Mesa DRI Intel(R) HD Graphics 5500 (BDW GT2) Intel Open Source Technology Center 4.6 (Core Profile) Mesa 20.0.8 **Blender Version** Broken: version: 3.0.0 Alpha, branch: master, commit date: 2021-10-18 21:38, hash: `d7b231baa8` Worked: (newest version of Blender that worked as expected) **Short description of error** Hi when I try to use the remesh command on this object, Blender crashes **crash-report** ```lines # Blender 3.0.0, Commit date: 2021-10-18 21:38, Hash d7b231baa87e bpy.context.space_data.context = 'DATA' # Property bpy.context.object.data.remesh_mode = 'QUAD' # Property bpy.data.window_managers["WinMan"].(null) = False # Property bpy.data.window_managers["WinMan"].(null) = True # Property bpy.data.window_managers["WinMan"].(null) = True # Property bpy.ops.object.quadriflow_remesh(use_mesh_symmetry=False, use_preserve_sharp=True, use_preserve_boundary=True) # Operator QuadriFlow: Remeshing completed # Info bpy.ops.object.editmode_toggle() # Operator bpy.ops.object.editmode_toggle() # Operator bpy.ops.object.delete(use_global=False, confirm=False) # Operator Deleted 1 object(s) # Info bpy.context.space_data.recent_folders_active = 0 # Property bpy.ops.object.hide_view_set(unselected=False) # Operator bpy.ops.object.select_all(action='SELECT') # Operator bpy.data.window_managers["WinMan"].(null) = "bool-obj" # Property bpy.ops.object.move_to_collection(collection_index=0, is_new=True, new_collection_name="bool-obj") # Operator Objects moved to bool-obj # Info bpy.data.scenes["Scene"].(null) = True # Property bpy.data.scenes["Scene"].(null) = False # Property bpy.ops.object.convert(target='MESH') # Operator bpy.context.space_data.params.filename = "remesh-bug.blend" # Property Saved "remesh-bug.blend" # Info bpy.ops.outliner.item_activate(deselect_all=True) # Operator bpy.ops.outliner.delete(hierarchy=True) # Operator Saved "remesh-bug.blend" # Info bpy.context.object.data.remesh_mode = 'QUAD' # Property bpy.ops.object.quadriflow_remesh() # Operator # backtrace /home/pafu/Software/blender-3.0.0/blender(BLI_system_backtrace+0x20) [0xaf71120] /home/pafu/Software/blender-3.0.0/blender() [0x109e04b] /lib/x86_64-linux-gnu/libc.so.6(+0x3f040) [0x7fb4c2537040] /home/pafu/Software/blender-3.0.0/blender(AUD_Device_getRate+0x3d) [0x9ce251d] /home/pafu/Software/blender-3.0.0/blender(BKE_sound_scene_playing+0x20) [0x118d760] /home/pafu/Software/blender-3.0.0/blender(wm_event_do_handlers+0xa6) [0x149b416] /home/pafu/Software/blender-3.0.0/blender(WM_main+0x20) [0x1491620] /home/pafu/Software/blender-3.0.0/blender(main+0x31e) [0xf875ae] /lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xe7) [0x7fb4c2519bf7] /home/pafu/Software/blender-3.0.0/blender() [0x109a9ac] # Python backtrace ``` **Exact steps for others to reproduce the error** - Open .blend file - {nav Object Data Properties > Remesh > Quad > QuadriFlow Remesh} - Refer image for the settings {F11295360} {F11295382}

Paolo Furani commented

2021-10-19 22:18:24 +02:00

Added subscriber: @pafurijaz-4

blender-admin commented

2021-10-19 22:18:24 +02:00

#92484 was marked as duplicate of this issue

blender-admin commented

2021-10-19 22:18:24 +02:00

#92391 was marked as duplicate of this issue

Pratik Borhade commented

2021-10-20 08:16:11 +02:00

Added subscriber: @PratikPB2123

Pratik Borhade commented

2021-10-20 08:21:06 +02:00

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

Philipp Oeser commented

2021-10-20 09:27:27 +02:00

Added subscribers: @ZedDB, @Sergey, @neXyon, @lichtwerk

Philipp Oeser commented

2021-10-20 09:27:27 +02:00

Apparently we are confronted with a NULL Audio device (this might be because remesh is fiddeling with a scene in a background job [I recall something with remesh and the scene being the job owner and not the mesh object] -- but I have not checked further), so security check in sound_verify_evaluated_id is not enough it seems.

