This works by tricking the depsgraph into giving us a smaller list of objects to evaluate, with all the necessary objects + their dependencies at the start of the list.
On any complicated setup where non-object parameters need to be referred to (i.e. by drivers) to affect an object's transform, these optimisations will fail and the old (slower) method is still the best way (modify the ifdef and comment out the optimise depsgraph call to do so). However, we'll assume that these aren't too common in real productions, so things should be fine with these fixes. If there really is a need for both, then global options to control these things could follow.
Potential fix (the error would explain the gimbal case, not the normal case. Also, it doesn't explain why it worked from time to time on other platforms).
would convert from quat to matrix and back if the bone had any constraint,
but did not normalize the quat first as done in other places, giving a
sudden jump when starting transform on some bones with constraints. Two
changes:
* Normalize quaternion first.
* Only do this conversion on bones with limit rotation constraints,
instead of all bones with any constraint.
Made 'PoseLib', 'Pose Paste', and 'Transforms' use the active KeyingSet instead of a hardcoded one if there is an active KeyingSet and the 'Only Insert for Keying Set' option is enabled in the User Prefs.
Also, made sure that for transforms, the active KeyingSet is provided with the data being modified instead of having them retrieve this from the context (which may miss a few items).
---
While making the changes for pose paste, made pasting poses not destroy the existing properties on the bones if the buffer bones didn't have any properties to replace the old ones with. IMO, this seems a bit too destructive if they don't get replaced, but perhaps in some cases not removing causes some problems with bad poses?
- Rename option and flag to something more sane
- Add property to manipulator operator and set true by default
Confirm on Release can now be forced true or false per operator, in which case it won't use the default value (the user preference).
Add hidden property to be able to force confirm on release (drag immediately is a silly name) on or off.
Streamline method for adding common properties to transform operators.
- library data allows pointcache writing (hard to know how this should work long term so ifdef'd for now)
- changing the frame now updates the dupligroup objects
- BKE_ptcache_ids_from_object(), option to get the id's from duplis
note! scene_update_tagged() is called from the main() loop, and runs BKE_ptcache_quick_cache_all(), this could become a performance issue, especially with duplis, should probably not call BKE_ptcache_quick_cache_all() all the time, even when not playing back animation.
(commits 27856 by Campbell from render25 branch)
This commit removes the Old Track method (used to be found under Object -> Animation -> Track), with all existing instances of this being converted to Track To Constraints. In fact, while performing this removal, I found that this was supposed to have happened in version 2.27 already, but for some reason the options were left in, and this function managed to survive for a further decade.
I've left the tracking axes around still, since it seems some curve tools still use that. However, that usage should probably get faded out in future too?
Misc notes:
* Fixed compiling error with constaints from harkyman's Maintain Volume patch.
* Subversion of 2.52 now bumped up to .2
Only source/blender/editors/ dir, should not give errors on different platforms
Only removing: UI_*.h, ED_*.h, WM_*.h, DNA_*.h, IMB_*.h, RNA_*.h, PIL_*.h
After a few days of wrong turns and learning the finer points of RNA-type-subclassing the hard way, this commit finally presents a refactored version of the Keying Sets system (now version 2) based on some requirements from Cessen.
For a more thorough discussion of this commit, see
http://sites.google.com/site/aligorith/keyingsets_2.pdf?attredirects=0&d=1
------
The main highlight of this refactor is that relative Keying Sets have now been recoded so that Python callbacks are run to generate the Keying Set's list of paths everytime the Keying Set is used (to insert or delete keyframes), allowing complex heuristics to be used to determine whether a property gets keyframed based on the current context. These checks may include checking on selection status of related entities, or transform locks.
Built-In KeyingSets have also been recoded, and moved from C and out into Python. These are now coded as Relative Keying Sets, and can to some extent serve as basis for adding new relative Keying Sets. However, these have mostly been coded in a slightly 'modular' way which may be confusing for those not so familiar with Python in general. A usable template will be added soon for more general usage.
Keyframing settings (i.e. 'visual', 'needed') can now be specified on a per-path basis now, which is especially useful for Absolute Keying Sets, where control over this is often beneficial.
Most of the places where Auto-Keyframing is performed have been tidied up for consistency. I'm sure quite a few issues still exist there, but these I'll clean up over the next few days.
Revised the conditions under which motion paths get recalculated after transforms (when auto-keying is enabled). Now, the type of path display does not matter, but rather that the object/bone in question has any paths at all. This makes animating with these a much smoother experience.
Proper fix for this. Moving special mouse input stuff to custom callbacks (this also makes the per transform main functions a bit cleaner).
It also fixes the operator property (value) for shear and warp.
