Maniphest T80495

Cache/Baking seems to be broken (RB´s with harmonic field in zero gravity)
Closed, Resolved
Actions

Assigned To
Richard Antalik (ISS)
Authored By
Dennis Fassbaender (dafassi)
Sep 5 2020, 1:55 PM
Tags
  • BF Blender
Subscribers
Dennis Fassbaender (dafassi)
Richard Antalik (ISS)

Description

System Information
Operating system: Win10
Graphics card: 1080 ti

Blender Version
Broken: (example: 2.90.0, official)

Short description of error
It seems RB Cache/Baking is broken

Exact steps for others to reproduce the error
I attach a file. All objects insider are RB´s in ZERO gravity. A Sphere in the middle uses a harmonic fields to pull them all together.
Playing back (space) works fine - but once cached - or baked everything goes wild.

There is also no way to clear frames in memory. No matter if you delete all bakes, close/reopen the file, there are always frames in memory. Could be improved, too.

AntiGravityObjects.zip1 MBDownload

Related Objects

Event Timeline

Dennis Fassbaender (dafassi) created this task.Sep 5 2020, 1:55 PM
Richard Antalik (ISS) closed this task as Archived.Sep 10 2020, 5:54 AM
Richard Antalik (ISS) claimed this task.
Richard Antalik (ISS) added a subscriber: Richard Antalik (ISS).

There are dependency cycles reported in console (Window > Toggle System Console). I think it stems from rigidbody existing on object with force field. If I remove rigidbody, simulation works correctly. You will have to create some sort of workaround for this situation.

Technically this is duplicate of T80248.

Dennis Fassbaender (dafassi) changed the task status from Archived to Resolved.Sep 10 2020, 12:40 PM

Yes that worked! Thank you!
Is there any information about that forcefield-objects cannot be rigidbodies? A hint for users would be helpful.
I solved it by duplicating the object and used one as forcefield- the other as rb ;).

Richard Antalik (ISS) added a comment.Sep 11 2020, 2:06 AM

I am not sure if or how a documentation entry would be useful. There are nearly infinite configurations where you can get into same problem, basically object can not have influence on itself.

So what I could recommend is, that when you experience incorrect simulation, just look into console to see if there is something wrong. It would be perhaps better if these errors would be visible from UI, but there may be some development needed to optimize this a bit.

Dennis Fassbaender (dafassi) added a comment.Sep 11 2020, 8:45 AM

Yes i see! And it does not make sense to have an object influencing itself.
Good tip to look into the console! Never did this when having issues like this.

One thing that should be fixed is "deleting cache" - because there always are frames kept in memory, no matter what we do.
Can you please add a "clear cache" function that makes sure the physics-cache is clear?

Richard Antalik (ISS) added a comment.Sep 14 2020, 11:56 AM

There are some bugs reported for cache clearing I think. Cache should be cleared or updated automatically. If it isn't, it is a bug, so if it isn't reported already you can create report for that.