Blender Git Loki

Reverting commit 11974 as the setting in question, apparently causes major performance bottlenecks.

Bugfix #6923:

NLA was calculating extensions of strips in the wrong order. This was not an issue in many cases, although it is highlighted in the following case:

[strip 1 - 'hold' is on] ----------------------------
[strip 2 ]

Logically, the end of strip 1 would hold on until the start of strip 2. However, the start of strip 2 was extended to the end of strip 1 instead.

Patch #6794: Subdivide Multi for Armatures

This patch, by Juho Vepsalainen (BeBraw), introduces subdivide multi functionality for armatures. It lets you specify the number of divisions that selected bones should be divided into.

I've slightly optimised the code a bit, though the change shouldn't make much of a difference. I've also fixed a minor bug in the menu highlighting, due to duplicate menu event-codes.

-> Duplicate tool

Shift-D, Duplicate tool in editmode ported over.
This should work for mixed selections as well when
multiselect is fixed later....

-> Selection mode switching conversions

Switching between selection modes now converts selections. When
moving to a higher selection mode, selections are flushed 'upwards'
and when moving to lower selection modes selections are flushed 'downwards'.
Unlike old editmesh, selection modes are actually restrictive and elements not
represented by current mode cannot be selected.

Multiselect will take some thought with regards to flushing rules and how tools
should behave.....

== IPO Copy/Paste ==

When in EditMode for IPO-curves, keyframes are now pasted from the IPO-editor's copy/paste buffer instead of the entire curves being pasted. This makes it possible to 'move' keyframes from one IPO-curve to another.

* Only keyframes in the copy/paste buffer that are selected, are pasted
* All keyframes that are pasted, are pasted relative to the current frame, with the current frame being the location of the first pasted keyframe.
* Pasted keyframes replace exisitng keyframes if they occur at the same location.

update from stable

Materials setting "TraShad" is now on by default for all newly created materials.

fbx update from stable (parent/child hierarchy isnt flattened anymore)

* Tweaked the knife tool header info text, making mention of the previously hidden ability to use MMB to constrain to screen axes.

=Epydoc Fix=

epydoc changes for last commit weren't entirely clear. Reworked a paragraph
to be a bit clearer.

Fix compile warning.
The function newVectorObject is declare in the vector.h.

=Python Bugfix=
The python wrapper code for shape keys was really bad; whoever wrote it
(mis)read the wrong section of blender's codebase and got the totally wrong
idea. The code was definitely broken to the point where either it had to be
fixed for 2.45, or else the entire keyblock wrapper would have to be removed
from the stable branch. The fact that it didn't crash is just sheer luck;
the code assume mesh keys were MVerts, when in fact mesh keys are just
arrays of three-float vectors.

So shapekey data can now be editing directly, and is exposed as Mathutils.Vectors.
Also I updated the epydocs to explain how it all works now.

update from mingus

-> (de)select all

AKey now selects/deselects all in editmode for meshes.
Fixed record keeping of countall() for keeping track of
selected elements.

matrix to scale fixes from stable

* QMC Raytracing

This introduces QMC sampling for use in glossy reflections/refractions, soft raytraced shadows, and ambient occlusion.

This work includes many new features and speed-ups, so check out the nice docs here:

Glossy Reflection/Refraction
http://www.blender.org/development/current-projects/changes-since-244/glossy-reflectionrefraction/

Raytraced Soft Shadows
http://www.blender.org/development/current-projects/changes-since-244/raytraced-soft-shadows/

QMC Sampling
http://www.blender.org/development/current-projects/changes-since-244/qmc-sampling/

Many thanks to Brecht van Lommel for some initial code snippets and for reviewing the patch, and especially to Alfredo de Greef who gave me a lot of guidance and help along the way!

When Optimize keyframes was enabled, the comma's wernt written in the right place, oddly enough this still loaded in the FBX SDK.

-> Edge and Vertex Extrude

Added edge and vertex extrude (non-manifold). Dependant
upon selection mode.

Face extrude will follow along with a fix for face selection mode.

[#7299] Orbit around selected causes panning to be reversed

The actual "bug" is much more general than simple viewmove. In fact, any time the center (for viewmove, or transform, or ...) on which initgrabz is called was behind the camera (in perspective, then), all mouse motion where reversed.

What I added is a special handling case that reverts those situation to the default case (center = viewport offset).

This changes the behavior for those case to something much more predictable/useable, but I doubt anyone expected it to work incorrectly, so I'd say that's alright.

This covers other cases than transform and viewmove (which are the only ones I really tested), but I don't expect breakage elsewhere.

If anyone disagrees with the change, feel free to offer a better solution.