[Bf-committers] About Pose evaluation of child bones in armatures

[Bastien Montagne]

Ok, so here is a patch addressing the issues mentioned above (I choose 
the second solution, but rather using the first one is trivial change): 
http://projects.blender.org/tracker/download.php/9/498/29461/18881/bt29461_armature_pchan_pose_mode.patch

Basically, I factorized into a new function (pchan_to_pose_mat) the 
compute of two matrices, one to compute rot/scale of pchan in parent 
space, the other to do the same for location. That two matrices differ 
when either HINGE/NO_SCALE/NO_LOCAL_LOCATION options are enabled.

Then, I can use those two matrices in the pose solver 
(where_is_pose_bone) and armature_mat_pose_to_bone (used by snap to 
cursor/to grid). We probably lose a bit of performances, but imho the 
gain in readability is more than worth it!

It could probably be used by the transform code too (in 
add_pose_transdata) – but I’ve just had enough of matrices for now. ;)

Also, that new code most likely can be optimized a bit by 
power-matrix-users, and need a general cleanup/proofread/testing, but I 
did test it in quite some various situations, and it always worked for 
me so far.

Cheers,
Bastien

Le 30/12/2011 12:10, Bastien Montagne a écrit :
> Hi devs!
>
> I have spent quite some hours this week on those two bugs:
> http://projects.blender.org/tracker/index.php?func=detail&aid=27898 and
> http://projects.blender.org/tracker/?func=detail&aid=29461
>
> While searching and reading code about pose transforms (I don’t like
> matrices, and they don’t like me :/ ), I think I found some
> problems/incoherences in the way the Hinge/NoScale options are handled
> *for the position of the bone*. See also this small demo file
> (http://www.pasteall.org/blend/10564), with one child, unconnected bone,
> pose-translated, and a parent bone pose-scaled by a factor two.
>
> When you disable only Inherit Rotation, the bone moves back half of its
> Pose translation. This is because its position is evaluated fully in its
> parent rest space (instead of using parent rest rotation and parent pose
> scale).
>
> When you disable only Inherit Scale, the bone does not move. This is
> because its position is evaluated into its parent pose space (instead of
> using parent pose rotation and parent rest scale).
>
> So, imho, we should fix those incoherences… But how ? Do we always want
> to use the parent pose space for the child position (thus never moving
> the child bone when disabling/enabling the Inherit Rotation/Scale
> options) ? Or do we want to fully respect the logic, and make the
> pose-location of the child bone depend on the hinge/no scale options
> (i.e. use parent rest scale and/or rotation) ?
>
> Cheers,
> Bastien
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> http://lists.blender.org/mailman/listinfo/bf-committers