[Bf-taskforce25] List of problems/bugs with 2.5 UI re aramture animatio

[Benoit Bolsee]

Hi,

On Fri Jul 24 18:48:10 CEST 2009, Brecht Van Lommel wrote:
>> 2) Context is volatile
>Fixed.
>> 4) Changing the name of a Bone in pose mode ruins the pose channel
>Fixed.
>> 5) Set parent to Armature Deform ruins the vertex group
>Fixed.

Wow, that's quick. Thanks!

>> 7) changing a target position ruins the cache
>> - Dragging the target of a pose channel constraint causes the point
>> cache of the armature to be cleared. It ruins the simulation of the
new
>> IK solver, which uses a point cache to store the previous armature
>> position (the new solver is stateful). Why is it necessary to clear
the
>> point cache when dragging an object? It wasn't the case in 2.4.
>
>This is actually expected behavior? The point cache is supposed to be
>invalidated the moment something that it depends on changes. The point
>cache system could be made more fine grained, for example to detect
that
>only after a particular frame something changed, but this needs fairly
>major changes to the dependency graph code.
>
>Maybe it is clearing too much in this case, but I couldn't find what
>changed here compared to 2.4x?

Still, the exact same operation doesn't clear the cache in 2.49. I
didn't trace down the call stack in 2.49 to see where is the difference.


Anyway I understand the necessity to clear the cache since the change of
target position will have an effect not only on the subsequent frames
but also on the previous frames because of the interpolation. 
Though it's a bit disturbing when building an animation with the new IK
system since you don't get the correct pose as soon as you change a
target: instead of computing the new pose from the previous pose, it
must start from the rest pose. The only way to see the real effect of
the target change is to run the animation from the start. 
To fix this I may have to put some more refinement in the IK cache
system. Will think about it.

/Benoit