Blender 2.80 Crashes in Rigid Body World Sim #59607

New Issue

Matthias babisch · 2018-12-19T09:12:41+01:00

Matthias babisch commented

2018-12-19 09:12:41 +01:00

System Information
Operating system: Debian Buster
Graphics card: NVIDIA Corporation GP106 [GeForce GTX 1060 6GB] (rev a1)

Blender Version
Broken: blender-2.80-d46d8e831ce-linux-glibc224-x86_64

Short description of error
Segmentation Fault occured not in an obvious repeatable way. Got a Backtrace from gdb.

Exact steps for others to reproduce the error
I am playing around with Physics and Scripting. Dont know how to repeat the crash. It occured when running a simulation of a few stacked boxes in or before simulating the first frame.

[New Thread 0x7fff98cb6700 (LWP 7540)]

Thread 47 "blender" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffd3960700 (LWP 31163)]
0x0000000001d5e337 in btDbvtBroadphase::destroyProxy(btBroadphaseProxy*, btDispatcher*) ()
(gdb) bt

0 0x0000000001d5e337 in btDbvtBroadphase::destroyProxy(btBroadphaseProxy*, btDispatcher*) ()
1 0x0000000001d6cb02 in btCollisionWorld::~btCollisionWorld() ()
2 0x0000000001dc7a29 in btDiscreteDynamicsWorld::~btDiscreteDynamicsWorld() ()
3 0x00000000019da0f2 in RB_dworld_delete ()
4 0x0000000001971456 in BKE_rigidbody_validate_sim_world ()
5 0x00000000019725c0 in rigidbody_update_simulation ()
6 0x000000000197290b in BKE_rigidbody_rebuild_world ()
7 0x0000000001bade4c in DEG::deg_task_run_func(TaskPool*, void*, int) ()
8 0x0000000001b90310 in task_scheduler_thread_run ()
9 0x00007ffff7f7bfa3 in start_thread (arg=) at pthread_create.c:486
10 0x00007ffff723688f in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95

**System Information** Operating system: Debian Buster Graphics card: NVIDIA Corporation GP106 [GeForce GTX 1060 6GB] (rev a1) **Blender Version** Broken: blender-2.80-d46d8e831ce-linux-glibc224-x86_64 **Short description of error** Segmentation Fault occured not in an obvious repeatable way. Got a Backtrace from gdb. **Exact steps for others to reproduce the error** I am playing around with Physics and Scripting. Dont know how to repeat the crash. It occured when running a simulation of a few stacked boxes in or before simulating the first frame. [New Thread 0x7fff98cb6700 (LWP 7540)] Thread 47 "blender" received signal SIGSEGV, Segmentation fault. [Switching to Thread 0x7fffd3960700 (LWP 31163)] 0x0000000001d5e337 in btDbvtBroadphase::destroyProxy(btBroadphaseProxy*, btDispatcher*) () (gdb) bt - 0 0x0000000001d5e337 in btDbvtBroadphase::destroyProxy(btBroadphaseProxy*, btDispatcher*) () - 1 0x0000000001d6cb02 in btCollisionWorld::~btCollisionWorld() () - 2 0x0000000001dc7a29 in btDiscreteDynamicsWorld::~btDiscreteDynamicsWorld() () - 3 0x00000000019da0f2 in RB_dworld_delete () - 4 0x0000000001971456 in BKE_rigidbody_validate_sim_world () - 5 0x00000000019725c0 in rigidbody_update_simulation () - 6 0x000000000197290b in BKE_rigidbody_rebuild_world () - 7 0x0000000001bade4c in DEG::deg_task_run_func(TaskPool*, void*, int) () - 8 0x0000000001b90310 in task_scheduler_thread_run () - 9 0x00007ffff7f7bfa3 in start_thread (arg=<optimized out>) at pthread_create.c:486 - 10 0x00007ffff723688f in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95

Matthias babisch commented

2018-12-19 09:12:41 +01:00

Added subscriber: @mbabisch

Jacques Lucke commented

2018-12-19 10:47:05 +01:00

Added subscriber: @JacquesLucke

Jacques Lucke commented

2018-12-19 10:47:05 +01:00

Can you share the .blend file you used?
Does it happen every time you simulate these boxes?

