Page MenuHome

Keyed sliders instantly jumping back
Closed, ResolvedPublic

Description

*Blender Version**
Broken: 2.80 9b80c70fc5b

Short description of error
The usual behaviour now when changing the values of a keyed value slider is that the sldier turns orange to indicate that the values don't match with the animation anymore. Afterwards they can be keyed or they will jump back to their keyed values once the current frame changes.
With some shapekeys this does not happen. Instead the value jumps back instantly when changed. Even with autokey the value will jump back when sliding, completely ignoring the value input.

In this file there is a cube and a plane. The plane has the above described bug. When changing the value of the keyed shapekeys this should be clear. The current cube or any new object likely will not have the bug but at some point the bug will appear.
I couldn't find out how or why.

Event Timeline

Jacques Lucke (JacquesLucke) triaged this task as Confirmed, Medium priority.Dec 17 2018, 3:26 PM

I can reproduce it in this example but don't reallly understand what is happening. Were you able to reproduce it in any other file as well?

What I know is that when the value is changed, the depsgraph will be triggered and it overrides the value (twice). I don't know why changing the value of one shape key triggers the reevaluation of fcurves and changing the value of another does not.

@Jacques Lucke (JacquesLucke) At a random point while setting up the simple blend file the cube got the bug as well. Trying to reproduce it didn't work unfortunately.
I couldn't pinpoint it in other files as well.

Philipp Oeser (lichtwerk) raised the priority of this task from Confirmed, Medium to Confirmed, High.Jan 25 2019, 11:33 AM

The list of this phenomenom is quite worrying...
I am not sure if these all have the same roots (so atm. there are a lot of reports flying around), but this should be looked at sooner than later I think.
I tried gathering some of the reports in T60559 (pretty sure there are others as well), but will repeat them here:
T59904, T59992, T60318, T59178, T59495, T60598, T60830, the list keeps growing...

@Brecht Van Lommel (brecht), @Sergey Sharybin (sergey), @Alexander Gavrilov (angavrilov), @Bastien Montagne (mont29) : have an idea?
(I dare setting this to High Priority -- even though not even assigned yet...)