CC @ZedDB
CC @neXyon
CC @Sergey

But this also might not be the "real" problem, when doing something like the following (to avoid the crash in Audio)
P2517: T92355_snippet



diff --git a/source/blender/blenkernel/intern/sound.c b/source/blender/blenkernel/intern/sound.c
index f523c5e02bd..af61ec69535 100644
--- a/source/blender/blenkernel/intern/sound.c
+++ b/source/blender/blenkernel/intern/sound.c
@@ -994,6 +994,10 @@ int BKE_sound_scene_playing(Scene *scene)
     return -1;
   }
 
+  if (sound_device == NULL) {
+    return -1;
+  }
+
   /* In case of a "None" audio device, we have no playback information. */
   if (AUD_Device_getRate(sound_device) == AUD_RATE_INVALID) {
     return -1;

we are then getting here (which also does not sound pleasing): https://developer.blender.org/diffusion/B/browse/master/extern/quadriflow/src/subdivide.cpp$503

Apparently we are confronted with a NULL Audio device (this might be because remesh is fiddeling with a scene in a background job [I recall something with remesh and the scene being the job owner and not the mesh object] -- but I have not checked further), so security check in `sound_verify_evaluated_id` is not enough it seems. CC @ZedDB CC @neXyon CC @Sergey But this also might not be the "real" problem, when doing something like the following (to avoid the crash in Audio) [P2517: T92355_snippet](https://archive.blender.org/developer/P2517.txt) ``` diff --git a/source/blender/blenkernel/intern/sound.c b/source/blender/blenkernel/intern/sound.c index f523c5e02bd..af61ec69535 100644 --- a/source/blender/blenkernel/intern/sound.c +++ b/source/blender/blenkernel/intern/sound.c @@ -994,6 +994,10 @@ int BKE_sound_scene_playing(Scene *scene) return -1; } + if (sound_device == NULL) { + return -1; + } + /* In case of a "None" audio device, we have no playback information. */ if (AUD_Device_getRate(sound_device) == AUD_RATE_INVALID) { return -1; ``` we are then getting here (which also does not sound pleasing): https://developer.blender.org/diffusion/B/browse/master/extern/quadriflow/src/subdivide.cpp$503

Sergey Sharybin commented

2021-10-20 09:41:00 +02:00

remesh is fiddeling with a scene in a background job

Such tools are expected to localize data in the job's init function (which is run on the main thread) and do not fiddle around with data which is accessed by the main thread. If the remesh is indeed fiddeling with non-localized data it is to be addressed at a much higher priority than adding nullptr checks.

> remesh is fiddeling with a scene in a background job Such tools are expected to localize data in the job's init function (which is run on the main thread) and do not fiddle around with data which is accessed by the main thread. If the remesh is indeed fiddeling with non-localized data it is to be addressed at a much higher priority than adding nullptr checks.

Philipp Oeser commented

2021-10-20 10:25:33 +02:00

In #92355#1238493, @Sergey wrote:

remesh is fiddeling with a scene in a background job

Such tools are expected to localize data in the job's init function (which is run on the main thread) and do not fiddle around with data which is accessed by the main thread. If the remesh is indeed fiddeling with non-localized data it is to be addressed at a much higher priority than adding nullptr checks.

Sorry, this might have come across wrong, not sure quadriflow is actually altering anything scene related in the job (it was just my foggy memory of the job being owned by the scene [and not by to object -- but I also recall this had its reasons: progressbar etc]

> In #92355#1238493, @Sergey wrote: >> remesh is fiddeling with a scene in a background job > > Such tools are expected to localize data in the job's init function (which is run on the main thread) and do not fiddle around with data which is accessed by the main thread. If the remesh is indeed fiddeling with non-localized data it is to be addressed at a much higher priority than adding nullptr checks. Sorry, this might have come across wrong, not sure quadriflow is actually `altering` anything scene related in the job (it was just my foggy memory of the job being owned by the scene [and not by to object -- but I also recall this had its reasons: progressbar etc]

Joerg Mueller commented

2021-10-20 22:04:35 +02:00

Looks like a depsgraph/memory related bug where I cannot do much!?

Sebastian Parborg commented

2021-10-21 13:53:24 +02:00

The model has several zero length edges which quadriflow can't handle.
(Use "merge by distance" to remove these and it will not crash anymore)

I will have to update the mesh manifold check to check for these.

The reason why we get the backtrace to the sound system is because of race condition were instead of crashing, quadriflow calls exit() and blender frees all scene data while an other thread tries to access it.

The model has several zero length edges which quadriflow can't handle. (Use "merge by distance" to remove these and it will not crash anymore) I will have to update the mesh manifold check to check for these. The reason why we get the backtrace to the sound system is because of race condition were instead of crashing, quadriflow calls `exit()` and blender frees all scene data while an other thread tries to access it.

Paolo Furani commented

2021-10-21 14:15:30 +02:00

Hi the original model is this but with no boolean modifiers applied.
retopology-experiments.blend

Hi the original model is this but with no boolean modifiers applied. [retopology-experiments.blend](https://archive.blender.org/developer/F11361251/retopology-experiments.blend)

Paolo Furani commented

2021-10-21 14:31:57 +02:00

On Blender 2.83LST it works fine, the problem is in 293.lts and in Blender version 3.0  Below the attached file with two objects to do the tests and it works as seen from the video screen.

remesh-B2.83lts.blend

On Blender 2.83LST it works fine, the problem is in 293.lts and in Blender version 3.0  Below the attached file with two objects to do the tests and it works as seen from the video screen. ![Remesh-2.8LTS-OK.png](https://archive.blender.org/developer/F11361547/Remesh-2.8LTS-OK.png) [remesh-B2.83lts.blend](https://archive.blender.org/developer/F11361545/remesh-B2.83lts.blend)

Sebastian Parborg commented

2021-10-21 14:43:08 +02:00

The latest file you attached works in master as well (because it doesn't have zero length edges).