Can you share the .blend file you used? Does it happen every time you simulate these boxes?

Matthias babisch commented

2018-12-19 12:20:23 +01:00

construction.blend

I found no clear pattern of crash. It does not happen every time. The probabiity for a crash seemed to go up the longer i play around.
Creating, simulating and starting again. I suggest you loop that and see if you can find it.

I let this Blender run with gdb as soon as then instabilities where apparent. Thats where i got the Backtrace from.

I will upload the blend file from a zfs-snapshot. It is certain that i made changes to it without saving when it crashed.
Perhaps it helps you anyway. It contains a script called "Generator" which tries to build a tower of
bricks. (Which is more like walls in this state.) It removes all Objects in a collection and then creates objects in that collection.

I needed to create a Script "Clear" to remove the objects and make the file smaller (for upload).

I noticed the blender crash log in /tmp just now. It contains a backtrace too, but the file is newer and contains the backtrace of a newer crash. (8:46) I am not sure its connected and in what phase this crash happened. Since it seems similar i will paste it here:

bpy.ops.object.select_all(action='DESELECT') # Operator
bpy.ops.rigidbody.objects_add() # Operator
bpy.ops.object.select_all(action='DESELECT') # Operator
bpy.ops.rigidbody.objects_add() # Operator
bpy.ops.object.select_all(action='DESELECT') # Operator
bpy.ops.rigidbody.objects_add() # Operator
bpy.ops.text.run_script() # Operator

backtrace

/opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender(BLI_system_backtrace+0x1d) [0x1b8ffad]
/opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender() [0x1153c09]
/lib/x86_64-linux-gnu/libc.so.6(+0x378e0) [0x7f439837d8e0]
/opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender(ZNK16rbFilterCallback23needBroadphaseCollisionEP17btBroadphaseProxyS1+0x2a) [0x19ded1a]
/opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender(ZN28btHashedOverlappingPairCache18addOverlappingPairEP17btBroadphaseProxyS1+0x26) [0x1d696d6]
/opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender(ZN16btDbvtBroadphase11createProxyERK9btVector3S2_iPvssP12btDispatcherS3+0x2dc) [0x1d5f73c]
/opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender(_ZN16btCollisionWorld18addCollisionObjectEP17btCollisionObjectss+0xdc) [0x1d6c9dc]
/opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender() [0x1970db3]
/opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender() [0x19725d7]
/opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender(BKE_rigidbody_do_simulation+0x1aa) [0x1972aea]
/opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender() [0x1bade4c]
/opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender() [0x1b90351]
/lib/x86_64-linux-gnu/libpthread.so.0(+0x7fa3) [0x7f4399184fa3]
/lib/x86_64-linux-gnu/libc.so.6(clone+0x3f) [0x7f439843f88f]

Please note I am new to blender scripting and python. The scripts are quite horrible right now ?

