Fix T46599: Copy Rotation behaves erratically when Use Y is disabled
authorJoshua Leung <aligorith@gmail.com>
Mon, 26 Oct 2015 02:14:02 +0000 (15:14 +1300)
committerJoshua Leung <aligorith@gmail.com>
Mon, 26 Oct 2015 07:18:11 +0000 (20:18 +1300)
When the "Use Y" option in the Copy Rotation constraint is disabled, the constraint
behaves eratically when rotating all the target on all axes at the same time.
This is partially to be expected due to the way that euler rotations work
(i.e. the rotation orders stuff - you should use a rotation order based on most to
least important/significant rotations). Hence, by locking Y, you're causing accuracy
problems for Z.

What was not expected though was that changing the rotation orders on the objects
involved (for the record, it's the constraint owner that counts) did nothing.
It turns out that for objects, the rotation order settings were getting ignored!
This commit fixes this problem, and this particular case can be resolved by using
"XZY".

Notes:
* Since all object constraints were previously working on the assumption that they
  used XYZ (default) order, it is possible that this change may have the unintended
  consequence of changing the behaviour of some rigs which relied on the buggy
  behaviour. Hopefully this will be a rare occurrence.

source/blender/blenkernel/intern/constraint.c

index 906a9e4e48427a01eafdd0432c6cd0a3e6b918c0..85383cc8cf141e070773c4804e26f1ab07768f86 100644 (file)
@@ -133,7 +133,20 @@ bConstraintOb *BKE_constraints_make_evalob(Scene *scene, Object *ob, void *subda
                        if (ob) {
                                cob->ob = ob;
                                cob->type = datatype;
-                               cob->rotOrder = EULER_ORDER_DEFAULT; // TODO: when objects have rotation order too, use that
+                               
+                               if (cob->ob->rotmode > 0) {
+                                       /* Should be some kind of Euler order, so use it */
+                                       /* NOTE: Versions <= 2.76 assumed that "default" order
+                                        *       would always get used, so we may seem some rig
+                                        *       breakage as a result. However, this change here
+                                        *       is needed to fix T46599
+                                        */
+                                       cob->rotOrder = ob->rotmode;
+                               }
+                               else {
+                                       /* Quats/Axis-Angle, so Eulers should just use default order */
+                                       cob->rotOrder = EULER_ORDER_DEFAULT;
+                               }
                                copy_m4_m4(cob->matrix, ob->obmat);
                        }
                        else