I'm unsure what you are trying to say.

The latest file you attached works in master as well (because it doesn't have zero length edges). I'm unsure what you are trying to say.

Paolo Furani commented

2021-10-21 14:56:44 +02:00

In the first I applied the modifiers with Blender 3.0 and in the second I imported the scene in Blender 2.83, and I applied the modifiers and I ran the command remesh, this means that in Blender 3.0 and in Blender 2.93 when you apply the modifiers both generate a corrupted mesh is .
You can try because I attached the original file from which I imported the scene from 3.0 to test with Blender 2.83LST, I don't know what the problem is but it is not that one has to remodel or redo the objects it is evident that there is a problem.

In the first I applied the modifiers with Blender 3.0 and in the second I imported the scene in Blender 2.83, and I applied the modifiers and I ran the command remesh, this means that in Blender 3.0 and in Blender 2.93 when you apply the modifiers both generate a corrupted mesh is . You can try because I attached the original file from which I imported the scene from 3.0 to test with Blender 2.83LST, I don't know what the problem is but it is not that one has to remodel or redo the objects it is evident that there is a problem.

Sebastian Parborg commented

2021-10-21 15:12:12 +02:00

Did you try the latest 3.0 version?

For me applying the modifiers doesn't not create zero length edges in master.

Did you try the latest 3.0 version? For me applying the modifiers doesn't not create zero length edges in master.

blender-admin commented

2021-10-21 15:13:55 +02:00

This issue was referenced by 77104bf318

This issue was referenced by 77104bf31891229e7d18ab56506ce6f456491eb0

blender-admin commented

2021-10-21 15:13:55 +02:00

This issue was referenced by 10fb5cc58d

This issue was referenced by 10fb5cc58d210592ef0ae410e305be906ad5ce51

Sebastian Parborg commented

2021-10-21 15:17:15 +02:00

Changed status from 'Confirmed' to: 'Resolved'

Sebastian Parborg closed this issue

2021-10-21 15:17:15 +02:00

Sebastian Parborg self-assigned this 2021-10-21 15:17:15 +02:00

Sebastian Parborg commented

2021-10-21 15:19:11 +02:00

In either case, the issue with generating invalid meshes is an other issue. So the reported crash has been "fixed", so I'll close this.

Paolo Furani commented

2021-10-21 16:38:17 +02:00

But what do you say this is a crash problem which originated from a file where I applied the modifiers and then tried to use the remesh command, it's not another problem it's a problem with Blender 3.0

Paolo Furani commented

2021-10-21 17:18:29 +02:00

In #92355#1239161, @ZedDB wrote:
Did you try the latest 3.0 version?

For me applying the modifiers doesn't not create zero length edges in master.

I've tested now and it's crash this is the file made with Blender 3.0 the other for 2.83lts has a different boolean object when the one for the 3.0 has an object made by boolean union

The original model is this but with no modifiers applied.
retopology-experiments.blend

> In #92355#1239161, @ZedDB wrote: > Did you try the latest 3.0 version? > > For me applying the modifiers doesn't not create zero length edges in master. I've tested now and it's crash this is the file made with Blender 3.0 the other for 2.83lts has a different boolean object when the one for the 3.0 has an object made by boolean union The original model is this but with no modifiers applied. [retopology-experiments.blend](https://archive.blender.org/developer/F11361251/retopology-experiments.blend)

Sebastian Parborg commented

2021-10-21 17:28:17 +02:00

I'm saying that the invalid geometry generated by the boolean modifier is an other issue.
You reported that blender crashed when quadriflow was used on bad meshes with zero length edges.

I've added a check so the operator does not run on meshes with zero edge length edges now.
So the issue for this ticket is fixed and now closed.

I'm saying that the invalid geometry generated by the boolean modifier is an other issue. You reported that blender crashed when quadriflow was used on bad meshes with zero length edges. I've added a check so the operator does not run on meshes with zero edge length edges now. So the issue for this ticket is fixed and now closed.

Paolo Furani commented

2021-10-21 17:39:19 +02:00

In #92355#1239242, @ZedDB wrote:
I'm saying that the invalid geometry generated by the boolean modifier is an other issue.
You reported that blender crashed when quadriflow was used on bad meshes with zero length edges.

I've added a check so the operator does not run on meshes with zero edge length edges now.
So the issue for this ticket is fixed and now closed.

Ok so now I've to create an other report for this issue?

> In #92355#1239242, @ZedDB wrote: > I'm saying that the invalid geometry generated by the boolean modifier is an other issue. > You reported that blender crashed when quadriflow was used on bad meshes with zero length edges. > > I've added a check so the operator does not run on meshes with zero edge length edges now. > So the issue for this ticket is fixed and now closed. Ok so now I've to create an other report for this issue?

Sebastian Parborg commented

2021-10-21 17:58:36 +02:00

If someone else hasn't already reported this issue, then you should report it as an other issue, yes.