[construction.blend](https://archive.blender.org/developer/F6005289/construction.blend) I found no clear pattern of crash. It does not happen every time. The probabiity for a crash seemed to go up the longer i play around. Creating, simulating and starting again. I suggest you loop that and see if you can find it. I let this Blender run with gdb as soon as then instabilities where apparent. Thats where i got the Backtrace from. I will upload the blend file from a zfs-snapshot. It is certain that i made changes to it without saving when it crashed. Perhaps it helps you anyway. It contains a script called "Generator" which tries to build a tower of bricks. (Which is more like walls in this state.) It removes all Objects in a collection and then creates objects in that collection. I needed to create a Script "Clear" to remove the objects and make the file smaller (for upload). I noticed the blender crash log in /tmp just now. It contains a backtrace too, but the file is newer and contains the backtrace of a newer crash. (8:46) I am not sure its connected and in what phase this crash happened. Since it seems similar i will paste it here: bpy.ops.object.select_all(action='DESELECT') # Operator bpy.ops.rigidbody.objects_add() # Operator bpy.ops.object.select_all(action='DESELECT') # Operator bpy.ops.rigidbody.objects_add() # Operator bpy.ops.object.select_all(action='DESELECT') # Operator bpy.ops.rigidbody.objects_add() # Operator bpy.ops.text.run_script() # Operator # backtrace /opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender(BLI_system_backtrace+0x1d) [0x1b8ffad] /opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender() [0x1153c09] /lib/x86_64-linux-gnu/libc.so.6(+0x378e0) [0x7f439837d8e0] /opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender(_ZNK16rbFilterCallback23needBroadphaseCollisionEP17btBroadphaseProxyS1_+0x2a) [0x19ded1a] /opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender(_ZN28btHashedOverlappingPairCache18addOverlappingPairEP17btBroadphaseProxyS1_+0x26) [0x1d696d6] /opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender(_ZN16btDbvtBroadphase11createProxyERK9btVector3S2_iPvssP12btDispatcherS3_+0x2dc) [0x1d5f73c] /opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender(_ZN16btCollisionWorld18addCollisionObjectEP17btCollisionObjectss+0xdc) [0x1d6c9dc] /opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender() [0x1970db3] /opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender() [0x19725d7] /opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender(BKE_rigidbody_do_simulation+0x1aa) [0x1972aea] /opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender() [0x1bade4c] /opt/blender-2.80b/blender-2.80-d46d8e831ce-linux-glibc224-x86_64/blender() [0x1b90351] /lib/x86_64-linux-gnu/libpthread.so.0(+0x7fa3) [0x7f4399184fa3] /lib/x86_64-linux-gnu/libc.so.6(clone+0x3f) [0x7f439843f88f] Please note I am new to blender scripting and python. The scripts are quite horrible right now ?

Jacques Lucke commented

2018-12-19 12:26:24 +01:00

Did you ever run the Generator script? I wonder how you got the switch statement working which Python does not even have.

Did you ever run the `Generator` script? I wonder how you got the switch statement working which Python does not even have.

Matthias babisch commented

2018-12-19 12:31:38 +01:00

Oh well, my bad. The snapshot contained the save from yesterday. Which as you noted, obviously did not work. I must have fixed it without saving when it crashed.

Try this one:
import bpy
import mathutils
import random
from math import *

bpy.ops.rigidbody.world_add()

count=500

width=10
height=10

boxxs=0.8
boxys=0.4
boxzs=0.4

PI=3.141

for obj in bpy.data.collections['Generated-Blocks'].objects:

   bpy.data.objects.remove(obj)

baseoffset = mathutils.Vector((0,0,0))
upvector=mathutils.Vector((0,0,boxzs))
for side in range(1,2):

  sideoffset=mathutils.Vector((0,0,0))
  runvector=mathutils.Vector((0,0,0))
  rotationvector=mathutils.Vector((0,0,0))
  if side == 1: 
      sideoffset = mathutils.Vector((boxxs*width/2,boxxs*width/2,0.1)) + baseoffset
      runvector = mathutils.Vector((boxxs,0,0))
      rotationvector=mathutils.Vector((0,0,PI/2))
  elif side == 2:
      sideoffset = mathutils.Vector((-boxxs*width/2,boxxs*width/2,0.1)) + baseoffset
      runvector = mathutils.Vector((-boxxs,0,0))
      rotationvector=mathutils.Vector((0,0,0))
  elif side == 3: 
      sideoffset = mathutils.Vector((boxxs*width/2,-boxxs*width/2,0.1)) + baseoffset
      runvector = mathutils.Vector((0,boxxs,0))
      rotationvector=mathutils.Vector((0,0,PI/2))
  elif side == 4:
      sideoffset = mathutils.Vector((boxxs*width/2,-boxxs*width/2,0.1)) + baseoffset
      runvector = mathutils.Vector((0,-boxxs,0))
      rotationvector=mathutils.Vector((0,0,0))
  else:
      raise Exception('Invalid Case')

  for h in range(1,height):
      for w in range(1,width):
          source_obj=random.choice(bpy.data.collections['Source-Blocks'].objects)
          new_obj = source_obj.copy()
          new_obj.data=source_obj.data.copy()
          new_obj.animation_data_clear()
          new_obj.location=(0,0,0)
          new_obj.rotation_euler=rotationvector
          ol=sideoffset+w*runvector+h*upvector
          new_obj.location=ol
          bpy.data.collections['Generated-Blocks'].objects.link(new_obj)
          bpy.ops.object.select_all(action='DESELECT')
          new_obj.select_set(state=True)
          bpy.context.view_layer.objects.active=new_obj
          bpy.ops.rigidbody.objects_add()
          bpy.context.object.rigid_body.collision_shape = 'BOX'
          bpy.context.object.rigid_body.use_margin = True
          bpy.context.object.rigid_body.collision_margin = 0.001

