Blender Git Loki

- displacement fix, need to calc normals before displacing

- removed internal functions from effect header
- made particle system generation use its own RNG object,
keeps it safe against varying calculation order in the renderer

Still a funny bug where having a material on a particle object
changes the particle locations (in renderer, not in interface...
bizarro)

- Dupliframes mball didn't work anymore (should be added to scene_update)
- restored nicer drawing for duplicator wireframe. Draws before selected,
so you can see the duplicated object better

- mixed up RNG type definition

- added RNG abstract random object rng_{new/free/seed/get{Int,Double,Float}}
to avoid use of global generator. at the moment the renderer owns the
number generator and this is important for retaining render consistency.

- fix some error handling in blend loader, correct error codes
were not always returned.

Some bad use of non-public API. Changed the calls to BIF_undo_push() and BIF_undo() instead

build.c + .h going
point.c + .h coming
/me wonders if there ara any MSVC6er

Little issues, combined commit!

- Bug #2857: Spin didn't create nice consistant normals
- removed unnecessary call to where_is_object() in init-render phase.
- Added DAG_scene_sort() calls when objects were removed (join cases)
- When using texture fonts, the file window header didn't display OK
- Saving a file didn't set the 'wait cursor' anymore (oldie!)

Regression file teapot.blend didn't render... this due to a hack removal of
a month ago (trying to cope with vertexcolor 'paint' and 'light'.)
Forgot to add the correct check for calculaing UV texture coords, needed to
have vertexcolor 'paint' working.

- warning fixes
- rare bug in calc_vertexnormals, fourth vertex normal was not flipped
correctly (found thanks to my new auto regression test script)

- switch displistmesh_to_mesh to just dupalloc the mface's
- added verteCos argument to sbObjectReset, lack of this was
causing softbody objects to not be initialized with deform
- made convert-to-mesh option call DAG_scene_sort, prevents
crashes due to obsolete object pointer

When entering editmode on Armature, the Pose now defaults to a correct
restposition, also for deform and bone childs.

Bug fix #2856

Changing 'active material' in Editing Buttons didn't update preview in
in Material Buttons.

NLA strip option "Add" accidentally also added constraint ipo values.
Result was characters going bezerk! By definition, pose blending now only
adds loc/rot/size, not constraint strength values.

Armature-Pose IK goodies!

- Added option to the IK buttons, to have it use the 'tip' as end of the
IK chain. I never really understood this old convention (IK didn't work
on the Bone itself).
Old files still will read OK though. But I made the "To Tip" a default
when adding new IK constraints.

- Hotkey CTRL+I: add IK, with option to have it adding an Empty target,
or use a selected Bone as target. With the new non-modal PoseMode, it
gives instant access to playing with the IK chain.

- Hotkey ALT+I: clears IK, on all selected Bones

- Hotkey ALT+C: clears Constraints on all selected Bones (incl IK)

GCC 4 didn't like static function after non static prototype. Moved prototype to c file and added static to it.

Removing "Custom" as a menu entry in 3D view header (ztonzy spotted that).

Missing include file from commits. (tsk)

Some stuff slipped in header_view3d.c too. Thanks to basse for spoting that quickly.