For replace strips, there is a rotation interpolation direction problem. (A character doing a Back Flip blended into a Back-Handspring action clip, video doesn't show this issue) Should we always interpolate along the shortest angle? Is there ever a case where the longer angle is desirable? This is complicated even more since the two strips are animating through the blend which means that even the shortest rotation direction may change over the blend. This problem should only apply to Replace strips. They define absolute poses while all the other blend modes define offsets.

Address feedback notes
Renamed StoredKeyframes to KeyframesCo so its more obvious what's stored.

• address feedback notes
• new keying pattern extracted to a class. Now key insertion appears similar to before the patch (easier to understand)

The patch affects the DSE selection behavior for all keys (masks, actions, grease pencil). IIRC, it was for consistency between the action editor modes.

A few notes.

Changed my mind. Its a one line patch to fix a typo. I don't think its necessary to spend more time improving the description.

• rebase
• apply review notes
• use BLI_assert_msg() instead of BLI_assert()

I agree to insert keys with the current value, not the animated one.

todo: rebase and improve description. Do problem/solution and why this solution is best.. Add media so its easier to understand at a glance.

TODO: rebase, add media, check that the bug still exists and the patch still works

TODO: rebase and check that it still works or whether it has been fixed already.

TODO: show media and change to design task. Make a patch for the simpler alternative that just fixes the issue that does not require further design considerations.

TODO: rebase, add media to description

TODO: improve description to explicitly explain the problems and why the solution is better.

Not ready for review until D10504: NLA: Keyframe Remap Through Upper Strips has been accepted.

commit: rB63d2980efa2f: NLA: Remove Hold resetting between Hold_Forward Behavior
Thanks Sybren and Ramil

There's a workaround for the core issue presented in the description: For these problematic channels, Blender already allows you to create action groups with useful names. In practice, I always do so anyways. When the collection is modified and the animation breaks, we just have to use a small python script to update fcurve datapaths to sync with the action group name. Of course if you change collection element names and change their order, then .. GGs. But, again, in practice, this hasn't been a problem.

Reason: Will be included after the Action Bake rewrite.

Reason: Will be included after the Action Bake rewrite.

Reason: Will be included after the Action Bake rewrite.

In D14534#390206, @Thibault de Villèle (thibaulltt) wrote:

we'll have to see if it doesn't make more sense to also have an option to crop to a whole number of repats, to have action extrapolation fill in the blank at the end.

• For BKE_object_apply_mat4(), send use_compat=False so the rotation is predictable.

Thanks for making the proposal.

Abandon reason: patch combined into D10504: NLA: Keyframe Remap Through Upper Strips since this one is a bit useless otherwise.

The old TODO's for failure on keyframe remapping for quaternions (cases 4 and 5) will no
longer be part of this patch. I'm not sure why I previously thought they should be. Those issues
existed beforehand already and can be fixed more cleanly in a different patch.

Not ready for review until D10101: Cleanup: Nla: Refactor nla.c functions is accepted.

rebase

• rebase
• upload visual examples of patch

rebase

@Bastien Montagne (mont29) This diff fixes the issue but I don't know the library override code base well enough to know if everything is alright. Feel free to commit if its good enough.

It will be an update to original parent menu, replacing the 3 parenting options (Parent, Parent (Keep Transform), Parent (Without Inverse)) with the new two options.

Can some artists give this a try? @Brad Clark (RiggingDojo)

Rebase

I'm not running into that issue. Can you send the blend file?

Alright, I'll get to it by the end of the week.

Thanks Philipp.

• readded use of has_object_motion_paths() and pose variant for efficiency when many armatures are visible. To set the recalc flag means we iterate through all bones of that armature, see (anim_motion_paths.c) animviz_get_object_motionpaths(). So now, we only set the recalc flag if the armature has motion paths.
• cleanup copy-pasted comments.

As to whether we should add an Extrapolate Backward, given that I have not found a use case, I don't think it would improve anything for anyone. Though, I'd be happy to add the support after being given a concrete practical use case.

How about placing the properties in the (N) panel under the tab "Action"? It could be shown for the active channel's action. That way it can be used in all* other editors, including the Graph Editor.

Thanks for taking another look Sybren and Met.

Thanks sybren