Oh well, my bad. The snapshot contained the save from yesterday. Which as you noted, obviously did not work. I must have fixed it without saving when it crashed. Try this one: import bpy import mathutils import random from math import * # bpy.ops.rigidbody.world_add() count=500 width=10 height=10 boxxs=0.8 boxys=0.4 boxzs=0.4 PI=3.141 for obj in bpy.data.collections['Generated-Blocks'].objects: ``` bpy.data.objects.remove(obj) ``` baseoffset = mathutils.Vector((0,0,0)) upvector=mathutils.Vector((0,0,boxzs)) for side in range(1,2): ``` sideoffset=mathutils.Vector((0,0,0)) runvector=mathutils.Vector((0,0,0)) rotationvector=mathutils.Vector((0,0,0)) if side == 1: sideoffset = mathutils.Vector((boxxs*width/2,boxxs*width/2,0.1)) + baseoffset runvector = mathutils.Vector((boxxs,0,0)) rotationvector=mathutils.Vector((0,0,PI/2)) elif side == 2: sideoffset = mathutils.Vector((-boxxs*width/2,boxxs*width/2,0.1)) + baseoffset runvector = mathutils.Vector((-boxxs,0,0)) rotationvector=mathutils.Vector((0,0,0)) elif side == 3: sideoffset = mathutils.Vector((boxxs*width/2,-boxxs*width/2,0.1)) + baseoffset runvector = mathutils.Vector((0,boxxs,0)) rotationvector=mathutils.Vector((0,0,PI/2)) elif side == 4: sideoffset = mathutils.Vector((boxxs*width/2,-boxxs*width/2,0.1)) + baseoffset runvector = mathutils.Vector((0,-boxxs,0)) rotationvector=mathutils.Vector((0,0,0)) else: raise Exception('Invalid Case') ``` ``` for h in range(1,height): for w in range(1,width): source_obj=random.choice(bpy.data.collections['Source-Blocks'].objects) new_obj = source_obj.copy() new_obj.data=source_obj.data.copy() new_obj.animation_data_clear() new_obj.location=(0,0,0) new_obj.rotation_euler=rotationvector ol=sideoffset+w*runvector+h*upvector new_obj.location=ol bpy.data.collections['Generated-Blocks'].objects.link(new_obj) bpy.ops.object.select_all(action='DESELECT') new_obj.select_set(state=True) bpy.context.view_layer.objects.active=new_obj bpy.ops.rigidbody.objects_add() bpy.context.object.rigid_body.collision_shape = 'BOX' bpy.context.object.rigid_body.use_margin = True bpy.context.object.rigid_body.collision_margin = 0.001 ```

Jacques Lucke commented

2019-01-15 12:13:13 +01:00

@mbabisch: Please bring the script in a form that allows me to easily load it in Blender. Either upload it as .py or preferably use the proper markup to show display code in a post here.
Also please check if you are still experiencing the bug the newest version of Blender.

@mbabisch: Please bring the script in a form that allows me to easily load it in Blender. Either upload it as .py or preferably use the proper markup to show display code in a post here. Also please check if you are still experiencing the bug the newest version of Blender.

Jacques Lucke commented

2019-03-07 15:03:15 +01:00

Changed status from 'Open' to: 'Archived'

Jacques Lucke closed this issue

2019-03-07 15:03:15 +01:00

Jacques Lucke self-assigned this 2019-03-07 15:03:15 +01:00

Jacques Lucke commented

2019-03-07 15:03:15 +01:00

More than a week without reply or activity. Due to the policy of the tracker archiving for until required info/data are provided.

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

Blender 2.80 Crashes in Rigid Body World Sim #59607

backtrace

bpy.ops.rigidbody.world_add()