Rigid body stops simulating when an object is selected #64710

New Issue

Luja Rajbhandari · 2019-05-16T15:54:01+02:00

Luja Rajbhandari commented

2019-05-16 15:54:01 +02:00

System Information
Operating system: Windows-10-10.0.17763 64 Bits
Graphics card: Intel(R) UHD Graphics 630 Intel 4.5.0 - Build 25.20.100.6373

Blender Version
Broken: version: 2.80 (sub 64), branch: master, commit date: 2019-05-15 23:47, hash: 02a7e57d9c
Worked: (optional)

Rigid body stops simulating when an object is selected. I really don't know if its a bug but it doesn't happen in 2.79. Also, it does not happen with other simulation like cloth, smoke, etc. I I've attached the screen recorded video below
2019-05-16 19-32-09.mp4

**System Information** Operating system: Windows-10-10.0.17763 64 Bits Graphics card: Intel(R) UHD Graphics 630 Intel 4.5.0 - Build 25.20.100.6373 **Blender Version** Broken: version: 2.80 (sub 64), branch: master, commit date: 2019-05-15 23:47, hash: `02a7e57d9c` Worked: (optional) Rigid body stops simulating when an object is selected. I really don't know if its a bug but it doesn't happen in 2.79. Also, it does not happen with other simulation like cloth, smoke, etc. I I've attached the screen recorded video below [2019-05-16 19-32-09.mp4](https://archive.blender.org/developer/F7041156/2019-05-16_19-32-09.mp4)

Luja Rajbhandari commented

2019-05-16 15:54:01 +02:00

Added subscriber: @LujaRajbhandari

Garry R. Osgood commented

2019-05-17 15:23:25 +02:00

Added subscriber: @grosgood

Garry R. Osgood commented

2019-05-17 15:23:25 +02:00

I see this too.

Running Blender in a shell with --depsgraph option shows a depsgraph update at the frame corresponding to the mouse-click select. As a matter of course, that depsgraph update exhibits USER_EDIT change sources, and copy_on_write and layer_update logical functions are placed in the update list. I surmise (but I have not established a causality chain) that transform calculations are invalidated as a side effect, disrupting a cache entry at the mouse click on frame T. By the
nature of simulations, if data are not available for frame T, then T+1, T+2... can't be calculated, giving rise to the original observation. The depsgraph evaluation time also drops dramatically. For about 100 objects in the simulation I experimented with, depsgraph update for me ran about 0.02 seconds on average before the offending mouse selection attempt. Subsequent to that, depsgraph evaluation dropped an order of magnitude to 0.002 seconds, on average. Of course, no position or rotational transformations were taking place, gravity was suspended, etc. One could see in the depsgraph dumps to the shell that transform functions were being called, but I surmise they were virtually no-ops, returning immediately upon being called. I did not step through a typical call to confirm this.

There seems to be a trivial workaround. Should I hold off selecting anything while the cache is being built, then, after a cache is wholly in place, I can select objects with the animation running and the animation doesn't freeze. The simulation has actually finished at that point and the animation is running off the cache.

I see this too. Running Blender in a shell with --depsgraph option shows a depsgraph update at the frame corresponding to the mouse-click select. As a matter of course, that depsgraph update exhibits USER_EDIT change sources, and copy_on_write and layer_update logical functions are placed in the update list. I surmise (but I have not established a causality chain) that transform calculations are invalidated as a side effect, disrupting a cache entry at the mouse click on frame T. By the nature of simulations, if data are not available for frame T, then T+1, T+2... can't be calculated, giving rise to the original observation. The depsgraph evaluation time also drops dramatically. For about 100 objects in the simulation I experimented with, depsgraph update for me ran about 0.02 seconds on average before the offending mouse selection attempt. Subsequent to that, depsgraph evaluation dropped an order of magnitude to 0.002 seconds, on average. Of course, no position or rotational transformations were taking place, gravity was suspended, etc. One could see in the depsgraph dumps to the shell that transform functions were being called, but I surmise they were virtually no-ops, returning immediately upon being called. I did not step through a typical call to confirm this. There seems to be a trivial workaround. Should I hold off selecting anything while the cache is being built, then, after a cache is wholly in place, I can select objects with the animation running and the animation doesn't freeze. The simulation has actually finished at that point and the animation is running off the cache. ```

Sergey Sharybin was assigned by Sebastian Parborg

2019-06-11 17:37:03 +02:00

blender-admin commented

2019-06-12 12:11:49 +02:00

This issue was referenced by 27441c7557

This issue was referenced by 27441c75573874c91a4fb71da993e7c7fb306a2c

Sergey Sharybin commented

2019-06-12 12:12:56 +02:00

Changed status from 'Open' to: 'Resolved'

Sergey Sharybin closed this issue

2019-06-12 12:12:56 +02:00

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

Rigid body stops simulating when an object is selected #64710